Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
IN DEGREE PROJECT MECHANICAL ENGINEERING,SECOND CYCLE, 30 CREDITS
, STOCKHOLM SWEDEN 2018
QUALITY ASSURANCE THROUGH SMART ANGLE MONITORINGIMPROVEMENT OF TIGHTENING IN SCANIA’S CAB AS-SEMBLY AND IMPLEMENTATION INTO AN INDUSTRY 4.0 BASED SYSTEM
CARLOS CERVANTES
ESMAEIL NIK ARMAN
KTH ROYAL INSTITUTE OF TECHNOLOGYSCHOOL OF INDUSTRIAL ENGINEERING AND MANAGEMENT
QUALITY ASSURANCE THROUGH
SMART ANGLE MONITORING
IMPROVEMENT OF TIGHTENING IN SCANIA’S CAB AS-
SEMBLY AND IMPLEMENTATION INTO AN INDUSTRY
4.0 BASED SYSTEM
CARLOS CERVANTES
ESMAEIL NIK ARMAN
Master of Science Thesis TPRMM 2018
KTH Industrial Engineering and Management
Production Engineering
SE-100 44 STOCKHOLM
ii
Abstract
There is currently an increasing need for manufacturing companies to achieve flexible,
smart and reconfigurable processes in order to address a dynamic and global market (Santos,
et al., 2017). Scania being a worldwide truck manufacturer aims to bring its expertise in truck
production into an Industry 4.0 based system.
The bolt tightening process in the assembly workshop in Scania Oskarshamn presents
promising possibilities for implementation of an Industry 4.0 based system. By monitoring
angle during the tightening process the system is able to identify deviations occurring in the
process or machine (Bickford & Nassar, 1998). An accurate angle interval can be calculated
by studying the relationship between torque and angle using the linear regression method
(Pennsylvania State University, 2018). In this project the angle intervals have been improved
to be able to perform efficient monitoring.
The monitoring process creates the possibility for Scania to implement a smart system
able to identify, analyze and eliminate deviations in real time production. This is achievable
by performing Statistical Process Control (SPC) using the data obtained from production
(Gejdoš, 2015). In this project the Process Capability (Cpk) of the tightening process was im-
proved by 365%.
An actual smart process should be able to automatically perform the analysis of data,
monitor the life of the machine, identify deviations and support the decision making process
(Weihrauch, et al., 2018). To achieve this a proposal is presented to connect the tightening
machines to a statistical analysis software able to present data efficiently to involved person-
nel. The improvement of Cpk and analysis of life of the machine presented in this project
prove that it is possible to implement the torque control with angle monitoring technique into
an Industry 4.0 system.
iii
Sammanfattning
Det finns för tillfället ett ökat behov för tillverkande företag att uppnå flexibla, smarta
och rekonfigurerbara processer för att kunna hantera en dynamisk och global marknad (San-
tos, et al., 2017). Scania är en världsomfattande lastbilstillverkare som strävar efter att för-
medla sin kompetens inom lastbilsproduktion till ett Industri 4.0-baserat system.
Skruvdragningsprocessen i monteringsverkstaden i Scania Oskarshamn erbjuder lovande
möjligheter att implementera ett Industri 4.0-baserat system. Genom att övervaka vinkeln un-
der åtdragningen, med hjälp av ett korrekt vinkelintervall, kan systemet identifiera avvikelser
som uppstår i processen eller maskinen (Bickford & Nassar, 1998). Ett exakt vinkelintervall
kan beräknas genom att studera förhållandet mellan vridmoment och vinkel med en linjär
regressionsmetod (Pennsylvania State University, 2018). I detta projekt har vinkelintervallet
förbättrats för att möjliggöra en effektiv övervakning.
Övervakningsprocessen skapar en möjlighet för Scania att implementera ett smart system
som kan identifiera, analysera och eliminera avvikelser i realtidsproduktion. Detta kan uppnås
genom att utföra statistisk processkontroll (SPC) med hjälp av data som erhållits från produkt-
ionen (Gejdoš, 2015). I detta projekt förbättrades processkapabiliteten (Cpk) för åtdragnings-
processen med 365%.
En smart process bör i praktiken automatiskt kunna analysera data, övervaka maskinens
livslängd, identifiera avvikelser och stödja beslutsprocessen (Weihrauch, et al., 2018). För att
uppnå detta, presenteras ett förslag för att ansluta åtdragningsmaskinerna till en statistisk ana-
lysprogramvara som kan presentera data effektivt för berörd personal. Förbättringen av Cpk
och analys av maskinens livslängd som presenteras i detta projekt visar att det är möjligt att
implementera momentstyrningen med vinkelövervakningsteknik i ett Industri 4.0-system.
iv
Acknowledgments
First off we would like to thank all the people without whom this project would have not
been possible. We would like to extend our most sincere thanks and appreciation to Scania
CV AB in Oskarshamn as Södertälje for giving us the opportunity to perform this project.
We want to thank all the people from the MCEA department, maintenance personnel and
operators for supplying valuable information, taking their time to support us and making this a
great experience.
We want to thank everyone at the Department of Production Engineering (IIP) at the
Royal Institute of Technology (KTH) for the valuable knowledge that was provided to us and
which made the thesis possible.To our supervisor Jonny Gustafsson for his guidance during
the thesis.
Lastly we cannot thank enough our supervisor at Scania, Kerim Hakim without whom we
would not have had the opportunity to work this project. His support and knowledge has been
of great value to us and it has been excellent working together during the thesis.
v
To my parents Blanca Barraza and Francisco Cervantes who have been the greatest sup-
port and without whom I would not be the person I am today. To my brothers, thanks for al-
ways being there for me. And to my oldest brother for his support in this journey of moving to
Sweden. Thank you all for everything.
Carlos Cervantes.
vi
To Ferdos
whom this journey could not be possible without her pure and tremendous support.
To Liana, Diana and Mehrsam
for bringing joy and enthusiasm to my life, in a way I never experienced before.
Esmaeil Nik Arman
vii
Nomenclature
NTG New Truck Generation
C joints Critical joints / Safety Critical Joints
L joints Legal Requirements / Legal Joints / Joints essential for function
ERP Enterprise Resource Planning
MES Manufacturing Execution System
SCADA Supervisory Control and Data Execution
HMI Human Machine Interface
PLM Product Lifecycle Management
PSB Plant Service Bus
DSS Decision Support System
SPCS Smart Process Control System
SPC Statistical Process Control
ANOVA Analysis of Variances
SNR Signal-to-Noise Ratio
OA Orthogonal Array
DOF Degrees of Freedom
viii
Table of figures.
Figure 1. Representation of Scania’s position in the market and the possibilities to implement
its process into an Industry 4.0 based system. ......................................................................... 12
Figure 2. Graphical representation of the Torque control tightening technique. ................... 14
Figure 3. Graphical representation of the Torque control with Angle monitoring tightening
technique. Tolerance limits (red) are defined for Angle and Torque is controlled. ................ 15
Figure 4. Graph representing the Torque vs Angle curve of one tightening and the quadratic
relationship between the two. ................................................................................................... 16
Figure 5. Graphical representation of the deviations that occur in any process. And a
comparison of current condition to an improved situation. ..................................................... 17
Figure 6. Flowchart representing a customized procedure of the Taguchi method. ............... 18
Figure 7. Picture of a regular pistol machine. ......................................................................... 20
Figure 8. Picture of a regular Angle machine. ........................................................................ 21
Figure 9. Picture of a smart Angle machine. ........................................................................... 21
Figure 10. Picture of steering cabinets. ................................................................................... 22
Figure 11. Graphical representation of the tightening process in Scania and the tools
involved in the process. ............................................................................................................ 22
Figure 12. Graphical example of the Torque control with Angle monitoring technique, on the
left is a tightening that has achieved both the proper Torque and Angle, while on the right
only the proper Torque has been achieved. ............................................................................. 23
Figure 13. Representation of Scania’s Digital system PISA. Separated in different software
layers. ....................................................................................................................................... 24
Figure 14. Graph representing the spread of 50 tigthenings of the joint A. In red deviations
are represented. ........................................................................................................................ 27
Figure 15. Capability analysis calculations using the calculated Angle interval compared to
production data. ....................................................................................................................... 28
Figure 16. Ishikawa diagram, with a list of the possible causes defined during brainstorming
sessions. .................................................................................................................................... 30
Figure 17. Graphical representation of operators influence in the process. Two different
operators input is presented. In green a proper operation is presented and in red an operation
not following the standards. ..................................................................................................... 31
Figure 18. Results from a 5 Why? root cause analysis. ........................................................... 31
Figure 19. Main effects plot for Angle (SNR: Nominal-is-Best). ............................................. 34
ix
Figure 20. Interaction plot for Angle. ...................................................................................... 35
Figure 21. Graph representing the spread of 50 tigthenings of the joint B. In red deviations
are represented. ........................................................................................................................ 37
Figure 22. Graph and results from capability analysis run onthe joint B compared to the
calculated Angle interval. This analysis was performed on tightenings after machine was
serviced. .................................................................................................................................... 38
Figure 23. Graphs from capability analysis run on the joint B. On the right is the Cpk
analysis with the calculated Angle interval. On the right the Cpk analysis with the Angle
interval defined by Scania. ....................................................................................................... 39
Figure 24. Flow of data in an ideal SPCS in Scanias tightening process. .............................. 41
Figure 25. Position of software involved in tightening within Scania’s architectural system. 42
Figure 26. Flowchart of the current flow of data in Scania’s tightening process. .................. 43
Figure 27. Flowchart of data flow of the tightening process considering proposal by Mahesh
et al (2017). .............................................................................................................................. 44
Figure 28. Flowchart representation of a proposed DSS into Scania’s system. ..................... 49
Figure 29. Graph showing programming conditions for the tightening tool. Retreived from
(Atlas Copco Industrial Technique AB, Publication Date Not Identified). .............................. 50
Figure 30. Flowchart of the process of calculating an Angle interval in Scania’s system. ..... 51
Table of tables.
Table 1. Experimental design using L8(2^4) orthogonal array using real level of parameters.
.................................................................................................................................................. 33
Table 2. Response table for Angle for SNR: Nominal-is-Best. ............................................... 34
Table 3. Analysis of variance for SNR (Angle). ...................................................................... 35
Table 4. List of necessary functions for a SPCS. ..................................................................... 40
Table 5. List of activities that a SPCS should perform. ........................................................... 40
x
Table of contents
Abstract ...................................................................................................................................... ii
Acknowledgments ..................................................................................................................... iv
Nomenclature ............................................................................................................................vii
Table of figures. ....................................................................................................................... viii
Table of tables. ...........................................................................................................................ix
Table of contents ........................................................................................................................ x
1. Introduction .................................................................................................................. 12
1.1 Background ............................................................................................................ 12
1.2 Problem statement and delimitations ..................................................................... 13
1.3 Purpose and project deliverables............................................................................ 13
2. Literature review .......................................................................................................... 14
2.1 Tightening process ................................................................................................. 14
2.2 Calculating the Torque-Angle relationship (Linear Regression) ........................... 15
2.3 Statistical Process Control ..................................................................................... 16
2.4 Design of experiment ............................................................................................. 17
2.5 Industry 4.0 ............................................................................................................ 18
2.6 Conclusions ............................................................................................................ 19
3. Assembly process in Scania ......................................................................................... 20
3.1 Power tools ............................................................................................................. 20
3.2 Tightening process ................................................................................................. 22
3.3 Production Information System Architecture ........................................................ 24
3.4 Software related to the tightening process ............................................................. 24
4. Case of study 1: Joint A ............................................................................................... 26
4.1 Data collection ....................................................................................................... 26
4.2 Analysis and results ............................................................................................... 27
5. Case 2: Joint B .............................................................................................................. 37
5.1 Analysis and results ............................................................................................... 37
5.2 Capability Analysis ................................................................................................ 37
6. Proposal for Industry 4.0 .............................................................................................. 40
6.1 Smart process control system ................................................................................. 40
6.2 Current state analysis ............................................................................................. 42
6.3 Proposal .................................................................................................................. 43
7. Mindset Strategy. ......................................................................................................... 45
8. Limitations and challenges ........................................................................................... 46
xi
9. Conclusions .................................................................................................................. 47
10. Future work. ............................................................................................................... 48
10.1 Decision Support System ................................................................................... 48
11. Appendix ................................................................................................................... 51
References ................................................................................................................................ 52
12
1. Introduction
1.1 Background
There is currently an increasing need for manufacturing companies to achieve flexible,
smart and reconfigurable processes in order to address a dynamic and global market. To meet
these demands an initiative called Industry 4.0 has been presented by the German Federal
Government (Santos, et al., 2017). The goal of Industry 4.0 is to convert the regular machines
to self-aware and self-learning machines. In order to improve their overall performance and
maintenance management, real time data monitoring as well as to hold the instructions to con-
trol production processes (Santos, et al., 2017).
Scania is a world-leading provider of transport solutions, including trucks and buses for
heavy transport applications. Scania's production system aims to meet customer needs while
achieving increased profitability, growth and competitiveness (Scania, 2018). This is the rea-
son why Scania aims to bring its expertise in truck production into an Industry 4.0 based sys-
tem (Figure 1).
In Scania CV AB in Oskarshamn truck cabs are being produced. The cab factory consists
of four workshops: Press, Body, Paint and Assembly Workshops (Scania, 2018). The bolt
tightening process in the assembly workshop in Scania Oskarshamn presents promising pos-
sibilities for implementation of an Industry 4.0 based system.
In the assembly process, joints that are considered as a necessity for the safety and func-
tionality of the truck or are considered legal requirements are known as Critical and Legal
(C/L) joints, these are of outmost importance in the assembly process since any deviation can
lead to major costs for the company. C/L joints are tightened using “smart” machines, which
are able to control Torque and monitor Angle of turn of the bolt during the process.
Figure 1. Representation of Scania’s position in the market and the possibilities to implement its pro-
cess into an Industry 4.0 based system.
13
To be able to efficiently monitor Angle in the tightening process, tolerances need to be
defined, a high and low tolerance creates what is known as an Angle interval. The Angle
monitoring data obtained from the process can be analyzed and deviations can be identified,
as it is explained in this thesis, consequently making it possible to provide real time feedback
of deviations to software and personnel. This creates the possibility to have a more proactive
and smart process.
1.2 Problem statement and delimitations
After conducting interviews and discussions with the tightening and maintenance per-
sonnel it is known that, control of the clamping force necessary for the Critical and Legal
joints is currently calculated through the Torque control and Angle monitoring method. An
Angle interval in Scania’s process is defined but it is too wide. While the tightenings are with-
in operational tolerances, the wide limits make the monitoring process unable to identify de-
viations and possible improvements.
Moreover the lack of an accurate monitoring process makes the work currently done
more active than proactive. The infrastructure and power tools currently used make it possible
for Scania to implement Angle monitoring into an Industry 4.0 based system. However the
possibilities, limitations, needs (in digital and physical infrastructure) and challenges must be
identified to make this a reality. This project was limited by certain aspects:
This project is focused on smart tools because only smart tools have sensors able to
monitor Angle during tightening.
It is not within the scope of the project to consider more tightening techniques other
than Torque with Angle monitoring.
It is not within the scope to consider thread forming joints.
This project is focused on C/L joints because only these joints use smart tools in their
process.
It is not within the scope of the project to perform design activities and the design
specifications must be followed.
1.3 Purpose and project deliverables
The goal of this project is to define a method for the calculation of an accurate Angle
interval. Identification of feasibility and benefits of the implementation of the Angle monitor-
ing process into an Industry 4.0 based system while involving relevant personnel into an In-
dustry 4.0 mindset.
1. Creation of a local method/tool for determination of accurate angular intervals, and a
plan for implementation into an Industry 4.0 based system.
2. Implementation of smart Angle monitoring and involvement of relevant personnel into
an Industry 4.0 based system.
14
2. Literature review
This section will provide a summary and explanation of relevant concepts in the tighten-
ing of bolted joints and the importance of it in any assembly process. A comparison between
different tightening techniques is presented in order to demonstrate the importance of the
method developed in this project. Finally some concepts from Industry 4.0 that are relevant
for an implementation of smart Angle monitoring are also presented.
2.1 Tightening process
As explained by Bickford (1995). the tightening of bolted joints consists on the creation
of a force able to keep two or more parts together in order to permit the functionality or
movement necessary. This force is created by applying Torque into a bolt, which turns and
screws the two parts together. As the bolt reaches its designed Torque and Angle of turn val-
ues (which is referred to as Angle from now on) it elongates and creates tension which is also
known as clamping force. Therefore in order to avoid failure of the joint, and consequently
economic costs, a correct clamping force is necessary. The correct amount of clamping force
is achieved by a correct design but mostly by the tightening technique used in the process and
performed by the operator as has been concluded by Bickford et al. (1998).
The behavior and life of a joint depends very much on the magnitude and stability of the
clamping force (Atlas Copco, 2018). In automotive assembly tightening techniques of joints
are used to prevent the joints from separating while under service conditions. In current as-
semblies this process is done by an operator using a power tool that performs the job efficient-
ly and is also able to monitor the process.
2.1.1 Torque control
Torque control (Figure 2) is the most common way to control clamping force. A Torque
is defined and the power tool applies Torque on the bolt to create a clamping force. The pow-
er tool controls and measures Torque and tightens until the Torque is inside acceptable toler-
ances. The sole creator of the clamping force in this method is Torque and no other variable is
considered (ASSEMBLY Magazine, 2002).
Figure 2. Graphical representation of the Torque control tightening technique.
15
Although Torque control performs a good enough job there is no way to be 100% certain
that the desired clamping force will be achieved. And as Bickford et al. (1998) concludes
Torque measurements that are not backed up with simultaneous Angle of turn measurement
cannot be totally relied upon.
2.1.2 Torque control with Angle monitoring
Figure 3. Graphical representation of the Torque control with Angle monitoring tightening technique.
Tolerance limits (red) are defined for Angle and Torque is controlled.
For C/L joints Torque must be controlled while the Angle of turn is monitored (Figure 3).
The monitoring of Angle will provide a check that the tightening process was performed cor-
rectly.
In this method Torque is applied and the sensors will measure the Angle through which
the bolt is turned. As the defined Torque value is achieved, the system will evaluate the final
value of the Angle. Tolerances for Angle are defined within an interval where the expected
clamping force is achieved.
If the final Angle value measured is found within the defined tolerances the tightening
will be considered as correct. However if deviations occur, the process will take far too much
(or too little) Torque to arrive within the specified tolerances. This will be defined by the
system as an incorrect tightening (Bickford, 1995).
When the importance of the joint being tightened is high one must assure that the clamp-
ing force achieved is the correct one. The most convenient tightening technique in these cases
is Torque control with Angle monitoring.
2.2 Calculating the Torque-Angle relationship (Linear Regression)
In order to calculate a correct clamping force several methods have been presented by
Bickford et al. (1998). One of them considers that the clamping force is being created as a
function of the increase in both Angle and Torque simultaneously. Therefore it is concluded
that by analyzing the Torque and Angle relationship, and calculating their contribution to the
process it is possible to calculate a correct Angle interval.
16
Figure 4. Graph representing the Torque vs Angle curve of one tightening and the quadratic relation-
ship between the two.
From looking at Figure 4, where one measurement of Torque (x axis) against Angle (y
axis) during tightening was taken (directly from monitoring), it is observable that Angle pre-
sents a relationship with Torque. By using the linear regression method it is possible to quan-
tify the relationship.
This method tries to prove that a percentage of the variation in the “response Y or Angle”
is explained by the variation in the “predictor X or Torque”. Then an equation can be defined
which is used to calculate an Angle interval in which we will find, with a 95% certainty, the
nominal value of Torque (Pennsylvania State University, 2018).
By calculating an Angle interval using this method, the process should be able to identify
deviations from the optimal process. The deviations that occur in any process should be elim-
inated to achieve a high quality product as defined by Gejdoš P. (2015).
2.3 Statistical Process Control
Statistical Process Control or SPC states that the basic issue in a quality oriented process
is to what level we are able to satisfy customer’s expectations. A product which should be
suitable for use should be produced in a stable process (Gejdoš, 2015).
The basic principle of improving processes is based upon the assumption that the varia-
bility of quality index values have two types of causes as shown in Figure 5:
17
Figure 5. Graphical representation of the deviations that occur in any process. And a comparison of
current condition to an improved situation.
● Random causes (common causes) are a permanent part of the process and influence all
process components. They create a wide variety of individually un-identifiable causes,
from which each slightly contributes towards overall variability.
● Assignable causes (special causes) are causes which are not a permanent part of the
process, they do not influence all process components but occur as a consequence of
specific circumstances (Gejdoš, 2015).
2.4 Design of experiment
The experiment can help to prioritize the cause of deviations and illustrates the best com-
bination of different levels of each cause (Dean, et al., 2017). Since the goals of the project
are valid and applicable improvement plans and not just statistical analysis of an isolated con-
dition, a complete cycle of planning, performing, analyzing data and interpreting the results is
required (Jiju & Jiju Antony, 2001).
The nature of the problem that is trying to be solved involves complicated causes of de-
viations with possible interactions, cross-relating different functional areas within the factory
which is another necessity for designing an experiment (Limon-Romero, et al., 2016). There
are different methods to design an experiment, including widely used factorial and fractional
factorial design and innovative methods of Taguchi design. In conditions where a large num-
ber of controllable and uncontrollable causes of deviations ( which now on we call parame-
ters) are involved. Factorial and fractional factorial design illustrate some drawbacks as (1)
they result to more time and cost (2) two design of the same experiment might show different
results (3) interpreting data can become difficult as there is no clear and standard instruction
on how to design an experiment (Roy, 2010). However the Taguchi design works on the basis
of calculating a specific measure called SNR (Signal-to-Noise Ratio) which demonstrates the
ratio between a process mean and its variation. Higher the value of SNR leads to lower devia-
tion in process aimed to be optimized.
18
Figure 6. Flowchart representing a customized procedure of the Taguchi method.
Thus the Taguchi method serves the goal of reducing the deviation in tightening process,
re-tightenings and stop times. By selecting the most suitable combination of parameters ( both
controllable and uncontrollable) it helps to reduce the variation in Angle. Taguchi method
also employs specific set of orthogonal arrays (OAs) which is a table containing set of num-
bers where each set of number can be dedicated to a specific experimental design. learning
how to use OAs is the key to learn the Taguchi experimental design (Limon-Romero, et al.,
2016). The customized process of performing the Taguchi experiment is illustrated in the Fig-
ure 6 (Limon-Romero, et al., 2016) (Yusoff, et al., 2011).
2.5 Industry 4.0
Industry 4.0 is a current trend in manufacturing. It is based on the integration of technol-
ogies for the collection and analysis of real time data in order to increase the efficiency. This
is achieved through collaborative processes, services and human-machine interfaces (Santos,
et al., 2017).
Smart machines in any assembly process enable the rapid generation and collection of
process data into big databases. As proposed by Saurabh P. (2018) the data obtained from real
time processes can be analyzed to identify patterns. This makes the system able to detect de-
viations, perform health monitoring, prediction of failure and send live feedback to the ma-
chines in the process (Weihrauch, et al., 2018).
19
For this purpose machine learning has had a great impact on monitoring systems for pat-
tern detection (Yusoff, et al., 2011). A recent trend in machine learning is Deep Learning
which has surfaced as a method for detection of patterns using raw signals as input data. Deep
learning is based on large data representations. Hinton et al. (1999) divides deep learning in:
● Supervised learning: A function is defined to map an input to an output based on ex-
ample input-output pairs this information is labeled as training data.
● Unsupervised learning: a function is inferred to describe an input to an output.
2.6 Conclusions
It is a fact that an accurate Angle interval should be able to measure if the tightening pro-
cess has been successful (Bickford, 1995). This creates the possibility to detect problems un-
noticeable by Torque control alone. Therefore an accurate Angle monitoring process and the
concept of SPC and Industry 4.0 provide guidelines to use Angle monitoring as a Smart Pro-
cess Control System (SPCS) (Weihrauch, et al., 2018). However the possibilities of imple-
menting a smart Angle monitoring system to improve Scania’s process need to be identified.
20
3. Assembly process in Scania
Scania’s procedures and standards define that bolted joint is the most commonly used
bond type in Scania. Bolted joints must withstand both static and dynamic forces, and they
have to work for a long time in a difficult environment.
In Scania’s assembly plant in Oskarshamn, truck cab components are assembled by oper-
ators that perform the tightening operations according to instructions using hand held power
tools. The selection of bolts, tools, process plan and tightening control technique for each in-
dividual joint is done depending on design specifications and on the degree of importance
(C/L joints). All of this information is provided in internal standards that work as rules to se-
lect the proper specifications (Scania, 2017). Torque is calculated according to internal stand-
ard specifications, which provides the Torque interval values calculated by design for each
specific type of joint (C/L) and bolt.
3.1 Power tools
Most of the industries, including the automotive industry, use hand held power tools in
their assembly process. A power tool is a nutrunner that is driven by an external power source
such as air or electricity. It makes the workers perform their operations efficiently. Power
tools also provide the possibility of feedback from the process which makes them better than
manual tools (Desoutter Industrial Tools, 2018). Different types of electric hand-held tools are
used during the assembly process, depending on the importance of the joint. Here is a sum-
mary of the most common tools used in Scania’s process:
3.1.1 “Regular” pistol machine
A regular pistol machine (Figure 7) is only able to control Torque and cannot monitor the
process. These tools are only used for joints with low Torque demand, lower than 8 Nm
(Atlas Copco, 2018).
Figure 7. Picture of a regular pistol machine.
21
3.1.2 “Regular” Angle machine
Figure 8. Picture of a regular Angle machine.
A regular Angle machine (Figure 8) only monitors Torque but unlike the regular pistol
machine this one is designed for a medium Torque, usually between 8 Nm and 24 Nm alt-
hough it can be used up to 40 Nm (Desoutter Industrial Tools, 2018).
3.1.3 Smart Angle machine
A smart Angle machine (Figure 9) is one that has strain gauges. It has programmable
speed and is capable of monitoring both Torque and Angle. Smart machines are more ergo-
nomic and enhance the quality of the joint. They are able to produce a higher Torque (up to 58
Nm) without affecting ergonomics (Atlas Copco, 2018).
Smart power tools used in Scania are always connected to a steering cabinet (Figure 10)
which sends feedback data from the tightening process to Toolsnet. Toolsnet is an Atlas Cop-
co software able to collect process data. It is also able to present statistics on the steering cab-
inet software using continuous data analysis. There exist a number of different manufacturers
who are producing such kind of tools. Major manufacturers include Bosch, Hitachi, Atlas
Copco, Desoutter, etc.
Figure 9. Picture of a smart Angle machine.
22
Figure 10. Picture of steering cabinets.
Since this smart machines are the only machines available that can monitor Angle, all of
the research done in this work is based on the performance, operation and functions of this
type of machine.
3.1.4 Fixtured Spindle Machine
A fixtured spindle machine is a set of one or more smart machines integrated for applications
that require specific positioning on assembly, involve multiple fastening points or require a
Torque higher than the recommended Torque for a handheld smart machine (Atlas Copco,
2018).
3.2 Tightening process
In the current assembly process while most of the power tools can identify a OK/NOK tight-
ening, only smart machines are capable of connecting to Scania’s current digital framework.
Figure 11. Graphical representation of the tightening process in Scania and the tools involved in the
process.
23
Before performing a tightening the operator must, as shown in Figure 11, scan the work
order (1) in each of the cabs. The scanner reads order specifications for the specific cab and
sends this specifications to the steering cabinet (2) which (according to programming) sends
commands to the smart machine (3) to perform the tightening.
The smart machine is programmed to send an OK/not OK (4) signal with colored lights
(green for OK red for not OK) for the operator to know if the tightening has been performed
correctly. If it is not performed correctly the operator knows to repeat the tightening and if
this takes more time than expected a line stop signal is activated by the operator.
The smart machine’s sensors that monitor the entire process send feedback to the steering
cabinet which stores it and sends process data to Toolsnet (5) which is able to analyze, store
and show data from the tightenings. Finally the tool can be programmed to send an alarm
each time a not OK tightening occurs.
3.2.1 Tightening control technique
As defined in Scania’s standards, most of the tightenings done in the assembly process
are controlled with the Torque control method, since it is the easiest to apply and almost all of
the tools available are capable of performing this method.
But when a joint is considered as C/L a Torque control with Angle monitoring method
(Figure 12) is the mandatory approach since Angle monitoring is necessary a smart machine
is the tool used by the operator to perform the tightening. The use of a smart tool and the deci-
sion of using the Torque with Angle monitoring technique are due to the necessity to assure a
better quality tightening in this type of joints.
Since currently there is no method inside Scania to calculate an Angle interval, the win-
dow for Angle monitoring is calculated by the maintenance department based on prior experi-
ence. To observe the performance of the process a bell curve is graphed using information
gathered from production. Higher and lower Angle limits are then defined so that the process
is able to operate at a high Process Capability (Cpk).
Figure 12. Graphical example of the Torque control with Angle monitoring technique, on the left is a
tightening that has achieved both the proper Torque and Angle, while on the right only the proper
Torque has been achieved.
24
3.3 Production Information System Architecture
Figure 13. Representation of Scania’s Digital system PISA. Separated in different software layers.
Scania has named its digital system Production Information System Architecture (Figure
13). It is a hierarchical system used to structure information management from production
through four different levels of production-related IT systems:
The base level is the production equipment and functions, components and communica-
tions. The second level Supervisory Control and Data Execution (SCADA) is managed by
DIDRIK. DIDRIK adds functionalities such as production calendar, quality assurance, KPI
monitoring, tact display as well as visualization of the production system at the Human Ma-
chine Interface (HMI) level. This system builds an efficient automation layer and finds a way
to meet the requirements that are common. DIDRIK takes care of the communication and
real-time visualization of key parameters. The third, MES level inside Scania is managed by
EBBA. It is a platform for managing production order execution, presenting assembly instruc-
tions, deviation handling and production follow-up. MES is concerned with production, quali-
ty, inventory and maintenance. This level is also in charge of data management. In Scania it is
the connection between MONA and DIDRIK.
The final level is the ERP. It is the tool for PLM, inside Scania this is done through the
software named as MONA, which connects to EBBA. There are also some other independent
systems in this level like MAXIMO and ACTA. This is the way information is transferred
through each of the levels in Scania’s digital architecture (Mahesh & Umer, 2017). However
for the tightening process all of the smart machines are managed by software outside of Sca-
nia’s digital architecture.
3.4 Software related to the tightening process
From our research of the digital structure currently used in the assembly shop it is known that,
since the smart machines are product from the supplier, the software used to manage them is
25
also a supplier’s product. A list of different software and their descriptions are presented as
follows.
3.4.1 Tools talk
Tools talk is an Atlas Copco software used in Scania for programming the controllers or steer-
ing cabinets of the tightening tools used in production. This software sends orders to the steer-
ing cabinet to follow. Speeds and tolerances for Torque and Angle are programmed through
this software (Atlas Copco, 2018).
3.4.2 Toolsnet
As mentioned in the tightening process the steering cabinets send monitoring information
from the process to a software known as Toolsnet. Toolsnet is an Atlas Copco software for
data collection and has some statistical analysis tools.
With this software historical data, statistics and capability indexes of Torque can be accessed
at any time via a standard web browser such as Microsoft Internet Explorer. This software
provides information on every tightening related to a specified period or product. It also pro-
vides a result database that provides access to critical information of final results in the pro-
duction (Atlas Copco, 2018).
3.4.3 Maximo
Maximo is a PLM system that supports the inventory and purchase of non-automotive prod-
ucts, including machines, spare parts, etc. It also works as the maintenance planner software
of machines at Scania. The system is used in all of Scania’s production units (Scania, 2017).
3.4.4 ACTA
ACTA is an offline tool used by the maintenance department to schedule and record calibra-
tion dates of the tightening tools used in the process. It is found only on one computer in the
assembly shop and can only be accessed by maintenance personnel (Scania, 2017).
26
4. Case of study 1: Joint A
From our research of Scania’s assembly process it is known that the joint A is one of the
most important joints in the cab. It is considered a C/L joint. The assurance of a high quality
tightening for this joint is necessary to avoid deviations from happening.
The joint is tightened using a handheld “smart” Angle machine manufactured by Atlas
Copco which is connected to a steering cabinet that sends process data to Toolsnet. In toolsnet
it is possible to look at the behavior of the joints.
4.1 Data collection
We collected data from a six month period (approx 10000 tightenings) from Toolsnet.
The result report from the smart tool used in the joint A was extracted and from the report an
N number of tightenings were picked randomly (to assure reliability). N is defined with the
following empirical equation:
𝑵 = √𝟏𝟎𝒍𝒏 𝒙 𝟒
We created the equation based on the following considerations:
Variable 𝑥 represents number of tightenings performed. For ensuring that influ-
ence of random parameters including (but not limited to) operator shift, environ-
ment temperature, tightening machine battery drain and bolt batch are properly
reflected into calculations, we decided that minimum number of 400 tightenings
should have been performed before being able to use the equation.
The equation provides enough number of samples required to perform the analy-
sis (Johnson, et al., 2011).
Even if number of tightenings increases significantly, number of samples will in-
crease on much slower pace and remain on a limit to be easily processed by any
statistical software.
Then we extracted the trace analysis data for each of the N tightenings. From the behav-
ior of the tightening we are only interested in the elastic region of the tightening up until the
final Torque and Angle values [Angle >= 0 ; Torque >= nominal Torque]. Then we gathered
all the data into one Excel file containing three different columns of information: 𝐴𝑛𝑔𝑙𝑒,
𝑇𝑜𝑟𝑞𝑢𝑒 and 𝑇𝑜𝑟𝑞𝑢𝑒2.
All of these steps are done only to gather the information from the software and facilitate
the interface into Minitab which is the statistical analysis software used during this project.
The data collection process for this case of study took around two days because of the limita-
tions of software and access to data. It is important to explain the data collection process to
emphasize the need of automation and smart processes. Automation of this process is further
explained in Chapter 5 of this report.
(1)
27
4.2 Analysis and results
4.2.1 Linear regression
We defined that Linear regression was the best option for the calculation of an accurate
Angle interval since it calculates the least error using other regression methods like the least
squares method. We then performed a linear regression analysis using the previously men-
tioned data of N tightenings. The result of the analysis shows that as Torque (T) increases
Angle (θ) also increases this relationship is better explained by:
𝜽 = 𝒃𝟎 − 𝒃𝟏 × 𝑻 + 𝒃𝟐 × 𝑻𝟐
All the values used for calculation and results we obtained from analysis are considered
as confidential within Scania, that is why no actual data is presented in this report.
After finding that the relationship between predictor and response exists it is possible to
predict with a 95% confidence that a determined “Torque” value will occur in a “Angle” in-
terval of:
[𝜽𝑳𝒐𝒘 ; 𝜽𝑯𝒊𝒈𝒉]
The prediction of an Angle interval using this method assures that the tightening process
will be performed as expected and increase the quality of tightenings currently performed in
the assembly process.
However during the linear regression analysis we also observed that although the rela-
tionship exists the process presents a substantial amount of deviations from optimal behavior
as observed in Figure 14.
Therefore we concluded that before being able to implement the calculated Angle inter-
val into production a capability analysis is necessary to define if the current process is able to
meet the customer demands within this tolerances.
Figure 14. Graph representing the spread of 50 tigthenings of the joint A. In red deviations are repre-
sented.
(2)
28
4.2.2 Capability Analysis
Figure 15. Capability analysis calculations using the calculated Angle interval compared to produc-
tion data.
For the capability analysis (Figure 15) the previously gathered data of results reported
from Toolsnet is extracted into an Excel file, so that the statistical analysis software is able to
calculate Cpk using the calculated Angle interval values:
The capability analysis values and results are considered as confidential within Scania.
However in the graph we can notice that the process presents high number of variation from
the nominal Angle value. In an optimal process all of the bars in the graph should be found
inside the lower and higher values, showing a tighter curve.
From this analysis we concluded that the current process was not capable of performing
within the Angle interval, and that a high number of tightenings would be consider out of
specifications. This leads to an increase in stop time in the line. Thus it is necessary to per-
form further analysis of the deviations.
As we know from the literature review the high variability that is occurring in the process can
cause a decrease in quality. It is possible to ensure and improve quality of the tightening pro-
cess via operative quality management which includes all methods and activities focusing
upon monitoring processes and removing causes of non-conformity and defects (Gejdoš,
2015).
29
4.2.3 Problem Solving
For the analysis of the deviations due to assignable causes Toyota’s 8 steps for practical
problem solving was chosen to be able to identify the root causes of the problem and assure
that the deviations do not occur again (Goldsmith, 2014).
1) Clarification of the problem
High number of deviations are happening in the tightening process, this means that the
process is not stable and complicates the implementation of a proper monitoring system
through Angle interval.
2) Breakdown of the problem
From a brainstorming session with involved Scania personnel the possible causes are
listed into an Ishikawa diagram as shown in Figure 16:
Based on feasibility, possibility to control, data availability and reach of the project, the
possible causes of deviations were narrowed down. Then an initial observation of the actual
process and cross reference to data indicated that the highest source of deviation in the pro-
cess comes from operator influence.
To investigate the causes of deviation in more detail a second observation was planned,
recording all actions of operator, in addition of recording the condition of the tightening tool,
joint parts and cab. During the observation, the unique identifier of each cab was recorded on
a log sheet, corresponding to the time of the tightening and the rest of the observation parame-
ters mentioned earlier. The second observation was performed on 138 cabs. Results were
transferred to an Excel file, categorized into two group of controllable and uncontrollable
causes and arranged by different causes and their subsequent levels. Finally they were ana-
lyzed with the Minitab statistical software.
30
Figure 16. Ishikawa diagram, with a list of the possible causes defined during brainstorming sessions.
We chose ANOVA method to determine if causes of deviations are actually affecting the
Angle in the tightening process. Since causes of deviations appear randomly, the number of
observations were not the same for each cause. This led to choosing a one-way ANOVA
method (Durivage, 2015).
At the first step we did a normality test on the Angle data, resulting in a need of data
transformation into normal distribution (Johnson, et al., 2011). Johnson transformation was
chosen for this purpose (Minitab, 2018). The result of the one-way ANOVA indicates that all
of the controllable causes of deviations are statistically significant, and most of them are relat-
ed to how the operator holds the tool or how the joint reacts to the forces during the tighten-
ing.
We then concluded that the analysis performed for this joint should focus only on prob-
lems that occur because of operators influence (Figure 17). The first operator was able to per-
form within the specifications and following the standards and instructions defined. However
the operator two was not following the instructions and it is observable from the graph that
this lead to a substantial increase in the standard deviation of the process.
31
Figure 17. Graphical representation of operators influence in the process. Two different operators
input is presented. In green a proper operation is presented and in red an operation not following the
standards.
3) Set the target
We determined that the standard deviation could be reduced by 40% and Cpk increased
by 200%.
From the root cause analysis (Figure 18) it is identified that there is a need for more
competent personnel in the tightening process of C/L joints, to have a more stable process.
4) Analyze the root cause
Figure 18. Results from a 5 Why? root cause analysis.
32
5) Possible solutions
An optimal solution should result in the minimum variation in Angle. Considering the
time and cost limitations on any suggested improvement plan, it is also important to know
which controllable causes of deviations are more significant than others in order to efficiently
sequence and run the improvement plans and monitor the results. To satisfy the mentioned
requirements, we designed an experiment using Taguchi method. Which we defined as the
best methd since there are multiple variables affecting the process and there is no recorded
data on how much variation each of them create.
6) Implementation
Activities 1 and 2 in Taguchi design method (Figure 6) have already been done in steps 1
to 4 of the Toyota’s 8 steps problem solving method. Due to the nature of the production and
limited time available to run the experiment, we excluded uncontrollable parameters from the
experiment and only controllable parameters have been focused on.
Four controllable parameters were identified “Loose holding of the tightening tool”,
“shaft play during tightening”, “applying extra force on tightening tool at the end of tighten-
ing process” and “ Use of dirty gloves by the assemblers”. Also during the previous observa-
tions it we noticed that some parameters might interact with each other which is better to be
investigated in order to provide better insight about tightening process. possible interactions
are:
“Loose holding of the tightening tool” × “Shaft play during tightening”
“Loose holding of the tightening tool” × “applying extra force on tightening tool at the
end of tightening process”
Each of these parameters have 2 levels of “yes” and “no”. Orthogonal array calculated by
using Minitab 18 software resulted to the OA of L8(2^4) which means four parameters , each
in two levels resulting to 8 separate combinations of parameters (which we now call them
cases). The Taguchi design can be seen in Table 1.
Several factors playing a role on deciding how many cabs should be included in each
case. On one side there is an essential need for accuracy of data, especially comparing number
of cabs in each case to the total number of cabs produced on annual basis. On the other side
time constraint needs to be considered. The current number of cabs per each case is consid-
ered to serve both delimitations.
33
Table 1. Experimental design using L8(2^4) orthogonal array using real level of parameters.
Another method to define the number of cases is by calculating the degrees of freedom.
Utilizing following procedure, we concluded the same number of cases (Neseli, 2014).
1- Index each parameter from 1 to 𝑛
levels of paramaters is represented by 𝐿 and we calculate 𝑃1 as:
𝑷𝟏 = ∑(𝑳𝒊 − 𝟏)
𝒏
𝒊=𝟏
2- Index each interaction from 1 to 𝑚 first, for each interaction we calculate 𝐼 as.
𝑰𝒎 = (𝑳𝒊 − 𝟏)(𝑳𝒊 − 𝟏)
then we calculate 𝑃2 as:
𝑷𝟐 = ∑ 𝑰𝒎
𝒎
𝒋=𝟏
3- Degrees of freedom can be calculated by equation 6:
𝑫𝑶𝑭 = 𝑷𝟏 + 𝑷𝟐 + 𝟏
7) Monitoring
When performing the experiment, we prepared the procedure and risk assessment and re-
ceived approval by the responsible personnel. Prior to the experiment we held a meeting with
the line operators to properly explain the purpose of experiment and calrify what is expected
from each operator to do in each of the cases.
Experiment procedure had been followed according to the plan, and we recorded required
data in a log sheet. We had to repeat he experiment on about 30 of the cabs due to uncertainty
of having the exact combination of controlled parameters as required.
For calculating SNR, three strategies are available: lower-is-better, higher-is-better and
nominal-is-best (Limon-Romero, et al., 2016). Since it is required to minimize the deviation
around mean Angle, we chose “nominal-is-best” strategy. Figure 19 show the result of SNR
analysis. Horizontal line illustrates no effect on process and with increase of the line slope, the
(3)
(4)
(5)
(6)
34
magnitude of parameters importance increases. Additionally a response table can be included
to calculate rank of each parameter as in Table 2. Response table calculates Delta (∆). Higher
Delta value illustrates the higher importance and hence higher magnitude of parameter effect
of process performance.
Figure 19. Main effects plot for Angle (SNR: Nominal-is-Best).
Table 2. Response table for Angle for SNR: Nominal-is-Best.
Similar to the plot of SNR, the plot of standard deviation was generated illustrating the
same rank as in Table 2. In order to utilize SNR to study the interactions, interaction plot was
generated which is illustrated in Figure 20. For interpreting the interaction plot, parallelism of
lines is very important. If lines are crossing each other it illustrates significant importance of
the interaction. If lines are just nonparallel, it illustrates interaction, but not as significant as
crossed lines. Parallel lines means no interaction between parameters (Neseli, 2014). It can
then be concluded that there is strong interaction between the parameters that were initially
expected.
35
Figure 20. Interaction plot for Angle.
Next activity was an ANOVA analysis. Since there are equal number of observations in
each case, multiple parameters (causes of deviations) affecting tightening Angle and interac-
tion between parameters is in our interest, multi-way ANOVA is required (Durivage, 2015).
First we performed a normality test on Angle data with satisfactory results. As it is illustrated
in Table 3 there is a strong interaction between “Loose holding of the tightening tool” and
“applying extra force on tightening tool at the end of tightening process”. The analysis also
indicates that most important parameters to consider are “applying extra force on tightening
tool at the end of tightening process” followed by “Loose holding of the tightening tool”.
Table 3. Analysis of variance for SNR (Angle).
As for comparing the experiment results with our target, we calculated standard deviation
of Angle and Cpk for all the cases. Caes 8 illustated the best values by reducing standard de-
viation by 66% (compared to 40% target) and increasing Cpk by 365% (compared to 200%
target).
8) Standardize
From a visit to the basic skills department the people in charge of training explained to us
the way training of the operators for the tightening process is performed. We then concluded
that the training performed is sufficient and covers all the points that have been identified as
critical. However it came to our attention that there is no monitoring on how the operator ac-
tually performs while working on the line. Moreover there is a lack of reinforcement of the
36
training of operators and this is due to the mindset of the personnel and management. The
focus on producing as much as possible causes a lack of commitment on operator’s training
and causes the deviations occurring in the actual process.
We propose that a mindset strategy to involve operators and management into the tight-
ening process, explaining how tightening affects both quality and production, is implemented.
This would increase the conscience of the personnel on the importance of performing a proper
tightening, thus increasing the quality of the process and the product.
37
5. Case 2: Joint B
The second case of focus for this project involved the joint B which uses a fixtured one
spindle machine. This machine was chosen because it is used in the tightening of the joint B
which is considered as a C/L joint because its role is essential for the functionality of the cab.
Recently the machine was serviced and the data for this period is available in Toolsnet data-
base. Because of the functionalities of the machine (theoretically) the operation of this ma-
chine is not affected by the operator’s influence unlike the joint A.
5.1 Analysis and results
The same data collection process as in the previous case was applied and analyzed in the
statistical software and the Angle interval was calculated.
[𝜽𝑯𝒊𝒈𝒉 ; 𝜽𝑳𝒐𝒘]
Linear regression analysis also showed that for this joint, the relationship is better and the
process is more stable compared to te joint A.
It is observable from the analysis and Figure 21 that this joint presents a better relationship
between Torque and Angle. It also shows how the process behaves without the operator’s
influence.
Although the process is more controlled in this joint, it still presents a high number of
deviations in the final Angle.
Figure 21. Graph representing the spread of 50 tigthenings of the joint B. In red deviations are repre-
sented.
5.2 Capability Analysis
From a preliminary capability analysis (Figure 22) using the calculated Angle interval it
was defined that a lower percentage, compared to the joint A, of tightenings would be consid-
ered as NOK if the calculated Angle interval was implemented.
38
Figure 22. Graph and results from capability analysis run on the joint B compared to the calculated
Angle interval. This analysis was performed on tightenings after machine was serviced.
As in the previous case this information is classified for Scania so it was impossible for
us to share. However we can observe in the graph that although the behavior of the process is
more stable than the previous case there are still deviations from the process, and possible
improvements that can help lower the standard deviation and improve Cpk can be found.
5.2.1 Testing the calculated Angle interval
During the time we were working in the project there occurred a breakdown in this spe-
cific machine. We are not able to discuss the nature of the breakdown since this is considered
as confidential information. However this created the possibility for us to test how the previ-
ously calculated Angle interval would have been capable of identifying that there was a prob-
lem with the machine before the breakdown occurred. By doing a capability analysis of the
time prior to the identification of the breakdown of the machine it is possible to test how the
calculated Angle interval would have been able to identify the problem with anticipation in
Figure 23:
39
Figure 23. Graphs from capability analysis run on the joint B. On the right is the Cpk analysis with
the calculated Angle interval. On the right the Cpk analysis with the Angle interval defined by Scania.
By comparing the two graphs and the values of Cpk and standard deviation we were able
to observe how the process would have performed when using the calculated Angle interval.
It was noticeable that a high number of deviations where occurring 2 weeks prior to the fail-
ure. Since the Angle interval defined by Scania [LSL ; USL] was too wide the process was
not able to identify that the standard deviation had doubled from the optimal process. Howev-
er if the calculated Angle interval was used the process would have automatically detected
this problem. From calculations we know that the defect rate would have increased by a 200%
this would have made it obvious for the involved personnel that the machine was not perform-
ing at its optimal conditions and service could have been scheduled before the breakdown
occurred which would have saved time and money for Scania.
We concluded from this second case that by using an efficient Angle interval it is possi-
ble to monitor the life of the machine and that it is possible to schedule service and calibration
dates from the Angle monitoring data.
40
6. Proposal for Industry 4.0
In this chapter current state analysis and a proposal for implementation into an Industry
4.0 based system, considering that the next step in the project is to present ideas for the im-
plementation of the Angle monitoring process. Also enabling technologies, needs, possibili-
ties, advantages, disadvantages, limitations and future work are presented.
6.1 Smart process control system
In order to create a conceptual model for developing a Smart Process Control System
(SPCS) Scania’s requirements where identified and compiled in Table 4 by comparing current
and target implementation of a smart monitoring process. The suggestions from operators and
personnel involved in tightening were taken into consideration.
System necessary function
Calibration scheduling
Maintenance Scheduling
Feedback to operator
Deviation analysis
Quality checks
Feedback to tightening and maintenance personnel
Table 4. List of necessary functions for a SPCS.
Using the previously mentioned requirements a conceptual model for a Smart Process
Control System can be created considering Table 5:
SPCS
The system shall share data automatically
The system shall continuously analyze data and present results efficiently
The system shall identify deviations using SPC.
The system shall alert person responsible depending on the deviation that occurs.
The system shall schedule tool calibration and service automatically
The system shall trace products and link relevant process and quality data to it.
The system shall support the decision making process.
Table 5. List of activities that a SPCS should perform.
From this analysis two main goals are defined with the implementation of an Industry 4.0
based system in the Angle monitoring process. First to automate the process for the calcula-
tion of a proper Angle interval. Second to perform a continuous real-time smart monitoring of
the tightening process. For both goals there are two main activities, data collection and data
analysis.
As defined by Weihrauch et al. (2018) an efficient SPCS consists of four main features:
Data from the live virtual representation is used to generate manufacturing and quality rec-
ords. Algorithms for simulation and prediction of future states use data from and to determine
possible effects of events or decisions. The Decision Support System (DSS) finally uses data
from all sources to provide a holistic view of the current situation, past events and possible
future developments.
41
Figure 24. Flow of data in an ideal SPCS in Scanias tightening process.
It is important to point out that although a live virtual representation of the process is a
main feature for an efficient SPCS, it is not considered in this report since there is no virtual
twin of the production and the creation of it is out of the research scope.
Now that the ideal SPCS has been defined, and with the system requirements, activities
and delimitations, it is possible to map the ideal process of correlations between involved
tools (Figure 24).
To be able to have a smart tightening process the system should be able to extract the
monitoring data from tightening tools. This data should be transferred automatically into a
software that is able to perform advanced statistical analysis. The next step is to run the statis-
tical analysis which varies depending on the goal. Next is a brief proposal of this system for
both goals: calculation of Angle interval and smart Angle monitoring.
6.1.1 Calculating the Angle interval
When calculating a proper Angle interval the data from the tightenings (Torque and An-
gle) needs to be analyzed by linear regression. Using the calculated Angle interval a capability
analysis needs to be run to evaluate feasibility of implementation. Deviations occurring in the
process are identified and analyzed to eliminate them (decision making process). Finally all
the data must be recorded into a database for future work. The process of Angle interval cal-
culation is better explained in Figure 31 in the appendix. It is good to clarify that we per-
formed this calculation for two joints. However Scania requires that this process is done for
each and everyone of the C/L joints found in the assembly line and also for new joints. This is
the reason why this process should be automatized.
6.1.2 Smart Angle monitoring
On the other hand, for real time process monitoring, data is being continuously analyzed
by a statistical software, so the data from tightenings must be extracted without delay into a
42
statistical analysis software. The results from Statistical Process Control can then be analyzed.
Then a tool for automatic decision making is necessary. This tool should be able to, by look-
ing at results from SPC and maintenance (MAXIMO), calibration (ACTA) and quality rec-
ords, make decisions to finally make changes in production or alert people involved in the
process (team leader, tightening or maintenance departments) of possible deviations and solu-
tions.
Besides the previously mentioned things, maintenance (MAXIMO), calibration (ACTA)
and quality tools should be accessible by the decision making software which can then, after
SPC analysis, provide live feedback and make changes. This system should work as a loop to
assure continuous improvement and monitor performance of the system.
6.2 Current state analysis
Figure 25. Position of software involved in tightening within Scania’s architectural system.
As mentioned before all of the software involved in the programming, maintenance, cali-
bration and monitoring of the tightening processes in Scania work independently without a
connection to Scania’s system (Figure 25) or between each other, this is because the software
is made by the supplier of the machines (Atlas Copco). This is the reason why, as explained in
the data collection part of this report, data is extracted from Toolsnet into an Excel file, before
it is possible to do a statistical analysis of the process.
Toolsnet works as the database where all of the information that occurs during tightening
process is found. Although it is able to perform a capability analysis, it can only perform this
with Torque data. Even though Angle data is collected Toolsnet is not able to perform analy-
sis. Currently the analysis of Angle monitoring done is by looking at the Torque Angle curves
found in Toolsnet when a problem in production occurs.
The possibility of an SPCS implementation is limited by the lack of connections between
the tools used currently in Scania as shown in Figure 26.
43
Figure 26. Flowchart of the current flow of data in Scania’s tightening process.
Most of the data transfer between the involved tools is currently performed manually and
there is no tool on PISA system that is able to collect and send relevant data to other tools.
Calibration, maintenance and quality checks scheduling are currently implemented using in-
dependent software solutions with no automated interface to the PISA system.
Finally, all decisions depend on a limited number of people who have enough knowledge
to interpret the monitoring, maintenance, calibration and quality data. No virtual representa-
tion or simulation data is used in the decision making process.
6.3 Proposal
There is an alternative approach proposed by the digital factories department in Söder-
tälje (Mahesh & Umer, 2017) to use a Plant Service Bus (PSB) system in order to establish
communication with the power tools. This connection makes it possible to extract data auto-
matically into a database which can connect to a statistical software for analysis of the data.
Finally after analyzing the data it can be presented in an efficient way to the involved person-
nel according to their needs.
6.3.1 Data gathering and statistical analysis
Data would then be extracted directly from the steering cabinets using the PSB service as
shown in Figure 27. Then another service which is able to subscribe to the PSB known as the
Hadoop data lake works as a database that is able to connect to Zeppelin. Zeppelin is a statis-
tical analysis tool that can perform SPC and present data efficiently for personnel to interpret
(Mahesh & Umer, 2017).
44
Figure 27. Flowchart of data flow of the tightening process considering proposal by Mahesh et al
(2017).
1) Identification of needs
Since this kind of system has already been developed by Scania the needs for implemen-
tation would be mostly about infrastructure. The system requires that every steering cabinet is
connected to an adaptor able to send data to the database.
Other requirement is the training of personnel in this kind of software. This can be fixed
if Scania creates a tool easy to access by the involved personnel. However some level of train-
ing would still be necessary.
2) Economic analysis
The requirements previously mentioned would lead to some costs but the major costs that
have been identified are the major changes that need to be done in the actual production line.
Since every steering cabinet should have it’s own adapter the activity to perform this would
require people to perform, time and possibly stop time necessary.
Finally the IT department or someone involved in the process should also be able to con-
figure all the connections to make sure that everything is correct. Also the system should be
able to handle massive amounts of data (one day of production approx. 350 cabs).
45
7. Mindset Strategy.
One of the main problems while implementing a new technology, system or making
changes into a current system is the involvement of personnel. It is the personnel involved in
the process who bring the system to life (Liker, 2013).
An Industry 4.0 based system focuses on the efficient use of all tools available in the pro-
cess to support the personnel involved to continuously improve their work (Vaidya, et al.,
2018). However as mentioned before there are some limitations for the implementation of a
smart Angle monitoring system.
From interviews with the involved personnel and research performed we know that the
greatest problem is the lack of involvement. We then defined that if this method is explained
to the management it would be easier to implement in the line. The personnel in the line is
open for an implementation of this kind of system. However, management needs to accept
this implementation plans. If the benefits and ease of implementation is explained, using as
example this project, the involvement of management will increase since it improves the
product and the brand as it helps improving Scania's core values of customer first and elimi-
nation of waste.
Moreover, as Scania's main objective is to "provide the customer with profitable and sus-
tainable transport solutions to contribute to the success of their businesses" the examples pre-
sented in this thesis prove that the implementation of a system like this eliminates both waste
of over working and increases the life of the truck, thus increasing customer's profitability.
Presenting all this benefits in this manner to the management would increase the interest
in involving all the personnel in an Industry 4.0 based system. Thus reducing the time in
which it would be possible to implement.
46
8. Limitations and challenges
During the project several limitations where identified that are necessary to overcome be-
fore considering a successful implementation of an Industry 4.0 based system. The challenges
met during the project, considering Table 7 in chapter 6, are discussed below to provide an
insight on the feasibility of implementation of a smart Angle monitoring process.
To begin with, the biggest problem encountered was the fact that the tools connect direct-
ly to Toolsnet. Toolsnet provides some insight into the process. However it was insufficient
for the research performed during this project. Moreover, Toolsnet is unable to share data
automatically and the data extraction process, that needs to be done manually, is tedious and
time consuming. This makes personnel reluctant to work with the method presented in this
project and makes the involvement of personnel into a SPC driven system more difficult.
This same problem was encountered with several tools involved with the smart machines.
This makes the decision making process more difficult and makes it impossible to analyse the
behavior of the process during the occurrence of the events. Moreover the solutions to prob-
lems that previously occurred are not documented or are not accessible by all personnel. Most
of this information is only known by the personnel involved. The current system does not
efficiently support the decision making process.
There is no statistical analysis tool inside Scania’s PISA system. The lack of an analysis
tool makes the system dependent on external tools. If the proposed system were to be imple-
mented, a major restructuring of the assembly shop may be necessary.
47
9. Conclusions
This paper is mainly focused on the calculation of an Angle interval. The research proves
that when properly calculated, the Angle interval works as an efficient monitoring tool that is
able to identify deviations from the process and monitor the life of the machine. The method
is a suitable base for the definition of calibration and maintenance times and it would help
Scania implement a continuous improvement system into the tightening process.
While the method defined here works for the calculation for the Angle interval. It is no-
ticeable that the process is not controlled (as shown in chapter 4) and before being able to
implement the calculated Angle interval into other joints an analysis of the deviations must be
performed by the involved personnel.
Results from the experiments performed during this project show that a eight step prob-
lem solving methodology is an effective way to identify and eliminate deviations in the pro-
cess. This allows for the Angle interval to perform a proper monitoring while increasing the
capability and quality of the process to finally reduce problems and costs for Scania overall.
In theory Angle monitoring can also be implemented into an Industry 4.0 based system.
However the research of Scania’s system done in chapter 5 shows that the main limitation for
an immediate implementation is the dependence on suppliers. Suppliers software and lack of
connections to other tools make the data collection and analysis process both tedious and time
consuming.
The automation of data extraction from Toolsnet would eliminate the aforementioned
challenges and the analysis would depend only on simple tasks in an statistical software
(Minitab). This kind of tools are able to present results and analysis efficiently enough for a
person with enough knowledge to interpret.
There is also the possibility for Scania to consider other options other than Toolsnet
which should be designed to fulfill Scania’s necessities. Moreover if Scania aims to eliminate
dependence on supplier software, tools inside the PISA system should be created. It is im-
portant to point out that tools are being developed to perform different purposes but there is
currently no tool inside PISA system able to perform statistical analysis or a DSS. The devel-
opment of proposed tools would give a great advantage to Scania CV AB, increasing process
capability, quality of tightenings, reduce time in problem solving, implement continuous im-
provement system and reduce workload on personnel involved.
The concept presented in chapter 4 (PSB) would seem, for the moment, as the most ap-
propriate approach to create a SPCS. If the interfaces are done correctly the involved person-
nel should only receive the data that concerns them and would make data interpretation and
decision making process easier. However the amount of time, personnel and resources that
need to be allocated for the implementation of a system like this may be too high for an im-
mediate implementation.
Another limitation with this approach is that the system needs to be capable of collecting
huge amounts of data and at the same time be able to connect without delay to the analysis
software. This may not be possible at the moment since the production in Scania Oskarshamn
is substantial and the amount of data gathered may be too much for the system to handle.
48
10.Future work.
As mentioned before, if Scania aims for the implementation of an SPCS three main tools
have been identified for this purpose:
● Database: where relevant tightening data can be found and presented in effective man-
ner and can connect automatically to other tools.
● Statistical Analysis: Tool that can extract automatically data from the database and
perform continuous analysis.
● Decision support system: Tool to ease the process of decision making and in the future
even make decisions automatically.
The next step in the design of a SPCS is the creation of a DSS for automatic feedback
system. Since there is no virtual or simulation data, this is done by collecting information
from different activities that affect the process (maintenance, quality, calibration and process
data) and creating a database. A database can be used as the basis for a DSS by mapping the
necessary information for each department and involved tools e.g. MAXIMO.
10.1 Decision Support System
Previously in this paper, proposals for the first two tools have been identified but the
SPCS should also support the decision making process. For this purpose a Decision Support
System (DSS) is proposed as future work.
Information is the foundation of a robust decision making process and is a requirement
for efficient reaction on events. A decision making process requires (Weihrauch, et al.,
2018):
● A complete picture of the current situation in manufacturing to understand the event
and its context.
● Records of the past manufacturing process, as well as data about previously occurred
events.
Figure 28 shows a flow chart of a basic Decision Support System proposed for Scania
CV AB. The DSS monitors the process constantly and is searching for deviations. If a devia-
tion has been identified, the system decides if there is enough data to proceed with the DSS. If
enough data is not available the decision making process needs to be performed manually by
using a problem solving approach. If the system has enough data it analyzes the data and tries
to find similar situations that have occurred in the system before (decision alternatives) and
searches to see if there is a solution that is feasible for this specific deviation. This process
should also take into consideration data from MAXIMO and ACTA. When finding one feasi-
ble choice the DSS proposes preselected solutions, and the user selects one of the solution
alternatives or defines a new one. Finally, the decision is implemented by the system.
49
Figure 28. Flowchart representation of a proposed DSS into Scania’s system.
The concept of machine learning proves that since the data measurement and analysis is
presented in statistics e. g. Cpk, SPC, Torque vs Angle curve, etc. this data can be analyzed to
identify patterns so the system is able to automatically detect faults (Fukunaga, 1990).
10.1.1 Pattern recognition
The proposal for the use of pattern recognition is based on the idea that recognition by a com-
puter is the process of generating an appropriate representation for information and transform-
ing it into another representation.
A pattern is any data received from the object by an input device and entered into a recogni-
tion system. Translated into tightening terms it is the data that has been extracted from the
machine and repeats itself over a time period (Fukunaga, 1990).
For the implementation of a pattern recognition tool two main algorithms are necessary:
● Algorithms for extracting the boundary and region of a pattern.
● Algorithms for recognizing the original object based on knowledge of the object, its
patterns, and how it is used.
Finally the concept of unsupervised learning (Hinton & Sejnowski, 1999) defines that after a
pattern has been recognized an output can be automatically mapped for that specific pattern
thus making it possible for the DSS to make a decision by itself without human input (Hinton
& Sejnowski, 1999).
50
Figure 29. Graph showing programming conditions for the tightening tool. Retreived from (Atlas
Copco Industrial Technique AB, Publication Date Not Identified).
This concept can also be adapted into the tightening process. Results from experiments per-
formed in this project show that deviations occurring in the process present a certain behavior.
If every deviation that occurs in the production line is mapped, patterns can be recognized
over time. Then the smart tools used for C/L joints are able to send different alert signals de-
pending on the area of the curve where a NOK happens as seen in Figure 29.
If a pattern recognizes a deviation, the machine is able to send an alert directly to the per-
son involved or to the DSS and then a solution can be implemented without delay in the pro-
duction line.
In theory, the creation of a DSS is possible. But first there must be an improvement in
traceability of deviations and events occurring in the tightening process, (maintenance, cali-
bration, deviations and quality checks need to be properly recorded and studied). After acquir-
ing sufficient data a DSS based on pattern recognition will become a possibility.
In conclusion the research performed during this project proves that the monitoring pro-
cess presents promising possibilities and is suitable for an implementation into an Industry
4.0 based system. However, some organizational improvements need to be performed first. If
process, organization and personnel involvement is improved this implementation is most
definitely possible by using the tools and methodologies mentioned in this report. However, it
is important to mention that while the implementation is possible, it is a gradual process that
will occur over a long period of time.
51
11 Appendix
Figure 30. Flowchart of the process of calculating an Angle interval in Scania’s system.
52
References
ASSEMBLY Magazine, 2002. www.assemblymag.com. [Online]
Available at: https://www.assemblymag.com/articles/83305-choosing-the-optimum-fastening-
system
[Accessed March 2018].
Atlas Copco Industrial Technique AB, Publication Date Not Identified. Focus60/Focus61
User Guide, Publication Place Not Identified: Atlas Copco Industrial Technique AB.
Atlas Copco, 2018. Cordless Assembly Tools. [Online]
Available at: https://www.atlascopco.com/en-uk/itba/products/assembly-solutions/electric-
assembly-tools/cordless-assembly-tools
[Accessed March 2018].
Atlas Copco, 2018. Fixtured assembly solutions. [Online]
Available at: https://www.atlascopco.com/en-uk/itba/products/assembly-solutions/fixtured-
assembly-solutions
[Accessed March 2018].
Atlas Copco, 2018. ToolsNet 4000. [Online]
Available at: https://www.atlascopco.com/en-uk/itba/products/assembly-solutions/Error-
proofing-solutions/ToolsNet-4000
[Accessed May 2018].
Atlas Copco, 2018. ToolsTalk 2. [Online]
Available at: https://www.atlascopco.com/en-uk/itba/products/assembly-solutions/electric-
assembly-systems/toolstalk-2
[Accessed May 2018].
Bickford, J. H., 1995. An introduction to the design and behaviour of bolted joints. 3rd ed.
New York: Marcel Dekker.
Bickford, J. H. & Nassar, S. eds., 1998. Handbook of Bolts and Bolted Joints. New York:
Marcel Dekker.
Dean, A., Voss, D. & Draguljić, D., 2017. Design and Analysis of Experiments. 2nd ed.
s.l.:Springer.
Desoutter Industrial Tools, 2018. Electric Assembly Systems. [Online]
Available at: https://www.desouttertools.com/tools/2/electric-assembly-systems
[Accessed March 2018].
Durivage, M. A., 2015. Practical engineering, process, and reliability statistics. s.l.:American
Society for Quality (ASQ).
Fukunaga, K., 1990. Introduction to Statistical Pattern Recognition. 2nd ed. Boston:
Academic Press.
Gejdoš, P., 2015. Continuous Quality Improvement by Statistical Process Control. Procedia
Economics and Finance, Volume 34, pp. 565-572.
53
Goldsmith, R. H., 2014. Toyota's 8-Steps to Problem Solving. s.l.:CreateSpace.
Hinton, G. & Sejnowski, T. J., 1999. Unsupervised learning: Foundations of neural
computation. Computers & Mathematics with Applications, 38(5-6), p. 526.
Jiju, A. & Jiju Antony, F., 2001. Teaching the Taguchi method to industrial engineers. Work
Study, 50(4), pp. 141-149.
Johnson, R., Freund, J. & Miller, I., 2011. Miller & Freund´s Probability and statistics for
engineers. 8th ed. Upper Saddle River, NJ: Pearson Prentice Hall..
Liker, J. K., 2013. The toyota way : 14 management principles from the world´s greatest
manufacturer, New York: McGraw-Hill.
Limon-Romero, J. et al., 2016. Application of the Taguchi method to improve a medical
device cutting process. The International Journal of Advanced Manufacturing Technology,
87(9), pp. 3569-3577.
Mahesh, B. & Umer, M., 2017. Smart Connected Power Tools; An Industrial Implementation
of Event Driven Architecture, Stockholm: s.n.
Minitab, 2018. Data transformations. [Online]
Available at: https://support.minitab.com/en-us/minitab/18/help-and-how-to/quality-and-
process-improvement/capability-analysis/supporting-topics/distributions-and-transformations-
for-nonnormal-data/data-transformations/
[Accessed April 2018].
Neseli, S., 2014. Optimization of process parameters with minimum thrust force and Torque
in drilling operation using Taguchi method. Advances in Mechanical Engineering, Volume
2014.
Pennsylvania State University, 2018. Regression Methods (STAT 501). [Online]
Available at: https://newonlinecourses.science.psu.edu/stat501/
[Accessed March 2018].
Roy, R. K., 2010. A Primer on the Taguchi Method. 2nd ed. Place of publication not
identified: Society of Manufacturing Engineers (SME).
Santos, C. et al., 2017. Towards Industry 4.0: an overview of European strategic roadmaps.
Procedia Manufacturing, Volume 13, pp. 972-979.
Scania, 2017. Internal Standards, Södertälje: Scania CV AB.
Scania, 2018. Om Scania Oskarshamn. [Online]
Available at: https://www.scania.com/productionunitoskarshamn/sv/home/om-scania-
oskarshamn.html
[Accessed May 2018].
Scania, 2018. Scania at a glance. [Online]
Available at: https://www.scania.com/group/en/scania-at-a-glance
[Accessed May 2018].
54
Vaidya, S., Ambad, P. & Bhosle, S., 2018. Industry 4.0 – A Glimpse. Procedia
Manufacturing, Volume 20, pp. 233-238.
Weihrauch, D., Schindler, P. A. & Sihn, W., 2018. A Conceptual Model for Developing a
Smart Process Control System. Procedia CIRP, Volume 67, pp. 386-391.
Yusoff, N., Yusup, S. & Ramasamy, M., 2011. Taguchi's parametric design approach for the
selection of optimization variables in a refrigerated gas plant. Chemical Engineering Research
and Design, 89(6), pp. 665-675.
TRITA ITM-EX 2018:559
www.kth.se