UNDERSTANDING ROUGH MILL YIELD THROUGH
THE ANALYSIS OF THE INTERACTION BETWEEN
LUMBER CHARACTERISTICS AND
PROCESSING PARAMETERS
A Thesis
Submitted to the Faculty
of
Purdue University
by
Charles Clement
In Partial Fulfillment of the
Requirements fo r the Degree
of
Doctor of Philosophy
May 2002
ii
I would like to dedicate this work to the two principle women in my life, my
mother, Susan Pamela Stewart Clément, and my wife, Susan Jane Eckelman Clément.
Thank you for having been there for me, for having inspired me to go on, to better myself,
and to not settle for second best. I only hope I can follow your example and let my true-
self shine.
iii
ACKNOWLEDGEMENTS
I am deeply grateful to Rado Gazo for chairing my advisor committee and
providing guidance and support. He helped me learn, not only by letting me make
mistakes, but also by focusing my attention when I got distracted.
To Robert Beauregard, my thanks are twofold. First I would like to thank him for
inviting me to participate in this project, and for facilitating my transition to unfamiliar
territory. Second I would like to thank him for sharing his expert knowledge and
experience with me.
To the members of my committee – Dan L. Cassens, and Anton Sumali – thank
you! Your advice and guidance were critical.
Ed Thomas of the USDA Forest Service deserves thanks for guiding me through
rough mill simulations, and for helping me solve problems that would occur, in the most
expedient fashion. To Torsten Lihra and the people at Forintek Canada Corp., thank you
for your support, both professional and financial. My thanks to Senco and TLB for sharing
their lumber, and Kennebec for sharing their expertise.
Finally, to my immediate family, Susan, Nicolaus, Jules, Simon, Annie…Pamela,
and friends, Eric, Eva, Francisco, Henry, Ike, Valeria; thank you for believing in me.
iv
PREFACE
This study focuses on the difference in yield between NHLA-graded white birch
lumber sawn from conventional logs and short- length logs. Conventional logs are of such a
size and have sufficiently few defects that they can be sawn into NHLA lumber. Short-
length logs are considered by many as too short and/or of too small a diameter to
effectively produce an economic yield in NHLA grade lumber, and are often classified as
pulpwood quality timber. This lumber type will be termed short-length lumber for the
purpose of this study. It should be noted that all lumber used in this study was NHLA
graded and that the comparison of lengths refers to the effects of the timber from which it
was sawn and the quality therein.
v
TABLE OF CONTENTS
Page
LIST OF TABLES ........................................................................................................... viii
LIST OF FIGURES..............................................................................................................x
LIST OF ABBREVIATIONS........................................................................................... xiii
ABSTRACT......................................................................................................................xiv
1 INTRODUCTION....................................................................................................... 1
1.1.2 Objectives ......................................................................................................xiv
2 LITERATURE REVIEW............................................................................................ 4
2.1 Species Description............................................................................................. 5 2.2 Species Distribution............................................................................................. 5 2.3 Availability of the species .................................................................................... 6 2.4 Uses..................................................................................................................... 7 2.5 Database .............................................................................................................. 7 2.6 Grading.............................................................................................................. 14 2.7 Processing.......................................................................................................... 19 2.8 Simulation Programs ......................................................................................... 22 2.9 Yield .................................................................................................................. 27 2.10 Effect of lumber length...................................................................................... 28 2.11 Summary........................................................................................................... 29
References ................................................................................................................. 31
3 WHAT IS THE YIELD OF SHORT-LENGTH WHITE BIRCH LUMBER? .......... 42
Abstract ..................................................................................................................... 42 3.1 Introduction....................................................................................................... 44 3.2 Methodology ..................................................................................................... 45
3.2.1 Sample Material............................................................................................. 45 3.2.2 Board Grading ............................................................................................... 46 3.2.3 Database ........................................................................................................ 47
vi
Page
3.2.4 Cutting order .................................................................................................. 50 3.2.5 Simulation Parameters................................................................................... 53
3.2.5.1 ROMI-RIP simulation parameters:........................................................ 53 3.2.5.2 ROMI CROSS simulation parameters: .................................................. 53
3.3 Results and Discussion...................................................................................... 54 3.3.1 Database ........................................................................................................ 54 3.3.2 Yield .............................................................................................................. 57
3.3.2.1 Conventional- vs. Short-Length............................................................. 59 3.3.2.2 Rip-First vs. Crosscut-First.................................................................... 64
3.4 Conclusion......................................................................................................... 66 References ................................................................................................................. 69
4 WHITE BIRCH LUMBER USED IN THE PANEL INDUSTRY............................ 73
Abstract ..................................................................................................................... 73 4.1 Introduction....................................................................................................... 75 4.2 Methodology ..................................................................................................... 76
4.2.1 Sample material............................................................................................. 76 4.2.2 Cutting Order................................................................................................. 77 4.2.3 Rough Mill Processing .................................................................................. 79
4.2.3.1 ROMI-RIP simulation parameters:........................................................ 79 4.2.3.2 ROMI CROSS simulation parameters: .................................................. 79
4.3 Results and Discussion...................................................................................... 79 4.3.1 Yield .............................................................................................................. 79
4.3.1.1 Total Yield ............................................................................................. 80 4.3.1.2 Primary Parts ......................................................................................... 82
4.3.1.2.1 Conventional vs. short-length .......................................................... 82 4.3.1.2.2 Rip-first vs. crosscut-first ................................................................. 82
4.3.1.3 Salvage Parts ......................................................................................... 83 4.3.1.3.1 Conventional vs. short-length .......................................................... 83 4.3.1.3.2 Rip-first vs. crosscut-first ................................................................. 83
4.3.2 Part Size Distribution..................................................................................... 84 4.3.2.1 Conventional vs. short-length ................................................................ 84 4.3.2.2 Rip-first vs. crosscut-first....................................................................... 87
4.3.3 Correspondence Analysis .............................................................................. 91 4.3.4 Lumber grade, processing method and lumber length:
Relationship to component distribution ......................................................... 92 4.4 Conclusion......................................................................................................... 96
References ................................................................................................................. 98
5 THE EFFECT OF MANUFACTURING DEFECTS ON YIELD .......................... 100
Abstract ................................................................................................................... 100 5.1 Introduction..................................................................................................... 102
vii
Page
5.2 Methodology ................................................................................................... 103 5.2.1 Sample material........................................................................................... 103 5.2.2 Board Grading ............................................................................................. 103 5.2.3 Database ...................................................................................................... 104 5.2.4 Cutting order ................................................................................................ 105 5.2.5 ROMI-RIP simulation parameters:.............................................................. 107
5.3 Results and Discussion.................................................................................... 107 5.3.1 Spike Marks................................................................................................. 108 5.3.2 Conveyor Marks .......................................................................................... 112 5.3.3 Pressure Roller Stain.................................................................................... 115 5.3.4 Drying Checks ............................................................................................. 115 5.3.5 Machine Gouge ........................................................................................... 117 5.3.6 Machine Burn .............................................................................................. 117 5.3.7 All Defects Combined ................................................................................. 118
5.4 Conclusion....................................................................................................... 118 References ............................................................................................................... 120
6 Conclusion............................................................................................................... 121
APPENDICES
Appendix A: Creation of White birch database ........................................................... 122 Board Digitizing ...................................................................................................... 122 Defect type definitions ............................................................................................. 124
Natural Defects .................................................................................................... 124 Manufacturing Defects ........................................................................................ 125
Board Grading......................................................................................................... 129 Appendix B: Creating data files for simulation – Computer database.......................... 131 Appendix C: Simulation Software ............................................................................... 141
ROMI-RIP............................................................................................................... 141 ROMI-CROSS........................................................................................................ 149
Appendix D: Incidence of defects ................................................................................ 156 Clear surface area .................................................................................................... 156 Incidence of Defects ................................................................................................ 156 Observations ............................................................................................................ 159
VITA ............................................................................................................................... 160
viii
LIST OF TABLES
Table Page
Table 3.1. White birch database characteristics .............................................................. 48
Table 3.2. List of digitized defects, their simulation program name and number equivalents and their status ............................................................................ 49
Table 3.3. USDA "Easy" cutting order (Adapted from Steele et al. (1999))................... 51
Table 3.4. USDA "Tough" cutting order (Adapted from Steele et al. (1999)) ................ 51
Table 3.5. Furniture cutting order................................................................................... 52
Table 3.6. Defect frequency........................................................................................... 55
Table 3.7. Defect area..................................................................................................... 56
Table 3.8. Yield (%) results for rip- first and crosscut-first rough mills according to grade and cutting order as a function of lumber length .............................. 58
Table 3.9. Rip-first and Crosscut-first yield (%) results by lumber length according to grade and cutting order with wane and void filtered out ............................ 62
Table 4.1. White birch database characteristics .............................................................. 77
Table 4.2. Primary and salvage component yield (%) results by lumber type for Panel cutting order processed by a rip-first or crosscut- first rough mill ......... 81
Table 5.1. Database characteristics ............................................................................... 105
Table 5.2. Furniture cutting order................................................................................. 106
Table 5.3. Mechanical defect frequencies (# / m2) on white birch lumber .................... 110
Table 5.4. Average area of mechanical defects (cm2/m2).............................................. 111
ix
Table Page
Table 5.5. Yield decrease (%) by grade and lumber length for different types of mechanical defects for Furniture cutting order ............................................ 113
Table 5.6. Yield decrease (%) by grade and lumber length for different types of mechanical defects for Panel cutting order .................................................. 114
Appendix Table
Table D-1. Frequency of defects by defect type and by source.................................. 157
Table D-2. Clearwood percentage and defect area by defect type and wood source.. 158
x
LIST OF FIGURES
Figure Page
Figure 3.1. Rip-first yield: Conventional versus Short-length lumber .......................... 60
Figure 3.2. Crosscut-first yield: Conventional versus Short- length lumber.................. 61
Figure 3.3. Conventional-length yield: rip-first versus crosscut- first rough milling..... 65
Figure 3.4. Short- length yield: rip- first versus crosscut-first rough milling ................. 65
Figure 3.5. ROMI-CROSS cutup using Panel cutting order......................................... 67
Figure 3.6. ROMI-RIP cutup using Panel cutting order ............................................... 67
Figure 4.1. Part size distribution for Select grade lumber with a) Conventional-length, rip- first; b) Short-length, rip- first; c) Conventional- length, crosscut- first; d) Short-length, crosscut-first.............................................. 86
Figure 4.2. Part size distribution for No. 1C grade lumber with a) Conventional-length, rip- first; b) Short-length, rip- first; c) Conventional- length, crosscut- first; d) Short-length, crosscut-first.............................................. 88
Figure 4.3. Part size distribution for No 2AC grade lumber with a) Conventional-length, rip- first; b) Short-length, rip- first; c) Conventional- length, crosscut- first; d) Short-length, crosscut-first.............................................. 90
Figure 4.4. Correspondence analysis scatter plot for lumber grade, processing method, and lumber type ........................................................................... 92
Figure 4.5. Correspondence analysis between lumber type and processing method for Select lumber ........................................................................... 93
Figure 4.6. Correspondence analysis between lumber type and processing method for No. 1 Common lumber............................................................ 94
Figure 4.7. Correspondence analysis between lumber type and processing method for No. 2A Common lumber ......................................................... 96
xi
Figure Page
Figure 5.1. Picture depicting a spike mark................................................................. 108
Figure 5.2. Picture depicting a conveyor mark........................................................... 112
Appendix Figure
Figure A-1. Placement of the board on digitizing table ............................................... 123
Figure A-2. Positioning of Crooked Boards ................................................................ 123
Figure A-3. Enclosing defects in a rectangle ............................................................... 125
Figure A-4. Breakdown of large spike knot rectangle into series of smaller rectangles................................................................................................. 126
Figure A-5. “Field” of check ....................................................................................... 127
Figure A-6. Typical Crook marking ............................................................................ 127
Figure A-7. Heartwood Marking................................................................................. 128
Figure A-8. Digitizing Face 2 ..................................................................................... 128
Figure B-1. Random width lumber database opening screen...................................... 132
Figure B-2. Defect filter screen................................................................................... 133
Figure B-3. Board plot screen. .................................................................................... 134
Figure B-4. View Defect Coordinates Screen ............................................................. 135
Figure B-5. Export files screen. .................................................................................. 136
Figure B-6. Opening screen of ROMI CROSS crook-removal program..................... 138
Figure B-7. Board selection screen............................................................................. 139
Figure C-1. Sample grades and rules in the Part Grade Editor.................................... 142
Figure C-2. Cutting Order Editor showing sample cutting order ................................ 143
Figure C-3. Process control window........................................................................... 145
Figure C-4. Salvage length and width editing window ............................................... 146
Figure C-5. Salvage Length Modification window..................................................... 147
xii
Appendix Figure Page
Figure C-6. Processing option main edit window ....................................................... 150
Figure C-7. Cutting order definition window .............................................................. 150
Figure C-8. Part size, quantity, schedule, and type editing window............................ 151
Figure C-9. Cutting specifications window showing processing options .................... 152
Figure C-10. Primary part defect acceptance menu ...................................................... 153
xiii
LIST OF ABBREVIATIONS
No. 1C = No. 1 Common
No. 2AC = No. 2A Common
No. 3AC = No. 3A Common
bf = Board foot (1 ft. x 1 ft. x 1 in.)
xiv
ABSTRACT
Clement, Charles Ph.D., Purdue University, May 2002. Understanding Rough Mill Yield Through the Analysis of the Interaction Between Lumber Characteristics and Processing Parameters. Co-Major Professors: Rado Gazo, and Robert Beauregard
The purpose of the present study was to investigate the potential use for white birch
lumber, a species that is readily available in Eastern Canada, but underutilized because of
physio-morphological characteristics that make its end-use uncertain. For the purpose of
this study, 13.16 m3 (5,574 bf) of conventional and short- length lumber were used. The
effects of lumber length (conventional and short), grade (Select, No.1 Common, No.2A
Common), cutt ing order (Furniture, Panel, USDA Easy and USDA Tough) and processing
method (rip-first and crosscut-first) on yield were analyzed.
Highly significant yield differences of 8.8% for Select and 10.3% for No.
2A Common were observed between conventional and short-length lumber. These
differences can be explained by shorter average length and the increased presence of wane
and void. There is little difference in yield, when processing No.1 Common lumber.
Crosscut-first processing generates, on average, a 4.2% higher yield than rip- first
processing. Lumber grade, processing method and lumber type are the three variables that
explain most of the variability in component production.
Analysis of the incidence of manufacturing defects indicated that drying checks had
the largest impact on yield, reducing yield by 5.9% for the Furniture cutting order and 6.4%
for the Panel cutting order. No. 2A Common lumber was most affected due to
physiological properties of the boards, i.e. presence of heartwood and juvenile wood, which
make drying more difficult. Spike mark lowers yield by about 3% for either cutting order,
xv
but they occur only in mills that use ring debarkers, and mostly on high-grade external
boards. Pressure roller stain affected yield by less than 2%, and affected the smaller-sized
boards because the defect location offers less flexibility to cut the defect out. Machine burn
reduced yield by 0.6% and 0.7% for the Furniture and Panel cutting orders, respectively,
and it appears to affect conventional- length lumber more due to the dynamics of handling
longer- length boards. Conveyor marks reduced yield by 0.6% and 0.8% for the Furniture
and Panel cutting orders. Machine gouge affected yield by 0.5% for both cutting orders,
and affected short- length lumber more.
1
1 INTRODUCTION
Since 1967 digitized databanks have been used to store information concerning
defects found in rough lumber. This information was originally recorded to determine yield
for different species by using sophisticated – at the time – hardware and software. As
computer technology became more accessible, code was written that would allow the user
to set certain parameters such as arbor type, kerf, prioritization strategy, and salvage
operations. These parameters allowed the rough mills to simulate their operations more
accurately and judge how modifying the setup would influence yield. Hence the use of a
database was essential in generating reproducible – and therefore comparable –
simulations.
Several simulations have been done with regards to standard NHLA-grade lumber.
These simulations were concerned with the effect on crook on yields when gang ripping
narrow lumber (Gatchell 1990) and the potential the effects of grade quality on crosscut-
first and rip-first yield (Gatchell et al. 1996, and Gatchell et al. 1983), and within-grade
quality differences (Gatchell and Thomas 1997). Also, several simulations have been done
where different arbor type configurations were evaluated (Steele and Lee 1994, and
Gatchell 1991), and edging and trimming practices (Kline et al. 1993, and Regalado et al.
1992). These simulations would compare the effects of a specific cutting order on yield but
a general population sampling was generally considered.
Although total available lumber stock is not decreasing, the quality of the logs is
(Luppold 1994). Therefore the average price of NHLA-grade logs is increasing. In order
for the furniture manufacturers to be able to compete in today’s market, they process
2
shorter-length logs, even though they have a lesser yield, because they are readily available
and therefore, more economical.
Little work has been done on the characteristic initial length of rough lumber with
regards to yield. Only Wiedenbeck (1992) and Wiedenbeck and Araman (1995) have done
work studying the implantation of short-length red oak in rough mills. This study will
study three aspects of the use of short- length lumber in rip- first and crosscut-first rough
mills: 1) compare the yield of short- length to that of conventional-length white birch; 2)
determine optimal use of white birch; 3) determine a relationship between manufacturing
defects and lumber type.
This study involved the preliminary step of creating a digitized white birch
database, in which grade, length, width, crook, and all surface defects were included. The
database was then used to analyze the incidence of defects in terms of defect frequency
(occurrence per square meter) and average defect size (square centimeters per square
meter), in order to characterize the species as a whole and to differentiate any differences
between lumber sources (i.e. short-length vs. conventional- length).
1.1.2 Objectives
The general goal of this study was to determine the remanufacturing potential of
white birch. This was accomplished by following specific objectives:
1. determine yield of lumber with rip- first and crosscut-first simulation software
and analyze the effects of lumber type, processing method and cutting order on
yield;
3
2. analyze the optimal component distribution of white birch through the use of a
Panel cutting order and identify the main factors that influence distribution
variability; and
3. measure the effect of manufacturing defects on yield.
This dissertation consists of a literature review, followed by three self-standing
articles that have been reviewed and will be submitted for publication in a refereed journal
– each dealing with one of the above-mentioned specific objectives, a general conclusion
and a series of appendices detailing the creation of the white birch database, datafile
preparation, rough mill simulation setup and a detailed description of the incidence of every
defect that was digitized.
4
2 LITERATURE REVIEW
Over the past decade, interest in white birch has grown immensely. Manufacturers
realize that there is a significant untapped resource available, waiting to be exploited.
Where once white birch was used only for ice cream sticks, tongue depressors, chopsticks
and toothpicks, it is now being used for furniture and flooring manufacturing. However,
little is known about the lumber characteristics of the species. Grade rules, such as those
set by the National Hardwood Lumber Association, determine general parameters
prescribing minimum-sized clear cuttings, and admissible defects, but these rules are set to
qualify the general use of lumber.
The US Forest Service has established grading rules for logs destined to be used for
factory lumber. The first step is to position as many defects as possible one face. The log
grade is then based on the worst of the three remaining faces. Other factors that influence
grade are the diameter inside bark, length of log, length of clear cutting, the maximum
number of cuttings permitted, and the portion of log length required in clear cuttings
(Cassens and Fischer 1992). The minimum log size for factory logs is 8 feet long and 8
inches in diameter (Vaughan et al. 1966, Petro and Calvert 1990), anything less degrades
the log for use in the pulp and paper industry, thus the term pulpwood.
Pulpwood logs do not meet the minimum size requirements, in length or diameter,
to be factory-graded (Petro and Calvert 1990). It is deemed that half of the white birch
pulpwood could be retrieved and sawed for value-added components (Giguère 1998). Since
this lumber is unlikely to meet the NHLA lumber requirements, being most of the time of
too small dimensions, the rough mills must buy the lumber on the basis of in-house grading
rules that are designed to best match the resource and the rough mill needs. This allows
5
more enterprising sawyers to look into alternate sources of lumber, namely pulpwood logs,
in order to keep production costs down. While little is known about the ability of these
grades to fulfill the various cutting orders, this study focuses only on NHLA-graded lumber
from both conventional and short- length logs.
The creation of a digitized white birch database will allow the rough mills to
maximize the use of the available lumber and this at a minimum cost. The database should
be designed in such a way as to be able to provide an assessment of rough mill yields
comparing the use of conventional-length and short- length sourced lumber.
2.1 Species Description
White birch is found in boreal forests. It can grow to be 15 to 25 meters tall and
reach a DBH between 30 – 60 centimeters (Brockman 1968). A mature tree’s bark is white
and will easily peel off the tree trunk in horizontal strips (Marie-Victorin 1964, Brockman
1968). White birch will live to be 150-years old, however it will reach maturity at about
seventy years of age (Fowells 1965, Marie-Victorin 1964). The wood is straight-grained,
appearing rather fine and tight. Early wood, late wood, and the year rings can be clearly
distinguished. The heartwood is dark-hued.
2.2 Species Distribution
North American white birch has a transcontinental distribution from east to west
and can be considered a typical boreal forest species (Quigley and Babcock 1969, Fowells
1965).
This species needs light in order to grow and therefore it will usually grow only one
generation (Hyvarinien 1968). It will characterize any changes in tree cover. Thus, it will
6
be found in mixed forests in clearings caused by a disturbance (Hyvarinien 1968) or in pure
stands resulting from forest fires (Lortie 1979). According to Lortie (1979), white birch
covers vast territories mainly in association with fir and white spruce.
2.3 Availability of the species
In North America there are large quantities of white birch available throughout the
Eastern parts of Canada and the Northeastern and Great Lake States of the United States
(Verkasalo 1990). A large amount of it is currently unused because its quality is deemed
inferior. According to the Minnesota Department of natural resources (Martodam 1982),
there were over 1,820,000 m3 available for harvest and only 28 percent of that volume was
being used. To produce NHLA lumber in the province of Québec for example, there is a
gross merchantable volume of 381,152,000 m³ of white birch (MNRQ 1996). On a
sustainable basis, that province can cut, more than 5,230,000 m³ per annum – of which
3,230,000 m³ are of sawing quality – yet only 1,398,230 m³ of that amount (Giguère 1998)
are being allocated for processing on public land – with 587,200 m³ reserved for future use.
This leaves at least 3,245,000 m³ per annum of available timber, of both saw-quality and
pulpwood quality, on public land only, and for that province only.
European white birch (Betula pubescens), which greatly resembles American white
birch makes up a considerable hardwood potential in Northern Europe, Baltic Countries,
Russia, Belarus, and Poland. In the Scandinavian countries, there is a growing interest in
the use of white birch in value-added wood products. In Finland, birch accounts for 15 %
of all timber resources (Verkasalo 1996).
7
2.4 Uses
White birch is used commercially for veneer, plywood, and pulpwood. It is easily
worked and takes finishes and stains readily. Furniture, cabinets, and numerous specialty
items are made from birch lumber (USDA Wood Handbook 1999, Brière 1992, Panshin
and de Zeeuw 1980).
Tree chips are used for pulp and paper manufacture, and composites. White birch is
also commonly used as fireplace and wood stove fuel (BWCA 1999).
White birch has excellent machining – especially rotary – cutting, gluing and
surfacing properties (Brière 1992) however, it splits easily, especially when screws or nails
are used (USDA Forest Service 1953).
In addition to use for veneer, plywood, and lumber, white birch is used in some
amounts for ice cream sticks, toothpicks, tongue depressors, and turned products such as
spools, bobbins, small handles, and toys (USDA Wood Handbook 1999). Birch can also be
used for joinery, furniture of solid wood and parquet (Brière 1992).
2.5 Database
The main purpose of a lumber database is to store the physio-morphological
properties of a species. This allows the end user to analyze the lumber characteristics that
describe the species in terms of defect occurrence and defect frequency. Another feature of
the database is that it allows the user to simulate and compare rough mill processes by
always using the same reference data. This should provide users with confidence that
difference in results between different scenarios is attributable to varying process
parameters and not to variation in lumber used.
8
With such a large resource available many producers are looking into using white
birch for value-added products. The species has not been widely used so far due to
morphological and physical properties such as low density, small diameter, crook, and bark
pockets that, when compared with more broadly used species such as hard maple and red
oak, make it more difficult to obtain clear pieces of lumber.
In recent years, because of increasing scarcity and price of traditional hardwood
species, the furniture, cabinetry, and flooring industries in Europe, Northern US
(Minnesota) and Eastern Canada have geared towards using white birch.
Little is known about how to get the most efficient use of the lumber. One of the
major factors influencing its optimal use is the emergence of dimension plants that
manufacture components directly for the rough mill. In this case, the components meet
rough mill standards instead of NHLA grade rules, which makes it very difficult to obtain
an objective classification of the lumber. Even with standard-sized lumber, the quality of
the cuttings, within an NHLA grade, is not determined (Gatchell et al. 1993). Therefore,
the creation of a digitized database is of interest. Such a database allows the manufacturers
to qualify and quantify the incidence of defects and determine an expected yield.
Several databases have been created and are available to the public. There are
databanks on hard maple (Schumann and Englerth 1967a, _ 1967b), walnut lumber
(Schumann 1971), alder lumber (Schumann 1972), ponderosa pine (McDonald et al. 1981),
red oak (Gatchell et al. 1998, Wiedenbeck et al. 1994, Gatchell et al. 1992, Harding 1991,
Nordin et al. 1990, and Lucas and Catron 1973), yellow poplar (Osborn et al. 1992 and
Gilmore et al. 1984), radiata pine (Gazo et al. 1998). All these databases have the
following features:
The defects considered consist of characteristics judged as being either
mechanically unsound or unpleasing to the eye (e.g. knots, checks, mineral streaks, bark
9
pockets, stain decay) – according to NHLA (1998), Western Wood Products Association
(WWPA 1991), or ma nufacturer-specific grading rules.
These databases define the types of defects encountered and the location of each
defect, i.e. xy coordinates and side of board.
The two variables used to describe these characteristics are usually average number
and average size of defects found per surface area (Harding et al. 1993, Gazo et al. 1998).
Thomas (1962) developed a data collection technique that made use of a 12-inch
wide by 16 feet long clear acetate film on which areas of 1 inch wide by 3 inches long were
outlined and numbered from 1 to 768. The film would be superimposed over a board. Each
defect was recorded by type and the numbered rectangular area would locate it. 35,000 bf
of oak, yellow poplar, and hard maple were digitized in this manner. This data was then
used to estimate optimal yield for a crosscut-first rough mill.
Then, other databases were created in order to determine dimension yields from:
4/4 hard maple lumber (Schumann and Englerth 1967a, ___1967b), black walnut lumber
(Schumann 1971), alder lumber (Schumann 1972), and in 1973 Lucas and Catron created a
comprehensive defect data bank for No. 2A Common oak lumber. Gilmore et al. created a
yellow poplar database in 1984. The above databases all used the YIELD (Wodzinski and
Hahm 1966) program to determine the yield of various-size cuttings.
The ponderosa pine databank (McDonald et al. 1981) was created, and tested by
OPTYLD (Giese and McDonald 1982) in order to establish a yardstick against which to
measure the maximum cutting yield of a board.
10
With the advent of personal computers, the use of databases to obtain yield results
became more widely accessible and the creation of such databases to study the incidence of
defects in various species became standard.
Nordin et al. created the first randomly selected and statistically analyzed red oak
(quercus spp.) databank in 1990. In 1993, Harding et al. increased the size of the database
in order to obtain approximately 2,000 board feet (bf) in each of six grades (FAS, F1F,
SEL, No. 1C, No. 2AC, and No. 3AC). The purpose in creating the red oak database was to
quantify defect types in hardwood lumber used for furniture production. This later database
was used as a basis to provide information for a rip- first / crosscut-first simulation software
(RIPX) that would compare the respective value and volume yields statistically (Harding
and Steele 1997).
A total of 13,263 board feet (1,929 boards) came from four furniture rough mills
whose lumber came from twenty-one rough mill suppliers from Mississippi. The quantity
of the different lengths and widths was selected in order to be representative of what was
manufactured. All defects were digitized (i.e., unsound knot, sound knot, wane, worm
holes, grub holes, holes, checks, decay, mineral streaks, bark pockets, splits, and pith)
including mineral streaks, which are not considered an NHLA grade-defect. This defect
was accounted for because it is objectionable in most furniture. One important feature of
this study was the inclusion of the actual amount of crook contained in a board. This
consideration is particularly important because Nordin et al. (1990) had found that 85.3
percent of boards had some degree of crook. Nordin et al. (1990) and Harding et al. (1993)
agreed that knots, checks, wane, stain, and holes were the major defect types based on the
number of southern red oak boards affected.
Because of the abundance of yellow poplar and its underutilization the USDA
Forest Service created the West Virginia Yellow-Poplar Lumber Defect Database in 1992
(Osborn et al. 1992). This database contained the description of 627 boards totaling almost
3,800 board feet. Only FAS, F1F, 1 Common, and 2A Common were used because very
11
little yellow poplar was sold as SAP (i.e. sapwood only), Selects, or No. 3A Common. One
observation that was made was that the mean cutting surface area of No. 2A Common
boards was greater than the minimum required for FAS. This was due to FAS or F1F
boards that were downgraded to No. 1 Common due to stain or discoloration, which –
although limited in FAS and F1F – are not limited in the lower grades of yellow poplar.
The goal was to produce a database from which potential users could draw sub samples to
represent their own board distributions.
One interesting feature of this databank was the use of sub grades based on the
percentage of the board surface measure contained within the cuttings. There were three
sub grades for No. 1 Common, No. 2A Common and No. 3A Common NHLA grades.
The defects considered for the yellow-poplar database were: bark pocket; cross
break; check; crook; discoloration (mineral) – dark; discoloration (mineral) – light; hole;
knot, loose; knot – tight (includes burls); knot, unsound; pith; rot; stain, heavy; shake; stain,
light; split, and; wane. Osborn et al. (1992) did not analyze the incidence of defects for this
database. All the defects were recorded in a ¼-inch scale in order to increase processing
speed.
In 1992, Gatchell et al. (1992b) developed a 4/4 red oak lumber databank according
to NHLA rules from which sample boards could be drawn upon to meet the needs of the
user. Emphasis was on No. 1 Common and No. 2A Common lumber because a major
limitation of most rough mill yield studies was that the specific quality level of each board
is graded by a certified grader, the grader looks only for the minimum requirements of the
grade in question. Therefore, if data bank users unknowingly compare the high end of one
grade with the low end of another, then incorrect conclusions can result. The databank
consisted of 1,578 boards representing 10,712 board feet (16.8% FAS, 14.6% Selects,
34.8% No. 1C, 33.8% No. 2AC). The lumber was analyzed and the results compared using
conventional NHLA grading rules versus NHLA grading rules where the maximum
number of grading cuttings was allowed. The relaxing of the maximum number of cuttings
12
rule dramatically increased the number of high quality No. 1 Common (145%) and No. 2A
Common (287%).
Wiedenbeck et al. released a databank, in 1994, on 426 (1,140 bf) short – less than
8 foot long – 4/4 red oak lumber to serve as an addendum to Gatchell et al. 1992 red oak
databank. Although short lumber had little market at the time, the reason for the creation of
a short board databank was to study the effects of lumber length on rough mill productivity
and profitability (Wiedenbeck 1992, Wiedenbeck and Araman 1995).
Wiedenbeck and et al. (1995) published a report describing the quality
characteristics of the Appalachian red oak databank and they discussed the range of quality
possible within each grade and analyzed the most important quality characteristics by
grade. They discovered that No. 1 Common and No. 2A Common lumber was under-
graded by 23 and 35 percent respectively i.e., those boards clear-face cutting percentages
meet the minimum requirement for the next higher grade.
The mean defect areas for the four grades were: FAS – 1.2 %, Selects – 2.3 %, No.
1 Common – 6.8 %, and No. 2A Common – 9.8 percent. Most (89 %) of the total defect
area of these boards consisted of unsound defect. Wane, unsound knots, and bark pockets
were the three major defects found in red oak lumber.
Twenty-five percent of the No. 1 Common and No. 2A Common boards in the
databank contained ½ inch or more of crook. Yield studies have shown that this level of
crook can significant ly lower primary part yield and total part yield in gang-rip-first rough
mill processing.
In 1995, a study of 392 (7,245 bf) New Zealand cloned radiata pine random-width
boards was created. Gazo et al. (2000) analyzed the incidence of defects and later,
Beauregard et al. (1999) studied clonal variation. Defects such as intergrown knots,
13
partially intergrown knots, tight knots, loose knots, holes, spike knots, pith, bark pockets,
resin pockets, needle fleck, and wane were analyzed. The boards were graded into
Moldings, Factory Select, No. 1 Shop, No. 2 Shop, No. 3 Shop, and Finger Joint Common
Shop using the Western Lumber Grading Rules (WWPA 1991).
In their comparison of pruned versus non-pruned logs, Gazo et al. (1998) noticed
that bark pockets and blemishes occurred most frequently in pruned butt logs and that the
largest average size defect was area of needle fleck. Most common defects in unpruned
logs were intergrown knots, partially intergrown knots, loose knots, and bark
pocket/blemish. In this case, the largest average size defect were intergrown knots. It also
appeared that in boards from pruned logs only 52% ±22% of boards without knots were
clear of other defects. This observation challenged the assumption that a pruned tree will
only grow clear wood.
Since 1994, the NHLA made several revisions to its grading rules, which are found
in the 1998 rulebook (NHLA 1998). First and Seconds grades were combined into FAS.
FAS ONE FACE (F1F) was made a standard grade. FAS first foot rule was modified to
allow oversize knots. The option to grade Selects using cuttings with sound backs was
eliminated. FAS wane rules were changed, etc.
Gatchell et al. (1998) proceeded to expand the existing USDA Forest Service red
oak databank in order to encompass these changes. 1,400 new boards were added resulting
in a complete database of 3,487 boards that total 20,021 board feet. The grades covered by
the databank were: FAS (25%), F1F (13%), Selects (5%), No. 1 Common (29%), No. 2A
Common (23%), and No. 3A Common (5%). In order to reduce the negative effects of
crook on rip-first yields (Gatchell 1990, 1991), the boards had to have less than ¼ inch of
crook. In their study, Gatchell and Thomas (1997) learned that 21.1 percent of the boards
classified as No. 1 Common were of FAS stock, while 10.8 percent of No. 2A Common
could meet No. 1 Common requirements.
14
Databases are useful sources of information for the wood-products industry because
they can be used in computerized programs to help train lumber graders. They are also used
in rough mill simulations to estimate yield.
2.6 Grading
In 1971, Hallock and Galiger wrote a computer program to grade hardwood lumber.
Certain limitations made widespread commercial application difficult. For instance, it used
NHLA grading rules for standard grades, which are not readily adaptable to the numerous
exceptions allowed for many species. Furthermore, the algorithms used consider only one
side of the board, whereas both faces of the board are evaluated in the actual grading
process. This factor is especially critical when grading lumber that may potentially grade
Select.
Little happened until 1989 when Klinkhachorn et al. (1989a) took the above
limitations into consideration and wrote a modular program that had a flexible grading
procedure and could easily be integrated with other software i.e., lumber processing
software. The program would first evaluate the board to determine its surface measure,
length, and width. Areas declared as non-clear wood – defects – are then considered, where
the defective areas of the board are described mathematically as being rectangles that
enclose the periphery of the defect. The lower left-hand and upper right-hand corner’s
coordinates are used to note the position of the defective region. Each rectangular region is
coded to identify the defect type and the face on which it appeared. Larger defects, such as
wane, were divided into smaller rectangular regions in order to use as much clear wood as
possible in the analysis.
The program was written to interface with computer vision systems therefore
sufficient flexibility was required to handle regions identified as defective. Defects such as
small checks, knots, and burl could be identified as defects by the vision system when in
15
fact they were allowed by a particular set of grading rules. In order to allow for such
potential defects, a third planar view of the board was created onto which these defects
were placed. Thus, the defects that were allowed in clear cuttings were placed on the third
plane by changing the code tha t indicated the face on which the defect occurred.
A fourth planar view contained unsound defects from each side of the board. Thus,
unsound defects that occurred on the reverse face of the side being evaluated would be on
the visible plane. This allowed the program to consider the position of all unsound defects
when evaluating cuttings for either clear wood on one face or sound wood on the reverse
face.
Then a potential grade was assigned to each face and the computer assessed
existing clear area to determine the final grade assigned to the board.
Klinkhachorn et al. then used the above-mentioned program in 1989 to develop the
Hardwood Lumber Training (HaLT) computer program (Klinkhachorn et al.1989a). This
program allowed inexperienced graders to obtain much needed practice so that the
decisions these graders made would be proper and accurate.
Schwehm et al. (1990) used the above-mentioned Automated Hardwood Lumber
Grading Program to determine the grade of each board based on the NHLA Rules (1998) in
their Hardwood lumber Remanufacturing (HaRem) program. This program checked
different possibilities for edging and trimming each board to determine if the grade could
be improved. If so, it would then determine if the market value of the board had also
increased. Based on these measures, the program would make the determination of how the
board should be edged and trimmed.
In 1992, Klinkhachorn et al. (1992a) updated the HaLT program with HaLT2. In
this revised version, a board editor that allowed users to create a board and four different
16
board call ups (i.e., sequential, random order, serial position, and by board identification
number or name in order to improve and facilitate board grading learning).
Gatchell et al. (1992a) proceeded to write the Realistic Grading System (ReGS)
program in 1992 as an extension of the basic algorithm used to develop HaLT, HaLT2, and
HaRem lumber grading and remanufacturing training programs. One caveat of those
programs is that they were not designed to work on real boards i.e., they were designed to
use boards that are perfect rectangles. Therefore, when faced with a real board – in which
shrinkage or board tapering has occurred – a defect would be substituted for the gap. The
defect size limitations of FAS and Selects may easily be exceeded. Also, using the surface
measure of the enclosing rectangle may require more cutting units per grade that are
available in the board.
To get around this, ReGS created a new defect type to label the gap or space: void.
It was not a true defect because it occurred outside the board. Thus, width and length of
void are not evaluated in the same way as other defects (e.g. knots or wane).
Also, a timer was used to specify the maximum allowable interval to achieve a
solution. This feature was designed because in low-grade boards, for example, there will be
many large stepped defects that will take a long time to process. The timer allows the user
to accelerate the process by preventing the grading software to run through all possible
iterations.
One of the main disadvantages of the above grading systems is that they all use
rectangular modeling to define defects. Rectangular modeling of the defects eliminates a
substantial amount of clear wood from being considered in the grading process – especially
in the case of large defects. This is why Klinkhachorn et al. (1992b) enhanced their original
rectangular grading program to accept convex polygons. A convex polygon is a polygon in
which every point on a line segment joining two points within the boundary of the polygon
also lie within the polygon.
17
The program then proceeds to grade the board using the same logic as described
above. Although the randomness of defect occurrence, their shape and size, precludes a
quantitative analysis, there is a subtle advantage. The rectangular grading program looks
for the minimal solution to meet a grade in accordance with the NHLA rules. A polygonal
approximation of defects is an advantage if maximal solutions are desired, as in the case of
remanufacturing lumber for a higher grade and value.
It should be noted that a defect could also be represented by a series of stepped
rectangles to approximate the original defect shape. This approximation simplifies the
defect digitizing of the board. However, with stepped rectangles representing a large defect,
the number of defects increases – which increases the time to process the board. More
important, if a split is represented by a series of small rectangles, rules such as the split
divergence rule cannot be applied to the split but rather to each of its constituent rectangles.
The rule could be interpreted if it were a continuous defect. It should be noted that stepped
rectangles, when used, make it more complicated to assess the defect frequencies.
With the advent of remanufacturing of lower grade lumber to boards with a higher
overall value, Klinkhachorn et al. (1994) developed TRSys, a hardwood lumber grading
Training and Remanufacturing System.
The TRSys remanufacturing module is an enhanced version of that used by HaRem
where the TRSys uses a value-driven, division-based remanufacturing system that could
call for remanufacturing a single large board in to as many as four smaller boards. One or
more may be of the same or lower grade than the original board however; the constraint is
that the total value of all new boards be greater than that of the original board.
In New Zealand, Todoroki (1996) created FLGRADE to grade random-width
factory lumber according to Western Lumber Grading Rules (WWPA 1991). She used
dynamic programming to reduce computation time for discovering the optimal solution.
Although fast and accurate, the program is limited by its guillotine cutup process, which
18
renders some of the cuttings infeasible using traditional rip -first or crosscut-first rough mill
technology.
While each was successful to some degree, none provided fast and accurate grading
combined with remanufacturing of lumber. In 1998, Moody et al. wrote the Ultimate
Grading and Remanufacturing System (UGRS).
UGRS combined all the innovations of the above programs and increased the
grading flexibility by using a wide variety of grading procedures i.e., it can grade according
to air-dried, kiln-dried, or scant-widths. To provide maximum flexibility, UGRS stores
remanufacturing parameters such as lumber prices, remanufacturing costs (fixed cost and a
cost per lineal foot for rip and crosscuts), and grading rules in files so that modifications
can be made without modifying the program itself.
UGRS considers all defects as rectangles. For large defects it proceeds to encode
the defect as a series of smaller rectangles. One caveat of this procedure in older software
was that the defect would be broken down onto a series of smaller – different – defects.
This can be a grading problem in measuring defects such as split length or knot diameter
where the length and/or width of the entire defect is required. UGRS calculates the true
length and width of a “stepped defect” by calculating the total length and width of all
touching defects of the same type, thereby providing the benefits of “stepping defects” to
more accurately model the dimensions of the physical defect along with the true length and
width. It also calculates the slope of end splits that could affect grading of FAS lumber.
On the whole, UGRS obtained results that were the same or better than those
obtained by earlier grading programs. One of the main advantages of this system is that it
processes boards at least 50 times faster than earlier programs due to an enhanced cutting
unit algorithm.
19
All of the above programs were designed for use either as a learning/training tool or
for incorporation into an Automated Lumber Processing System (ALPS) (McMillin et al.
1984, Klinkhachorn et al. 1989b). The ALPS system integrated every processing step to
maximize yield and productivity.
2.7 Processing
In the rough mill there are basically two cutup procedures. Traditionally, the lumber
is crosscut first in order to maximize the desired cutting lengths between defect areas.
Then, the crosscut sections are ripped to predetermined cutting widths. A salvage operation
is then performed on the residual parts by crosscutting to smaller acceptable cuttings.
Today, however, many random-width cuttings can – and are – assembled and glued
to form the final cutting item. If edge gluing in all panels and for all cuttings is acceptable
then, another system to produce rough mill cuttings can be considered. All the lumber is
gang ripped to a predetermined width. These strips are then crosscut in order to remove
unwanted defects and obtain desired cuttings. Cuttings are edge glued into panels and
resawn to the desired final cutting width. The remaining random width strip is recycled into
the next panel to be glued up.
The cutting order determines the ideal cutup process. Manalan et al. (1980) defined
cutting orders as “a schedule of dimension parts where any one of these parts can be cut out
from this schedule during a given rough mill setup.” Cutting orders are thus the quantity,
size, and quality of parts to be cut in a rough mill. Despite the impact of the cutting order
on rough mill operations, it is often overlooked because of its impalpable nature.
Compared to other issues relating to rough mill operations (e.g. kerf, arbor
configuration, cutup optimization), cutting orders are a largely overlooked issue. Araman et
al. (1982) proceeded to describe cutting order requirements for different dimension part
20
producers by having twenty furniture makers and twelve kitchen cabinet makers partake in
a “parts requirements” survey. Through this survey, Araman et al. (1982) listed detailed
cutting order part size distributions for 5 major product lines. The product lines were: 1)
solid furniture, 2) veneered furniture, 3) upholstered furniture, 4) recliners, and 5) kitchen
cabinets. The companies gave information on the rough part requirements for the most
frequently produced items. The manufacturers provided information on their order of
materials consisting of length, width, thickness, quality, and number of parts required per
item. The authors then analyzed and sorted through the thousands of individual parts. The
Hardwood Dimension Manufacturers Association quality definitions were used
(HDMA 1961)1. The rough part requirements and the nominal length/width distribution
were then shown for each product line.
Due to the huge amount of different parts when length, width, thickness, part
quality, and product type are considered separately, Araman et al. (1982) proposed a new
system to produce dimension parts called Standard-size Hardwood blanks. The authors
suggested to produce glued blanks out of massive hardwood strips. The blanks would then
be used to cut the necessary parts from individual boards.
Four-quarter- inch clear solid wood furniture parts were found to be more evenly
distributed in length and width than 4/4-inch clear kitchen cabinet parts. More than 50
percent of the parts in the kitchen cabinet production were equal or shorter to 25 inches.
Little work has been done relating the effects of cutting order to yield output.
Obviously, the smaller the cuttings, the higher the yield but one must also consider the
effect of the cutup process.
In the mid-70’s, furniture production and hardwood lumber prices were rising and
this despite the fact that there was no shortage of hardwood timber. The problem was that
1 The Hardwood Dimension Manufacturers Association (HDMA) later changed its name to National Dimension
Manufacturers Association (NDMA) and recently to Wood Component Manufacturers Association (WCMA).
21
high-grade timber was becoming scarce while low-grade hardwood was readily available
(Luppold 1994). Having a small diameter and/or being too short prevented logs from
meeting top grade requirements. Reynolds and Gatchell (1979a) developed a program that
was designed to process logs that were between 8 and 12 inches in diameter. This “low-
grade” lumber could then be used for furniture parts, pallets, pulp, and/or energy. In order
to obtain the highest yield, the trees had to be harvested as bolts. These bolts would then be
sawed as cants instead of conventional lumber. The cants were then sawed to short boards,
which were made into rough dimension stock.
Production technology development has made it possible to manufacture dimension
parts for furniture and panels from white birch boards of smaller sizes than used earlier for
saw milling. One such development is the creation of a processing technique called
System-6 (Reynolds and Gatchell 1979b, Reynolds and Araman 1983, Reynolds et al.
1983, Reynolds and Hansen 1984, Reynolds 1984, ___ 1985, Hassler et al. 1995). System-
6 is geared specifically to the production of blanks and eliminates the production of grade
lumber as an intermediary step.
To convert small birch logs to furniture blanks, six-foot cants (round-edged
sections) are sawn from whole logs and then re-sawn into 3- to 4-inch-wide boards. After
drying, the boards are ripped and cut to lengths as needed to remove defects. Defect-free
pieces are then sorted by length and glued edge to edge to form blanks. System 6 is a low
cost operation (Hansen and Reynolds 1984).
In contrast to System 6, McMillin et al. designed the Automated Lumber
Processing System (ALPS) (McMillin et al. 1984). This system was conceived to produce
optimal lumber yield by using computer processing throughout the whole procedure. In this
ideal system, the logs are scanned in order to determine the location of internal knots and to
establish the log geometry. With this information, the computer then positions the lo g in
order to maximize grade or value yield. However, many boards will still contain defects
(knots, wane, stain, checks, etc.) that must be removed.
22
Once dried and surfaced, the boards are scanned for defects. A computer then
identifies the defects, their location, and defines the board geometry. ALPS then maps the
defect data and figures the best cutting pattern for each board in order to obtain maximum
yield.
The parts are then cutup using a high-power laser according to the best cutting
pattern. Finally, the parts are sorted for size. Residue material is chipped and used for fuel.
This system improves yield between 12.6 percent and 22.9 percent (Ruddell et al. 1990)
over conventional sawing due to several factors. One of these factors is the reduction of
waste brought about by operator fatigue and inexperience (Klinkhachorn et al. 1989b).
Another factor is greater cutting yields obtained by laser cutup. A laser can be positioned
anywhere on the board and can cut any arbitrary shape (Klinkhachorn et al. 1989b,
McMillin et al. 1984). Another advantage of using a laser is the narrow kerf (between
0.020 in. – 0.025 in.) that results from this type of cut (Klinkhachorn et al. 1989b,
McMillin et al. 1984).
2.8 Simulation Programs
One of the main advantages of creating a digitized database is the ability to use it
with rough mill simulation software. Indeed, today’s software offers great flexibility in
setting up either a rip -first or crosscut- first mill. However, this was not always the case.
With the advent of the digital computer, Thomas (1962) used a computer-based
simulation program to estimate the full potential lumber-yield. This allowed the elimination
of operator decisions as a variable affecting yield, which in turn would create reliable yield
estimates that could be used for analysis in order to determine the best – most cost-efficient
– way of operating. To do this, Thomas (1962) developed a program that would simulate a
crosscut- first rough mill.
23
In order to vary the emphasis on cutting length, Thomas designed a weighting
function that could adjust anywhere between absolute yield and maximum cuttings size.
The program determined the combination of cutting lengths that would maximize yield.
Although designed as a crosscut- first rough mill simulator, the program did not consider
the sawing kerf nor did it account for the path the tool would travel i.e., it did not actually
crosscut- first or rip-first but obtained the cuttings using a cookie-cutter technique, which
increased yield but rendered some of the cuttings infeasible.
Wodzinski and Hahm (1966) developed YIELD in order to take into account the
kerf loss and infeasible cuttings unaccounted for by the Thomas computer program by
assigning a ¼-inch kerf and verifying that the kerf lines did extend to one of the board
edges but not into adjacent clear cuttings.
This program used a Cartesian coordinate system with ¼ inch discrete units where
the lower-left and upper-right coordinates were used to identify the location of the defect
area. However, it utilizes the same weighting function as the program developed by
Thomas (1962).
It would determine which process – i.e. rip-first or crosscut-first – was most
efficient by comparing the number of operations required to extract the cuttings in either
case, and select ing the one with the least amount. The program could also determine the
total area taken up by defects and the percentage and was used to do so by the U.S. Forest
Products Laboratory in order to determine the potential dimension stock yields for different
cutup processes (i.e., for hard maple (Schumann and Englerth, 1967a and 1967b), black
walnut (Schumann 1971), and red alder (Schumann 1972)).
While YIELD models a single-saw operation, the use of multiple-blade ripsaws
required the creation of a program that could appropriately simulate a rip - first rough mill.
Hence, Stern and McDonald proceeded to develop RIPYLD (Stern and McDonald 1978).
The program gang-rips a board into optimal-width strips and then proceeds to remove the
24
defects through a crosscut operation and a potential salvage operation. MULRIP (Stern
1978) was an unpublished evolution of RIPYLD. It ascertained the best combination of
specified widths by which a board should be ripped i.e., it modeled an all-movable blade
ripsaw.
In 1982, Giese and McDonald wrote OPTYLD (1982) as a multiple-rip program
that included salvage cuttings in its yield estimates. OPTYLD differed from YIELD in
how the program located clear areas i.e., OPTYLD would scan the length and width of a
cutting area, searching for defects. If no defects were found, the area was clear. This
process simulated very closely the action of an automatic defect scanner. Another
difference in the programs was that OPTYLD could maximize yield as a function of value
of the cuttings or clear area, whereas YIELD maximized the largest clear area. Therefore,
instead of only trying to maximize the largest clear cutting area, the program could
maximize the value of the board. In 1983, Giese and Danielson wrote CROMAX (1983) to
simulate a crosscut-first rough mill; the algorithm was based on OPTYLD.
Brunner developed CORY in 1984. The program simulated – like YIELD – single-
bladed sawing. However, it was the first program to allow the user to perform either rip -
first or crosscut- first rough milling. The main difference between CORY and simulation
programs like YIELD, RYPYLD, OPTYLD, and CROMAX was in the algorithm used to
maximize yield. The above-mentioned programs all used the enumeration process. This
process establishes all possible kerf lines, calculates the yield for all possible sawing
combinations, and selects the combination with the highest yield. CORY, on the other
hand, used a “divide and conquer” algorithm. This strategy proceeded to identify the kerf
lines that would give the highest yield and then evaluated each one. Thus, the program
examined the whole board, cut it in two along the best kerf line, and then analyzed each
section individually until only one clear-cutting remained.
Another feature that CORY possessed was part prioritization. For instance, the
program could either optimize for value – where the more desirable parts are given a higher
25
dollar value, or by using a Lengthx x Width prioritization formula where x could either be 1
(maximum yield) or 2 (long cuttings). In 1990, Maristany et al. added an exponential
weighting factor (Lengthwf x Width) that allowed them to fine tune the prioritization of
parts. Although not necessarily optimal, the heuristic used by CORY outperformed the
algorithm used by YIELD in both cutting yield (between 2.7 and 4.2 percent more) and in
execution time (63 times faster) (Brunner 1984).
In order to assess different gang-rip first possibilities, Hoff et al. created a modified
version of MULRIP (Stern 1978) called GR-1ST (Hoff et al. 1991). They added a movable
outer blade and three different saw arbor options to provide optimum gang-rip-first
solutions. These features made GR-1ST more compatible with industry practice where the
rough mills then used fixed saw arbors, variable saw arbors, and equally spaced arbors. The
algorithm in the program was not very sophisticated nor was it very fast. This shortcoming
was addressed in AGARIS where Thomas et al. (1994) improved the user interface and
enhanced the existing code by modifying the salvage algorithm.
Harding (1991) developed RIP-X, a program that determined the yields of current
and least-cost grade mixes for both the crosscut-first system and the rip-first system.
Statistical comparisons of the crosscut- first and rip- first yields could be made, and a linear
programming model was incorporated into the software to determine the least-cost grade
mix.
Carnieri et al. (1993) created a heuristic to find a near-optimal cut-up solution. The
advantage of this method was the processing speed, however this model only worked on
boards that contained zero or one defect.
Thomas (1995a, 1995b) developed ROMI RIP 1.0 in order to simulate more
realistically a modern rip-first rough mill. It featured a random width and random length
counter; a salvage option that could use either the primary parts or a salvage-specific list;
six different arbor setups; and six different prioritization strategies. In 1999, ROMI RIP
26
was upgraded to version 2.0 (Thomas 1999a, ___1999b). The software could cut Clear
Two Face (C2F), Clear One Face (C1F), and Sound Two Face (S2F) simultaneously
instead of only one part quality at a time. It now integrated a part scheduling and
replacement mechanism where one defined the desired number of part sizes and how to
replace those part sizes once the requirements were met. ROMI RIP 2.0 can produce glued-
up panels that are specified in the cutting order. The program supports seven different arbor
types, namely a fixed-blade with either fixed or movable fence arbor, fixed-blade-best-feed
arbor, best-spacing-sequence arbor, best-spacing-sequence with movable outer blade arbor,
all-blades-movable arbor, selective-rip arbor. It has the ability to cut to sizes specified to
the nearest 1/16- inch or millimeter, adjustable rip and chopsaw kerf sizes, and the capacity
to process up to 600 parts sizes in the cutting order. Its user interface and data output were
very flexible and were conceived to be adaptable to new processing technology and needs.
In 1997, Thomas wrote ROMI CROSS, a crosscut-first version of ROMI RIP that
included similar features. Both ROMI RIP and ROMI CROSS use innovative part
prioritization formulae (Thomas 1995a, ___1995b, ___ 1997, ___ 1999a, ___1999b) that
can optimize for part value – static or dynamic; for area – Length×Width, Length2×Wid th,
or Length2×Width×NEED; or for dynamic part prioritization – Simple Dynamic Exponent
or Complex Dynamic Exponent.
When doing value-based prioritization, one can assign a value for each part size.
While this strategy will prioritize higher-value parts – whose values are user-defined – it
will not consider the need for part quantities. One way of doing this is by decreasing the
parts dollar value each time a part is cut i.e., for a demand of N parts, the value of the part
is reduced 1/N each time a part is cut.
Three strategies are available for area-based part prioritization. The Length×Width
and Length2×Width will prioritize for area and for long parts respectively, without
considering part quantity. The Length2×Width×NEED strategy adds that essential cutting
27
order element, the amount of parts that are needed. When a part is cut the need for that part
size is reduced by one thereby, the part-priority decreases.
Dynamic part prioritization strategies assign each part size a priority based on its
size and on the required quantity. Thomas (1995a) developed a Simple Dynamic Exponent
(SDE) and a Complex Dynamic Exponent (CDE). The SDE prioritizes part length and
width equally by using a single weighting factor (WF) (SDE = (Length×Width)WF) where
the weighting factor considers need for the part. The CDE strategy weighs the relative
importance of each part in the cutting order e.g., if there were a few hard to obtain parts that
were needed then the program would hunt for the most opportunistic time to obtain them.
The CDE generates two different weighting factors for length and width respectively (CDE
=widthlength W FW F WidthLength × ). This system still prioritizes longer parts however it will also
prefer wider parts for different cuttings of the same length.
2.9 Yield
The yield obtained from lumber varies according to several factors (Anon. 1985).
These factors include: (1) lumber grade, (2) lumber grading rules, (3) lumber size, (4)
drying quality, (5) the cutting order (Araman 1978 and Buehlmann 1998), (6) part quality,
(7) the type of rough mill (i.e., crosscut-first or rip-first) used (Mullin 1990, Gatchell 1987,
Gatchell et al. 1983, Hallock and Giese 1980, Araman 1978, and Hall 1978, Hall et al.
1980). Within a rip -first rough mill, the arbor type (fixed, movable fence, mo ving outer
blade, and all movable blades) and the kerf will influence the optimization of the boards
cutting units and limitation of waste, (8) operator skill and motivation, and (9) grade mix
and sorting (Gazo 1994).
Although smaller cutting order parts increase yield (Buehlmann 1998), it is the
selection of crosscut-first or rip - first that will affect component distribution. Hall et al.
(1980) sorted yield according to cutting order and to cutup process. Their cutting order was
28
comprised of six cutting lengths: 13, 26, 31, 37, 43, and 49 inches long. Overall, it was
found that there was no significant difference in yield whether the lumber was crosscut or
ripped first. However, a significant difference in yield was observed in the 49-inch long
pieces when the effect of the cutup process on each of the different lengths was analyzed.
Also, the rip-first process produced more of the extreme lengths while the crosscut- first line
produced more of the middle cutting lengths. Hallock and Giese (1980a) performed a
similar comparison and they concluded that one could expect higher overall cutting yields
from all grades of lumber using rip- first cutup method. Studies by Araman (1978) and
Lucas and Araman (1975) showed that rip -first yields were higher than crosscut-first but
only when the ripped strips were finger-jointed or when the longest possible random
lengths were cut.
2.10 Effect of lumber length
While cutting order, type of rough mill, grades, grade mix and sorting were
explored; little has been done on effect of length. Wiedenbeck (1992) studied the
feasibility of using short-length lumber (four to eight feet) in the rough mill by analyzing
the costs involved in its processing. Hamner et al. (2002) studied the effect of length
within NHLA grade lumber.
Wiendenbeck (1992) studied the different aspects involved in using short- length
lumber in the furniture and cabinet industries including the differences in lumber
characteristics between lumber length groups, and the effect of lumber length on random
width dimension yields.
In their study, Hamner et al. (2002) analyzed the effect of eight- to sixteen-foot
long lumber on yield. Their study focused on rip- first rough mill yield using ROMI-RIP
2.0, the USDA Forest Service 1998 Red Oak Database (Gatchell et al. 1998), and the
USDA Forest Service “Hard” and “Easy” standard cutting orders (Gatchell et al. 1999).
29
The “Hard” cutting order requires longer parts and parts cannot be glued-up into panels
while the “Easy” cutting order has smaller parts and glue -up is permitted. Results from
simulations indicate that the longer lumber has a higher yield by about 5 percent.
2.11 Summary
From the above observations it is apparent that the most efficient way of evaluating
a species remanufacturing potential is by using a computer. To do this, a digitized database
must be created. In this database, one must find all characteristic features and this is
because of the evolving nature of the value-added products market i.e., what might
currently be acceptable for one manufacturer/market may or may not be for another. What
is currently acceptable may or may not be in the future.
Prior work has characterized many species (hard maple, walnut, alder, red oak,
yellow poplar, ponderosa pine, southern yellow pine, and radiata pine) in order to obtain
yield where a certain surface of lumber was digitized and then processed using a program
that would proceed to successively place clear cuttings in a board until no more cuttings
could be obtained for a specific lumber grade.
Nordin (1990), Harding et al, (1993) and then Gazo et al. (1998) utilized the
database to analyze the incidence of defects in terms of defect frequency and average defect
area. This gave insight into the type of defect that occurred most frequently, their average
size, and how these defects will affect lumber yield – according to a specific cutting order.
Subsequent research using the above-mentioned databases has allowed studies of
the effect on crook (Gatchell 1990), various arbor configurations (Gatchell 1991), sorting,
throughput, and machine utilization (Gazo 1995), comparisons of the effects of crosscutting
before gang-ripping (Gatchell et al. 1996), within-grade quality differences (Gatchell and
Thomas 1997), and the inclusion of character marks into cutting order allowances
30
(Buehlmann et al. 1998a, ___1998b). Wiedenbeck et al. (1995) analyzed the potential use
of short- length lumber in the rough mill however no work has yet been done with regards
to comparing the yield obtained from processing conventional-length logs and short- length
logs with regards to a specific process or cutting order. This information is of interest
because of the increasing cost of acquiring conventional- length lumber, no matter what the
species.
31
References
Anon. 1985. Improving yield. Furniture design and manufacturing. 57(12).
Araman P.A. 1978. A comparison of four techniques for producing high-grade furniture core material from low-grade yellow-poplar. USDA Forest Service, Research Paper NE-429. Northeastern Forest Experiment Station, Broomall, PA.
Araman, P.A., C.J. Gatchell, and H. W. Reynolds. 1982. Meeting the solid wood needs of the furniture and cabinet industries: standard-size Hardwood blanks. USDA Forest Service, Research Paper NE-494. Northeastern Forest Experiment Station, Broomall, PA.
Beauregard, R., R. Gazo, M.O. Kimberley, J. Turner, S. Mitchell, and A. Shelbourne. 1999. Clonal Variation in the Quality of Radiata Pine Random Width Boards. Wood and Fiber Science 31(3):222-234.
Brière, J. 1992. Critère techniques d’utilisation du bouleau blanc. Unpublished Master’s Thesis, Laval University, Québec, Canada.
Brockman, C.F. 1968. Trees of North America. Golden Press, New-York, NY, 1968.
Brunner, C.C. 1984. CORY - a computer program to determine furniture cutting yields for both rip-first and crosscut-first sawing sequences. Unpublished doctoral dissertation, Virginia Polytechnic Institute and State University, Blacksburg, VA.
Buehlmann, U. 1998. Understanding the relationship of lumber yield and cutting bill requirements: a statistical approach. Unpublished doctoral dissertation, Virginia Polytechnic Institute and State University, Blacksburg, VA.
Buehlmann, U., D.E. Kline, and J.K. Wiedenbeck. 1998a. Cutting bill requirements and lumber yield: The influence of cutting bill requirements on lumber yield. Annual meeting of the Forest Products Society. Technical Forum Presentation, Merida, June 23, 1998.
32
Buehlmann, U., J.K. Wiedenbeck, and D.E. Kline. 1998b. Character-marked furniture: potential for lumber yield increase in rip-first rough mills. Forest Products Journal 48(4):43-50.
BWCA (Boundary Waters Canoe Area). 1999. Betula papyrifera. http://www.rook.org/earl/bwca/nature/trees/betulapap.html. (8 November 1999)
Carnieri, C., G.A. Mendoza, and W.G. Luppold. 1993. Optimal cutting of dimension parts from lumber with a defect: a heuristic solution procedure. Forest Products Journal 43(9):66-72.
Cassens, D.L, and B.C. Fischer. 1992. Hardwood log grades and lumber grades: Is there a relationship? Purdue University. Department of Forestry and Natural Resources. FNR-84. 4 p.
Fowells, H.A. 1965. Silvics of forest trees of the United States. USDA Forest Service. Agricultural handbook no. 271, Division of timber management research, Washington. 762 p.
Gatchell, C.J., R.B. Anderson, and P.A. Araman. 1983. Effect of gang-ripping width on C1F yields from No. 2 Common Oak Lumber. Forest Products Journal 33(6):43-48.
Gatchell, C.J. 1987. Rethinking the design of the furniture rough mill. Forest Products Journal 31(3):8-14.
Gatchell, C.J. 1990. The effect of crook on yields when processing narrow lumber with a fixed arbor gang ripsaw. Forest Products Journal 40(5):9 -17.
Gatchell, C.J. 1991. Yield comparisons from floating blade and fixed arbor gang ripsaws when processing boards before and after crook removal. Forest Products Journal 41(5):9-17.
Gatchell, C.J., P. Klinkhachorn, and R. Kothari. 1992a. ReGS—a realistic grading system. Forest Products Journal 42(10):37-40.
Gatchell, C.J., R.E. Thomas, and E.S. Walker. 1992b. 1992 Data bank for red oak lumber. USDA Forest Service, Research Paper NE-669. Northeastern Forest Experiment Station, Radnor, PA.
33
Gatchell, C.J., J.K. Wiedenbeck, and E.S. Walker. 1993. A red oak data bank for computer simulations of secondary processing. Forest Products Journal 43(6):38-42.
Gatchell, C.J., J.K. Wiedenbeck, and E.S. Walker. 1995. Understanding that red oak lumber has a better and worse end. Forest Products Journal 45(4):54-60.
Gatchell, C.J., J.K. Wiedenbeck, and E.S. Walker. 1996. The effects of crosscutting before gang-ripping on dimension part yields from No. 1 and 2A Common red oak lumber. Forest Products Journal 46(2):61-66.
Gatchell, C.J., and R.E. Thomas. 1997. Within-grade quality differences for 1 and 2A Common lumber affect processing and yields when gang-ripping red oak lumber. Forest Products Journal 47(10):85-90.
Gatchell, C.J., R.E. Thomas, and E.S. Walker. 1998. 1998 Data bank for kiln -dried red oak lumber. USDA Forest Service, General Technical Report NE-245. Northeastern Forest Experiment Station, Radnor, PA.
Gatchell, C.J. R.E. Thomas, E.S. Walker. 1999. Effects of preprocessing 1 common and 2A common red oak lumber on gang-rip-first rough-mill dimension part yields. Forest Products Journal. 49(3):53-60.
Gazo, R. 1994. Rough Mill Analysis and Modeling. Unpublished doctoral dissertation, Mississippi State University, MS.
Gazo, R., P.H. Steele. 1995. Rough Mill Analysis Model. Forest Products Journal 45(4):51-53.
Gazo, R, S. Mitchell, and R. Beauregard. 1998. Development of a database and its use to quantify incidence of defects in radiata pine random width boards. New Zealand Journal of Forestry Science 28(1).
Gazo, R. R. Beauregard, M. Kimberley, D. McConchie. 2000. Incidence of defects by tree characteristics in radiata pine random-width boards. Forest Products Journal. 50(6):83-89.
34
Giese, P.J., and K.A. McDonald. 1982. OPTYLD – a multiple rip-first computer program to maximize cutting yields. USDA Forest Service, Research Paper FPL-412. Forest Products Laboratory, Madison, WI.
Giese, P.J., and J.D. Danielson. 1983. CROMAX. A Crosscut-First Computer Simulation Program to Determine Cutting Yield. USDA Forest Service, General Technical Report FPL-38. Forest Products Laboratory, Madison, WI.
Giguère, M. 1998. Guide du sciage des billons de feuillus durs [A guide to sawing short-log hardwood (in French)]. Direction of the Forest.Products Development, Ministry of Natural Resources, Government of Québec. 27 p.
Gilmore, R.C., S.J. Hanover, and J.D. Danielson. 1984. Dimension yields for yellow-poplar lumber. USDA Forest Service, General Technical Report FPL-41. Forest Products Laboratory, Madison, WI.
Hall, S.P. 1978. The effects of one cross-cut-first line and one rip first -line on rough mill costs and yields. Unpublished master’s thesis, Virginia Polytechnic Institute and State University, Blacksburg, VA.
Hall, S.P., R.A. Wysk, E.M. Wengert, and M.H. Agee. 1980. Yield distribution and cost comparisons of a crosscut-first and a gang-rip-first rough mill producing hardwood dimension stair parts. Forest Products Journal 30(5):34-39.
Hallock, H., and L. Galiger. 1971. Grading hardwood lumber by computer. USDA Forest Service, Research Paper FPL 157. Forest Products Laboratory, Madison, Wis.
Hallock, H., and P. Giese. 1980. Does Gang Ripping Hold the Potential for Higher Clear Cutting Yields. USDA Forest Service, Research Paper FPL 369. Forest Products Laboratory, Madison, Wis.
Hamner, P, B. Bond, J.K. Wiedenbeck. 2002. The effects of lumber length on parts yield in gang-rip-first roughmills. Forest Products Journal (in press).
Hansen, B.G., and H.W. Reynolds. 1984. System 6 Alternatives: An Economic Analysis. USDA Forest Service, Research paper NE-551. Northeastern Forest Experiment Station, Broomall, PA.
35
Harding, O.V. 1991. Development of a decision software system to compare rip-first and crosscut-first yields. Unpublished Doctoral dissertation, Mississippi State University, MS.
Harding, O.V., P.H, Steele, and K. Nordin. 1993. Description of defects by type for six grades of red oak lumber. Forest Products Journal 43(6):45-50.
Harding, O . V., P. H. Steele. 1997. RIP-X: Decision software to compare crosscut-first and rip-first rough mill systems. Wood Science and Technology 31(1997):367-381.
Hassler, C.C, P.A. Araman, S.A. Sinclair, and H.W. Reynolds. 1995. A normative analysis of a System-6 mill. Forest Products Journal 35(11/12):43-48.
HDMA (Hardwood Dimension Manufacturers Association). 1961. Rules for Measurement and Inspection of Hardwood Dimension Parts, Hardwood Interior Trim and Moldings, Hardwood Stair Treads and Risers. 5th Editio n. Nashville, TN.
Hoff, K.G., E.L. Adams, and E.S. Walker. 1991. GR-1ST: PC Program for Evaluating Gang-Rip-First Board Cut-Up Procedures. USDA Forest Service, General Technical Report NE-150. Northeastern Forest Experiment Station, Radnor, PA.
Hyvarinien, M.J. 1968. Paper birch – its characteristics, properties and uses – a review of recent literature. USDA Forest Service. Research Paper NC-22.
Kline, D.E., C. Regalado, E.M. Wengert, F.M. Lamb, and P.A. Araman. 1993. Effect of hardwood sawmill edging and t rimming practices on furniture part production. Forest Products Journal 43(3):22-26.
Klinkhachorn, P., C.J. Schwehm, C.W. McMillin, and H.A. Huber. 1989a. HaLT: a computerized training program for hardwood lumber graders. Forest Products Journal 39(2):38-40.
Klinkhachorn, P., J.P. Franklin, C.W. McMillin, and H.A. Huber. 1989b. ALPS: yield optimization cutting program. Forest Products Journal 39(3):53-56.
Klinkhachorn, P., C. Gatchell, C.W. McMillin, R. Kothari, and D. Yost. 1992a. HaLT2—an enhances lumber grading trainer. Forest Products Journal 42(10):32-36.
36
Klinkhachorn, P., R. Kothari, D. Yost, and P. Araman. 1992b. Enhancement of the computer lumber grading program to support polygonal defects. Forest Products Journal 42(10):41-46.
Klinkhachorn, P., R. Kothari, R. Annavajjhala, and C.W. McMillin. 1994. TRSys: a hardwood lumber grading Training and Remanufacturing System. Forest Products Journal 44(9):68-72.
Lortie, M. 1979. Arbres, forets et perturbations naturelles au Québec. Presses de l’Université Laval, Québec.
Lucas, E.L., and L.R.R. Catron. 1973. A comprehensive defect data bank for No. 2 Common oak lumber. USDA Forest Service, Research Paper NE-262. Northeastern Forest Experiment Station, Upper Darby, Pa.
Lucas, E.L., and P.A. Araman. 1975. Manufacturing interior furniture parts: a new look at an old problem. USDA Forest Service, Research paper NE-334. Northeastern Forest Experiment Station, Upper Darby, Pa.
Luppold, W.G. 1994. Are perceived shortages of hardwood lumber real? The Northern Logger & Timber Processor 43(3):12-14,48.
Manalan, B.A., G.R. Wells, and H.A. Core. 1980. Yield deviations in hardwood dimension stock cutting bills. Forest Products Journal 40(1):40-42.
Marie-Victorin, Frère. 1964. La flore laurentienne. Les presses de l’Université de Montréal. Montréal. 925 p.
Maristany, A.G., C.C. Brunner, and J.D. Anderson. 1990. Effect of an exponential weighting function on random width dimension yield. Unpublished Report. Oregon State University, Corvallis, Oregon.
Martodam, D. 1982. White birch resource analysis: evaluation of Minnesota white birch potential use in a System 6 manufacturing process. State of Minnesota Department of Natural Resources, Division of forestry. Grand Rapids, MN.
37
McDonald, K.A., R.O. Woodfin, and P. J. Giese. 1981. 5/4 Ponderosa pine Shop grade cutting yields. USDA Forest Service, Research Paper FPL 394. Forest Products Laboratory, Madison, WI.
McMillin, C.W., R.W. Conners, H.A. Huber. 1984. ALPS—A potential new automated lumber processing system . Forest Products Journal 34(1):13-20.
MNRQ (Ministry of Natural Resources, Government of Québec). 1996. Ressources et industrie forestière. Portrait statistique [Forest Resource and Industry. A statistical portrait (in French)]. Ministry of Natural Resources, Government of Québec. 27 p.
Moody, J, C.J. Gatchell, E.S. Walker, P. Klinkhachorn. 1998. An introduction to UGRS: The ultimate grading and remanufacturing system. Forest Products Journal 48(9):45-50.
Mullin, S. 1990. Why switch to rip first? FDM – Furniture Design & Manufacturing. 62(9):36-42.
National Hardwood Lumber Association (NHLA). 1998. Rules for the measurement and inspection of hardwood and cypress. Memphis, TN.
Nordin, K. 1990. Incidence of defects in red oak lumber. Unpublished Master’s Thesis, Mississippi State University, MS.
Osborn, L.E., C.J. Gatchell, and C.C. Hassler. 1992. West Virginia Yellow-Poplar Lumber Defect Database. USDA Forest Service, General Technical Report NE-660. Northeastern Forest Experiment Station, Radnor, PA.
Panshin, A.J., and C. de Zeeuw. 1980. Textbook of Wood Technology. McGraw-Hill. Pp 722.
Petro, F.J., and W.W. Calvert. 1990. How to grade hardwood logs for factory lumber. Forintek Canada Corp., Eastern Laboratory. Ottawa, Ontario. 69p.
Quigley, K.L., and H.M. Babcock. 1969. Birch timber resources of North America. In: Birch symposium proceedings, USDA Forest Service. Northeastern Forest Experiment Station. P6-14.
38
Regalado, C., D.E. Kline, P. A. Araman. 1992. Optimum edging and trimming of hardwood lumber. Forest Products Journal 42(2):8-14.
Reynolds, H.W., and C.J. Gatchell. 1979a. Marketing low-grade hardwoods for furniture stock – a new approach. USDA Forest Service, Research Paper NE-444. Northeastern Forest Experiment Station, Broomall, PA.
Reynolds, H.W., and C.J. Gatchell. 1979b. New Technology for Low-Grade hardwood Utilization: System 6. USDA Forest Service, Research Paper NE-504. Northeastern Forest Experiment Station, Broomall, PA.
Reynolds, H.W., and P.A. Araman. 1983. System 6: making frame-quality blanks from white oak thinnings. USDA Forest Service, Research Paper NE-520. Northeastern Forest Experiment Station, Broomall, PA.
Reynolds, H.W., P.A. Araman, C.J. Gatchell, and B.G. Hansen. 1983. System 6 use to make kitchen cabinet C2F blanks from small-diameter, low-grade red oak.. USDA Forest Service, Research Paper NE-525. Northeastern Forest Experiment Station, Broomall, PA.
Reynolds, H.W. 1984. System 6: rough mill operating manual. USDA Forest Service, Research Paper NE-542. Northeastern Forest Experiment Station, Broomall, PA.
Reynolds, H.W., and B.G. Hansen. 1984. A sample plant design for System 6. USDA Forest Service, General Technical Report NE-87. Northeastern Forest Experiment Station, Broomall, PA.
Reynolds, H.W. 1985. System 6: Chips Versus Blanks Program. USDA Forest Service, General Technical Report NE-106. Northeastern Forest Experiment Station, Broomall, PA.
Ruddell, S., H.A. Huber, and P. Klinkhachorn. 1990. A comparison of two roughmill cutting models. Forest Products Journal 40(5):27-30.
Schumann, D.R., and G.H. Englerth. 1967a. Yields of random width dimension from 4/4 hard maple lumber. USDA Forest Service, Research Paper 81. Forest Products Laboratory, Madison, WI.
39
Schumann, D.R., and G.H. Englerth. 1967b. Dimension stock yields of specific width cuttings f rom 4/4 hard maple lumber. USDA Forest Service, Research Paper FPL-85. Forest Products Laboratory, Madison, WI.
Schumann, D.R. 1971. Dimension yields from Black Walnut lumber. USDA Forest Service, Research Paper FPL-162. Forest Products Laboratory, Madison, WI.
Schumann, D.R. 1972. Dimension yields from alder lumber. USDA Forest Service, Research Paper FPL-170. Forest Products Laboratory, Madison, WI.
Schwehm, C.J., P. Klinkhachorn, C.W. McMillin, and H.A. Huber. 1990. HaRem: Hardwood lumber Remanufacturing program for maximizing value based on size, grade, and current market prices. Forest Products Journal 40(7/8):27-30.
Steele, P.H., S. Lee. 1994. Yield comparisons of furniture parts for three gang-ripping systems. Forest Products Journal 44(3):9 -16.
Stern, A.R. 1978. MULRIP. USDA Forest Service, Unpublished computer program. Forest Products Laboratory, Madison, WI.
Stern, A.R., and K.A. McDonald. 1978. Computer optimization of cutting yield from multiple-ripped boards. USDA Forest Service, Research Paper FPL-318. Forest Products Laboratory, Madison, WI.
Thomas, R.E., C.J. Gatchell, and E. S. Walker. 1994. User’s guide to AGARIS: Advanced GAng RIp Simulator. USDA Forest Service, General Technical Report NE-192. Northeastern Forest Experiment Station, Radnor, PA.
Thomas, R.E. 1995a. ROMI RIP: ROugh MIll RIP-first simulator user’s guide. USDA Forest Service, General Technical Report NE-202. Northeastern Forest Experiment Station, Radnor, PA.
Thomas, R.E. 1995b. ROMI RIP: ROugh MIll RIP-first simulator. USDA Forest Service, General Technical Report NE-206. Northeastern Forest Experiment Station, Radnor, PA.
40
Thomas, R.E. 1997. ROMI-CROSS: ROugh Mill CROSScut-first simulator. USDA Forest Service, General Technical Report NE-229. Northeastern Forest Experiment Station, Radnor, PA.
Thomas, R.E. 1999a. ROMI RIP 2.0 user’s guide: ROugh MIll RIP-first simulator. USDA Forest Service, General Technical Report NE-259. Northeastern Forest Experiment Station, Radnor, PA.
Thomas, R.E. 1999b. ROMI RIP Version 2.0: A New Analysis Tool for Rip-First Rough-Mill Operations. Forest Products Journal 49(5):35-40.
Thomas, R.J. 1962. The rough-end yield research program . Forest Products Journal 12(11):536-537.
Todoroki, C.L. 1996. Grading Random-Width Lumber by Computer. New Zealand Jour nal of Forest Science 25(3):367-378.
USDA Forest Products Laboratory. 1953.Nailing dense hardwoods. USDA Forest Service. Technical note 247. Forest products laboratory, Madison, WI.
USDA Forest Products Laboratory. 1999. Wood Handbook: Wood As an Engineering Material. Forest Products Society. Madison, WI.
Vaughan, C.L., A.C. Wollin, K.A. McDonald, and E.H. Bulgrin. 1966. Hardwood log frades for standard lumber. USDA Forest Service, Research Paper FPL 63. Forest Products Laboratory, Madison, WI. 52 p.
Verkasalo, E. 1996. Evaluating the potential of European white birch (Betula pubescens) for veneer and plywood by timber and wood quality. From Proceedings: IUFRO, Kruger national Park, South Africa: August 26-31 1996.
Wiedenbeck, J.K. 1992. The potential for short length lumber in the furniture and cabinet industries. Ph.D. dissertation, Virginia Polytechnic Institute and State University. Blacksburg, VA.
Wiedenbeck, J.K., C.J. Gatchell, and E. S. Walker. 1994. Data Bank for Short -Length Red Oak Lumber. USDA Forest Service, Research Paper NE-695. Northeastern Forest Experiment Station, Radnor, PA.
41
Wiedenbeck, J.K., and P.A. Araman. 1995. Rough mill simulations reveal that productivity when processing short lumber can be high. Forest Products Journal 45(1):40-46.
Wiedenbeck, J. K., C. J. Gatchell, and E. S. Walker. 1995. Quality characteristics of Appalachian red oak lumber. Forest Products Journal 45(3):45-50.
Wodzinski, C., and E. Hahm. 1966. A computer program to determine yields of lumber. USDA Forest Service, FPL Unnumbered Publication. Forest Products Laboratory, Madison, WI.
WWPA. 1991. Western Lumber Grading Rules 91. Western Wood Products Association. Portland, Oregon.
42
3 WHAT IS THE YIELD OF SHORT-LENGTH WHITE BIRCH LUMBER?
Abstract
This study analyzes the potential use of short- length (less than 8-foot- long) white
birch lumber in the furniture industry. A database of random width and length white birch
boards containing information on all grade defects was developed for use in the simulation.
For the purpose of this study, 13.16 m3 (5,576 bf) of NHLA-graded lumber were used
including conventional length lumber (2.73 m3/1,157 bf of Select, 2.15 m3/911 bf
No.1 Common and 2.06 m3/873 bf No.2A Common), and short-length lumber (2.27
m3/962 bf of Select, 2.29 m3/970 bf of No.1 Common and 1.66 m3/703 bf of
No.2A Common). No FAS lumber was included. The effects of lumber length, grade,
cutting order and processing method on yield were analyzed. ROMI-RIP and ROMI-
CROSS simulation software were used to model two traditional processing methods, rip -
first and crosscut-first, respectively. Four cutting orders, Furniture, Panel, USDA Easy and
USDA Tough were simulation-processed.
Highly significant yield differences of 8.8% for Select and 10.3% for
No. 2A Common were observed between conventional and short-length lumber. These
differences can be explained by: a) a shorter average length (i.e. the longer conventional-
length lumber offers a greater number part combinations) and, b) the increased presence of
wane and void in short-length lumber. Results indicate, however, that there is little
difference in yield, when comparing No.1 Common short-length to conventional-length
lumber with appropriate cutting orders. Results also indicate that crosscut- first rough
milling generates, on average, a 4.2% higher yield than rip - first rough milling. This
analysis is of special interest to a value-added industry faced with scarcity and increasing
cost of high quality lumber.
43
Keywords: White birch, short-length, conventional length, rough mill, crosscut-first, rip-
first, yield, cutting order, grade
44
3.1 Introduction
The objective of this study is to evaluate the effect of lumber length on yield in
manufacturing furniture components. This evaluation is made using three grades (Select,
No.1 Common, No.2A Common) of white birch (Betula papyrifera, Marsh.) with two
processing methods (rip-first, crosscut- first) and four industry-related cutting orders.
Inventory data indicates that white birch is the last commercially available species,
on a sustainable basis, for industry expansion in Québec, Canada. There are over
5,300,000 m3 available (Giguère 1998, MNRQ 1996) per year, yet large quantities of this
species are left standing because the stems are generally considered too small to be an
economically viable source for conventional hardwood sawmills. In general, the log
diameter is either too small or the length is too short for traditional machinery (Bingham
1976) or the lumber produced is not covered by conventional grading rules (Wiedenbeck et
al. 1994).
Because of the increased demand for hardwoods over recent years, traditional
hardwoods are becoming scarce. This shortage has increased prices to the point that
previously non-profitable merchantable timber operations can now be considered, and
sawyers are fitting their production to the needs of furniture and other component
manufacturers. Since most of the components needed are of small dimension or panel parts
(Araman et al. 1982), a number of sawmills are tailoring their production to meet
customer-specific needs instead of sticking to a conventional standard. Increasing numbers
among them produce “in-house” graded parts to match end user requirements.
In the past, questions about yield, processing methods, parts distribution, etc. were
answered by computer modeling tools that utilized databases of digitized lumber (Anderson
et al. 1993, Araman 1977, Buehlmann et al. 1998, Gatchell et al. 1993, Gatchell et al.
1995, Gazo and Steele 1995, Harding 1991, Harding et al. 1993, Steele and Gazo 1995,
45
Steele and Lee 1994, Steele et al. 1999, Wiedenbeck et al. 1995, Wiedenbeck 2001).
Similar techniques can help answer similar questions about white birch. To do this, a
computer database was built on a sample consisting of 5,576 board feet contained in 1,613
boards of digitized random width and random length white birch. The data acquisition was
based on the methodology applied by Gazo et al. (1998) and Harding (1991).
Traditionally, lumber is graded on the poor face; however, some manufacturers use
only the best face for their products (e.g. tabletops or flooring); therefore, information on
what grade they purchase does not tell the whole story with regards to yield or cost per part.
Using the database in combination with rough mill simulation software such as ROMI-RIP
(Thomas 1999) and ROMI-CROSS (Thomas 1997) allows us to better fit the lumber grade
to the manufacturer’s needs.
3.2 Methodology
3.2.1 Sample Material
The boards selected for this study were required to show a range of qualities typical
of what is currently available in Québec. Two sawmills were chosen. The first sawmill
processes conventional logs into National Hardwood Lumber Association (NHLA) grade
lumber, whereas the second processes short-length logs into NHLA and “house” grades of
lumber. Petro and Calvert (1990) describe a grading system for conventional logs – these
are logs that are of such a size and have sufficiently few defects that they can be sawn into
NHLA lumber. A large number of clear cuttings in lengths of 8 feet or more can typically
be obtained from these boards. On the other hand, short-length logs do not conform to the
criteria defined by Petro and Calvert (1990) because they are too short, too crooked, of too
small a diameter, or present a combination of these characteristics. These logs, often
46
having a length between 4 and 8 feet are increasingly considered fit for sawing, but are
generally classified as pulpwood (Calvert 1965).
Table 3.1 tallies the number of boards analyzed per grade for each of the two
sawmills included in the study. One sawmill, located at Senneterre, Québec, provided 659
boards (6.94 m3/2,941 bf) of conventionally sawn white birch. The other, located at Ste-
Monique, Québec, provided 954 boards (6.22 m3/2,635 bf) of short- length white birch
lumber for a total of 1,613 (13.16 m3/5,576 bf) random width and length boards. The
lumber from both sawmills came from comparable mixed hardwood-softwood stands
distinctive of the Laurentian shield. All boards were dried in a commercial kiln using high
temperature schedule No. 23 from Cech and Pfaff (1980) and surfaced, on both faces, at
Forintek Canada Corp., Québec, to allow easier defect identification for digitizing. The
digitizing took place at Purdue University, West Lafayette, IN.
3.2.2 Board Grading
A large volume of random width and random length hardwood factory lumber
produced in Québec is used in furniture, cabinetry and flooring industries. This lumber is
graded using the National Hardwood Lumber Association lumber grading rules (NHLA
1998). Under these rules, the lumber is graded according to the potential recovery of clear
cuttings that can be obtained. In order to determine the lumber grade, areas of potential
clear cuttings are considered. The size of cuttings, number of cuts allowed, percentage of
clear cutting area on the entire board and size of board define the NHLA grading rules for
Factory Lumber. As the boards are intended for subsequent remanufacturing into flooring,
furniture and cabinetry, individual cuttings must satisfy both size and quality criteria.
47
Under the NHLA rules, the lumber is graded into six factory lumber grades,
namely, FAS, F1F, Select, No.1 Common, No.2A Common, and No. 3A Common. FAS,
F1F, and No. 3A Common were not considered for this analysis because they are not used
in the furniture-parts market segment under study. There are four basic requirements, one
of which is percentage of clear cutting area of entire board. Select grade boards have 83%
clear area on good face, whereas No.1 Common boards require only 66.7% and No. 2A
Common boards require 50% of clear wood. A detailed account of the grading rules is
given in the NHLA grading rulebook (NHLA 1998).
Prior to digitizing, all the boards were manually graded by an experienced grader
both before and after surfacing in order to insure that the grade quality was accurate. All
lumber was graded according to the NHLA grading rules.
3.2.3 Database
A database of 2.73 m3 (1,157 bf) random width and random length boards
containing information on all grade defects was developed. The lumber used for this study
consisted of 2.73 m3 (1,157 bf) of Select, 2.15 m3 (912 bf) No.1 Common, and 2.06 m3
(873 bf) No.2A Common and, 2.27 m3 (962 bf) Select, 2.29 m3 (970 bf) No.1 Common,
and 1.66 m3 (703 bf) No.2A Common NHLA-graded lumber. Table 3.1 lists the quantity,
average width, length, and maximum crook, for each of the above-mentioned lumber
grades.
48
Table 3.1. White birch database characteristics
Grade Volume
(bf / m3) Number
of Boards
Width Average
(m)
Length Average
(m)
Crook Average
Max. (mm)
Clear wood (%)
---------------------------- Conventional ---------------------------- Select 1,157 / 2.73 183 0.165
(0.040) 3.560 (0.258)
7.9 (5.2)
92.7 (4.3 )
No. 1C 911 / 2.15 241 0.141 (0.032)
2.475 (0.415)
6.6 (3.8)
90.9 (7.6 )
No. 2AC 873 / 2.06 235 0.140 (0.027)
2.456 (0.368)
7.2 (4.5)
89.3 (9.6 )
---------------------------- Short-length ---------------------------- Select 962 / 2.27 312 0.134
(0.030) 2.120 (0.246)
5.5 (3.8)
91.1 (7.6 )
No. 1C 970 / 2.29 292 0.152 (0.032)
2.030 (0.405)
5.2 (3.3)
91.3 (9.8 )
No. 2AC 703 / 1.66 350 0.124 (0.027)
1.490 (0.347)
4.5 (2.6)
90.9 (8.2 )
Standard deviation in parentheses
Table 3.2 lists the defects that were digitized, the correspondence with ROMI-RIP
and ROMI-CROSS defects, their identification number, and their status in the simulations.
Certain defects were filtered out of the database because they were acceptable on both sides
of the component or because they did not have to do with the species characteristics but
rather with processing (i.e. manufacturing defects). All sound knots and stain were
considered acceptable on the back side only and were defined accordingly in the rough mill
simulation software.
49
Table 3.2. List of digitized defects, their simulation program name and number equivale nts and their status
Defect Type Corresponding ROMI Defect Type
ROMI-RIP No.
ROMI-CROSS
No.
Status *
Natural Defects Bark pocket All bark pockets 1099 5 x
Burl Burl 13 23 a Compression failure Callus wood 21 41 a
Crook Void 2 n/a x Decay Decay 4 n/a x
Heartwood Bud trace with bark/check 20 40 a Hole All grub holes/Holes 1199 9 x
Loose knot All unsound knots 1299 13 x Mineral streak Sapstain/mineral streak 22 42 a
Open knot All unsound knots 1299 13 x Pin knot Pin worm hole (1/16”) 211 46 a
Pith Pith 3 2 x Shake Shake 5 n/a x
Sound knot All sound knots 1599 25 p Spike knot All unsound knots 1299 13 x Split knot All unsound knots 1299 13 x
Split Split 24 44 x Stain Incipient decay 18 38 p Void Void 2 n/a x
Wane Void 2 n/a x Twisted grain Burl 13 23 a
Mechanical Defects Drying check Surface check 14 24 a
Pressure roller stain Sticker stain 19 39 a Conveyor mark Mechanical damage 1 1 a Machine burn Sticker stain 19 39 a
Machine gouge Mechanical damage 1 1 a Spike mark Machine snipe/tearout 17 37 a
*Status: x = unacceptable on either side; a = acceptable on both sides;
p = acceptable on poor side
50
3.2.4 Cutting order
Four different cutting bills, USDA Easy (Table 3.3), USDA Tough (Table 3.4), Furniture
(Table 3.5), and Panel were used to best estimate the effect of lumber length and grade on
yield. The USDA cutting bills were taken from Steele et al. (1999). Cutting order
characteristics important for interpreting the results included the total number of parts, the
average length and width of the parts, and the board footage of parts required by the cutting
order. The Easy cutting order (Table 3.3) has an average length of 545 mm and width of
56.5 mm that is shorter and narrower then the Tough cutting order (987-mm-long and 76.5-
mm-wide) (Table 3.4). The Furniture (Table 3.5) and Panel cutting bills are from actual
Canadian furniture manufacturers using white birch lumber in their operations. The
Furniture cutting order was obtained from a rough mill that produced pre-cut components
and panel parts for several furniture plants. This cutting order has an average length of
803 mm and width of 36.2 mm. The specified cutting order is representative of the
production of buffet and hutch types of dining room furniture where relatively large pieces
are needed.
51
Table 3.3. USDA "Easy" cutting order (Adapted from Steele et al. (1999))
Widths (mm)
Length 44 51 57 95 114 127 133 (mm) 254 2 311 1 4 330 4 343 1 368 2 1 381 11 20 2 476 8 4 8 521 2 1 9 533 3 572 8 629 23 13 23 705 10 718 14 4 800 2
Table 3.4. USDA "Tough" cutting order (Adapted from Steele et al. (1999))
Widths (mm)
Length 51 70 89 108 (mm) 381 7 4 5 457 2 635 5 5 737 8 838 6 965 5
1143 12 1270 8 12 4 1524 2 1829 3 6
52
Table 3.5. Furniture cutting order
Widths (mm)
Length 25 32 38 44 51 57 64 70 76 (mm) 362 5 7 387 36 8 3 2 1 1 1 5 451 42 10 4 2 1 1 1 514 57 13 5 3 2 1 1 10 584 9 2 1 1 20
768 29 7 3 2 1 1
914 49 11 5 3 2 1 1 5
1073 51 12 5 3 2 1 1 8 35 1175 8 4 1
1245 24 6 2 1 1 1 4
1295 13 3 1 1
1346 19 4 2 1 1
The Panel cutting order comes from a plant that produces solid wood panels of
specific lengths. To overcome the inability of ROMI-RIP Version 2 in producing solely
panel parts, it was decided to divide the 25 to 114 mm (1 to 4.5 inches) width range in 6.3
mm (0.25-inch) increments and request an infinite number of each component.
For the Panel cutting order, an infinite demand of all combinations of the following
widths and lengths was used, namely, widths of 25, 32, 38, 44, 51, 57, 64, 70, 76, 83, 89,
95, 102, and 114 mm and lengths of 445*, 546*, 749*, 940, 991, 1041, 1092, 1143, 1245,
1372, and 1549 mm.
* Salvage length
53
3.2.5 Simulation Parameters
The following parameters were used for the rip-first and crosscut-first simulations.
These settings were designed to obtain the highest possible yield and are based on the best
available rough mill technology.
3.2.5.1 ROMI-RIP simulation parameters:
• Arbor type: All-blades movable arbor with 6 spacings; • Kerf: 4 mm; • Prioritization strategy: complex dynamic exponent (CDE); • Part prioritization: updated constantly for all cutting orders except for Panel cutting
order, which was never updated; • Salvage cuts: Made to primary part dimensions, except in Panel cutting order,
where three lengths were salvage-specific.
3.2.5.2 ROMI CROSS simulation parameters:
• Primary yield maximization method: Crosscuts optimized for best length fitting to board features;
• Kerf: 4mm; • Prioritization strategy: complex dynamic exponent (CDE); • Part prioritization: updated constantly for all cutting orders except for Panel cutting
order, which was never updated; • Salvage cuts: Made to primary part dimensions, except in Panel cutting order,
where three lengths were salvage-specific.
54
3.3 Results and Discussion
3.3.1 Database
Table 3.6 and Table 3.7 list the defect frequency and defect area respectively. It
should be noted that due to the subjectivity in the identification of certain defects, it is
estimated that these data are 95 percent accurate, i.e., there is a 5 percent chance of
misidentifying or not recording a defect.
In Table 3.6, one notices an increased occurrence of bark pockets, sound knots, and
unsound knots with a decrease in grade quality. Table 3.7 indicates that these same defects
and decay occupy increasing surface area as quality diminishes. It also appears that short-
length lumber has more knots, in general, than conventional-length lumber. This is due to
the characteristics of the logs, where the short-length lumber comes from small-diameter
trees, which have not had time to outgrow their loss of branches. Short-length lumber also
has more wane and void than conventional- length lumber, due to the smaller log diameter.
As expected, the results in Table 3.7 establish that the better grades have more clear wood.
However, this table does not illustrate defect location, which is of utmost importance when
calculating yield. It should be noted that the clear wood area (%) represents the ratio of
board clear area (board area minus total defect area) to board surface area.
55
Table 3.6. Defect frequency
Bark Pocket Sound Knot Unsound Knots --------------- (No./m²) ---------------
1.9 0.0 0.4 Conv. (5.6) (0.2) (1.1) 1.8 0.7 0.5 Se
lect
Short
(4.5) (2.2) (1.8) p-value 0.42 0.00** 0.20
5.4 0.2 1.4 Conv. (10.1) (0.9) (2.8)
3.8 1.0 2.0
No.
1C
Short (6.2) (2.4) (3.3)
p-value 0.01** 0.00** 0.01**
7.9 0.2 2.2 Conv. (15.3) (1.0) (3.4)
7.4 0.8 6.8
No.
2A
C
Short (15.8) (2.8) (6.8) p-value 0.36 0.00** 0.00**
Standard deviation in parentheses *Significant difference (α<0.05) **Highly significant difference (α<0.01)
Short- length lumber has less average maximum crook (Table 3.1) than
conventional- length lumber. This feature is directly related to the average length of the
lumber where the shorter pieces will have similar unit-quantities (mm crook / m length),
however, the actual crook will be less, in absolute terms.
Table 3.7. Defect area
Clear Wood Bark Pocket Decay Pith Shake Split Stain Wane/Void Sound Knot Unsound Knots (%) ---------------------------------------------------- (cm²/m²) ----------------------------------------------------
92.7 3.5 2.7 0.3 0.0 4.2 9.6 44.7 0.0 2.5 Conv. (4.3) (17.2) (19.8) (4.6) (0.0) (19.6) (118.4) (87.2) (0.0) (9.3) 91.1 9.1 6.7 1.8 0.0 2.1 28.5 213.9 0.4 1.7 Se
lect
Short (7.6) (108.1) (96.8) (31.5) (0.0) (16.1) (499.9) (276.5) (1.9) (12.4) p-value 0.42 0.19 0.24 0.21 0.50 0.11 0.26 0.00** 0.00** 0.23
90.9 20.9 7.1 0.0 0.3 16.6 69.9 159.4 0.3 9.6 Conv. (7.6) (94.15) (76.1) (0.0) (4.2) (103.5) (993.9) (230.2) (1.8) (31.8) 91.3 10.8 13.6 0.0 0.1 8.3 126.9 191.6 0.9 16.6
No.
1C
Short (9.8) (48.2) (111.8) (0.0) (1.3) (69.6) (1093.3) (256.0) (4.0) (61.1) p-value 0.89 0.06 0.21 0.50 0.27 0.15 0.26 0.06 0.01** 0.04*
89.3 32.7 73.7 0.5 0.4 12.8 70.3 177.7 0.4 16.9 Conv. (9.6) (100.4) (466.5) (5.4) (4.5) (64.1) (704.8) (263.4) (3.4) (35.2) 90.9 17.7 43.8 0.8 0.9 10.4 11.6 258.9 1.5 47.5
No.
2A
C
Short (8.2) (49.4) (350.0) (13.4) (9.4) (62.0) (73.9) (450.4) (9.8) (73.9) p-value 0.58 0.02* 0.20 0.35 0.20 0.33 0.10 0.00** 0.03* 0.00**
Standard deviation in parentheses *Significant difference (α<0.05) **Highly significant difference (α<0.01)
57
3.3.2 Yield
Table 3.8 shows yield results from 20 simulation replications for 2 lumber lengths,
3 grades, 4 cutting orders and 2 processing methods. Based on standard deviation
estimates of initial yield, simulations were replicated 20 times in order to verify
significance. However, due to the high variability in yield for the USDA cutting orders,
additional simulations were performed in order to determine significant differences. For
the Select grade USDA Tough cutting order with conventional- and short- length lumber,
65 and 80 simulations were needed respectively. For No.1 Common grade, 150
simulations were necessary for the USDA Easy with short- length lumber and the USDA
Tough with both conventional- and short-length lumber. The number of simulations
required was determined using the equation for sample size determination, taking into
account a standard deviation estimate (based on 20 simulations), in order to be able to
detect a one percent difference between two populations (Devore 1999). All comparisons
are highly significant (α ≤ 0.01) unless otherwise noted.
The USDA Easy cutting order has a higher yield when compared to the USDA
Tough cutting order. This result is explained by the greater selection of short and narrow
components (Steele et al. 1999). The higher yield obtained by the Furniture cutting order
versus the Panel cutting order is also explained by the greater selection of short components
as shown in the cutting order description.
The white birch database allowed us to examine the effect of grade, lumber length,
cutting order and processing method on yield
Table 3.8. Yield (%) results for rip- first and crosscut-first rough mills according to grade and cutting order as a function of lumber length
Cutting order USDA Easy USDA Tough Panel Furniture Rip-
first Crosscut-
first p -
valueb Rip-first
Crosscut-first
p -valueb
Rip-first
Crosscut-first
p -valueb
Rip-first Crosscut-first
p -valueb
conv. 64.4 (4.7)
67.7 (2.3)
0.01** 61.4 (4.0)
64.0 (2.3)
0.00** 71.8 (0.3)
78.3 (0.3)
0.00** 70.3 (0.7)
71.7 (1.2)
0.00**
Sele
ct
short 55.3 (4.7)
59.4 (5.3)
0.01** 45.2 (5.7)
53.6 (4.5)
0.00** 63.0 (0.5)
71.6 (0.5)
0.00** 65.5 (1.0)
65.5 (1.0)
1.00
p -valuea
0.00* 0.00** 0.00** 0.00** 0.00** 0.00** 0.00** 0.00**
conv. 48.8 (6.1)
60.0 (3.7)
0.00** 34.4 (8.7)
39.7 (8.1)
0.00** 62.6 (0.5)
70.2 (0.6)
0.00** 64.1 (1.1)
64.3 (1.3)
0.56
No.
1C
short 56.8 (4.4)
59.5 (2.8)
0.00** 33.9 (11.5)
33.4 (8.3)
0.67 62.1 (0.5)
66.6 (0.8)
0.00** 63.9 (1.0)
61.7 (1.2)
0.00**
p -valuea
0.59 0.00** 0.00** 0.00** 0.00** 0.55 0.70 0.00**
conv. 44.4 (3.7)
54.4 (3.6) 0.00** 20.5
(7.0) 24.7 (8.3) 0.94 57.5
(0.3) 64.0 (0.5) 0.00** 56.3
(2.2) 59.6 (1.2) 0.00**
No.
2A
C
short 35.9 (4.2)
42.5 (3.2) 0.00** 8.3x
(0.6) 13.5x (0.6) n/a 49.4
(0.6) 54.5 (0.6) 0.00** 47.2
(0.8) 47.0 (0.7) 0.46
p-valuea 0.00** 0.00** 0.00** 0.00** 0.00** 0.00** 0.00** 0.00**
Standard deviation in parentheses **Highly significant (α≤0.01) *Significant (α≤0.05)
p-valuea for comparison between conventional- and short-length lumber p-valueb for comparison between rip -first and crosscut -first xCutting order not filled
59
3.3.2.1 Conventional- vs. Short-Length
The primary objective of this study was to compare the yield obtained from
conventional- length and short- length lumber. In previous studies (Wiedenbeck and
Araman 1995, Wiedenbeck et al. 1995), significantly higher yield results were observed
when components were produced from conventional- length lumber compared to short-
length red oak lumber. The same observation ho lds true for white birch.
When yield was compared between short-length and conventional- length lumber,
the conventional- length lumber always had a significantly higher (α ≤ 0.01) yield than the
short- length lumber for Select and No.2A Common grades. The yield differences ranged
from 4.9%, when a rip- first rough mill processed Select grade lumber using the Furniture
cutting order, to 16.2%, when the same rough mill using the same grade lumber was
processed using the USDA Tough cutting order. On average, conventional-length lumber
had a 9.8% higher yield when ripped-first and 9.2% when crosscut first. There was also
greater variability in rip-first processing as demonstrated by a standard deviation of 3.5
versus 2.4 for crosscut-first.
Results for No.1 Common grade lumber were surprising. Figure 3.1 indicates that
in a rip- first rough mill no significant difference was observed in yield, between
conventional- or short- length No.1 Common lumber, for producing the Furniture and
USDA Tough cutting orders. Short- length lumber had significantly higher yield by 7.9%
when processed with the USDA Easy cutting order. Only when processing the Panel
cutting order did conventional- length lumber have significantly (α ≤ 0.01) higher yield but
the difference was small, only 0.5 percent in yield.
60
**Highly significant (α≤0.01) *Significant (α≤0.05)
Figure 3.1. Rip- first yield: Conventional versus Short-length lumber
For the crosscut-first rough mill, Figure 3.2 illustrates that conventional- length
lumber had a significantly (α ≤ 0.01) higher yield of 4.1%, with a standard deviation of 1.9,
for all cutting orders except the USDA Easy cutting order processing No. 1 Common
lumber where no significant difference was observed.
61
*Significant (α≤0.05) **Highly significant (α≤0.01)
Figure 3.2. Crosscut- first yield: Conventional versus Short- length lumber
Two factors help explain the decrease in yield for short-length lumber compared to
conventional- length lumber. The first is the shorter average length, which reduces the
number of component combinations that can be sawn out of a single board, limiting
maximum use of available lumber. The second is wane or void. As shown by Table 3.7
and Table 3.8, when wane and void occupy a much greater surface there is a larger
difference in yield between the two types of lumber. This increased amount of wane and
void comes from a different edging policy practiced in the short- log sawmill, because of
different log characteristics. Sawmills must extract lumber from smaller diameter timber
and in so doing they are subject to a greater amount of wane. They tend to allow for more
wane on boards in order to be able to recover more components from the resulting lumber.
The increase in wane/void areas allows increased absolute volume of output in components
but effectively reduces the yield because it generates a larger overall board surface. This
policy allows for higher volume recovery, on a tree level, but it probably has a detrimental
effect on throughput and productivity at the rough mill.
62
To verify the impact of wane, simulations were completed where wane and void
were filtered out of the database (Table 3.9). Table 3.9 indicates the yield results for the
Furniture and Panel cutting orders. A noticeable reduction in yield difference between
short- length and conventional- length lumber was observed when comparing Select grade
lumber – from an average of 6.9% for the Furniture and Panel cutting orders with
wane/void to 4.2% without for rip-first and from 6.2% to 3.0% for crosscut-first.
The average width and length for conventional- and short-length No. 1 Common
lumber (Table 3.1) were quite similar. The elimination of wane and void actually favored
the short-length lumber by 0.5% when ripped-first. In crosscut-first, the yield difference
went from 3.0% to 1.1% in preference of conventional-length lumber.
Table 3.9. Rip- first and Crosscut-first yield (%) results by lumber length according to grade and cutting order with wane and void filtered out
Rip-first Crosscut-first Cutting order / Panel Furniture Panel Furniture
Grade conv. short conv. short conv. short conv. short
Select 71.4 (0.6)
67.5 (1.2)
72.3 (0.3)
66.8 (0.6)
73.0 (0.9)
69.8 (0.9)
80.4 (0.2)
77.6 (0.6)
No.1 C 65.4 (0.8)
66.4 (1.4)
65.1 (0.4)
65.2 (0.6)
67.3 (1.6)
66.7 (0.9)
74.1 (0.4)
72.5 (0.6)
No.2A C 60.1 (1.3)
58.3 (0.9)
59.5 (0.4)
54.2 (0.9)
67.6 (0.9)
60.9 (1.1)
75.4 (0.5)
60.7 (0.6)
Standard deviation in parentheses
The yield difference results were mixed for No. 2A Common lumber. In every
case, yield was increased when wane/void was eliminated but the impact increased yield
differences in all cases but for the ripped- first Panel cutting order. One also observes that
the yield for No. 2A Common conventional- length-lumber exceeds slightly that of No. 1
Common, which would indicate that wane/void is the main defect that contributed to
lumber degrade, since the average width and length are comparable.
63
The high standard deviation between simulations in the USDA cutting orders
indicates that they are probably not adapted to white birch. The average width and length
of white birch lumber in our database are smaller than those of the red oak database
(Gatchell et al. 1998) for which the USDA cutting orders were designed. The red oak
database was built in the South-East of the U.S. where it is the resource of choice to the
hardwood furniture, cabinetry, and casket industries. The survey upon which was based
the development of the USDA cutting orders (Araman et al. 1982) was also made in this
region. This suggests that those cutting orders are better suited to the red oak resource.
The Furniture and Panel cutting orders, derived from Eastern Canada industries, include on
average shorter and narrower parts that are more appropriate for shorter and narrower white
birch boards. This might explain why we obtained less variability with the Furniture and
Panel cutting orders than with the USDA Easy and Tough cutting orders.
Although the present study allows us to conclude that lower yields should be
expected from short- length lumber when compared to conventio nal, several issues remain
to be dealt with. In a context where conventional lumber is increasingly scarce and
significant volumes of short- length lumber could be generated, the question arises as to
what is the limit of economic utilization of short-length lumber. Short-length lumber
should be expected to be cheaper than conventional which offsets, to a point, the yield
decrease. Further studies should be devised to define the thresholds of economic usability
of short- length white birch lumber.
Also, since short-length lumber has not been produced for a long time, one can
think that there is room for optimization both in sawing strategies, including edging and
trimming policies, and in grading. This study presents results from a database of
conventiona l-sourced lumber and short-length-sourced white birch lumber, both NHLA
graded. The yield results suggest that NHLA grade-rules might be too stringent for short-
length lumber, thus, certain sawmill and rough mills are agreeing on in-house grade rules
that better match the lumber quality and its end-use.
64
These in-house grades are in no way as definite as the NHLA grades. These grades
can be, and are, actually being further refined, based on a sawyer/industrial-user
relationship. More specific in-house grades can, in effect, be defined for specific uses such
as furniture, cabinetry, and flooring. One can think that increased yields could be expected
from such specific grades defined to meet the narrower needs of a specific user. Through
this approach, it is possible to devise economically sustainable ways of using short- length
white birch in the various hardwood using industries. This will necessitate the
multiplication of specific in-house grades that would be appropriate to specific clients or
industria l-usage segments. This too leads to future research in the area of grading of non-
conventional lumber resources.
3.3.2.2 Rip-First vs. Crosscut-First
Figures 3.3 and 3.4 compare yield between rip - first and crosscut-first processing for
either conventional-length or short-length lumber. It should be noted that comparisons
involving the USDA Tough cutting order with No.2A Common short- length lumber were
excluded because the cutting order requirements could not be met.
The results show that for white birch, crosscut-first rough milling generally
produces a significantly higher (α ≤ 0.01) yield than rip- first. An exception to this is that
no significant difference in yield was found when processing conventional- length lumber
using the Furniture cutting order with No.1 Common lumber (Figure 3.3), and when
processing short- length lumber using the USDA Tough (No.1 Common) and Furniture
cutting orders (Select, No.2A Common respectively) (Figure 3.4).
65
*Significant (α ≤ 0.05) **Highly significant (α ≤ 0.01)
Figure 3.3. Conventional- length yield: rip-first versus crosscut- first rough milling
*Significant (α ≤ 0.05) **Highly significant (α ≤ 0.01)
Figure 3.4. Short-length yield: rip- first versus crosscut-first rough milling
66
According to the cutting orders used in this study, crosscut- first generates on
average a 4.7% higher yield, with the highest difference (11.2%) occurring when using the
USDA Easy cutting order with No.1 Common lumber. Figures 3.5 and 3.6 illustrate a
typical example of yield difference when processing panel parts from an identical board.
These differences in yield can be explained by the characteristics of white birch.
As shown in Table 3.1, the lumber is narrow and contains crook. According to
Wiedenbeck (2001), and Gatchell (1991), these two properties favor crosscut-first rough
milling.
3.4 Conclusion
Although short-length lumber contains less crook than conventional-length lumber,
it does contain more wane and void defects due to the original log diameter. This
combined with the smaller board length affects lumber yield. Thus, conventional length
lumber generally produces a higher yield than short- length lumber. Select grade
conventional- length lumber resulted in an 8.8% higher yield, on average, and No. 2A
Common lumber had a 10.3% average higher yield. No. 1 Common lumber had, on
average, comparable yield results, where in one case, short-length lumber had a higher
yield. This indicates that No.1 Common short-length lumber can produce a similar or
better yield than conventional length lumber when using the Furniture, USDA Easy, and
USDA Tough cutting orders to rip -first and the USDA Easy cutting order to crosscut-first.
67
Figure 3.5. ROMI-CROSS cutup using Panel cutting order
Figure 3.6. ROMI-RIP cutup using Panel cutting order
68
It was also noted that crosscut- first achieved on average a 4.2% better yield than
rip-first rough milling. This property was associated to the characteristics of Northeastern
white birch, which produces narrow boards that generally contain crook. These two
characteristics combined reduce the rip -first processes flexibility in producing long clear
components and therefore reduces its yield.
Further studies will need to consider economical benefits of using short- length
lumber in the grade mix as well as the advantages of tailoring the lumber grades to the end-
users needs
69
References
Anderson, R. B., R. E. Thomas, C. J. Gatchell, and N. D. Bennett. 1993. Computerized
Technique for Recording Board Defect Data. USDA Forest Service, Research Paper
NE-671. Northeastern Forest Experiment Station, Radnor, PA.
Araman P. A. 1977. Use of Computer Simulation in Designing and Evaluating a Proposed
Rough Mill for Furniture Interior Parts. USDA Forest Service, Research Paper NE-
361. Northeastern Forest Experiment Station, Upper Darby, PA.
Araman, P. A., C. J. Gatchell, and H. W. Reynolds. 1982. Meeting the solid wood needs of
the furniture and cabinet industries: standard-size Hardwood blanks. USDA Forest
Service, Research Paper NE-494. Northeastern Forest Experiment Station, Broomall,
PA.
Bingham, S. A. 1976. Managing and using our hardwood resources. Proc. Fourth Annual
Hardwood Symposium of the Hardwood Research Council. 1976.
Buehlmann, U., J. K. Wiedenbeck, and D. E. Kline. 1998. Character-marked furniture:
potential for lumber yield increase in rip-first rough mills. Forest Products Journal
48(4):43-50.
Calvert, W. W. 1965. Le surrendement et son importance. [The overrun and its importance
(in French)]. Forintek Canada Corp., Eastern Laboratory, Ottawa in: Forêt
Conservation. 5 pp.
Cech M.Y., and F. Pfaff, 1980. Kiln Operator’s Manual for Eastern Canada. Special
Publication SP504ER, Eastern Laboratory, Forintek Canada Corp, Ste-Foy, Qc. 185
pp.
70
Devore, Jay L. 1999. Probability and Statistics for Engineering and the Sciences,
5th Edition. Duxbury Press, California Polytechnic State University, San Luis Obispo,
CA p. 750
Gatchell, C. J. 1991. Yield comparisons from floating blade and fixed arbor gang ripsaws
when processing boards before and after crook removal. Forest Products Journal
41(5):9-17.
Gatchell, C. J., J. K. Wiedenbeck, and E. S. Walker. 1993. A red oak data bank for
computer simulations of secondary processing. Forest Products Journal 43(6):38-42.
Gatchell, C. J., J. K. Wiedenbeck, and E. S. Walker. 1995. Understanding that red oak
lumber has a better and worse end. Forest Products Journal 45(4):54-60.
Gatchell, C. J., R. E. Thomas, and E. S. Walker. 1998. 1998 Data bank for kiln-dried red
oak lumber. USDA Forest Service, General Technical Report NE-245. Northeastern
Forest Experiment Station, Radnor, PA.
Gazo, R., and P. H. Steele. 1995. Rough Mill Analysis Model. Forest Products Journal
45(4):51-53.
Gazo, R, S. Mitchell, and R. Beauregard. 1998. Development of a database and its use to
quantify incidence of defects in radiata pine random width boards. New Zea land
Journal of Forestry Science 28(1).
Giguère, Marc. 1998. Guide du sciage des billons de feuillus durs. [A Guide to Sawing
Short-Log Hardwood (in French)]. Direction of the Forest Products Development,
Ministry of Natural Resource, Government of Québec, 27 pp.
Harding, O. V. 1991. Development of a decision software system to compare rip -first and
crosscut-first yields. Unpublished doctoral dissertation, Mississippi State University,
MS.
71
Harding, O. V., P. H, Steele, and K. Nordin. 1993. Description of defects by type for six
grades of red oak lumber. Forest Products Journal 43(6):45-50.
MNRQ , 1996. Ressources et Industrie forestières. Portrait stastiques. [Resource and
Industry. A statistical portrait. (in French)]. Edition 1996, Ministry of Natural
Resource, Gov. of Québec. 142 p.
National Hardwood Lumber Association. 1998. Rules for the Measurement and Inspection
of Hardwood and Cypress. NHLA, Memphis, TN, p. 19
Petro, F.J., and W.W. Calvert. 1990. How to Grade Hardwood Logs for Factory Lumber.
Forintek Canada Corp. Eastern Laboratory, Ottawa. 64 p.
Steele, P. H., and S. Lee. 1994. Yield comparisons of furniture parts for three gang-ripping
systems. Forest Products Journal 44(3):9 -16.
Steele, P. H., and R. Gazo. 1995. A procedure for determining the benefits of sorting
lumber by grade prior to rough mill processing. Forest Products Journal 45(6):69-73.
Steele, P. H., J. Wiedenbeck, R. Shmulsky, and A. Perera. 1999. The Influence of Lumber
Grade on Machine Productivity in the Rough Mill. Forest Products Journal 49(9):48-
54.
Thomas, R. E. 1997. ROMI-CROSS: ROugh MIll CROSScut-first simulator. USDA Forest
Service, General Technical Report NE-229. Northeastern Forest Experiment Station,
Radnor, PA.
Thomas, R. E. 1999. ROMI RIP 2.0 user’s guide: ROugh MIll RIP-first simulator. USDA
Forest Service, General Technical Report NE-259. Northeastern Forest Experiment
Station, Radnor, PA.
Wiedenbeck, J. K., C. J. Gatchell, and E. S. Walker. 1994. Data Bank for Short-Length Red
Oak Lumber. USDA Forest Service, Research Paper NE-695. Northeastern Forest
Experiment Station, Radnor, PA.
72
Wiedenbeck, J. K., and P. A. Araman. 1995. Rough mill simulations reveal that
productivity when processing short lumber can be high. Forest Products Journal
45(1):40-46.
Wiedenbeck, J. K., C. J. Gatchell, and E. S. Walker. 1995. Quality characteristics of
Appalachian red oak lumber. Forest Products Journal 45(3):45-50.
Wiedenbeck, J. K., 2001. Deciding Between Crosscut and Rip -First Processing. Wood &
Wood Products 106(9):100-104
73
4 WHITE BIRCH LUMBER USED IN THE PANEL INDUSTRY
Abstract
This study examines parts distribution for lumber sawn from conventional- length
and short-length logs. Select, No. 1 Common, and No. 2A Common white birch lumber
was simulation-processed using both rip-first and crosscut-first processing methods with a
typical panel- industry cutting order. A white birch database was developed and used to
simulate crosscut- first and rip-first rough mills and determine the effects of the species
physio-morphological characteristics on yield. Two length-groups of lumber were used,
namely; conventional- length and short-length.
ROMI RIP and ROMI CROSS simulations show that conventional-length lumber
offers the greatest production flexibility because it is able to produce long and wid e
components. These components can be broken down into combinations of shorter length
parts. Short- length lumber produces a greater variety of components in order to maximize
part yield from the lumber. Crosscut-first lumber produces a higher volume yie ld owing to
salvage parts production, which is much higher than when the lumber is ripped first.
Correspondence analysis was used to determine that lumber grade, processing
method, and lumber length are the three variables that explain most of the variability in
component production. When variables were examined for each grade, it was determined
that lumber type has little influence on variability for Select and No. 1 Common lumber. It
does play an important role in the parts distribution of low grade lumber (No. 2AC),
however.
74
Keywords: White birch, short-length, conventional length, rough mill, crosscut-first, rip-
first, yield, cutting order, grade, comparative parts distribution,
correspondence analysis
75
4.1 Introduction
White birch is the last untapped hardwood lumber resource. This statement is
based on inventory statistics that indicating that large timber volumes are available for
processing on a sustainable basis (Giguère 1998, MNRQ 1996). To date, the physical
characteristics of the species make it a less than ideal candidate for value-added wood
products, but today’s technology and markets could make another analysis of the resource
worthwhile.
In the previous section, yield was calculated for a 13.16 m3 (5,576 board feet) white
birch database processed either by rip - first or crosscut-first rough milling according to four
different cutting orders. Two of the cutting orders were USDA Easy and USDA Tough
(Steele et al. 1999), a third was selected from a Québec component manufacturer, and the
fourth came from a furniture manufacturer producing panel parts used in white birch
tabletops.
Previous work has shown that lumber length has a direct effect on yield (Hamner et
al. 2002, Wiedenbeck 1992). Wiedenbeck (1992) studied the impact of using short- length
lumber in terms of yield and rough mill throughput. No significant yield difference was
found using a furniture case goods cutting order for crosscut-first or cabinet cutting order
when ripped-first. Throughput, in terms of parts processed per time unit, was higher for the
short- length lumber when crosscut-first due to inherently easier material handling
properties. No difference in processing speed was determined for rip -first processing.
When comparing the effects of length between short (7- 8-foot), medium length
(11- to 12-foot), and long (15- to 16-foot) NHLA-graded boards, Hamner et al. (2002)
noticed a direct relationship between length and yield, when ripped-first using either USDA
Easy or USDA Hard cutting orders.
76
Results in section 3 indicated that conventional-length lumber had a higher yield
than short- length lumber, except for No. 1 Common and equivalent lumber when ripping-
first Furniture, USDA Easy, and USDA Tough cutting orders and when crosscutting first
USDA Easy and USDA Tough cutting orders. The physical characteristics of the sample
population also indicated that crosscut- first processing would generate a higher yield than
rip-first due to the narrowness of the lumber and the presence of crook (Wiedenbeck 2001).
In order to better understand how to improve the yield and marketability of white
birch, this study focuses on the comparative distribution of part widths and lengths obtained
when cutting conventional- and short- length white birch lumber in Select, No. 1 Common
and No. 2A Common grades. This lumber was processed using rip -first and crosscut-first
rough milling and a local panel- industry cutting order that was without part quantity
restrictions to determine the optimal component distribution.
4.2 Methodology
4.2.1 Sample material
A previously developed white birch database, discussed in section 3 and in
Appendix A, consisting of 2.73 m3 (1,157 bf) of Select, 2.15 m3 (912 bf) No.1 Common,
and 2.06 m3 (874 bf) No.2A Common NHLA-graded conventional-length lumber and, 2.27
m3 (960 bf) Select, 2.29 m3 (970 bf) No.1 Common, and 1.66 m3 (702 bf) No.2A Common
NHLA-graded short- length lumber was used.
Table 4.1 characterizes the database by indicating the total volume analyzed,
average length, average width, the average maximum crook, and the clear surface area that
is free of defects along with the standard deviation associated with each.
77
Table 4.1. White birch database characteristics
Grade Volume
(bf / m3) Number
of Boards
Width Average
(m)
Length Average
(m)
Crook Average
Max. (mm)
Clear wood (%)
---------------------------- Conventional ---------------------------- Select 1157 / 2.73 183 0.165
(0.040) 3.560 (0.258)
7.9 (5.2)
92.7 (4.3 )
No. 1C 911 / 2.15 241 0.141 (0.032)
2.475 (0.415)
6.6 (3.8)
90.9 (7.6 )
No. 2AC 873 / 2.06 235 0.140 (0.027)
2.456 (0.368)
7.2 (4.5)
89.3 (9.6 )
---------------------------- Short-length ---------------------------- Select 962 / 2.27 312 0.134
(0.030) 2.120 (0.246)
5.5 (3.8)
91.1 (7.6 )
No. 1C 970 / 2.29 292 0.152 (0.032)
2.030 (0.405)
5.2 (3.3)
91.3 (9.8 )
No. 2AC 703 / 1.66 350 0.124 (0.027)
1.490 (0.347)
4.5 (2.6)
90.9 (8.2 )
Standard deviation in parentheses
4.2.2 Cutting Order
The objective of this study was to analyze the optimal component distribution of
conventional and short-length white birch in a rip-first or crosscut-first rough mill through
the use of a Panel cutting order and identify the main factors that influence distribution
variability. ROMI-RIP (Thomas 1999) and ROMI-CROSS (Thomas 1997) were used to
respectively simulate rip -first and crosscut- first processing.
78
The panel industry cuts fixed-length, random width strips between 25 and 114 mm
(1 and 4.5 inches), then proceeds with edge-gluing them together into specific-sized panels.
This mode of operating assures a high yield because length is the only constraining factor.
It also builds a high quality panel because the defects can effectively be cut out of the
strips.
Due to limitations in ROMI-RIP 2.1, a purely random-width cutting order could not
be defined. To get around this shortcoming, the following system was devised: the width
range was divided into fifteen 6.35-mm (¼- inch) increments between 25 and 114 mm (1 to
4.5 inch). According to Buehlmann (1998), small width-spacing minimizes the distortion
that could occur, particularly when quantity is not a factor, thus the proximity of the
different width ranges was small enough not to have a significant effect on yield i.e. the
presence or absence of one particular width would not affect yield significantly.
An advantage of specifying random wid th in this manner is that it allows the
components to be clearly identified and tallied, according to size (width and length),
enabling a graphical representation of the output.
Infinite demand of all combinations of the following widths and lengths was used,
namely, widths of 25, 32, 38, 44, 51, 57, 64, 70, 76, 83, 89, 95, 102, 108, and 114 mm and
lengths of 445, 546, 749, 940, 991, 1,041, 1,092, 1,143, 1,245, 1,372, and 1,549 mm. 445,
546 and 749 mm were salvage specific lengths.
To insure that the parts demand would be considered infinite/constant by the
simulation software, the “Parts Priorities” were set to be adjusted every 10,000 bf in
ROMI-RIP and 9,999 bf in ROMI-CROSS i.e., volumes so large that they would never be
met and therefore the part prior ities would remain constant.
79
4.2.3 Rough Mill Processing
4.2.3.1 ROMI-RIP simulation parameters:
• Arbor type: All-blades movable arbor with 6 spacings; • Kerf: 4 mm; • Prioritization strategy: complex dynamic exponent (CDE); • Part prioritization: never updated; • Salvage cuts were made to three salvage-specific lengths in addition to the primary
part dimensions.
4.2.3.2 ROMI CROSS simulation parameters:
• Primary yield maximization method: Crosscuts optimized for best length fitting to board features;
• Kerf: 4mm; • Prioritization strategy: complex dynamic exponent (CDE); • Part prioritization: never updated; • Salvage cuts were made to three salvage-specific lengths in addition to the primary
part dimensions.
4.3 Results and Discussion
4.3.1 Yield
Table 4.2 shows the average yield for primary and salvage components, and total
average yield obtained from 20 simulation replications for 2 lumber types, 3 grades, and 2
processing methods using the Panel cutting order. The number of 20 replications was
based
80
on standard deviation estimate of initial yield that was determined using the formula:
( )2
221,21,2
δβα stt
n nn −− +=
where:
n = Sample size
t = From t-distribution table, with n-1 degrees of freedom
s = sample standard deviation of yield results
α = Significance level set at 0.05
β = 1-Power of the test set at 0.10
δ = Detection level was set at 1%
4.3.1.1 Total Yield
From Table 4.2 it can be observed that the yield for conventional- length lumber
was always significantly higher (α=0.01) than that for short-length-lumber although yield
differences were small when processing No. 1 Common lumber – 0.5% when ripped- first
and 3.6% when crosscut-first. These small differences can be explained by examining the
average length and width for No. 1 Common lumber in Table 4.1, where the sizes are
comparatively similar.
Total yield for crosscut-first rough milling was always significantly higher
(α=0.01) than for rip -first processing. Wiedenbeck (2001) and Gatchell (1991) indicate
that crosscut first processing has a higher yield than rip-first when crooked and narrow
lumber is used. Table 4.2 indicates that the boards contain crook and that their average
width is small, which tends to explain the results.
Table 4.2. Primary and salvage component yield (%) results by lumber type for Panel cutting order processed by a rip- first or crosscut-first rough mill
Select No. 1C No. 2AC Rip-first Crosscut-
first p -valueb Rip-first Crosscut-first
p -valueb Rip-first Crosscut-first
p -valueb
conv. 64.4 (0.3)
66.8 (0.3) 0.00** 52.8
(0.5) 54.5 (0.6) 0.00** 47.3
(0.3) 46.8 (0.7) 0.01**
Prim
ary
short 55.0 (0.5)
55.7 (0.5) 0.06 52.9
(0.6) 50.5 (0.7) 0.00** 37.3
(0.7) 35.9 (0.8) 0.00**
p -valuea 0.00** 0.00** 0.40 0.00** 0.00** 0.00**
conv. 7.4 (0.2)
11.4 (0.4)
0.00** 9.8 (0.3)
15.7 (0.3)
0.00** 10.2 (0.3)
17.2 (0.4)
0.00**
Salv
age
short 8.0 (0.4)
16.3 (0.6) 0.00** 9.1
(0.5) 16.2 (0.4) 0.00** 12.1
(0.4) 18.6 (0.7) 0.00**
p -valuea
0.00** 0.00** 0.00** 0.00** 0.00** 0. 00**
conv. 71.8 (0.3)
78.3 (0.3)
0.00** 62.6 (0.5)
70.2 (0.6)
0.00** 57.5 (0.3)
64.0 (0.5)
0.00**
Tota
l
short 63.0 (0.5)
71.6 (0.5) 0.00** 62.1
(0.5) 66.6 (0.8) 0.00** 49.4
(0.6) 54.5 (0.6) 0.00**
p-valuea
0.00** 0.00** 0.00** 0.00** 0.00** 0.00**
Standard deviation in parentheses **Highly significant (α≤0.01) *Significant (α≤0.05)
p-valuea for comparison between conventional- and short-length lumber p-valueb for comparison between rip -first and crosscut -first
82
Analyzing the yield results from primary and salvage parts provides additional
insight into why crosscut- first rough milling had a higher total yield.
4.3.1.2 Primary Parts
4.3.1.2.1 Conventional vs. short-length
Conventional- length lumber had a significantly higher (α=0.01) yield of primary
parts in the order of 10%, on average, compared to short- length lumber for Select and No.
2A Common lumber (Table 4.2).
No. 1 Common lumber had no significant difference in yield between conventional-
length and short- length lumber, when ripped first. When crosscut- first, the yield difference
was only 4% (α=0.01) in favor of conventional- length lumber.
4.3.1.2.2 Rip-first vs. crosscut- first
Yield in primary parts was significantly higher (α=0.01) for crosscut- first rough
milling when processing conventional- length, Select and No. 1 Common lumber. Rip- first
rough milling had a significantly higher (α=0.01) yield when using short- length
No. 1 Common lumber and No. 2A Common short-length lumber. Rip-first yield was
significantly higher (α=0.05) when processing No. 2A Common conventional- length
lumber.
In all cases, the yield differences were small, ranging from 0.5% to 2.4%. The
lower quality lumber grades (short-length No. 1 Common and all No. 2A Common) had a
higher yield when ripped-first. These results indicate that both processing methods
83
generate approximately the same primary yield in primary parts when using a panel-
industry cutting order.
4.3.1.3 Salvage Parts
4.3.1.3.1 Conventional vs. short-length
Table 4.2 shows that short-length lumber had a small, but significantly higher yield
in salvage parts (α=0.01) than conventional-length lumber, with the following exceptions,
a) No. 1 Common, rip -first lumber where conventional-length lumber had a higher yield
then short- length, and b) Select, short- length lumber had a 4.6% higher yield when
crosscut- first.
The short-length lumber produced more salvage components because longer parts
were prioritized in the primary operation. The residual lumber was too short to meet the
primary components size requirements, which resulted in an increased amount of salvage
components.
4.3.1.3.2 Rip-first vs. crosscut- first
More dramatic differences in salvage yield were obtained when comparing rip- first
and crosscut-first rough mills, as indicated in Table 4.2. A factor contributing to the higher
salvage yield obtained with crosscut- first lies in the process’ cutting logic. With crosscut-
first, maximum width components are prioritized, whereas in rip -first, maximum length
components are given priority. When these respective logics are applied to narrow crooked
lumber, shorter wide components are obtained when crosscutting first, whereas long and
narrow components are obtained when ripping first (Gatchell 1991, Wiedenbeck 2001).
Because this cutting order had short, salvage-specific, component- lengths (445 mm, 546
84
mm, and 749 mm), the crosscut- first simulation program used these lengths to increase
yield significantly (α=0.01) by 4.0 to 8.3% more than rip- first processing.
4.3.2 Part Size Distribution
Figures 4.1, 4.2, and 4.3 give the part size distribution, in terms of relative
frequency for a panel industry cutting order with conventional- length and short- length
Select, No. 1 Common, and No. 2A Common lumber that was ripped-first and crosscut-
first.
4.3.2.1 Conventional vs. short-length
Figures 4.1, 4.2, and 4.3 show that yield in conventional- length lumber favors long
components (1,549 mm). There is a peak in the length of salvage components at 546 mm.
The component distribution was even between 51 mm and 108 mm with peaks at 25, 38,
and 114 mm. This last peak, at 114 mm, consisted mostly of long and wide components,
especially for conventional- length lumber, which indicates a certain component
manufacturing flexibility.
Short- length lumber has a similar distribution to conventional-length lumber, but
appears to produce more scattered distributions and tends to produce shorter cuttings. Like
conventional- length lumber, mostly narrow components were produced. The shift in
production from longer and wider to shorter and narrower components is attributable to the
smaller lumber size, which reduces the number of combinations that can be extracted.
Figure 4.1 indicates that conventional- length select grade lumber offers the most
flexibility in components produced because the long and wide components can be broken
85
down into any combination of sizes. Short- length lumber, on the other hand, produces a
variety of components in a wide range of lengths and widths.
No. 1 Common lumber showed a similar component spread for either conventional- length
or short- length lumber (Figure 4.2). This can be explained by the similarities of both
lumber types in the database (in width and length, cf. Table 4.1). The components length
distribution resembles that of Select grade lumber with peaks at 546 mm and at 1,549 mm,
however, the 1,549 mm peak is not as pronounced. Although the short-length lumber
produced fewer long components (1,549 mm), it did increase production of 1,143 and
1,372 mm long parts. Both conventional- length and short- length lumber (Figure 4.2)
favored narrow (25, and 32 mm) components with a slight production peak at 114 mm.
Short- length lumber (Figure 4.2b) produced mostly narrow-sized parts hen using rip - first.
Conventional- length No. 2A Common lumber (Figures 4.3a, 4.3c) produced
components following the same trend as Select and No. 1 Common grade lumber. Short-
length lumber (Figures 4.3b, 4.3d), had a peak at 1,143 mm in length. This is explained by
an average board length of only 1,490 mm (Table 4.1), which prevents the production of
any of the longest components. The parts produced, therefore, were mostly narrow with
few wide components. No. 2 A Common grade lumber should be used for short and
narrow components only.
86
a) b) c) d)
Figure 4.1. Part size distribution for Select grade lumber with a) Conventional-length, rip -first; b) Short- length, rip-first; c) Conventional-length, crosscut-first; d) Short-length, crosscut- first
87
4.3.2.2 Rip- first vs. crosscut- first
The rip-first operation tries to place all the defects in the narrowest strips in order to
produce the longest cuttings. The crosscut-first operation produces the widest components
by cutting out defects at appropriate lengths. The logic of these processes is demonstrated
in Figures 4.1, 4.2, and 4.3. The rip- first rough mill produces long and narrow
components. The crosscut- first rough mill has a similar component distribution although
biased towards slightly shorter and wider components, with a peak, salvage specific length,
at 546 mm. This result confirms findings from the previous section about additional
production of salvage parts when crosscutting.
In the case of select grade lumber (Figures 4.1a, 4.1b), the boards in the database
generally had some degree of crook in them (Table 4.1). Therefore, when ripping select
grade lumber, long and narrow components were produced because of the shape of the
board (Wiedenbeck 2001).
When crosscutting lumber, shorter components are favored because this process
maximizes the width of the cuttings. In the case of select grade lumber (Figures 4.1c,
4.1d), a crosscut-first rough mill will naturally produce wide components. Short
components are produced in the salvage operation.
88
a) b) c) d)
Figure 4.2. Part size distribution for No. 1C grade lumber with a) Conventional-length, rip -first; b) Short- length, rip-first; c) Conventional-length, crosscut-first; d) Short-length, crosscut- first
89
The presence of defects directly affects scatter. In No. 1 Common lumber
(Figure 4.2), the scatter increases into multiples of lengths that fit into average board
length. One can observe that in order to cut around the defects, a rip-first or a crosscut-first
rough mill must produce a greater variety of components.
A rip-first rough mill (Figures 4.2a, 4.2b) favors the production of narrow
components in general, and produces long components to maximize yield. A crosscut- first
rough mill (Figures 4.2c, 4.2d) produces various-sized components, but favors wide and
short components.
No. 2A Common boards have the same component-distribution trend as the other
grades. Owing to the increased occurrence of defects, the components produced when
ripped-first (Figures 4.3a, 4.3b) are mostly narrow and cover the entire range of lengths.
The component spread when crosscut- first (Figures 4.3c, 4.3d) remains scattered,
with wide components produced overall.
90
a) b) c) d)
Figure 4.3. Part size distribution for No 2AC grade lumber with a) Conventional- length, rip-first; b) Short- length, rip-first; c) Conventional- length, crosscut-first; d) Short- length, crosscut-first
91
4.3.3 Correspondence Analysis
Correspondence analysis was used to evaluate how the 12 different combinations of
variables (3 grades x 2 processing methods x 2 lumber types) affect the part distribution (n-
1 = 11 dimensions). This method of analysis is an exploratory and descriptive technique,
which uncovers, and describes graphically, the relationships between the dimensions in
large contingency tables (Clausen 1998, Greenacre 1993). It should be noted that if a
dimension represents less than 9.09% (1/n*100) of the systems variability, then it is
considered of random nature.
The analysis was performed on a single simulation run of the entire white birch
database (5,574 bf in 1,613 boards of select, No. 1 Common, and No. 2A Common). Only
one run for each grade and lumber length was necessary owing to the comparative nature of
the analysis.
The resulting 2-D plots are expressed in terms of dimensions, which in turn, must
be interpreted to represent one of the variables under analysis. Figure 4.4 shows the overall
relationship among the 12 variables to the 2 main dimensions. Dimension 1 explains most
of the systems variability at 35.03%. This dimension can be interpreted as representing the
lumber grade since Select grade is on the far right of the axis defined as Dimension 1.
No 1 Common is in the center, and No 2A Common in on the left. Dimension 2 explains
an additional 18.21% of the system’s variability and can be seen as representing the
processing method since all rip -first scores are located on the upper part and all crosscut-
first on the lower part. Grade and processing method combined explain 53.24% of the
systems variability; however, lumber grade is, by definition we could say, the main factor
affecting the component production variability. Since it is expected that different grades
will produce different part distributions, the following analysis repeats the correspondence
analysis procedure within each grade to see what two dimensions emerge as explanatory
variables.
92
Legend: Conv: Conventional-length lumber Short: Short -length lumber Sel: Select grade No1C: No. 1 Common grade No2AC: No. 2A Common grade RR: Rip-first process RX: Crosscut-first process
Figure 4.4. Correspondence analysis scatter plot for lumber grade, processing
method, and lumber type
4.3.4 Lumber grade, processing method and lumber length:
Relationship to component distribution
By removing grade as a variable, the total number of variables is reduced to 4
(2 processing methods x 2 lumber types). Thus, Dimension 1 is considered random if it
represents less than 33.33%, and the system (the contribution of the two first dimensions) is
deemed random if it represents less than 66.67% of the variability. Figures 4.5, 4.6, and
4.7 show correspondence analysis graphs for each grade.
For select grade lumber as a whole (Figure 4.5), Dimension 1 explains 46.73% of
the variation and can be interpreted as the processing method. Dimension 2 can be taken to
-1.0 -0.5 0.0 0.5 1.0Dimension 1 (35.03%)
-0.5
0.0
0.5
Dim
ensi
on
2 (
18
.21
%) Conv,No1C,RR
Conv,No1C,RX
Conv,No2AC,RR
Conv,No2AC,RX
Conv,Sel,RR
Conv,Sel,RX
Short,No1C,RR
Short,No1C,RX
Short,No2AC,RR
Short,No2AC,RX
Short,Sel,RR
Short,Sel,RX
93
-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6Dimension 1 (46.73%)
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
Dim
ensi
on
2 (
36
.43
%)
Conv,Sel,RRConv,Sel,RX
Short,Sel,RR
Short,Sel,RX
represent lumber length and explains 36.43% of the variation by itself. Combined
however, these two dimensions explain 83.16% of the variability in Select grade lumber.
When examining Dimension 1, one observes that rip -first rough milling is on the
positive side of the axis, which means that a rip-first rough mill produces more narrow (25
mm in width) and long (1,549 mm in length) components. The crosscut rough mill, on the
other hand, produces more wide parts (114 mm in width) and salvage-specific components
(445, 546, and 749 mm in length). The choice of rip -first or crosscut-first rough milling
plays a greater role in determining the part distribution than does lumber length when select
grade lumber is processed.
Legend: Conv: Conventional-length lumber Short: Short -length lumber Sel: Select grade RR: Rip-first process RX: Crosscut-first process
Figure 4.5. Correspondence analysis between lumber type and processing
method for Select lumber
94
Analysis of Dimension 2 indicates that conventional length lumber produces
essentially either long (1,549 mm) and wide (114 mm), or long (1,549 mm) and narrow (25
mm) components. Short- length lumber has much more scatter and produces a wide range
of components without any clear concentration. These observations are confirmed by
looking at the component distribution (Figure 4.1).
Dimension 1 explains 51.74% of the system variation for No. 1 Common lumber
(Figure 4.6). This dimension can be interpreted as representing the processing method.
Dimension 2 explains only 26.45% of the variation and can be interpreted as representing
lumber length. Once combined, both dimensions explain 78.19% of the variability. The
lesser importance of Dimension 2 is not surprising when the database characteristics are
Legend: Conv: Conventional-length lumber Short: Short -length lumber No1C: No. 1 Common grade RR: Rip-first process RX: Crosscut-first process
Figure 4.6. Correspondence analysis between lumber type and processing
method for No. 1 Common lumber
-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6Dimension 1 (51.74%)
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
Dim
ensi
on
2 (
26
.45
%)
Conv,No1C,RR
Conv,No1C,RX
Short,No1C,RRShort,No1C,RX
95
examined in Table 4.1. Average width was 7% greater for the short-length lumber and
average length was approximately 22% shorter for No. 1 Common lumber when compared
to width differences of 23% and 13%, and length differences of 68% and 65% in favor of
conventional- length lumber for Select and No. 2A Common grades.
When looking at the component distribution for Dimension 1, production of 25-
mm- and 32-mm-wide and 1,549-mm- long components was favored in the rip-first rough
mill. Crosscut-first rough milling produced more 114-mm-wide components and salvage
components. These patterns can be observed in Figure 4.2.
There was little difference in the production of components between convention-
length and short- length lumber with No. 1C lumber. This was expected, since Dimension 2
contributed little to the explanation of variability in component distribution.
In Figure 4.7, Dimension 1 explains 46.30% and can be interpreted as explaining
the influence of lumber length on variability when No. 2A Common lumber is processed.
Dimension 2 explains 37.73% of the variability and can be seen to represent the processing
method. Combined, these factors explain 84.03% of the variability within this grade. The
importance of the lumber length is explained by examining the database characteristics for
No. 2A Common lumber in Table 2. The difference in length is markedly important,
especially since the average length of the boards was less than the maximum cutting order
length.
In this case, Dimension 1 represents the lumber length. Conventional- length
lumber produces long (1,549 mm) components in various widths, whereas short-length
lumber produces short and narrow (25, 32, 38, and 44 mm in width) components.
96
Legend:
No2AC: No. 2A Common grade RR: Rip-first process RX: Crosscut-first process
Figure 4.7. Correspondence analysis between lumber type and processing method for No. 2A Common lumber
Dimension 2 has rip-first producing mostly narrow (25 mm in width) components
and crosscut-first having more scatter and covering a wider range of component sizes,
including salvage parts. These observations are confirmed when looking at Figure 4.3.
4.4 Conclusion
The Panel cutting order demonstrates where components are produced without
quantity constraints. Rip- first processing generally produces long narrow components,
whereas crosscut-first processing produces long and wide components. When rip-first and
crosscut- first processing are compared, the crosscut-first rough mill will produce wider
components, generate a more scattered output, and produce more salvage components.
-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6Dimension 1 (46.30%)
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
Dim
ensi
on
2 (
37
.73
%) Conv,No2AC,RR
Conv,No2AC,RX
Short,No2AC,RR
Short,No2AC,RX
97
Conventional- length lumber produced longer and wider components than short-
length lumber. This result was expected because the conventional- length lumber offered a
greater number of part-size combinations that could be fitted into each board.
Correspondence analysis indicated that lumber grade explained more than 35% of
the part-size distribution variability. Processing method explained more than 18%, and
both combined explained over 53% of the system variability. When each lumber grade is
analyzed separately, lumber length has some importance in explaining component
variability, especially with No. 2AC lumber. It should be noted that these conclusions
apply specifically to the white birch database or any database with similar length, width,
and crook characteristics.
98
References
Buehlmann, U. 1998. Understanding the relationship of lumber yield and cutting bill
requirements: a statistical approach. Ph.D. dissertation, Virginia Polytechnic Institute
and State University. Blacksburg, VA. 209 p.
Clausen, S.-E. 1998. Applied correspondence analysis: an introduction. Sage university
series: quantitative applications in the Social Sciences; No. 07-121. Sage
Publications. 69 p.
Gatchell, C. J. 1991. Yield comparisons from floating blade and fixed arbor gang ripsaws
when processing boards before and after crook removal. Forest Products Journal
41(5):9-17.
Giguère, M. 1998. Guide du sciage des billons de feuillus durs. [A Guide to Sawing Short-
Log Hardwood (in French)]. Direction of the Forest Products Development, Ministry
of Natural Resource, Government of Québec, 27 p.
Greenacre, M. J. 1993. Correspondence analysis in practice. Academic Press Inc. 195 p.
Hamner, P., B. Bond, and J Wiedenbeck. 2002. The effects of lumber length on parts yield
in gang-rip- first roughmills. Forest Products Journal (in press).
MNRQ. 1996. Ressources et Industrie forestières. Portrait stastiques. [Resource and
Industry. A statistical portrait. (in French)]. Edition 1996, Ministry of Natural
Resource, Gov. of Québec. 142 p.
Steele, P. H., J. Wiedenbeck, R. Shmulsky, and A. Perera. 1999. The Influence of Lumber
Grade on Machine Productivity in the Rough Mill. Forest Products Journal 49(9):48-
54.
99
Thomas, R. E. 1997. ROMI-CROSS: ROugh MIll CROSScut-first simulator. USDA Forest
Service, General Technical Report NE-229. Northeastern Forest Experiment Station,
Radnor, PA.
Thomas, R. E. 1999. ROMI RIP 2.0 user’s guide: ROugh MIll RIP-first simulator. USDA
Forest Service, General Technical Report NE-259. Northeastern Forest Experiment
Station, Radnor, PA.
Wiedenbeck, J. K. 1992. The potential for short length lumber in the furniture and cabinet
industries. Ph.D. dissertation, Virginia Polytechnic Institute and State University.
Blacksburg, VA. 255 p.
Wiedenbeck, J. K., 2001. Deciding Between Crosscut and Rip -First Processing. Wood & Wood Products 106(9):100-104
100
5 THE EFFECT OF MANUFACTURING DEFECTS ON YIELD
Abstract
This study analyzes the incidence of manufacturing defects – spike marks,
conveyor marks, pressure roller stain, drying checks, machine gouge, and machine burn in
terms of their occurrence, size and impact on yield – with regards to lumber length, lumber
grade and cutting order in the furniture industry. A database of 13.16m3 (5,576 bf) of
random width and length white birch boards along with two cutting orders, Furniture and
Panel, was used in ROMI–RIP simulation. Boards were either conventional- or short-
length.
Drying checks had the largest impact on yield, reducing yield by 5.9% for the
Furniture cutting order and 6.4% for the Panel cutting order. No. 2A Common lumber was
most affected due to physiological properties of the boards, i.e. presence of heartwood and
juvenile wood, which make drying more difficult. Spike mark lowers yield by about 3%
for either cutting order, but they occur only in mills that use ring debarkers, and mostly on
high-grade external boards. Pressure roller stain affected yield by less than 2%, and
affected the smaller-sized boards because the defect location offers less flexibility to cut the
defect out. Machine burn reduced yie ld by 0.6% and 0.7% for the Furniture and Panel
cutting orders, respectively, and it appears to affect conventional-length lumber more due
to the dynamics of handling longer- length boards. Conveyor marks reduced yield by 0.6%
and 0.8% for the Furniture and Panel cutting orders. Machine gouge affected yield by
0.5% for both cutting orders, and affected short- length lumber more.
101
Keywords: White birch, short length lumber, spike mark, conveyor mark, pressure
roller stain, drying check, machine gouge, machine burn, yield, cutting order, grade,
manufacturing defect.
102
5.1 Introduction
When processing wood by heavy equipment, wood can often be damaged. This
paper focuses on occurrence and impact on yield of such defects when processing
Northeastern white birch (Betula papyrifera, Marsh.).
There has been an increased demand for hardwoods over recent years and, because
of this, traditional hardwoods are becoming increasingly scarce. It is, therefore, important
for sawmills to make the best use of lumber. To do so, sawmills are looking at computer-
optimized equipment in order to improve consistency of quality and productivity. Such
approach is necessary in order to make better use of the resource. These efforts are
lessened, however, if the raw material is mishandled. From the moment the trees are felled,
care must be taken with handling, loading and unloading of the logs and lumber. Even
when operators follow appropriate procedures, mechanical damage caused by mishandling
can happen.
The objective of this paper was to assess the impact on rough mill yield of various
manufacturing defects – namely spike marks, conveyor marks, pressure roller stain, drying
checks, machine gouge, and machine burn – that can occur once logs enter the sawmill and
are processed into dry lumber. For this purpose, ROMI-RIP 2.10 (Thomas 1999), a rip-
first rough mill simulation program, will be used to determine the effect on yield of each
individual manufacturing defect, and the combined effect of all the defects together. The
analysis will be based on the relative difference in yield results between “no manufacturing
defects” and the defect under consideration.
103
5.2 Methodology
5.2.1 Sample material
The boards selected for this study were required to show a range of qualities typical
of what is currently available in northern Québec. Two sawmills were chosen to typify two
sawing techniques. The first sawmill processes conventional logs into National Hardwood
Lumber Association (NHLA) grade lumber, whereas the second processes short-length
logs. Petro and Calvert (1990) describe conventional saw logs – these are logs of sufficient
size and quality to be sawn into NHLA lumber. In this paper we refer to boards cut from
such logs as conventional- length lumber. A large number of clear cuttings in lengths of 8
feet or more typically can be obtained from these boards. On the other hand, short- length
logs are logs that do not conform to the criteria defined by Petro and Calvert (1990)
because they are too short, too crooked, of too small a diameter, or present a combination
of these characteristics. These logs often have a length between 4 and 8 feet and are
generally classified as pulpwood (Calvert 1965). However, they are increasingly
considered fit for sawing. We refer to boards cut from such logs as short- length lumber.
5.2.2 Board Grading
A large volume of random width and length hardwood factory lumber produced in
Québec is used in furniture, cabinetry and flooring industries. Both conventional- and
short- length lumber used in this study was graded using the National Hardwood Lumber
Association’s (NHLA 1998) lumber-grading rules. Following these rules, the lumber is
graded according to the potential recovery of clear cuttings that can be obtained by
combinations of ripping and cross cutting. Number of cuts that can be made is determined
through surface measure for each board. In order to determine the lumber grade, areas of
placements of potential clear cuttings are determined considering the location of natural
defects such as knots, wane, and checks. Manufacturing defects were tallied as natural
104
defect equivalents. A detailed account of the grading rules is given in the NHLA rulebook
(NHLA 1998).
Table 5.1 shows the number of boards analyzed per grade for each of the two
sawmills included in the study. One sawmill, located at Senneterre, Québec, provided 659
boards in 6.94 m3 (2,941 bf) of conventional-length white birch lumber. The other, located
at Ste-Monique, Québec, provided 954 boards in 6.22 m3 (2,635 bf) of short- length white
birch lumber for a total of 1,613 random width and length boards in 13.16 m3 (5,576 bf).
The lumber from both sawmills came from comparable mixed hardwood-softwood stands
distinctive of the Laurentian shield. All boards were dried in a commercial kiln using
Forintek’s high temperature drying schedule No. 23 (Cech and Pfaff 1980) and surfaced on
both faces at Forintek Canada Corp., Québec, to allow easier defect identification during
digitizing process.
5.2.3 Database
A database of 5,576 board feet (bf) random width and length boards containing
information on all grade defects was developed (Section 3 and Appendix A). Table 5.1
describes the database characteristics, i.e. the number and volume of boards that were
digitized along with the average width and length for each grade and lumber length. For
this study, 2.73 m3 (1,157 bf) of Select, 2.15 m3 (911 bf) No.1 Common, 2.06 m3 (873 bf)
No.2A Common NHLA-graded lumber; 2.27 m3 (962 bf) Select, 2.29 m3 (970 bf)
No.1 Common, and 1.66 m3 (703 bf) No.2A Common custom graded short-length lumber
were used.
105
Table 5.1. Database characteristics
Grade Volume (m3)
Volume (bf)
Number of
Boards
Width Average
(m)
Length Average
(m) Conventional
Select 2.73 1,157 183 0.165 3.560 No. 1C 2.15 911 241 0.141 2.475
No. 2AC 2.06 873 235 0.140 2.456
Short- length Select 2.27 962 312 0.134 2.120 No. 1C 2.29 970 292 0.152 2.030
No. 2AC 1.66 703 350 0.124 1.490
5.2.4 Cutting order
Two cutting orders, Furniture and Panel, were used in this study. The Furniture
and Panel cutting orders are from actual Canadian furniture industries using white birch
lumber in their operations. The Furniture cutting order (Table 5.2) was obtained from a
rough mill that produced pre-cut components and panel parts for several furniture plants.
This cutting order has an average length of 803 mm and average width of 36.2 mm. The
specified cutting order is representative of the production of buffet and hutch types of
dining room furniture.
106
Table 5.2. Furniture cutting order
Width (mm)
Length 25 32 38 44 51 57 64 70 76 (mm) 362 5 7 387 36 8 3 2 1 1 1 5 451 42 10 4 2 1 1 1 514 57 13 5 3 2 1 1 10 584 9 2 1 1 20
768 29 7 3 2 1 1
914 49 11 5 3 2 1 1 5
1073 51 12 5 3 2 1 1 8 35
1175 8 4 1
1245 24 6 2 1 1 1 4
1295 13 3 1 1
1346 19 4 2 1 1
The Panel cutting order is from a plant that produces solid wood panels of specific
lengths. This cutting order calls for parts of random width between 25 and 114 mm in a set
of specified lengths. Due to ROMI-RIP software restrictions when processing solely panel
parts, the 25-114 mm width interval was divided into fourteen discrete widths in 6.3-mm-
increments (¼-inch). This resulted in a order that consists of all combinations of 25, 32, 38,
44, 51, 57, 64, 70, 76, 83, 89, 95, 102, and 114 mm widths and 445, 546, 749, 940, 991,
1041, 1092, 1143, 1245, 1372, and 1549 mm lengths (445, 546 and 749 mm are salvage-
specific lengths). An infinite part quantity for each part size was specified.
107
5.2.5 ROMI-RIP simulation parameters:
• Arbor type: All-blades movable arbor with 6 spacings; • Kerf: 4 mm; • Prioritization strategy: complex dynamic exponent (CDE); • Part prioritization: updated constantly for all cutting orders except for Panel cutting
order, which was never updated; • Salvage cuts: Made to primary part dimensions, except in Panel cutting order,
where three lengths were salvage-specific.
5.3 Results and Discussion
Incidence of defects was analyzed by calculating average defect frequency (number
of defects per m2) and average defect size (area of each defect type cm2/m2) on a per board
basis. Defects included in this analysis were spike marks, conveyor marks, pressure roller
stain, drying checks, machine gouge, and machine burn. Defects are listed in the order in
which they occur during processing. Table 5.3 shows average defect frequencies and their
differences by grade and lumber length. Pressure roller stain and drying checks are not
listed here because their occurrence was sometimes so frequent that they were digitized as
one group. Table 5.4 shows average defect areas for the same.
Yield loss caused by defects, whether natural or manufactured, will manifest itself
during board processing in the rough mill. For the purposes of this study, ROMI-RIP
(Thomas 1999), a rip -first rough mill processing simulation software was used to estimate
the yield loss. First, the processing was simulated using boards without any manufacturing
defects present. Then, simulation was repeated while adding each of six manufacturing
defects to the existing database individually. Finally, the simulation was performed with
all manufacturing defects present. Tables 5.5 (Furniture cutting order) and 5.6 (Panel
cutting order) show 1) yield for lumber without any manufacturing defects, 2) yield
decrease for each defect individually, 3) yield decrease for all defects combined, and 4)
statistical differences between conventional- and short-length lumber for Select, No. 1C,
and No. 2AC lumber grades. Based on standard deviation estimates of initial yield,
simulations were replicated 20 times in order to verify significance.
108
5.3.1 Spike Marks
Spike marks (Figure 5.1) are defined as small (< 3mm), discolored spots caused by
excess pressure on the feed system at the debarker. This defect occurs when using ring
debarkers. While feeding logs into the debarker, the endwise conveyor cylinders must
exert sufficient pressure and grip to move frozen, wet, slippery, muddy, and misshaped logs
forward without slippage. In order to prevent slippage, operators sometimes set pressure
on conveyor cylinders too high, especially in winter, when processing frozen logs.
Figure 5.1. Picture depicting a spike mark
Both conventional- length and short-length lumber in this study came from sawmills
using a ring debarker. The occurrence of spike marks in short- length lumber (Tables 5.3
and 5.4) was likely caused by mis-adjustment of conveyor cylinder pressure in this
sawmill, because short- length logs are more difficult to handle. Contributing to this result is
a fact that on average, in small logs, more lumber comes from the zone close to the bark
109
when compared to proportions of such lumber from larger logs. Incidence of spike marks
was higher in Select grade than in No. 1C or No. 2AC grades. This can be explained by the
location of Select grade lumber on the outside perimeter – which is where there are the
fewest defects – but also where the pressure cylinders from the ring debarker apply
pressure.
In short- length lumber, spike marks lowered yield by 3.4% for the Furniture cutting
order (Table 5.5) and 3.0% for the Panel cutting order (Table 5.6). No occurrence of this
defect was found in the conventional- length lumber. Spike marks are of concern for two
reasons. The first is that they are barely visible in rough lumber, but they become obvious
after finishing coat has been applied to the final product. The second is that they occur
mostly in Select lumber, where an average 5.4% yield decrease was observed, compared to
2.2% for No. 1C and 1.9% for No. 2AC (Tables 5.5 and 5.6). This affects lumber
desirability from such a mill and makes it necessary to be extremely vigilant during
processing.
Spike marks can be controlled, to an extent, by use of sharp spikes and appropriate
cylinder pressure. Cylinder pressure on newer systems can be adjusted on a per log basis.
This procedure along with appropriate maintenance can significantly decrease the spike
mark occurrence. In a follow-up study at this short- length sawmill, it was found that spike
marks were likely caused by inadequate pressure at the debarker feed system. After
customers expressed their concern, the appropriate pressure adjustment and maintenance
procedures were applied and largely solved the problem.
Table 5.3. Mechanical defect frequencies (# / m2) on white birch lumber
Grade Lumber type Spike Mark
Conveyor Mark
Machine Gouge
Machine Burn
----------------------- # / m2 ----------------------- Conv. 0.0
(0.0) 0.2 (0.7)
0.1 (0.4)
0.1 (0.6)
Short 3.7 (9.2)
3.1 (6.7)
0.2 (1.0)
0.3 (1.4) Se
lect
p-value 0.00** 0.00** 0.00** 0.07 Conv. 0.0
(0.0) 0.7 (2.4)
0.1 (1.0)
0.1 (0.7)
Short 2.0 (6.5)
2.6 (6.3)
0.3 (1.9)
0.3 (1.2) N
o. 1
C
p-value 0.00** 0.00** 0.06 0.01** Conv. 0.0
(0.0) 0.8 (2.7)
0.1 (0.8)
0.2 (1.0)
Short 0.9 (4.3)
3.1 (7.5)
0.3 (1.9)
0.2 (1.1) N
o. 2
AC
p-value 0.00** 0.00** 0.04* 0.45 Standard deviation in parentheses *Significant difference (α<0.05) **Highly significant difference (α<0.01)
Table 5.4. Average area of mechanical defects (cm2/m2)
Grade Lumber length Spike Mark
Conveyor Mark
Pressure Roller Stain
Drying Check
Machine Gouge
Machine Burn -------------------------------------------- cm2 / m2 --------------------------------------------
Conventional 0.0 (0.0)
2.0 (9.1)
0.5 (3.8)
25.0 (56.2)
0.1 (0.9)
1.5 (9.1)
Short 6.2 (43.2)
8.8 (29.2)
33.6 (217.8)
41.4 (155.3)
8.1 (59.8)
8.4 (70.6) Se
lect
P-value 0.01** 0.00** 0.00** 0.05* 0.01** 0.04* Conventional 0.0
(0.0) 4.1
(17.2) 4.3
(20.3) 71.2
(255.0) 0.8 (5.6)
1.6 (13.8)
Short 1.0 (4.5)
5.9 (17.5)
0.1 (1.9)
123.9 (429.2)
3.1 (29.5)
8.4 (44.1) N
o. 1
C
P-value 0.00** 0.13 0.00** 0.04* 0.13 0.01** Conventional 0.0
(0.0) 4.3
(22.9) 6.5
(25.9) 87.0
(278.0) 4.4
(31.9) 1.4
(10.1) Short 0.2
(1.2) 6.4
(29.4) 2.3
(25.7) 213.9 (669.3)
6.1 (52.7)
1.7 (13.0) N
o. 2
AC
P-value 0.00** 0.17 0.03* 0.00** 0.31 0.39 Standard deviation in parentheses *Significant difference (α<0.05) **Highly significant difference (α<0.01)
112
5.3.2 Conveyor Marks
Conveyor marks occur when wood is torn away by a chain dog (Figure 5.2), and
are about 6 mm (¼-inch) in width. For all lumber grades, conveyor marks were more
frequent in shor t-length boards (Table 5.3). A short-log sawmills, to be economical, has to
process the lumber more quickly, which, probably, leads to more handling defects. Also,
short- length lumber, being of lesser size (Table 5.1) and weight, seems to be more difficult
to convey correctly using standard chain dogs, hence leading to more conveyor marks.
Conveyor mark incidence had in general more impact on yield in short-length than
in conventional- length lumber. The presence of conveyor marks affected yield by 1.1% for
the Furniture cutting order (Table 5.5) and 0.5% for the Panel cutting order (Table 5.6).
Conveyor marks reduced yield in short-length lumber most, which was to be expected due
to their large area of 8.8 cm2/m2 in short-length vs. only 2.0 cm2/m2 in conventional- length
lumber (Table 5.4).
Figure 5.2. Picture depicting a conveyor mark
Table 5.5. Yield decrease (%) by grade and lumber length for different types of mechanical defects for Furniture
cutting order
Select No. 1C No. 2AC Defect Type Conv. Short P-value Conv. Short P-value Conv. Short P-value
69.7 64.6 63.3 63.9 57.8 47.4 None Yield (%) (0.5) (0.9) (1.0) (1.5) (1.5) (0.8)
Spike 0.0 6.1** 0.00 0.0 2.7** 0.00 0.0 1.3** 0.00 mark
0.4 0.7 0.00 0.7 1.7** 0.00 2.7** 0.7 0.00 Conveyor mark
0.6* 3.0** 0.00 2.4** 0.2 0.00 6.9** 1.0* 0.00 Pressure roller stain
1.5** 4.0** 0.00 4.8** 8.1** 0.00 9.8** 9.0** 0.00 Drying Check
0.5 1.1* 0.00 0.7 0.2 0.00 0.2 0.4 0.04 Machine gouge
0.7* 0.6 0.26 0.5 0.4 0.42 0.8 0.7 0.24 Machine burn
2.6** 14.8** 0.00 7.3** 15.5** 0.00 12.8** 11.9** 0.04 All Numbers in bold italics represent % decrease in yield compared to no mechanical defects Standard deviation in parentheses *Significantly different from zero (α<0.05) **Highly significantly different from zero (α<0.01)
Table 5.6. Yield decrease (%) by grade and lumber length for different types of mechanical defects for Panel
cutting order
Select No. 1C No. 2AC Mech. def. Conv. Short P-value Conv. Short P-value Conv. Short P-value 71.2 63.5 62.1 62.2 57.6 50.2 None Yield
(%) (0.3) (0.5) (0.4) (0.5) (0.7) (0.7) 0.0 4.7** 0.00 0.0 1.8** 0.00 0.0 2.5** 0.00 Spike
mark 0.1 1.1** 0.00 0.8** 0.3 0.00 0.2 0.4 0.00 Conveyor
mark 0.4** 3.1** 0.00 2.5** 0.3 0.00 2.8** 1.8** 0.00 Pressure
roller stain 2.6** 4.8** 0.00 5.3** 8.2** 0.00 5.5** 14.2** 0.00 Drying
Check 0.0 0.9** 0.00 0.3 0.1 0.00 0.0 1.3** 0.00 Machine
gouge 0.1 1.2** 0.00 0.8** 0.5* 0.00 0.3 0.5 0.00 Machine
burn 3.3** 12.5** 0.00 8.7** 12.2** 0.00 8.8** 17.2** 0.00 All
Numbers in bold italics represent % decrease in yield compared to no mechanical defects Standard deviation in p arentheses *Significantly different from zero (α<0.05) **Highly significantly different from zero (α<0.01)
115
5.3.3 Pressure Roller Stain
Pressure roller stain, a brownish stain – less than 5 cm (2 inches) wide – across the
width of the board, is believed to be a chemical discoloration of wood, which sometimes
occurs during the air -drying or kiln drying, apparently caused by the application of
excessive mechanical pressure on wood (Chauret and Giroux 1999). Among potential
causes are debarker conveyor cylinder and other machine feed systems imposing excessive
mechanical strain on board surface. Contrary to spike marks, in pressure roller stain, the
wood fibers are not mechanically altered but rather chemically. Pressure roller stain
occurred mostly in Select short- length lumber. Again, this points to excessive debarker
conveyor cylinder pressure. Due to the same cause, this defect type is closely related to
spike marks, but it occurs a little deeper in the wood. The same methods applied to
decrease spike marks occurrence should also help reduce pressure roller stain.
Pressure roller stain affected yield by 2.3% in the Furniture cutting order (Table
5.5), and by 1.8% in the Panel cutting order (Table 5.6). Pressure roller stain affected
short- length lumber more when using Select grade lumber – by 2.4% and 2.7% for the
Furniture and Panel cutting orders respectively – but reduced the yield of conventional-
length lumber more for No. 1C and No. 2AC lumber. With the Furniture cutting order, the
absolute yield difference was of 2.2% for No. 1C lumber, and of 5.9% for No. 2AC lumber,
while the Panel cutting order provided absolute differences of 2.2% and 1.0% for No. 1C
and No. 2AC lumber, respectively. This behavior demonstrates the influence of different
cutting orders on yield, and how the selection of parts sizes is crucial to maximizing yield.
5.3.4 Drying Checks
The manufacturing defects that occupied the most board surface area were drying
checks (Table 5.4). Drying checks are a lengthwise separation of the wood that usually
116
extends across the rings of annual growth, and they occur when the moisture gradient is too
high, especially at the beginning of the drying schedule. This defect was more frequent in
No. 1C and No. 2AC short-length lumber. These lesser quality boards generally come
from a log center, have a larger proportion of juvenile wood and heartwood, and are more
prone to checking. These two types of wood cells are more difficult to dry and particular
attention must be taken in order to ensure ideal drying conditions. Also, there were
significant differences in check occurrence between short- length and conventional- length
lumber for all grades (Table 5.4). These may have come from the higher proportion of
juvenile wood in small logs that were processed into short- length lumber.
There are several ways of preventing checking. In the air-drying yard, use of pile
roofs, good yard layout and use of appropriately sized, uniformly thick stickers will
minimize the effects of drying too rapidly. Placing stickers close to the board ends when
stacking will prevent end checks from progressing into the board. In the dry kiln, starting
drying at low temperatures and high humidity can control checking of green lumber. When
kiln-drying air-dried stock, it is best not to steam the load initially.
Drying checks were the defect with the most yield-decreasing impact, reducing
average yield by 6.2% with the Furniture cutting order (Table 5.5) and 6.7% with the Panel
cutting order (Table 5.6). No. 2AC lumber was the most affected grade and this could be
related to the physiological properties of the boards as described above. Checks had the
greatest effect on the yield of short-length lumber. The yield difference was of 2.5% for
Select lumber and 3.3% for No. 1C when using the Furniture cutting order, and of 2.2%,
2.9%, 8.7% for Select, No. 1C, and No. 2AC respectively with the Panel cutting order.
These differences indicate that the drying of lumber sawn from lesser quality logs requires
greater care and attention. In short-length lumber, logs being on average smaller, the
proportion of juvenile wood and heartwood is on average higher which would lead to check
having a greater negative impact.
117
5.3.5 Machine Gouge
Machine gouge (snipe) is a depression across the width of a board due to the
machine cutting below the desired line of cut. This defect happens at the planer when
boards are not properly held in position by pressure rollers and occurred most frequently in
short- length lumber (Table 5.3). The shorter average length (Table 5.1) explains why the
boards might not have been firmly held at the machine infeed or outfeed. Average area of
machine gouge was much larger for short-length Select lumber (Table 5.4) than any other
grade.
Machine gouge reduced yield by 0.5% for the Furniture (Table 5.5) and by 0.4%
for the Panel (Table 5.6) cutting orders. The short- length No. 2AC boards were more
affected by this defect, when processing the Panel cutting order. This indicates that the
longer lumber was easier to hold firmly by planer pressure rollers, reducing the occurrence
of this type of defect. This defect had more negative impact on yield for short-length
Select lumber than on the corresponding conventional-length lumber. It lowered yield by
0.6% and 0.9% when processing short- length lumber with the Furniture and Panel cutting
orders respectively. This defect did not have a significant influence on yield for
conventional- length lumber.
5.3.6 Machine Burn
Machine burn is a darkening, or charring, of the wood due to overheating by the
machining knives when a piece is stopped in a machine. This defect also occurs mostly at
the planer. Caused by a pause in the feed, the knives either rub on the wood or, if dull, are
being forced into the work piece, increasing temperature in one spot and burning wood.
Machine burn was well controlled at the conventional- length sawmill (Tables 5.3 and 5.4).
Short- length lumber had more machine burn area in Select and No. 1C lumber (Table 5.4)
than the conventional- length lumber. This could be related to the average size (length and
width) of the lumber (Table 5.1), and probably lead to more handling problems.
118
Machine burn lowered yield by 0.6% for the Furniture cutting order (Table 5.5) and
0.6% for the Panel cutting order (Table 5.6). The effect of this defect on yield was low and
not statistically significant when No. 2AC lumber was processed using the two cutting
orders. Yield for the lower lumber grades was not as influenced by machine burn. At the
same time, the lower grades have shorter average lengths (Table 5.1). These two
observations point to more difficulties in feed longer lumber through the planner. In effect,
longer and wider lumber will tend to present more crook, cup and warp - all causes of
potential planer jams. The impact of machine burn was statistically significant for No. 1C
when processing the Panel cutting order but not the Furniture. There was a yield difference
of 1.1% when processing Select lumber with the Panel cutting order but that difference was
of only 0.1% when using the Furniture cutting order. This shows again different impacts of
the same defects on different cutting orders.
5.3.7 All Defects Combined
When all six defects were included in the simulation, a sizable yield reduction
occurred. Yield for conventional- length lumber was reduced by about 7% and for short-
length lumber by 14%. The smallest yield reduction of 2.6% was for conventional-length,
Select lumber when processing Furniture cutting order. Largest yield reduction of 17.2%
was for short-length, No. 2AC lumber when processing Panel cutting order. This indicates
a sizable potential for yield increase by focusing on process improvement.
5.4 Conclusion
Manufacturing defects have an influence on rough mill yield and investing in
reducing the occurrence of certain types is worth the effort, considering the easily improved
yield results that can be expected. Many manufacturing defects could be related back to
drying such as checks and warp that may cause other defects such as machine burn,
119
conveyor mark and machine gouge when it results in board handling problems, mostly at
the planer mill. According to results, it is most beneficial to pay attention to the dry-kiln
operation because that is where the greatest yield reduction occurs.
In this study, it was observed that short- length lumber was bearing more
manufacturing defects than conventional- length lumber. This is for several reasons. First, it
was observed that the shorter lumber was bearing more spike marks, which caused more
severe yield decrease at the rough mill but was thought to be due to the application of
excessive pressure at the conveyor cylinder feeding the debarker. Proper cylinder pressure
can be applied on a per log basis when using automated devices and performing
appropriate maintenance.
More drying related defects were also detected on short- length lumber, which had a
higher bearing on yield. This is due to the smaller size of logs producing short- length
lumber. Processing smaller logs makes it more frequent for graded lumber to bear the effect
of heartwood and juvenile wood. Appropriate care must be taken in drying to minimize
these defects.
Finally, many problems with short-length lumber are thought to be caused by
smaller weight of this lumber, which makes it more difficult to handle by and feed into the
various machines designed to process conventional- length lumber. All these problems can
be addressed and minimized to an extent but this study shows how important it is to
maintain good practices both in terms of drying and material handling at all steps.
120
References
Calvert, W. W. 1965. Le surrendement et son importance. [The overrun and its importance
(in French)]. Forintek Canada Corp., Eastern Laboratory, Ottawa in: Forêt
Conservation. 5 pp.
Cech M.Y., and F. Pfaff, 1980. Kiln Operator’s Manual for Eastern Canada. Special
Publication SP504ER, Eastern Laboratory, Forintek Canada Corp, Ste-Foy, Qc. 185
p.
Chauret, G., Y. Giroux, 1999. Érable à sucre taché, essais préliminaires (Stained sugar
maple, preliminary trials. In French). Forintek Canada Corp., Eastern Division,
Project report 1122. 14 p.
NHLA. 1998. Rules for the Measurement and Inspection of Hardwood and Cypress.
NHLA, Memphis, TN, 19 p.
Petro, F.J., and W.W. Calvert. 1990. How to Grade Hardwood Logs for Factory Lumber.
Forintek Canada Corp. Eastern Laboratory, Ottawa. 64 p.
Thomas, R. E. 1999. ROMI RIP 2.0 user’s guide: ROugh MIll RIP-first simulator. USDA
Forest Service, General Technical Report NE-259. Northeastern Forest Experiment
Station, Radnor, PA.
121
6 CONCLUSION
Three objectives were set forth for this study. In particular, the study’s objectives
were to determine yield of lumber using rip -first and crosscut -first simulation software
and compare the effects of lumber type, processing method and cutting order on yield;
estimate the remanufacturing potential of white birch in different industries through the use
of a quasi-random-width Panel cutting order and determine the principle factors that
influence component distribution; and measure the effect of manufacturing defects on
yield, and associate particular defects to lumber type/processing method. The conclusions
of this study are as follows:
1) Although short-length lumber contains less crook than conventional- length lumber, it
does contain more wane and void defects due to the original log diameter. This
combined with the smaller board length affects lumber yield. Thus, conventional
length lumber generally produces a higher yield than short-length lumber. Select grade
conventional- length lumber resulted in an 8.8% higher yield, on average, and No. 2A
Common lumber had a 10.3% average higher yield. No. 1 Common lumber had, on
average, comparable yield results, where in one case, short-length lumber had a higher
yield. This indicates that No.1 Common short-length lumber can produce a similar or
better yield than conventional length lumber when using the Furniture, USDA Easy,
and USDA Tough cutting bills to rip-first and the USDA Easy cutting order to crosscut-
first. It was also noted that crosscut-first achieved on average a 4.2% better yield than
rip-first rough milling. This was related to the characteristics of Northeastern white
birch, which produces narrow boards that generally contain crook. These two
characteristics combined reduce the rip - first processes flexibility in producing long
clear components and therefore reduce its yield.
122
2) The panel cutting order allowed us to observe where components are produced without
quantity constraints. Rip-first processing generally produces long narrow components,
while crosscut- first processing produces long and wide components. When rip - first and
crosscut- first processing are compared, we notice that the crosscut-first rough mill will
produce wider components, generate a more scattered output, and produce more
salvage components. Conventional-length lumber produced longer and wider
components than short- length lumber. This result was expected because the
conventional- length lumber offered a greater number of part-size combinations that
could be fitted into each board. Correspondence analysis indicated that lumber grade
explained more than 35% of the part-size distribution variability. Processing method
explained more than 18%, and both combined explained over 53% of the system
variability. When each lumber grade was analyzed separately, then lumber type had
some importance in explaining the component variability, especially with low-grade
lumber.
3) Manufacturing defects have an influence on rough mill yield and investing in reducing
the occurrence of certain types is worth the effort, considering the easily improved
yield results that can be expected. Many manufacturing defects could be related to
drying such as checks and warp that may cause other defects such as machine burn,
conveyor mark and machine gouge when it results in board handling problems, mostly
at the planer mill. According to results, it is most beneficial to pay attention to the dry-
kiln operation because that is where the greatest yield reduction occurs. It was
observed that short- length lumber was bearing more manufacturing defects than
conventional- length lumber. This is for several reasons. First, it was observed that the
shorter lumber was bearing more spike marks, which caused more severe yield
decrease at the rough mill but was thought to be due to the application of excessive
pressure at the conveyor cylinder feeding the debarker. Proper cylinder pressure can be
applied on a per log basis when using automated devices and performing appropriate
maintenance. More drying related defects were also detected on short- length lumber,
which had a higher bearing on yield. This is due to the smaller size of logs producing
123
short-length lumber. Processing smaller logs makes it more frequent for graded lumber
to bear the effect of heartwood and juvenile wood. Appropriate care must be taken in
drying to minimize these defects. Finally, many problems with short-length lumber are
thought to be caused by smaller weight of this lumber, which makes it more difficult to
handle by and feed into the various machines designed to process conventional- length
lumber. All these problems can be addressed and minimized to an extent but this study
shows how important it is to maintain good practices both in terms of drying and
material handling at all steps.
Further studies will need to consider economical benefits of using short- length
lumber in the grade mix as well as the advantages of tailoring the lumber grades to the end-
users needs.
It should be noted that the conclusions drawn from this study are only valid for
similar type lumber characteristics, namely length, width, and crook.
APPENDICES
122
Appendix A: Creation of White birch database
Board Digitizing
Board digitizing consisted of manually recording the board dimensions (width and
length), defect positions and defect types for each face of the board. It is important to
maintain consistency of positioning the boards, marking defects, and recording coordinates
when digitizing large number of boards. A set of rules described below was developed to
achieve consistent and accurate readings.
First, a digitizing table was built. The table consisted of a flat tabletop and two
guard rails along the length of the tabletop. An auto-adhesive plastic measuring tape was
attached near the rails for measuring the length readings (x-coordinate). A fixture that could
slide on the top of the rails along the length was constructed and another auto -adhesive
plastic measuring tape was attached to it for width readings (y-coordinate).
Each board was placed on the table that represents the [x,y] coordinate system. The
x-axis runs along the length of the board and y-axis along the width. Boards were placed on
the digitizing table and were pushed flush against the rail and square with the end of the
table as shown in Figure A-1.
123
Figure A-1. Placement of the board on digitizing table
Most boards, however, are not perfectly rectangular. If a board did not have a
straight sawn edge, it was placed so that it touched the rail with at least two points. Boards
with crook were placed so that there was an equal gap between the table and the board edge
at either end of the board. Tapered boards were placed so that one edge of the board was
flush against the edge of the table. Placement of crooked and tapered boards is illustrated in
Figure A-2.
Figure A-2. Positioning of Crooked Boards
Once the board was properly positioned on the table, the board dimensions and
defects were read. Length of the board was measured from the [0,0] point of the coordinate
Coordinate Origin Point (0,0)
Y
X
124
system to the most distant point at the other end of the board. Width of the board was
measured at board’s widest point.
Defect type definitions
The defect types are organized into two groups of twenty natural and six
manufactured defects. For the proper identification of the defects, the following definitions,
from NHLA and custom grading, were used:
Natural Defects
Bark Pocket - A patch of bark partially or wholly enclosed in the wood Burl - A burl is a swirl or twist in the grain of the wood, which usually occurs near a knot but does not contain a knot Check - A lengthwise separation of the wood that usually extends across the rings of annual growth Compression Failure - A distortion the board like a “glass worm” Crook - A distortion of a board in which there is a deviation edgewise from a straight line from end to end of the board Decay - The decomposition of wood substance by fungi Heartwood / Sapwood - Presence of both heartwood and sapwood in the same board Hole - A hole extending partially or entirely through the piece and attributable to any cause Loose Knot - A knot that is loose or likely to become loose in drying or machining. Generally includes any knot exceeding 12 mm (1/2”) in diameter that is fully enclosed in bark. A knot that has not more than half his perimeter separated from the surrounded wood by bark Mineral Streak - An olive to greenish-black or brown discoloration (of undetermined cause in hardwoods) Open Knot – Absence of wood inside the knot’s core Pin Knot - A knot that does not exceed 3.175 mm (1/8 inch) in average diameter Pith Fleck - Small mineral streaks that are of reddish color Pith - The central core of the stem consisting mainly of parenchyma or soft tissue Sound Knot - A knot that is fixed by growth, shape, or position, which remains firmly, fixed within the piece or a knot that is wholly intergrown with fibers of the surrounding wood
125
Spike Knot - A branch cut longitudinally by the plane of the face and extending to the edge of the piece but also including knots that would have been spike knots had they not been occluded Split Knot - Knot that is checked or split due to drying constraints Split - A lengthwise separation of the wood, due to the tearing apart of wood cells Stain - In hardwoods, wood stain is used to describe the initial evidences of decay Wane - The presence of the original underbark surface, with or without bark, on any face or edge of a piece of lumber
Manufacturing Defects
Conveyor Mark – Wood torn away by a spike, about ¼-inch in width Machine Burn - Darkening or charring of the wood due to overheating by the machining knives Machine Gouge - A groove across a piece due to the machine cutting below the desired line of cut Pressure Roller Stain – A brownish stain across the width of the board caused by pressure rollers – less than 2 inches in width Spike Marks - Very little discoloration spots Void - A part of the wood is torn out in dressing
Each surface defect was identified and marked with a pencil by containing the
defect in a smallest possible rectangle, as shown in Figure 3.
Figure A-3. Enclosing defects in a rectangle
LX,LY
UX,UY
Y
X
Rectangle Boundaries Defects
126
When marking spike knots and wane, a significant amount of clear wood is
included within the boundaries of the defect rectangle. To achieve a better representation of
these defects, the original rectangle defect is broken down into a series of smaller
rectangles. When breaking the larger rectangle defect into smaller ones, the width of
smaller rectangles was set to 6.4 mm (¼ inch). Breakdown of defect into rectangles is
illustrated in Figure 4.
Figure A-4. Breakdown of large spike knot rectangle into series of smaller
rectangles
Certain types of defect are grouped in clusters. In such case, whole area was
marked as that type of defect. The rectangle in such a case may overlap with other defects,
which are marked independently (Figure 5).
7 mm
Original Defect Series of Smaller Defects
Y
X
7 mm
127
Figure A-5. “Field” of check
Wane, void (missing wood), heartwood, crook, check, split, stain, and decay defects
were also marked as a series of rectangles. A typical board with a wane or void defect
rectangles is shown in Figure 6.
Figure A-6. Typical Crook marking
Crook DefectRectangles
30 cm
128
The number of the rectangles is dependent on the slope of the wane, void
heartwood, crook, check, split, stain, or decay. For example, if the heartwood was relatively
uniform along the length of the board, fewer rectangles were marked, as illustrated in
Figure 7.
Figure A-7. Heartwood Marking
Figure A-8. Digitizing Face 2
Once all the defects were identified and marked, the data was manually recorded in
an Access data sheet. The lower left and upper right corners identify the perimeter of every
board and every defect in the database. The x- and y-coordinates for each of the corners is
Coordinate Origin
Coordinate Origin
Face 2
Face 1
Face 1 Face 2
X
Y
X
Y
Heartwood DefectRectangles
7 mm
129
recorded. For the defects, the defect type and face of the board that the defect is on are
recorded as well.
Initially, the board was positioned on the table with face one up and against the
right rail. Face 1 was defined as the outside face of the board (towards the bark). The other
side of the board was defined as Face 2. After all the information from Face 1 was
recorded, the board was turned Face 2 up and pushed against the left rail as shown in
Figure 9. The coordinate origin has now moved as shown in the Figure 9 and the lower
right corner of the board became the [0,0] point. To keep the measurements from the Face
2 compatible with measurements from the Face 1, the coordinates of lower right and upper
left corners of the defect rectangles were recorded on Face 2.
In addition to all the recorded data the following board properties were also
recorded - crook (maximum deviation); presence of heart color, and surface checks. All the
measurements were recorded on a millimeter scale.
Board Grading
A large volume of random width and length factory lumber produced in Québec is
used for remanufacture. This lumber is graded using the National Hardwood Lumber
Association’s (NHLA) Lumber Grading Rules. Under this rule, the lumber is graded
according to the potential recovery of clear cuttings that can be obtained by combinations
of ripping and cross cutting. The sequence of these combinations is established for some
grades whereas others permit either rip -first or cross cut-first combinations. In order to
determine the lumber grade, areas of placements of potential clear cuttings are considered.
The NHLA Grading Rules for Factory Lumber are defined by the percentage of cuttings
that can be removed from boards. As the boards are intended for subsequent
remanufacturing into flooring and tabletops, individual cuttings must also satisfy both size
and quality criteria.
130
The NHLA Grading Rules the lumber is graded into six factory lumber grades –
FAS, F1F, Select, No. 1 Common, No. 2 Common and No. 3 Common. However, FAS and
F1F were not considered for this analysis because they are unused in the market segment
under study. The requirements are based on the percentage of potential cuttings that can be
obtained from the board. In general, a piece of Select quality is free from defects on both
sides of the board, whilst a piece of No. 1 Common quality can admit minor imperfections.
A detailed account of the grading rules is given in grading handbook (NHLA 1994).
Prior to digitizing, all the boards were manually graded according to the NHLA
Grading Rules by an experienced grader both pre- and post-processing in order to insure
that the grade quality was respected.
131
Appendix B: Creating data files for simulation – Computer database
The information about the board sources, board grades, board dimensions, defect
locations, and types were entered into a computer file. The format of this file was
developed with compatibility with other programs and minimum size – increased
processing speed – in mind. Once all the data were entered, they were checked for typing
errors and corrected. The file is in TAB delimited ASCII format with designated sdf
extension.
A user-friendly interface for displaying and sorting the database information,
plotting the boards, calculating various database parameters and exporting information into
file formats for use with existing modeling programs was developed. The interface was
written for the Windows operating environment using Visual Basic 6 programming
language. The program can be driven by either a mouse or through the menu system.
Once the database file is open from the initial screen (Figure B-1), the information
about the board number, the corresponding source, the board grade, width, length, total
number of defects and crook of the defect location on the board are displayed in this screen.
Next to individual board information is a check box. If this box is marked, the marked
board will be included in any display, count, or export operation on the database.
132
Figure B-1. Random width lumber database opening screen
The total number of boards in the database as well as the count of checked boards is
also given. Checking or unchecking of the “Invert Selection” button will cause inverting of
all individually checked boxes into unchecked status and vice versa. By selecting one of the
Board, Source, Crook X, Crook Y, Grade, Number of defects sorting options, the boards in
the database will be displayed according to the selected criteria.
Filter function (Figure B-2) displays all the defect types found in the database and
their total count. By checking or unchecking individual defect types, these defects will or
will not be included in further operations on the database. Different defect types are coded
using different colors. These colors correspond to color codes used when plotting the
boards.
133
Figure B-2. Defect filter screen.
Plot function (Figure B-3) plots the image of the selected board. A board is selected
by clicking on the appropriate board number before clicking on Plot button. Plot window
can be re-sized using the mouse to accommodate any monitor size. A relative scale is
provided next to the board image. Defects on either face one, face two or both faces of the
board can be plotted by selecting the appropriate option. Different defect types are plotted
in different colors. These colors correspond to color codes in filter function.
134
Figure B-3. Board plot screen.
View function (Figure B-4) allows viewing the coordinates of defect rectangles,
defect type and board face on a selected board. Board is again selected by clicking on a
board number prior to clicking on View button.
Export function (Figure B-5) allows selected boards and selected defects to be
saved in one of several available formats. These include ROMI-RIP and ROMI-CROSS
(USDA rip - first and crosscut-first simulation programs, (Thomas 1995, ______ 1997,
______ 1999)), CORY (Brunner 1984 crosscut-first and rip-first simulation program, RAM
(Rough Mill Analysis and Modeling program) (Gazo 1995)), FLGRADE (Todoroki 1996))
and sdf database format. For ROMI-RIP and CORY programs all the measurements are
converted to ¼-inch units. If the statistics option of the export function is selected, then a
comma delimited ASCII file is created. This file is best viewed as an Excel spreadsheet.
The file contains statistics on boards and defects previously selected. These include total
number of boards, minimum, maximum, and average board width and length and total
board volume
135
Figure B-4. View Defect Coordinates Screen
in cubic meters and board feet. The following information is also listed: board number,
width, length volume in board feet and cubic meters, total surface area, total defect area,
percentage of clear area, count of each of the thirty seven defect types and cumulative area
for each of the thirty seven defect types. Above mentioned data can be calculated for face
one or face two only, or for both faces combined.
When exporting board information into another format, boards are placed in the file
organized by the board number in the ascending order. Some modeling functions require
several files which contain the same boards but with random order of boards within each
file. This can be achieved by clicking on the Randomise option in export screen and by
specifying the number of files.
136
Figure B-5. Export files screen.
For these simulations, ROMI-RIP and ROMI-CROSS data files were created. In
order to minimize the size of the files, acceptable defects – burl, compression failure,
heartwood/sapwood, mineral streak, and pin knot – were filtered out of each file, while
mechanical defects – checks, pressure roller stain, conveyor mark, machine burn, machine
gouge, and spike mark – were selectively filtered out depending on what defect was under
analysis. Once the “filtered” data files were saved, then they were opened again to assure
that no residual data remained. The file is then checked to confirm that the defects were
filtered, and the Export function is selected.
It should be noted that when the lumber was digitized, crook was measured as part
of the board width. This was done in order to keep the defect coordinates within the
confines of the digitized area, however, the actual board-width needed to be corrected in
order not to underestimate yield. To create a ROMI-RIP compatible file, the Romi-RipW-
137
C (Width-Crook) is selected., and a “Standard” file is then created. The “Standard” option
is selected because the randomizing features did not function properly. The created file has
a dat extension and can be edited using a text editor. It is recommended to open each file
to verify that information in the header, namely the grade (Select, 1C, or 2AC) is correct.
The first lines should read as follows:
BIRCH 2C BOARD NUMBER 1532 UNK TOTAL NUMBER OF DEFECTS 16 MEASURED BOARD WIDTH 119 GRADING: 0-0 0-0 0- 0 122-1215 10
where “BIRCH” is the lumber; “2C” grade; BOARD NUMBER 1532; TOTAL NUMBER
OF DEFECTS 16; “MEASURED BOARD WIDTH 119” is actual board width without
crook; “0- 0 122-1215” are lower left and upper right size coordinates, “10” indicates
metric measuring scale; “UNK, GRADING: 0-0, 0-0” are of no consequence.
Also, the file name should characterize the file contents. For these simulations, the
file names consisted of the number of boards, lumber length type, grade, defect(s) under
study, and file number. The following is a sample file name: tscm where t was used to
indicate the sawmill name – TLB – which saws short- length lumber; s is the grade, Selects;
cm is an abbreviation used to indicate that the effect on yield of conveyor marks was under
analysis
To create a ROMI CROSS file, the Romi-Cross option is selected. The creation of
a ROMI CROSS file is a two step process.
1) ROMI RIP file is first created and then it is converted using a “Vector” program to
create a vbd file. The resulting file can – and should – be verified using a text editor to
make sure that the board heading reads like this:
138
1039 10 3733 112 109 17 S
where 1039 is board number; 10 indicates metric measurement scale; 3733 is length, 112 is
total board width (including crook); 109 is actual board width; 17 is number of defects on
board and; s is grade (Select, in this case). If in any of the files the header has random
numbers instead of a grade, then the computer should be restarted and the files should be
re-created. The ROMI CROSS data files must then be processed using the Romi Cross
Conversion (Figure B-6) program that removes crook from the board width.
2) Once the files are added (Figure B-7) then the Convert File command is selected is
chosen and the crook is removed from the board width. The new files will have the letter
“c” added to their original name, it is therefore important that the original name contain
only seven letters to prevent any DOS conflicts if the file name has more than 8 letters in it.
Figure B-6. Opening screen of ROMI CROSS crook-removal program
139
Figure B-7. Board selection screen
When the process has been repeated for every grade and lumber length type, then a
series of randomized replicates can be created using the mix-mstr and vbd-mix programs
for ROMI-RIP and ROMI-CROSS, respectively, and a batch file. The batch file must
contain the executable (i.e. mix-mstr or vbd-mix), the original file name, and file-to-be-
created name. For these simulations, a certain number of boards were selected; this was
done by adding “/nx” where x is the number of boards that are desired. This last parameter
allows the program to randomize the x first boards. In order to insure that all boards could
be chosen, the data files were completely randomized first and then the x first boards were
randomized. Sample command lines are listed below.
For ROMI-RIP files Mix-mstr sscmc sscm01 Mix-mstr sscm01 15sscm01 /n150
140
For ROMI-CROSS files vbd-mix sscmc sscm01 vbd-mix sscm01 15sscm01 /n150
where sscm represents lumber length group (s=conventional-length and t=short-length),
lumber grade, and defect under analysis (conveyor mark).
Repeating these lines and incrementing the file number allows the user to create
random data files to be used with the USDA’s rough mill simulation software.
141
Appendix C: Simulation So ftware
Rough mill simulation program that lets users adjust rough mill parameters in order
to calculate a yield for those settings. This software is useful in estimating the effect of
processing method (rip-first or crosscut-first), lumber grade, lumber size (width and/or
length), defects, as well as manufacturing tools, and prioritization strategies. For this study,
the USDA’s ROMI-RIP and ROMI-CROSS programs were used because of they have
been verified and validated in industry and are available to the general public. A detailed
description of how the software works is included with the ROMI-RIP (Thomas 1999), and
ROMI-CROSS (Thomas 1997) therefore the following will describe the setup for the
simulations rather than the workings of the programs. Certain screen captures will be used
to illustrate particular setups.
ROMI-RIP
ROMI-RIP is a rip - first simulation program that lets the user define a part quality,
create a cutting order, set up the arbor, set up the chopsaws, set up the overall processing
and control options, specify salvage part sizes (if any), select board data to process in order
to simulate the rough mill and analyze the results.
The first step in preparing a ROMI-RIP simulation is defining the Part Grades. For
these simulations, the Readsdf program allowed us to filter out defects that were considered
“acceptable” by industry. This allowed us to filter out Burl, Compression failure,
Heartwood, Mineral streak, and Pin knot. The program also let us filter out manufacturing
defects: Drying check, Pressure roller stain, Conveyor mark, Machine burn, Machine
gouge, and Spike marks. Definitions of each defect can be found in the Methodology
section of this dissertation. Therefore, our Part Grades were designed to generate clear-one-
face lumber, where sound knots and stain were acceptable on the back side. In order to let
ROMI-RIP examine both sides of the board and increase processing efficiency, two mirror-
142
grades were input. Figure C-1 depicts what the Part Grade Editor should look like. Grades
0 and 1 are defaulted by the program. Grade 2 and 3 are similar, except that the acceptable
defects are on the opposite side of the board (BACK and FACE, respectively). These
grades, when input into the cutting order will allow the program to analyze both sides of
the lumber independently, and allow it to cut out a component from one side and then flip
the board over, if need be, to cut out another component. This instance could occur in the
presence of stain and spike marks; or mineral streak and sound knot.
Figure C-1. Sample grades and rules in the Part Grade Editor
A Rerip grade is necessary if the rough mill is designed to use a salvage operation.
This grade will not be defined in the cutting order; however, the program is designed to
look for this grade when cutting salvage components. Unlike the primary parts, only one
grade can be defined for salvage parts.
143
Figure C-2. Cutting Order Editor showing sample cutting order
Once the part grades have been defined, we are ready to define our cutting order.
The simula tions that were run for this study consisted solely of solid parts, i.e. no panel
parts were defined.
144
To create a cutting order one must have the width, length, quantity, and quality of
the desired components. These characteristics are defined in the methodology section of
this work and a screen shot of the easy cutting order is shown in figure C-2.
Figure C-2 demonstrates the typical cutting order settings used for this study. The
two separate Part Grades are what allow us to instruct the program to examine each side of
the board independently. Part scheduling is only used when the number of part lengths or
widths exceeds the capacity of the sorting system, this was not our case, therefore we used
the default “1” setting. The Cutting Order Editor also allows the user to define a Part
Prioritization Strategy. For these simulations, the Complex Dynamic Exponent was
selected because it uses lumber the most efficiently and produces a close to optimal yield.
Default Part Prioritization Parameters were used because they have been determined to be
efficient general values.
It should be noted that the cutting order file (*.rip) is a text file and can be edited in
any text editor. The parameters are easy to distinguish and one can modify the cutting order
easily by copying the appropriate data.
The next step is setting up the options that will be used to process the cutting order.
There are four option areas dealing with the arbor, chopsaw, process control, and salvage
operation.
The arbor setup for these simulations consisted of an All-Blades-Movable type,
with six blades, each with a 4-mm kerf. This option was chosen because it is very efficient
and it simplified setup by not having to adjust the arbor in function of lumber type, lumber
grade, or cutting order.
145
Figure C-3. Process control window
The chopsaws are used to crosscut the lumber and therefore are only have
adjustments for kerf and endtrim. Kerf was kept at 4 mm and we made no allowance for
endtrim.
The Process Control (Figure C-4) allows us to adjust process settings. Our database
is in metric units, therefore, we chose millimeters as the Processing units. Since we are
using a dynamic part prioritization strategy, we want to update the part count constantly.
The next two settings, Primary operations avoid orphan parts and Salvage cuts to
cutting order requirements instruct ROMI-RIP not to cut excess primary parts without first
146
determining whether the area can be salvage ripped to obtain a narrower cutting order part.
None of our cutting orders allow for panel parts, therefore, we do not select the Random
width strip parts okay in panels.
The final setting, Board cutup optimization step controls how many random widths
are examined. Since the cutting orders do not allow for random-width components, this
setting was arbitrarily set to 1.
Figure C-4. Salvage length and width editing window
147
In ROMI-RIP we can define salvage part sizes in several ways. The salvage parts
in all the cutting orders used for this study used the same widths as in the cutting order and
therefore the Salvage widths to primary part widths was chosen.
The lengths in our cutting order used the same primary part lengths except for the
Panel cutting order, which had three salvage-specific lengths. In order to include those
lengths in the processing, we must select the Salvage lengths cut to fixed salvage lengths.
The Salvage Length Modification window (Figure C-5) allows us to indicate all – up to 15
– the lengths that we need to cut. In this particular instance, the 445.0, 546.0, and 749.0
lengths were cut only in the salvage operation while the other 8 were cut in both the
primary and salvage operations.
Figure C-5. Salvage Length Modification window
148
Once these steps have been completed and the files have been saved, we are ready
to proceed with the simulations. Due to the large number of simulations that needed to be
accomplished, a DOS environment batch file (*.bat) was written using a text editor. The
command line reads as follows:
ROMI-RIP Cutting_order grade data_file Output_filename
The names in italic are variables where the Cutting order and grade were created in
the previous step. The data file are *.dat files in which the board information (grade,
length, width, defect location) is found, and the Output file name is the name to which the
simulation output information will be written.
Once the simulations complete, we can examine most of the results with a text
editor except the plot files (*.plt) that show the user how the boards were cut out and are
examined using the View.exe program. The *.out files summarizes the processing options
used in the analysis, yield summary tables, and the cutting order results.
The summary information about the processing options is useful to verify that the
settings were appropriate, and that no errors infiltrated the simulation. The next step is to
verify the cutting order results to verify that all the cutting order requirements were met; if
they were not, then the cutting order must be adjusted appropriately. The yield results let
the user know what a rough mill would expect to produce in similar conditions. The yield is
broken down per grade into primary, excess primary, salvage, excess salvage, and total
yield. The excess primary or salvage yield comes from components that were produced but
not on the cutting order. This occurs because the program creates a square matrix of all the
widths and lengths that are required and then cuts components to fill the matrix, therefore,
these yield results are generally subtracted from the total yield. There is an exception to
this case, salvage specific components are not listed in the cutting order and as such are
excess. To add to the complexity of the analysis, in the case where salvage specific parts
are requested, actual excess components may be produced. In order to assure oneself of the
149
“real” yield results, one must manually compile the surface area of salvage specific
components and divide this amount by the total surface in the data file. This information is
in a *.sas file that can be imported into a spreadsheet and then sorted and compiled.
ROMI-CROSS
ROMI-CROSS is a crosscut-first simulation program that allows the user to specify
optimization strategies, part qualities, kerf sizes, specify salvage part sizes (if any), select
board data to process in order to simulate the rough mill and analyze the results.
In ROMI-CROSS the user prepares the cutting order first. As a shortcut, the user
can input all the lengths and widths that will be used in the cutting order by selecting Edit
Options (Figure C-6), then Lengths or Widths modification, where he will input all the
primary part sizes, there is no need to input salvage sizes at this point. The user then
selects Cutting order (Figure C-5), then open or create Cutting order; New – he will then
enter cutting order name. The program then offers the option to “Create cutting order using
currently defined part lengths and widths?” When Yes is selected, ROMI-CROSS then
proceeds to create a matrix of all the size combinations, which the user then proceeds to
Modify (Figure C-7) by either deleting the component or changing the specified quantity
from 0 to the desired amount. Part scheduling was set to 1 as with ROMI-RIP. The size
specifications for salvage parts are defined by selecting Other, Don’t drop back to random
widths when no feasible fixed widths are available, and entering the determined salvage
Widths and salvage Lengths.
150
Figure C-6. Processing option main edit window
Figure C-7. Cutting order definition window
151
Figure C-8. Part size, quantity, schedule, and type editing window
The user must now select his part prioritization method (Figure C-7) using the
Weighting method. To keep the simulations as comparable as possible, the same Complex
Dynamic Exponent strategy was selected in order to reduce the number of variable in the
system.
After editing the cutting order and having saved the changes, the user returns to the
Edit Options menu and selects Process (Figure C-9).
152
Figure C-9. Cutting specifications window showing processing options
This Sawing Specifications allow the user to change the part optimization
strategies, end-trim specification, kerf sizes, part qualities, and the unit of measure. The
goal of these simulations was to determine what optimal yield could be expected from
white birch, therefore, we Optimize crosscuts for defect and clear area fitting instead of
best length fitting only because the latter does not consider the influence of defects on part
yields.
The scanner optimization length specifies at which frequency the board length is
reviewed/examined. The optimization length was set to 0 to optimize the entire board,
regardless of length.
The boards were not end-trimmed
153
Figure C-10. Primary part defect acceptance menu
The Defect options and Salvage specifications allow the user to define the part
quality one proceeds to define the part qualities but first the user must define what the parts
are: clear two-face Cuttings, clear one-face Cuttings, or sound two-face Cuttings. Since
we already filtered out the defects that were acceptable on both sides of the board, we are
producing clear one-face Cuttings.
The primary part-grade defect acceptance is not as flexible as with ROMI-RIP 2.
Only one grade can be defined for clear one- face cuttings. Sound knots and Stain (defects
number 25 and 38 respectively) were selected as acceptable on the back side (Figure C-10).
The same options are offered for the Salvage part grade when making the Salvage
specifications.
Measurement units, Kerf, and back gauge priorities, were set the as the in rip - first
simulations. Primary operations aVoid orphan parts and saLvage cuts to cutting order
requirements were selected to optimize yield, just like in the rip -first simulation settings.
154
Once the cutting order has been edited and saved then the simulations can occur. A
batch file (*.bat) can be written in order to process a large number of simulations. The
command lines can be written in a text editor and should look like this:
ROMI-X Datafile Output_filename +B +LCutting_order +OOutput_filename
The names in italic are variables where the Cutting order was created in the
previous step. The data file are *.vbd files in which the board information (grade, length,
width, defect location) is found, and the Output file name is the name to which the
simulation output information will be written. The arguments +B, +L, and +O (O not zero)
indicate that this is a Batch mode command line, the cutting order file, and the Cutting
order Output file name. If no output file name is specified, the results will be stored in a
file with the same name as the cutting order but with an *.out extension, which is useless if
the user intends on doing several replicates with the same cutting order – each simulation’s
results will be successively overwritten.
As with ROMI-RIP, the simulation output files are text editable, except the plot
files (*.plt) that show the user how the boards were cut out. The *.out files summarizes the
processing options used in the analysis, yield summary tables, and the cutting order results.
The summary information about the processing options is useful to verify that the
settings were appropriate, and that no errors infiltrated the simulation. The next step is to
verify the cutting order results to verify that all the cutting order requirements were met; if
they were not, then the cutting order must be adjusted appropriately. The yield results let
the user know what a rough mill would expect to produce in similar conditions. The yield is
broken down per grade into primary, excess primary, salvage, excess salvage, and total
yield. The excess primary or salvage yield comes from components that were produced but
not on the cutting order. This occurs because the program creates a square matrix of all the
widths and lengths that are required and then cuts components to fill the matrix, therefore,
these yield results are generally subtracted from the total yield. There is an exception to
155
this case, salvage specific components are not listed in the cutting order and as such are
excess. To add to the complexity of the analysis, in the case where salvage specific parts
are requested, actual excess components may be produced. In order to assure oneself of the
“real” yield results, one must manually compile the surface area of salvage specific
components and divide this amount by the total surface in the data file. This information is
in a *.sas file that can be imported into a spreadsheet and then sorted and compiled.
It should be noted that the Windows NT environment is case sensitive and that the
extensions of all files used should be in small case to insure compatibility; as a shortcut,
one can open a command prompt, go to the appropriate directory, and use the rename
command e.g. ren *.DAT *.dat, which will actually rewrite the extension of all DAT files
to lower case.
156
Appendix D: Incidence of defects
In order to analyze the occurrence of defects in random width and length white
birch lumber, two variables were considered - frequency of defects and average area of the
defects. The defect frequency is defined as the number of occurrences of a defect per
square meter of board surface area. The number of defects of each type for each board was
recorded and divided by the board surface area. The board surface area was calculated as
length of the board multiplied by largest width of the board. Frequency of defects was
calculated for all defect types except wane, void, heartwood / sapwood, crook, check, stain
and decay since the total size of these defects is unknown per se. The defect area is defined
as the area (cm2) of a defect type per square meter of board. The average defect area was
calculated for all the defect types, including wane. Checks were not always considered
individually. In some instances – especially in heartwood – the frequency and area of
checks were recorded aggregately as an area affected by check occurrence. When
calculating frequency and area of the defects, only defects on face 1 (worse face) were
considered.
Clear surface area
Percentage of the clear surface area was calculated as the ratio of board clear area
(board surface area minus total defect area) to board surface area.
Incidence of Defects
The first objective of this project was to create the database of digitized, random
width and length and length white birch boards. The total volume of boards in the database
is 17.86 m3 (7,571 bf).
157
Table D-1. Frequency of defects by defect type and by source
Defect frequency (No. defects/m2 board) Defect type \ Conventional-legth Short-length
Source S 1C 2C 3 C S 1C 2C 3 C Bark Pocket 6.245 5.031 8.475 0.560 5.565 3.919 8.228 8.373
Burl 4.017 6.294 11.850 0.389 1.292 0.855 1.088 1.348 Compression
Failure 0.228 0.331 0.382 0.337 0.932 0.444 1.088 1.161
Hole 0.206 0.071 0.037 0.570 0.054 0.011 0.107 0.057 Loose Knot 0.735 0.472 0.862 1.451 1.002 0.655 1.501 1.606
Mineral Streak
3.922 3.448 2.870 4.491 2.593 2.466 2.298 3.182
Open Knot 0.231 0.094 0.172 0.415 0.420 0.311 0.781 0.530 Pin Knot 0.311 6.677 6.677 10.433 4.416 3.330 6.252 6.953
Pith 0.860 0.118 1.943 1.943 0.305 0.144 0.322 0.860 Pressure
Roller Stain 0.086 1.039 0.043 0.468 0.086 0.022 0.153 0.086
Sound Knot 2.036 0.307 0.542 0.463 1.432 1.399 0.996 2.036 Spike Knot 0.789 0.083 0.561 0.198 0.433 0.322 0.582 0.789 Split Knot 3.606 1.665 8.697 3.606 4.260 1.898 0.947 7.469
Split 0.509 0.590 0.415 0.509 0.305 0.277 0.460 0.287 Conveyor
Mark 0.087 0.106 0.026 0.087 0.305 0.433 0.230 0.143
Machine Burn
0.185 0.094 0.250 0.185 0.261 0.122 0.322 0.430
Machine Gouge
0.211 0.201 0.363 0.211 0.080 0.056 0.046 0.100
Spike Marks 0.000 0.000 0.000 0.000 1.979 1.554 0.950 1.075
158
Table D-2. Clearwood percentage and defect area by defect type and wood source
Defect area (cm2/m2) Conventional-length Short-length
Defect Type \ Source S 1 C 2 C 3 C S 1 C 2 C 3 C Clearwood area (%) 53.56 65.13 59.31 47.86 60.01 52.82 49.74 50.34
Bark pocket 0.16 0.15 0.26 0.21 0.18 0.24 0.24 0.22 Burl 0.16 0.27 0.47 0.00 0.03 0.01 0.04 0.03
Check 1.38 1.04 1.48 2.59 3.89 4.01 5.78 5.28 Compression Failure 0.05 0.13 0.03 0.02 0.24 0.23 0.11 0.54
Crook 0.09 0.00 0.01 0.09 0.02 0.03 0.00 0.05 Decay 1.08 0.67 1.30 2.02 0.36 0.15 1.03 0.39
Heartwood 41.86 31.22 36.39 43.91 32.51 39.33 41.05 38.54 Hole 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.01
Loose Knot 0.03 0.02 0.03 0.04 0.05 0.03 0.07 0.07 Mineral Streak 0.20 0.03 0.03 0.05 0.29 0.22 0.41 0.56
Open Knot 0.01 0.00 0.01 0.01 0.02 0.02 0.03 0.02 Pin knot 0.01 0.00 0.01 0.01 0.00 0.01 0.00 0.00
Pith 0.15 0.02 0.02 0.46 0.11 0.15 0.05 0.25 Check 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Sound knot 0.01 0.01 0.02 0.02 0.02 0.01 0.02 0.05 Spike knot 0.02 0.02 0.01 0.05 0.05 0.05 0.07 0.09 Split knot 0.19 0.13 0.16 0.41 0.25 0.12 0.56 0.42
Split 0.09 0.12 0.15 0.08 0.06 0.10 0.07 0.03 Stain 0.75 0.71 0.04 1.95 1.37 2.16 0.33 2.54 Wane 0.02 0.04 0.03 0.03 0.03 0.00 0.05 0.03
Conveyor mark 0.01 0.01 0.01 0.00 0.01 0.04 0.00 0.00 Machine burn 0.01 0.01 0.01 0.02 0.15 0.02 0.12 0.26
Machine gouge 0.04 0.04 0.04 0.06 0.01 0.00 0.00 0.01 Pressure roller stain 0.03 0.07 0.06 0.00 0.08 0.00 0.00 0.01
Void 0.11 0.16 0.13 0.09 0.22 0.24 0.22 0.26 Spike mark 0.00 0.00 0.00 0.00 0.03 0.01 0.01 0.01
159
Observations
This study has two objectives: create a digitized white birch database and analyze
the incidence of defects. First, the methodology used to create the database is outlined then
the database itself is presented.
The second objective was the analysis of incidence of defects in white birch random
width and length boards. The presence of heartwood and sapwood is considered a defect
for certain end-use products such as Select quality floorboards and its presence will only
affect yield in that one instance. The most frequent defects in the boards from the
conventional- length- log sawmill are pin knots and bark pockets. The largest average size
defects are the areas of checks. The most frequent defects in the boards from short-length-
log sawmill are bark pockets and the largest area size defects are the areas of checks – it
should be noted that these incidence of defects is for the sample as a whole and does not
segregate conventional- length from short- length logs. Later analysis made those
distinctions.
VITA
160
VITA
Charles Clement
Business Address
University of Tennessee Tennessee Forest Products Center 2506 Jacob Dr. Knoxville, TN 37966 Phone: (865) 946-1125 Fax: (865) 946-1109 E-mail: [email protected]
Education
2002 DOCTOR OF PHILOSOPHY Forestry and Natural Resources Purdue University, West Lafayette, Indiana
2001 MASTER OF SCIENCE Wood Science Laval University, Québec, Québec
1996 BACHELOR OF SCIENCE Wood Science Laval University, Québec, Québec
161
PRACTICAL EXPERIENCE 1999-2002 Research assistant
Department of Forest & Natural Resources Purdue University, Knoxville, TN
1998 Lecturer Department of Wood and Forest Sciences Laval University, Québec, QC, Canada
1994 Teacher`s Assistant
Department of Wood and Forest Sciences Laval University, Québec, QC, Canada
1996 Intern, Lumber Drying
Forintek Canada Corp. Ste-Foy, QC, Canada
1993 Intern, Quality Control
Chantiers de Chibougamau Chibougamau, QC, Canada