Final Report

A Project Report on

“Developing a System for Improving Effectiveness of Process Control and Risk Minimization at End Cap Welding Process”

ByG SAIKIRAN

Indian Statistical Institute, KolkataM.Tech (Quality Reliability and Operation Research)

1st Year, Registration No – QR1109

Under the Guidance ofProf. A L N Murthy

(ISI, SQC-OR Unit, Hyderabad)

&

Dr D. S. SettySr. Manager

(Nuclear Fuel Complex, Hyderabad)

Statistical Quality Control & Operation UnitIndian Statistical Institute

1

Hyderabad

Certificate

This is to certify that Mr. G.SAIKIRAN, Ist year student of M Tech. (QR-OR), Indian Statistical Institute, Kolkata has undergone a project titled “Developing a System for Improving Effectiveness of Process Control and Risk Minimization at End Cap Welding Process” at Nuclear Fuel Complex (NFC), Hyderabad from 7th May 2012 to 25th July 2012, under my guidance in partial fulfillment of the course requirements.

Prof A L N MurthyIndian Statistical InstituteHyderabad.

2

Acknowledgement

I take this opportunity to express my deep sense of gratitude to Prof. A.K.Chakraborty, Indian Statistical Institute Kolkata for giving me the opportunity to do the project in ISI- Hyderabad and sincere thanks to my guide Prof. A L N Murthy, SQC-OR Unit, ISI-Hyderabad, for his timely & invaluable guidance and unstinted co-operation throughout my project work. His vast knowledge, experience, thoroughness, patience and simple behavior have greatly inspired me.

I am very thankful to Dr. D. S. Setty, Senior Manager, Fuel Assembly – Pressurized Heavy Water Reactor of Nuclear Fuel Complex (NFC), Hyderabad and the management of NFC for providing me an opportunity to work under them. Despite of his very busy schedule, the way Dr. Setty helped me to carry out my experiment and to clarify my doubts has inspired me a lot. His dedication and effort to the work is always a great learning during this period of time.

I would like to express my deep gratitude to Mr. S K Pathak, for their inspiration and their interest to carry out this project to a success and all the workers, Supervisors of NFC who has helped me in no of ways in the Shop-Floor.

Special thanks to the Ms G Swetha, Scientific Officer C, NFC without her the journey would have been difficult to complete. The way she helped me to understand the technical aspects of the process is really great. I hope she will continue his hard work to achieve new heights in her career.

I would like to thank all the other Professors, Administrative Officer in ISI-Hyderabad for their inspiration and encouragement. I am also thankful to all the Office-Staffs at ISI-Hyderabad for their help.

G. SaikiranM.Tech (QR-OR), ISI

3

Table of Contents

About the Organization:..............................................................................................................................5NUFAP........................................................................................................................................................7

End Cap Welding Process.......................................................................................................................9Resistance Butt Welding (RBW) Process...............................................................................................9Quality Control:.....................................................................................................................................13

Back ground to the Project........................................................................................................................15Problem Statement.....................................................................................................................................15Objectives of Study:..................................................................................................................................15Scope of the Study.....................................................................................................................................16

Approach...............................................................................................................................................16Details of the Study...................................................................................................................................16

Data Collection......................................................................................................................................16Analysis.....................................................................................................................................................18

Pareto Analysis......................................................................................................................................18Analysis of UT values for Machine C...................................................................................................20

X bar – S chart for the Machine C.....................................................................................................21ANOVA for UT Values on Machine C.............................................................................................24ANOVA of Log Standard deviation for Machine C..........................................................................31Histogram of UT values of Machine C.............................................................................................35Probability Plot..................................................................................................................................36Process Capability Analysis for Machine C......................................................................................37Metallography Test Data Analysis....................................................................................................38P- Chart..............................................................................................................................................38Existing Process Capability:..............................................................................................................39

Analysis of Machine D..........................................................................................................................40ANOVA for UT – Machine D...........................................................................................................42Process Capability Analysis for Machine D......................................................................................50Metallography Test Data Analysis of Machine D.............................................................................51

Analysis of Machine E:.........................................................................................................................53Analysis of UT- Machine E...............................................................................................................53Process Capability Analysis for Machine E......................................................................................65Metallography data Analysis of Machine E......................................................................................66

Conclusion and Results:............................................................................................................................68Recommendations.....................................................................................................................................70FMEA (Failure Mode and Effect Analysis)..............................................................................................71

Introduction:..........................................................................................................................................71Developing Process FMEA for End Cap Welding................................................................................71PROCESS OF CONDUCTING FMEA:...............................................................................................72ANALYSIS:...........................................................................................................................................74

Pareto Analysis..................................................................................................................................74CONCLUSION:........................................................................................................................................75Process Flow of ECW................................................................................................................................77FMEA table...............................................................................................................................................78

4

Nuclear Fuel Complex

About the Organization:

The Nuclear Fuel Complex (NFC), established in the year 1971 is a major industrial unit of Department of Atomic Energy, Government of India. The complex is responsible for the supply of nuclear fuel bundles and reactor core components for all the nuclear power reactors operating in India. It is a unique facility where natural and enriched uranium fuel, zirconium alloy cladding and reactor core components are manufactured under one roof starting from the raw materials.

The Fuel India is pursuing an indigenous three stage Nuclear Power Program involving closed fuel cycles of Pressurized Heavy Water Reactors (PHWRs) and Liquid Metal cooled Fast Breeder Reactors (LMFBRs) for judicious utilization of the relatively limited reserves of uranium and vast resources of thorium. PHWRs are used in the first stage of Indian nuclear Power program which uses zirconium alloy material as clad & Natural uranium dioxide as fissile fuel. In addition, India is operating two Boiling Water Reactors (BWRs) for the last 30 years. The zircaloy clad enriched uranium oxide fuel elements and assemblies for these reactors are fabricated at NFC starting from imported enriched uranium pellets.

Uranium Refining and ConversionThe raw material for the production of PHWR fuel in NFC is Magnesium Di-uranate (MDU) popularly known as 'Yellow Cake'. The MDU concentrate is obtained from the uranium mine and milled at Jaduguda, Jharkhand, operated by Uranium Corporation of India Limited (UICL). The impure MDU is subjected to nitric acid dissolution followed by solvent extraction and precipitation with ammonia to get Ammonium Di-uranate (ADU). By controlled calcination and reduction, sinterable uranium dioxide powder is formed which is then compacted in the form of cylindrical pellets and sintered at high temperature to get high density uranium dioxide pellets. For BWRs, in case of enriched uranium hexafluoride import it is subjected to pyro hydrolysis and converted to ammonium di-uranate, which is treated in the same way as natural ADU to obtain high density uranium dioxide pellets.

Zircaloy Production The source mineral for the production of zirconium metal is zircon (zirconium silicate) available in the beach sand deposits of Kerala, Tamil Nadu and Orissa and is supplied by the Indian Rare Earths Ltd. Zircon sand is processed through caustic fusion, dissolution, solvent extraction (to remove hafnium), precipitation and calcination steps to get zirconium oxide. Further, the pure zirconium oxide is subjected to high temperature chlorination, reactive metal reduction and vacuum distillation to get homogeneous zirconium sponge. The sponge is then briquetted with alloying ingredients and multiple vacuum arc melted to get homogeneous zircaloy ingots which are then converted into seamless tubes, sheets and bars by extrusion, pilgering and finishing operations.

5

Fuel Fabrication For PHWR fuel, the cylindrical UO2 pellets are stacked and encapsulated in thin walled tubes of zirconium alloy, both ends of which are sealed by resistance welding using zircaloy end plugs. A number of such fuel pins are assembled to form a fuel bundle that can be conveniently loaded into the reactor. The fuel bundles for PHWR 220 MW and PHWR 500 MW consist of 19 and 37 fuel pins respectively. For BWRs, two types, namely 6x6 and 7x7 array fuel assemblies are fabricated.

Seamless Tubes, FBR Sub-assemblies and Special MaterialsThe Stainless Steel Tubes Plant and Special Tubes Plant at NFC produce a wide variety of stainless steel and titanium seamless tubes for both nuclear and non nuclear applications. NFC is supplying sub-assemblies and all stainless steel hardware including tubes, bars, sheets and springs for the operating FBTR and the forthcoming PFBR. The Special Materials Plant at NFC manufactures high value, low volume, high purity Special Materials like tantalum, niobium, gallium, indium etc., for applications in electronics, aerospace and defense sectors.

Fabrication of Critical Equipment A notable feature at the Nuclear Fuel Complex is that, apart from in-house process development, a lot of encouragement is given to the Indian industry for fabrication of plant equipments and automated systems. Major sophisticated equipments fabricated in-house at NFC include the slurry extraction system for purification of uranium, high temperature (1750oC ) pellet sintering furnace, vacuum annealing furnace, cold reducing mill, split spacer and bearing pad welding machines, automatic tube cleaning station, etc. In addition to this, several services like vacuum arc melted alloys production, seamless tube extrusion and finishing, production of tools, NDT services, etc., are undertaken.

Self RelianceThe Nuclear Fuel Complex is an outstanding example of a successful translation of indigenously

developed processes to production scale operations. The strong base of self-reliance in the crucial area of nuclear fuel and core components is a great asset to the country in not only supporting the nuclear power program but also in developing a large number of allied and ancillary industries.

ScopeThe NFC has different types of production facilities which include the Zirconium Oxide Plant for processing of Zircon to pure Zirconium oxide; the Zirconium Sponge Plant for conversion of Zirconium oxide to pure sponge metal; facilities for reclamation of zircaloy mill-scrap; the Zircaloy Fabrication Plant for producing various zirconium alloy tubes and also sheet, rod and wire products; the Uranium Oxide Plant for processing crude uranium concentrate to pure uranium di-oxide powder; the Ceramic Fuel Fabrication Plant for producing sintered Uranium oxide pellets and assembling of the fuel bundles for the PHWRs.

The Enriched Uranium Oxide Plant for processing of imported enriched uranium hexafluoride to enriched uranium oxide powder; the Enriched Uranium Fuel Fabrication Plant for producing enriched UO2 pellets and the fuel assemblies for the BWR reactors; and a plant for fabrication of components and

6

sub assemblies for Fast Breeder Reactors. A Special Materials Plant for producing a number of electronic grade high purity materials for supplies to the Electronic Industry and plants producing stainless steel seamless and other special tubes have also been set up in this complex

NFC has strong team of research scientists and engineers for carrying out the industry needed research and process development activities. NFC has designed, developed and exported critical equipments to the nuclear and non nuclear establishments. The research activities helped NFC to achieve highest process recoveries and lowest waste generations. It has strong commitment towards nature and developed various green processes in the chemical plants.

The Nuclear Fuel Complex is an outstanding example of a successful translation of indigenously developed processes to production scale operations. The strong base of self-reliance in the areas of nuclear fuel and core components manufacturing is a great asset to the country in not only supporting the nuclear power program but also in developing a large number of allied and ancillary industries.

NUFAP

NUFAP is the part of NFC where all the fabrication worked is being carried out. The processes which are being carried in are as follows:

Machining Spacer Pad Welding Bearing Pad Welding (For Outer Elements) Graphite Coating and Vacuum Baking Despatch to NUOFP for pellet loading End cap welding Element Machining Degreasing Back filling and Helium leak test Element inspection Element assembling and End plate welding

The Process flow sheet of the NUFAP is as shown below.

7

Rejects

8

Tube Machining

Spacer pad Welding

Graphite Coating and Vacuum Baking

Bearing pad Welding (Outer Elements)

Pellet LoadingStart of Active

material handling

End Cap Welding Process

Element Machining Process

Degreasing

Back filling and HLT

Element inspection

Element Assembling and End Plate

Welding Process

Separation of pellets and tubes

Accepted pellets

Recycle of rejected tubes / elements

Tube Cleaning

End Cap Welding Process

This is the process of study in the project, after the pellets are loaded into tubes the lots are sent back to the plant for further processing of the elements. In ECW after filling the element with Helium both the ends are welded with profile caps.

The details description about each of the sub processes is described in the FMEA section of this report.

The Welding process adopted is Resistance butt welding the details of the process is described below.

Resistance Butt Welding (RBW) Process

Resistance Butt welding process is a solid phase welding process wherein the joint is achieved between the two components butting each other in the same axis. One component is fixed in the stationary electrode and the other component is placed in the movable electrode. These two electrodes are connected to secondary side of a power transformer. The two components are pressed together by applying constant squeeze force through a pneumatic system. This pneumatic squeeze pressure is regulated through the pressure gauge fitted on the machine. Prior to welding and during welding process, helium gas is purged in and around the weld joint for achieving both cover gas and inert atmosphere requirements. After the stabilization of squeeze force, high current at low voltage is passed over a specified time period through the secondary circuit of the welding machine with the help of a programmable weld controller. Weld controller is a device, which controls the sequence and duration of the complete resistance welding cycle consists of four basic steps as follows:

(i) Squeeze time (ii) Weld time(iii) Hold time(iv) Off time

9

Fig.1. Principle of resistance butt welding

Squeeze time is the interval between the initial application of electrode force on the work and the first application of current. During this period, squeeze pressure stabilization is achieved on the components to be welded and high points on the contact surfaces are collapsed due to localized yielding of the joining materials. This results in increase of contacting area.

Weld time is the duration of weld current passage, is controlled by electronic, mechanical, manual, or pneumatic means. Times ranges from one half cycles (1/100 sec.) for very thin sheets to several seconds for thick plates. For the capacitor or magnetic type of stored energy machines, the weld time is determined by the electrical constant of the system. The basic welding cycle shown in Fig. consists of one or more of the following features to improve the physical and mechanical properties of the weld zone:

1. Pre compression force to seat the electrodes and work pieces together. 2. Pre heat to reduce the thermal gradient and to achieve intimate contact between the butting surfaces before the start of weld current. 3. Forging force to consolidate the weld nugget.4. Quench and temper times to produce the desired weld strength properties in hardened alloy steels. 5. Post heat to refine the weld grain size in steels. 6. Current decay to retard cooling on aluminum. In some applications, the welding current is supplied intermittently during weld time; it is on during heat time and ceases during cool time.

10

Fig. 2 Basic Welding Cycle (Single Impulse)

Hold time - the time during which force is maintained to the work after the last impulse of current ends; during this time, the weld nugget solidifies and is cooled until it has adequate strength.

Off time - the time during which the electrodes are off the work and work is moved to the next weld location; the term is applied where the welding cycle is repetitive.

Advantages of Resistance Butt Welding Process

The advantages of resistance butt-welding process are as follows:

1. Suitable for automation and mass production.2. Generation of consistent weld quality.3. Less heat affected zone (HAZ)4. Higher chemical purity of the weld joint.5. Independent of operator skill.6. No filler material requirement.7. Very less or Zero Defect generation after qualification of the process.8. Cost effective process.9. Light machinery with high precision weld controllers is an added advantage.10. Suitable for wide range of materials joining due to its higher flexibility.

Defects in Resistance Butt Welding Process

The resistance butt-welding joints (fuel tube to end cap) are subjected to high temperature, pressures and neutron flux environment. It is very essential to ensure weld joint quality otherwise; inherent flaws in the weld joint will lead to fuel failures. Weld defects in the end cap joints results in the release of highly radioactive fission products into the Primary Heat Transport (PHT) system also called as coolant of the nuclear power reactor. This leads to removal of failed fuel assembly from the reactor much early resulting in loss of money, time and sometimes resulting in reactor shut down. Defects are categorized mainly on the nature of formation namely, dimensional and metallurgical.

11

Dimensional Defects

Dimensional defects in resistance welding joints are associated with misalignment of components and dimensional variations of the components.

a) Eccentricity: If the axis of tube and end plug is not in same line during welding, eccentricity defect is generated in the weld joint, resulting in formation of lack of fusion line along the weld interface also called as non-fusion line. Eccentricity up to 50 microns is allowed in normal production.b) Less or More Weld Upset: During resistance welding process, if the material deformation (upset) is less or more than optimized either due to heat or squeeze force variations during welding process results in change of welded product’s length. This result in defect generation in the subsequent operations carried out on the fuel element. More or less upset leads to lack of fusion and metal expulsion defect formation. This also generates bad end profile defect during element machining operation.c) Tube or Weld Slippage: In resistance welding process, one component (Tube) is fixed firmly in the electrode (4 Jaw split Collet) and other part (End cap) is positioned in the movable electrode. Slippage of tube from the fixed electrode (collect) is due to either less collet pressure or less tube outside diameter than the collet inside diameter, results in generation of incomplete or weak weld.

Metallurgical Defects

The Lack of fusion or non-fusion line and phase transformation are the most common defects that occur in welding process. Non-fusion line defect is mainly due to the presence of impurities such as oxides, grease and other organic substances and solid foreign bodies that are left on the mating surfaces. As a rule, their presence at the interface causes incomplete metallic contact between the surfaces to be joined. Consequently, the progress of bond formation is impeded and various defects appear in the weld joint.

Incomplete fusion manifests itself as non-fusion line of a few tenths of a micrometer wide (in the direction perpendicular to the joint surface) and of a few tenths of square millimeter area. In some cases, unbonded spots may be filled with air, inert gas, oxides or contaminants that have not had time to diffuse into base material being joined. As a rule, unbonded spots are narrow, scattered all over the joint and do not merge to the external surface of the weld joint. Sometimes, the unbonded spots join to form into stringer, which appears on a micro section as intermittent lines running along the joint surface. Any stringer inclusion, which exists along the interface, may be extended outwards to the surface of the material being welded. They often open up into cracks during subsequent operation. It is essential that such flaws should be identified by visual or ultrasonic examination.

When zirconium alloy material is heated up to 8530C and above, it changes from alpha to beta phase resulting in change of structural and mechanical properties. Due to high currents, heat generation

12

increases and temperature reaches to melting point, resulting in metal expulsion or sparking defect generation at the interface. These defects are not acceptable for the present application, as it results in through and through defects, revealed by helium leak testing and UT methods. Leak rate beyond 1X10-8

Std. CC/Sec is not acceptable.

Quality Control:

In order to get assured about the quality of the weld the customer has set up a sampling plan. The following tests are conducted to ensure quality

1. Ultrasonic Test2. Metallography Test 3. Visual inspection.4. Helium leak Test

Sampling plan can be subdivided into 4 hierarchies

Qualification: Machine is to be qualified in any of the following instances:1) For new machine2) When machine is not in use for more than 6 months3) After major break down 4) Relocation of machine

Setup: Setup is performed in order to ensure that the weld parameters are set perfectly for a qualified machine. It is generally done at the start of a shift. In this case the machine should be in normal working condition.

Process: After every 40 welds, take one process element for Metallograpahy test.

Random: Four random elements (8 welds) from every lot from each machine for Ultrasonic test (UT).

Requalification: It is kept to ensure if a qualified machine is kept idle for a long period (6 months) then by requalification we ensure that the machine is still capable of making perfect welds.

The sampling plans for each test are as below:

SET UP (2 TUBE WELDS & 2 ELEMENT WELDS)

Step 1: Subject for UT& Metallography test.

Step 2: If Not Ok then REPEAT THE SETUP (AFTER CHECKING THE COLLET CONDITION &

ALIGNMENT)

Step 3: If Not Ok then take 10 welds after Checking Cleanliness of Tube and Plugs.

Step 4: f Not Ok then take 10 more welds after more welds after checking parameters like Preheat, Weld heat,

13

squeeze pressure etc

Step 5: If Not Ok take 30 weld check the number of weld cycles and shape of weld.

Step 6: If Not Ok take 30 more weld after checking Line voltage and circuit resistance.

Step 7: If Not Ok then take 100 welds and thoroughly check the machine.

Step 8: If Not Ok then take 300 welds for Machine qualification.

Step 9: If Ok at any of the steps then we can start production.

RANDOM SAMPLING PLAN FOR UT AT END CAP WELDING

Start production after acceptance of setup by UT and Metallography

Take 4 random elements from each lot

OK Not Ok

Accept the lot Subject this lot to 100% UT

Accept OK Reject Not OK

elements elements

If > 4 welds i.e. 10% of welds found rejected

in this lot, succeeding and preceding lots shall

be subjected to 100% UT

Accept ok elements Reject not ok elements

Procedure for UT of End Cap welds:

14

If > 4 welds (10% of welds) are found reject able in any lot, complete shift production shall be subjected to UT. If more

than 10% of the welds found rejected in complete shift production machine shall be stopped and is to be re-qualified

with 30 welds.

1. Calibrate the equipment with known defect standards and note down the peak height. 2. Confirm the peak height with known leaky element. 3. Subject 10% of the elements to UT. 4. Note down the peak height for each weld and mark the defective location. 5. Acceptance criteria: Welds with signal height equal to or greater than 70% of FSH (full-scale

height) shall be rejected

Back ground to the Project

The ECW is the most important process and any defect at this stage would severely affect the targets, dispatch and if any defects are found at later stages would affect reactor functioning. Though the rate of non conformities is very low but by the current system Corrective and preventive action takes a long time in view of this the senior executives felt a need to understand this situation well in advance either to QC or customer to report such quality issues. In order to develop a proactive risk assessment from proactive risk control it is thought that the best place to understand and take effective measures is at process and as well as during production. In view of this an urgent need is felt to develop such a system by examining the current process control as well as quality control information.

Problem Statement

“Developing a system for improving Effectiveness of process control and risk minimization at PHWR (Pressurized Heavy Water Reactor) – End Cap welding process.”

Objectives of Study:

The main Objectives of the project are as stated below: Understanding the current process control system. Identification of possible causes affecting the process stability. Study of CTQ’s and understanding the behavior using the statistical principles. Developing process control measures for predicting the behavior and risk assessment. Carrying out Process FMEA for identification of high risk areas and suggesting remedial

measures that minimize the risk.

Scope of the Study

15

Currently in the plant there are Seven machines where the end cap welding is carried out of which Machine C and D are essentially used for ECW of 37 E bundle and Machine E for ECW 19 E bundle while rest of the machines are used for ECW of both the elements depending on the production requirements. Therefore it is felt to concentrate our study only on Machine C, D, and E.

Approach

Studying the current process control procedures. Data collection for defects, UT and MT. Pareto Analysis on defects data to find Vital few and trivial many. ANOVA to find effects of Operator, shift and other process parameters on UT values. Finding the underlying distribution of UT. Evaluating the Process Capability for UT. Evaluating the Process Capability for MT. Process Map for ECW. Process FMEA to find the high risk areas in ECW. Conclusions. Recommendations.

Details of the Study

Data Collection

The following data has been collected for this study on the defects, Ultrasonic test (UT), Metallography test (MT) for ECW process.

Data on defects for 1 year (from April’11 to March ’12).

Monthly reports of QC (both UT and MT) from April’11 to March ’12.

UT values of the Random inspection from the log books of QC dept, Machine C for 2 months (April’12 & May’12), Machine D for 2 weeks (2nd March-15th March 2012), and Machine E for 1 month (December 2011).

Setup and Process values (MT) from the log books of QC dept.

The analysis is carried out on the data for UT and MT from random, process and setup inspections conducted by the quality control dept. of the ECW. The sampling plans of these tests are already stated above.

In view of the complexity of the product as well as the process being adopted for ECW and the existing controls being followed it is thought to be reasonable to assume the following:

The end caps used for welding are as per the design specifications. 16

The tubes are defect free.

The Machine has been correctly qualified.

The setup has been perfect.

The readings given by Ultrasonic testing machine are correct.

The quality of the weld is affected only because of the change in the parameters of the process.

The analysis for each machine has been done separately as the settings for each machine is different and also discussion and preliminary study with the operators and supervisors of the plant has helped in coming to this conclusion. Importantly each machines C, D, E handles different type of the elements, 19E and 37E bundle tubes as they vary in dimensions and design features are also different.

The following statistical methods were applied on each machine:

Pareto Analysis to determine the Vital few and trivial many defects

Using X bar - S chart to find the out of control points if any and the control limits for regular monitoring of the process are determined.

To find out the effects of the Operator, shift and also the influence of important parameters on the quality of the weld (UT values) using ANOVA, Main effect plot, Box plot for each of the factors.

ANOVA on the Log of Standard deviation for each subgroup to find the effect of operator, shift and also the influence of important parameters on variability of the weld quality, Main effect plot and box plot are also studied for each of the factors.

Fitting the distribution for the UT values and choosing the best fit distribution.

Evaluation of the existing process capability of each machine and making recommendations to improve the process capability.

Using P-chart to analyze the last one year data for non conformities in MT.

Evaluating the process capability on the basis of the MT data and making the recommendations to improve the process capability.

Conducting Process FMEA of ECW for identification of high risk areas and suggesting remedial measures that minimize the risk.

Analysis

Pareto Analysis

17

The data collected on stage wise defects of total plant have been subjected to Pareto analysis is done to identify the vital few stages where most defects are occurring as shown in Fig. 3, 4, 5. This exercise is repeated for both 19 Elements and 37 Elements in the ECW as shown in Fig. 6.

For 19 Elements

Total_19E 690 470 203 188 142 127 98 75Percent 34.6 23.6 10.2 9.4 7.1 6.4 4.9 3.8Cum % 34.6 58.2 68.4 77.8 84.9 91.3 96.2 100.0

Defects_ SDSPWSamplesBPWOthersUsed in RBDHTE C W

2000

1500

1000

500

0

100

80

60

40

20

0

Tota

l_19E

Perc

ent

Pareto Chart of Defects_

Fig.3 Pareto chart of defects for 19 E

For 37 Elements

Total_37E 1151 1105 253 209 169 154 138 97Percent 35.1 33.7 7.7 6.4 5.2 4.7 4.2 3.0Cum % 35.1 68.9 76.6 83.0 88.1 92.8 97.0 100.0

Defects_

3500

3000

2500

2000

1500

1000

500

0

100

80

60

40

20

0

Tota

l_37E

Perc

ent


18

Fig.4 Pareto chart of defects for 37 E

For Total Defects in 19 Elements and 37 Elements case:

Total19&37E 1795 1621 441 412 296 296 213 195Percent 34.1 30.8 8.4 7.8 5.6 5.6 4.0 3.7Cum % 34.1 64.8 73.2 81.0 86.6 92.3 96.3 100.0

Defects_SWS

D

Sample

sBP

W

Used

in RB

Others

DHT

E C W

6000

5000

4000

3000

2000

1000

0

100

80

60

40

20

0

Tota

l19&

37E

Perc

ent


Fig.5 Pareto chart of total defects for 19 E &37 E

It is evident from the Pareto chart that the ECW (End cap welding) and DHT (Double Head Turning) stage contribute to around 65% of the defects. The defects at the stage of ECW are more serious in view of its contribution to highest no. of defects and also any defect at this stage results in greater loss and most importantly cause greater customer dissatisfaction.

The Pareto chart for the defects in the End Cap Welding process as shown in Fig. 6.

19

Total 829 739 102 89 78Percent 45.1 40.2 5.6 4.8 4.2Cum % 45.1 85.4 90.9 95.8 100.0

DEFECT OtherSlippageSparkingUpset VariationUT

2000

1500

1000

500

0

100

80

60

40

20

0

Tota

l Defe

cts

Perc

ent

Pareto Chart of DEFECT ECW

Fig.6 Pareto chart of total defects in ECW

From the Pareto analysis it is evident that rejections due to UT and upset variation contribute to 85.4 % of the total rejections at ECW. So the primary focus should be on the reasons which cause rejections due to UT and Upset variations.

Analysis of UT values for Machine C

The analysis is carried out separately for each machine.

Machine C

This machine is used for welding the 37E bundle lots in the period of analysis.

X bar S chart is used to analyze whether process means are in control and control limits are set for the process as shown in Fig. 7 and 8.

X bar – S chart for the Machine C

20

Fig.7 X bar S chart for Machine C (existing process)

The initial X bar S chart suggests that there are many out of control points in the existing process and to set the control limits homogenization is required.

The X bar S chart after homogenization is as shown below.

The Control limits for the X bar chart are:

CL= 40.456, UCL= 44.615, LCL= 32.296

The Control limits for S chart are

CL=3.784, UCL= 6.868, LCL=0.700 as seen in the control chart below.

21

Fig.8 Homogenized X bar S chart for Machine C

Currently no control charts are being used for the data collected on UT. As control chart help in effective process control, it is therefore suggested to the plant officials to use X bar- S chart to check if the process is in control or not and if it goes out of control then necessary actions be taken to bring it to find out the assignable causes for that.

In order to understand the influence of Operator and other parameters the data that is collected on UT has been summarized day wise for understanding the process level as well as the variability.

Main Effect Plot

This plot shows the variation of the Mean UT, Standard deviation, Log Standard of each subgroup with the parameters i.e. operator, shift, squeeze pressure, collet pressure, pre heat and weld heat. The plots are as shown in Fig 9 and 10.

22

Fig.9 Main effects plot of Mean UT values

Observations: The mean of operator KSR is high compared to rest of the operators. The average UT value decreased with increase in the squeeze pressure up to 1.8kg/cm2

The average UT doesn’t change much with Collet pressure. Steep decrease in Mean UT value at preheats 5.6 kA.

Fig.10 Main effects plot of Standard deviation of UT values

23

Observations: Standard deviation is more for operators JB and SNT. Standard deviation is more in the General shift compared to other shifts. For squeeze pressure of 1.8 kg/cm2 the standard deviation is minimum. Least Standard deviation is observed for the preheat value of 5.6 kA. Standard deviation has increased moreover with the increase in the weld heat.

ANOVA for UT Values on Machine C

The Generalized Linear Model (GLM) ANOVA couldn’t be applied to the data as there is Rank deficiency due to empty cells. So to study the effect of Operator, Shift and each parameter of the process one way ANOVA is used on the UT data.

One-way ANOVA: UT versus Operator

Source DF SS MS F POperator 7 415.1 59.3 2.60 0.011Error 5808 132463.7 22.8Total 5815 132878.8 Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev ----+---------+---------+---------+-----BP 2056 40.468 4.390 (--*-)DB 344 40.235 4.107 (-----*-----)JB 1712 40.956 5.526 (--*--)JSRI 1144 40.643 4.537 (--*---)KSR 72 41.722 4.520 (-------------*------------)MLN 168 40.714 4.553 (--------*--------)RP 208 40.346 4.360 (-------*-------)SNT 112 41.134 4.691 (----------*----------) ----+---------+---------+---------+----- 40.00 40.80 41.60 42.40

Pooled StDev = 4.776

The hypothesis is rejected which implies there exists significant difference between the Operators. So Operator performance is different on this Machine C.

24

Fig.11 Box plot of Operators for UT values

From the box plot in Fig.11, except for operator KSR rest all operators are showing many outliers, Operator JB has much non conformity i.e. UT values greater than 70 one of the reasons for this could be the change in settings by operators.

One-way ANOVA: UT versus Shift

Source DF SS MS F PShift 3 175.2 58.4 2.56 0.043Error 5812 132703.6 22.8Total 5815 132878.8

Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev ----+---------+---------+---------+-----First 2376 40.868 5.061 (-*-)General 32 40.281 4.861 (----------------*---------------)Second 2880 40.540 4.615 (*-)Third 528 40.438 4.315 (---*---) ----+---------+---------+---------+----- 39.0 40.0 41.0 42.0


The hypothesis is rejected which implies there exists significant difference between the shifts. So there exists significant effect of shifts on UT values on Machine C.

25

Fig.12 Box plot of Shift for UT values

Though there are some visible differences in between Shifts as shown in Fig. 12, but any conclusion can’t be drawn as the no of data points supported by Third shift & General Shift are insufficient. So box plot of different operators across the shifts is drawn to have better conclusions about the performance.

Fig.13 Box plot of Operators and shift for UT values

26

From the box plot, Fig. 13 Operator KSR, SNT has consistent performance across the shifts while MLN behavior in the second shift is much better than in first. So rescheduling is required for this operator.

One-way ANOVA: UT versus Sq. Pressure

Source DF SS MS F PSq.Pressure 4 277.8 69.4 3.04 0.016Error 5811 132601.0 22.8Total 5815 132878.8

Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev ---------+---------+---------+---------+1.2 80 41.438 5.243 (-----------------*----------------)1.6 568 41.063 4.706 (-----*------)1.7 272 41.228 6.242 (--------*---------)1.8 4568 40.560 4.691 (-*-)1.9 328 40.759 4.561 (-------*--------) ---------+---------+---------+---------+ 40.80 41.40 42.00 42.60


The hypothesis is rejected which implies there exists significant difference in performance for different values of squeeze pressure on Machine C.

Fig.14 Box plot of Squeeze pressure for UT values

Though there is some visible difference in between squeeze but any conclusion can’t be drawn as the no

27

of data points supported by 1.2Kg/cm2 is insufficient and it has been observed that the every operators more or less uses same setting . So box plot of different operators for different squeeze pressure would give better picture.

Fig.15 Box plot of Squeeze pressure and Operator for UT values

From the box plot in Fig.15 Operator BP has more outliers with squeeze pressure of 1.8 while JB has shown many nonconformity too with 1.8 squeeze pressure. MLN has very inconsistent performance with 1.9 squeeze pressure.

One-way ANOVA: UT versus Collet Pressure

Source DF SS MS F PColletPressure 3 274.3 91.4 4.01 0.007Error 5812 132604.6 22.8Total 5815 132878.8

Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev ------+---------+---------+---------+---3.0 2072 40.472 4.262 (-----*-----)3.1 1256 40.493 4.670 (-------*------)3.2 208 40.688 4.539 (-----------------*------------------)3.4 2280 40.929 5.272 (----*-----) ------+---------+---------+---------+--- 40.25 40.60 40.95 41.30


28

The hypothesis is rejected which implies there exists significant difference in performance for different values of collet pressure on Machine C.

Fig.16 Box plot of collet pressure for UT values

From the box plot as in Fig.16 it is clear that the Collet pressure of 3.4 has very inconsistent behavior compared to the other collet pressures. This has to be addressed and the necessary action must be taken.

One-way ANOVA: UT versus Preheat

Source DF SS MS F PPreheat 4 141.6 35.4 1.55 0.185Error 5811 132737.3 22.8Total 5815 132878.8

Individual 95% CIs For Mean Based on Pooled StDev

Level N Mean StDev -+---------+---------+---------+--------4.5 248 40.585 4.222 (-----*-----)5.0 2720 40.647 4.687 (*-)5.5 2568 40.703 4.722 (-*-)5.6 56 39.161 3.515 (------------*-----------)5.7 224 40.862 6.916 (------*-----) -+---------+---------+---------+-------- 38.0 39.0 40.0 41.0


29

The hypothesis is accepted which implies that the change in the pre heat has no significant effect on the UT values.

One-way ANOVA: UT versus Weld heat

Source DF SS MS F PWeldheat 7 907.7 129.7 5.71 0.000Error 5808 131971.1 22.7Total 5815 132878.8 Individual 95% CIs For Mean Based on Pooled StDev

Level N Mean StDev -+---------+---------+---------+--------10.7 56 39.161 3.515 (------------*-----------)11.0 1632 40.333 4.399 (-*--)11.2 80 39.737 3.818 (---------*----------)11.5 1320 40.557 4.556 (--*-)11.6 1416 41.064 4.692 (--*-)12.0 96 40.635 4.742 (--------*---------)12.2 712 41.254 5.876 (---*--)12.3 504 40.371 5.155 (---*---) -+---------+---------+---------+-------- 38.0 39.0 40.0 41.0


There are 8 different weld heat settings used and from ANOVA it is clear that that the effect of weld heat is significant on the UT values.

Fig.17 Box plot of weld heat for UT values

30

It is clearly evident from Fig.17 that there exists difference in performance for the weld heat used but to have a clear understanding box plot, Fig.18 of UT values for Operator and weld heat is considered.

Fig.18 Box plot of weld heat and Operators for UT values

ANOVA of Log Standard deviation for Machine C

To understand the effect of different parameters like Operator, shift, squeeze pressure, collet pressure, pre heat, weld heat on standard deviation. In order to ensure that the underlying assumptions of ANOVA are valid a log transformation has been used for analyzing standard deviation.

One-way ANOVA: Log STD versus Operator

Source DF SS MS F POperator 7 0.2045 0.0292 1.04 0.399Error 719 20.1244 0.0280Total 726 20.3288

Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev --------+---------+---------+---------+-BP 257 0.5680 0.1514 (---*--)DB 43 0.5665 0.1481 (-------*--------)JB 214 0.5949 0.1906 (---*---)JSRI 143 0.5632 0.1728 (----*---)KSR 9 0.6020 0.1489 (-----------------*------------------)MLN 21 0.5988 0.1490 (-----------*-----------)

31

RP 26 0.6008 0.1338 (----------*----------)SNT 14 0.6398 0.1505 (--------------*-------------) --------+---------+---------+---------+- 0.540 0.600 0.660 0.720


The hypothesis is not rejected that implies that there doesn’t exist any significant difference in the log standard deviation of operators.

One-way ANOVA: Log STD versus Shift

Source DF SS MS F PShift 3 0.1134 0.0378 1.35 0.256Error 723 20.2154 0.0280Total 726 20.3288

Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev -+---------+---------+---------+--------First 297 0.5901 0.1740 (-*-)General 4 0.6547 0.1719 (---------------*----------------)Second 360 0.5672 0.1658 (-*)Third 66 0.5861 0.1417 (---*---) -+---------+---------+---------+-------- 0.50 0.60 0.70 0.80


The hypothesis is not rejected that implies that there doesn’t exist any significant difference in the log standard deviation in shifts.

One-way ANOVA: Log STD versus Sq. Pressure

Source DF SS MS F PSq.Pressure 4 0.2920 0.0730 2.63 0.033Error 722 20.0368 0.0278Total 726 20.3288

S = 0.1666 R-Sq = 1.44% R-Sq(adj) = 0.89%

Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev -+---------+---------+---------+--------1.2 10 0.7011 0.1156 (--------------*--------------)1.6 71 0.5987 0.1633 (-----*----)1.7 34 0.6092 0.2122 (-------*-------)1.8 571 0.5701 0.1666 (-*-)1.9 41 0.6098 0.1354 (------*------) -+---------+---------+---------+-------- 0.560 0.630 0.700 0.770


32

The hypothesis is rejected that implies that squeeze pressure affects the Log standard deviation.

Fig.19 Box plot of squeeze pressure for Log standard deviation for UT values

The box plot, Fig.19 clearly shows that squeeze pressure affects the Log Std with squeeze pressure of 1.8 showing many outliers; this has to be addressed to improve the productivity.

One-way ANOVA: Log STD versus Collet Pressure


Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev -----+---------+---------+---------+----3.0 259 0.5643 0.1469 (-----*-----)3.1 157 0.5918 0.1584 (------*-------)3.2 26 0.6146 0.1556 (------------------*-----------------)3.4 285 0.5814 0.1889 (----*-----) -----+---------+---------+---------+---- 0.560 0.595 0.630 0.665


33

The hypothesis cannot be rejected which implies that collet pressure doesn’t have any significant influence on the Log Standard deviation.

One-way ANOVA: Log STD versus Preheat


Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev +---------+---------+---------+---------4.5 31 0.6017 0.1215 (------*-------)5.0 340 0.5851 0.1583 (-*-)5.5 321 0.5669 0.1708 (-*-)5.6 7 0.5250 0.1373 (---------------*--------------)5.7 28 0.6259 0.2568 (-------*-------) +---------+---------+---------+--------- 0.400 0.480 0.560 0.640


The hypothesis cannot be rejected which implies that pre heat doesn’t have any significant influence on the Log Standard deviation.

One-way ANOVA: Log STD versus Weld heat

Source DF SS MS F PWeldheat 7 0.0959 0.0137 0.49 0.844Error 719 20.2329 0.0281Total 726 20.3288

Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev +---------+---------+---------+---------10.7 7 0.5250 0.1373 (---------------*--------------)11.0 204 0.5739 0.1563 (--*--)11.2 10 0.5897 0.0802 (------------*------------)11.5 165 0.5850 0.1557 (--*--)11.6 177 0.5668 0.1650 (--*--)12.0 12 0.5765 0.1806 (-----------*-----------)12.2 89 0.5965 0.2141 (----*---)12.3 63 0.5911 0.1768 (----*----) +---------+---------+---------+--------- 0.400 0.480 0.560 0.640


The hypothesis cannot be rejected which implies that weld heat doesn’t have any significant influence on the Log Standard deviation.

34

Histogram of UT values of Machine C

The histogram of the UT values from Machine C is as in the Fig.20 below.

Fig.20 Histogram of UT values

Observations:The histogram looks right skewed with long tail.

From the ANOVA it is clear that the influence of the operator, shift, and all parameters of the welding process exists on the UT values. The underlying distribution of the process can only be found out keeping this Operator, shift and all parameters of the process same, KSR has consistent behavior the distribution is found using his data.

35

Fig.21 Box plot of Operator KSR for UT values

The box plot, Fig.20 shows no outliers in the data and the probability plot is made to find the underlying distribution of the data.

Probability Plot

Fig.22 Probability Plot of Log normal distribution for Operator KSR

36

The data fits the log normal distribution as p-value, as shown in Fig.22 is greater than 0.05, so log normal distribution can be assumed as an appropriate distribution for UT and also to estimate process capability and non conformance if any.

Process Capability Analysis for Machine C

Existing process:

Fig.23 Process capability of Machine C (existing process)

37

Process Capability after removing the out of control points

Fig.24 Process capability of Machine C

From the above Fig.24 it can be seen that sigma level of Machine C on UT is 5.60 with a PPM of 0.01 which is very good for any CTQ even in six sigma parlance.

Metallography Test Data Analysis

The MT data is discrete in nature; the possible values are < 90, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, and 200. Since the data is discrete with less resolution it is difficult to do deeper analysis, so it recommended to improve the measuring system to draw more inference more from the data.

However the data has been analyzed using attribute control chart to understand the process capability with respect to non conformance.

P- Chart

The p chart is used to analyze the fraction of non conformities for one year (April’ 11- March ‘12); no non conformities are found so the process is in control.The Control limits for the p-chart, Fig.25 are UCL =0.003970, CL= 0.001322, LCL= 0

38

Fig.25 P chart for MT values Machine C

The p value for the machine C = 0.001322.This p value is used to calculate the existing process capability for attribute characteristics.

Existing Process Capability:

39

From the above Fig.26 it can be seen that PPM of machine C on MT is 1322 which is not good considering the severity of lack of fusion on the weld quality. The causes for lack of fusion have been identified in the FMEA more focus should be on the reasons which result in Lack of fusion.

The same analysis is repeated for the other two machines i.e. Machine D and Machine E

Analysis of Machine D

This machine is used for welding the 37E bundle lots in the period of analysis.

X bar S chart is used to analyze whether process mean is in control and the control limits are set for the process.

X bar S Control chart for Machine D:

Fig.27 X bars S chart for Machine D

Since there are some points out of the control limits the homogenization is required to determine the actual control limits of the process.

After Homogenization, Fig.28 the control limits for the X bar S chart are as below:

X bar chart are UCL = 46.47, CL = 41.07, LCL = 36.68 and

S chart are UCL = 8.910, CL = 4.909, LCL = 0.909

40

Fig.28 Homogenized X bar S chart for Machine D

Main Effect Plot

Fig.29 Main effect plot for Mean UT values Machine D

Observations: The mean UT value of Operator DBU and JB is low compared to other operators. The mean UT value of Third shift is very high compared to the other shifts. The Squeeze pressure was not varied by any operator for this machine.

41

Collet pressure is also not varied by any operator. Change in mean decreased due to change in pre heat. Mean UT is highest for the weld heat of 11.2 kA.

Fig.30 Main effect plot for Standard deviation of UT values Machine D

Observations: The Standard deviation of UT value is highest for the Operator KSR compared to other

operators. The Standard deviation is high for the third shift compared to other shifts. The Standard deviation decreased for by decrease in Preheat and is highest for the weld heat of

11.2 kA.

ANOVA for UT – Machine D


Source DF SS MS F POperator 5 113.5 22.7 0.78 0.564Error 2002 58282.5 29.1Total 2007 58395.9 Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev ------+---------+---------+---------+---DBU 632 40.891 4.671 (----*----)JB 72 40.764 4.900 (---------------*--------------)JSK 240 41.433 5.418 (--------*-------)JSRI 544 41.261 5.192 (-----*----)KSR 432 41.440 6.599 (-----*-----)

42

SNT 88 41.193 5.237 (-------------*-------------) ------+---------+---------+---------+--- 40.00 40.80 41.60 42.40


The hypothesis is accepted which implies that the change in the operator has no significant effect on the UT values. The box plot of the UT values with respect to Operator is as shown in Fig.31.

Fig.31 Box plot of Operators for UT values

Operator KSR and JSRI has many outliers compared to the other Operators and SNT has consistent values of UT with no outliers.


Source DF SS MS F PShift 3 214.5 71.5 2.46 0.061Error 2004 58181.4 29.0Total 2007 58395.9 Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev ---------+---------+---------+---------+First 728 41.078 5.396 (-------*------)General 640 40.944 5.025 (-------*-------)Second 312 41.160 5.245 (-----------*-----------)Third 328 41.902 6.138 (-----------*-----------) ---------+---------+---------+---------+ 41.00 41.50 42.00 42.50


43

The hypothesis is rejected which implies there exists significant difference between the shifts. So there exists significant effect of shifts on UT values on Machine D.

Box plot showing UT values for each operator is as shown in Fig.32.

Fig.32 Box plot of shifts for UT values

The difference between the shifts exists because of the Outliers mainly in the first and the Third shifts and the performance in the Second shift is very consistent.



Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev -+---------+---------+---------+--------5.4 80 41.938 8.905 (-------------------*-------------------)5.5 1928 41.151 5.199 (---*---) -+---------+---------+---------+-------- 40.80 41.40 42.00 42.60


The hypothesis is accepted which implies that the change in the pre heats used has no significant effect on the UT values.The box plot of the UT values with respect to Preheat is as shown in Fig.33.

44

Fig.33 Box plot of Pre heats for UT values

Although preheat value of 5.5 has more outliers with respect to 5.4 but it is difficult to comment on the performance as the data for 5.4 is very less compared to that of 5.5.



Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev -+---------+---------+---------+--------11.0 424 41.340 5.293 (--------------*--------------)11.2 632 41.484 5.485 (-----------*-----------)12.5 952 40.913 5.370 (---------*---------) -+---------+---------+---------+-------- 40.60 40.95 41.30 41.65


The hypothesis is rejected which implies there exists significant difference between the Weld heats. So there exists significant effect of Weld heats used on UT values on Machine D.

Box plot showing UT values for each weld heat value is as shown in Fig.34.

45

Fig.34 Box plot of weld heat for UT values

One-way ANOVA: LOG StD versus Operator

Source DF SS MS F POperator 5 0.2513 0.0503 3.09 0.010Error 245 3.9869 0.0163Total 250 4.2382 Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev ----+---------+---------+---------+-----DBU 79 0.6578 0.1025 (----*---)JB 9 0.6581 0.0986 (-------------*-------------)JSK 30 0.6639 0.1315 (-------*------)JSRI 68 0.7034 0.1360 (----*----)KSR 54 0.7379 0.1534 (-----*-----)SNT 11 0.7089 0.1007 (-----------*------------) ----+---------+---------+---------+----- 0.600 0.660 0.720 0.780


The hypothesis is rejected which implies there exists significant difference between the operators. So there exists significant effect of shifts on Log Standard deviation on Machine D.

Box plot showing UT values for each operator is as shown in Fig.35.

46

Fig.35 Box plot of Operators for Log Standard deviation for UT values

One-way ANOVA: LOG StD versus Shift

Source DF SS MS F PShift 3 0.0919 0.0306 1.92 0.127Error 247 3.9416 0.0160Total 250 4.0335 Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev -------+---------+---------+---------+--First 91 0.6661 0.1395 (--------*--------)Second 80 0.6918 0.1051 (---------*--------)Second 39 0.7057 0.0932 (------------*-------------)Third 41 0.7172 0.1571 (------------*------------) -------+---------+---------+---------+-- 0.660 0.690 0.720 0.750


The hypothesis is accepted which implies that the change in the shift used has no significant effect on the Log Standard deviation on Machine D.

One-way ANOVA: LOG StD versus Preheat

47

Source DF SS MS F PPreheat 1 0.0065 0.0065 0.38 0.536Error 249 4.2317 0.0170Total 250 4.2382 Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev ---+---------+---------+---------+------5.4 10 0.7154 0.2600 (---------------*---------------)5.5 241 0.6893 0.1229 (--*--) ---+---------+---------+---------+------ 0.650 0.700 0.750 0.800


The hypothesis is accepted which implies that the change in the pre heat used has no significant effect on the Log Standard deviation on Machine D.

One-way ANOVA: LOG StD versus Weld heat


Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev --+---------+---------+---------+-------11.0 53 0.6795 0.1137 (-------------*-------------)11.2 79 0.7059 0.1369 (----------*-----------)12.5 119 0.6849 0.1326 (--------*--------) --+---------+---------+---------+------- 0.650 0.675 0.700 0.725


The hypothesis is accepted which implies that the change in the weld heat used has no significant effect on the Log Standard deviation on Machine D.

Histogram of UT values of Machine D

48

The histogram of the UT values from Machine D is as in the Fig.36 below.

Fig.36 Histogram for UT values

Observations:The histogram looks right skewed with long tail.From the ANOVA it is clear that the influence of the operator, shift, and all parameters of the welding process exists on the UT values. The underlying distribution of the process can only be found out keeping this Operator, shift and all parameters of the process same, SNT has consistent behaviour so distribution is found using his data.

Fig.37 Box plot of Operators SNT for UT values

49

The box plot, Fig.37 shows no outliers in the data and the probability plot is made to find the underlying distribution of the data.

Fig.38 Probability Plot of SNT operator for Machine D

The data fits in the log normal distribution as p-value is close to 0.05, so log normal distribution can be assumed as an appropriate distribution for UT and also to estimate process capability and non conformance if any.

Process Capability Analysis for Machine D

Fig.39 Process Capability of Machine D (existing process)

50

Since the existing process has many out of control points as shown in Fig.39 the process capability has to be calculated by removing the out of control points.

Process Capability after removing the out of control points:

Fig.40 Process Capability of Machine D

From the above, Fig.40 it can be seen that sigma level of machine D on UT is 4.46 with a PPM of 4.08 which is

good for any CTQ under normal conditions but considering the severity of the defects or failure the aim should be to reach six sigma levels.

Metallography Test Data Analysis of Machine D

The MT data is discrete in nature; the possible values are < 90, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, and 200. Since the data is discrete with less resolution it is difficult to do deeper analysis, so it recommended to improve the measuring system to draw more inference more from the data. However the data has been analyzed using attribute control chart to understand the process capability with respect to non conformance

P- Chart

The p chart is used to analyze the fraction of non conformities for one year (April’ 11- March ‘12); no non conformities are found so the process is in control.The Control limits for the p-chart as shown in Fig.41 are:

51

UCL =0.00761, CL= 0.00270, LCL= 0

Fig.41 P chart of MT values for Machine D

The p value for the machine C = 0.00270.This p value is used to calculate the existing process capability.

Existing Process Capability:

52

From the above it can be seen that PPM of machine D on MT is 2700 which is not good considering the severity of lack of fusion on the weld quality. The causes for lack of fusion have been identified in the FMEA more focus should be on these and necessary action should be taken to decrease the PPM.

Analysis of Machine E:

Analysis of UT- Machine E

This machine is used for welding the 19 E bundle lots in the period of analysis.

X bar S chart is used to analyze whether process mean are in control and control limits are set for the process.

X bar S Control chart for Machine E:

Fig.43 X bar S chart for Machine E (existing process)

The initial X bar S chart suggests that there are many out of control points in the existing process and to set the control limits homogenization is required.

The X bar S chart after homogenization is as shown in Fig.44.

The Control limits for the X bar chart are: CL= 40.07, UCL= 46.25, LCL= 33.88

The Control limits for S chart are CL=3.802, UCL= 8.615, LCL=0 as seen in the Fig.44.

Homogenized X bar S Control Chart

53

Fig.44. Homogenized X bar S chart for Machine E

Main Effect Plot

Fig.45 Main effect Plot of Mean UT values of Machine E

Observations: The mean UT value of Operator DBU is very less compared to other operators.

54

The Squeeze pressure was not varied y any operator for this machine. Mean UT is highest at collet pressure of 3.1 kg/cm2. Change in mean due to change in pre heat is very less. Mean showed an upward trend with the increase in weld heat from 14.3 to 14.5.

Fig.46 Main effect Plot of Standard deviation of UT values for Machine E

Observations: Operator DBU has very low standard deviation compared to other operators. Standard deviation is highest at collet pressure of 3.1 kg/cm2. Standard deviation showed an upward trend with increase in pre heat. Standard deviation for all the values of weld heat is nearly same except for 14.5 where there is a

sudden increase in Standard deviation.

ANOVA for Machine E



Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev -------+---------+---------+---------+--BP 460 39.948 4.073 (--*--)

55

DBU 36 37.944 2.164 (---------*---------)J SRI 36 40.889 4.419 (----------*---------)JB 336 40.976 5.119 (--*--)JSRI 68 40.029 3.558 (-------*------)KSR 268 40.030 4.572 (---*---)MLN 136 39.625 4.220 (----*----)RP 48 40.542 3.770 (-------*--------)SAN 84 40.262 7.657 (-----*------) -------+---------+---------+---------+-- 37.5 39.0 40.5 42.0


The hypothesis is rejected which implies there exists significant difference between the Operators. So Operator performance is different on this Machine E.The Box plot , Fig.47 will give us a better idea of the difference in means.

Fig.47 Box Plot of Operators for Machine E

Operators BP, JB, KSR, MLN, and SAN have shown outliers in their UT values and JB has shown non conformities so the behavior of the process with JB as operator should be taken care.


Source DF SS MS F PShift 2 6.2 3.1 0.14 0.867

56

Error 1469 32095.6 21.8Total 1471 32101.8

Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev -------+---------+---------+---------+--First 684 40.114 4.195 (---*----)Second 752 40.246 5.090 (---*---)Third 36 40.167 4.205 (------------------*------------------) -------+---------+---------+---------+-- 39.20 40.00 40.80 41.60


The hypothesis is accepted which implies that the change in the shift has no significant effect on the UT values. The Box plot for different shifts is Fig.48.

Fig.48 Box Plot of Shifts for Machine E

The Second shift has many outliers compared to the other shifts.

One-way ANOVA: UT versus Collet Pressure

Source DF SS MS F PColletPressure 2 42.7 21.3 0.98 0.377

57

Error 1469 32059.2 21.8Total 1471 32101.8

Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev -+---------+---------+---------+--------3.0 828 40.306 5.030 (---*---)3.1 40 40.600 4.749 (-----------------*------------------)3.4 604 39.987 4.124 (----*---) -+---------+---------+---------+-------- 39.20 40.00 40.80 41.60


The hypothesis is accepted which implies that the change in the Collet pressure has no significant effect on the UT values. The Box plot for different collet pressures is in Fig.49.

Fig.49 Box Plot of Operators for Machine E

From the Box plot although it is evident that that collet pressure of 3.0 and 3.4 has many outliers while 3.1 has no outliers but it is not backed by sufficient data.


Source DF SS MS F PPreheat 1 24.4 24.4 1.12 0.290Error 1470 32077.4 21.8

58

Total 1471 32101.8 Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev -----+---------+---------+---------+----5.4 328 39.942 3.930 (---------------*----------------)5.5 1144 40.252 4.863 (--------*--------) -----+---------+---------+---------+---- 39.60 39.90 40.20 40.50


The hypothesis is accepted which implies that the change in the Pre heat has no significant effect on the UT values.

The Box plot for different Pre heat is as shown in Fig.50.

Fig.50 Box Plot of Pre heat for Machine E



Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev --+---------+---------+---------+-------14.0 256 40.305 4.147 (--*-)14.2 928 40.109 4.578 (-*)14.3 228 40.237 5.652 (--*--)

59

14.4 52 40.558 4.118 (------*-----)14.5 8 40.875 4.673 (---------------*----------------) --+---------+---------+---------+------- 38.0 40.0 42.0 44.0


The hypothesis is accepted which implies that the change in the Weld heat has no significant effect on the UT values. The Box plot, Fig.51 below shows how the UT values are distributed for different values of Weld heat.

Fig.51 Box Plot of Weld heat for Machine E

ANOVA of Log Standard deviation:

One-way ANOVA: LOG STD versus Operator


Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev ------+---------+---------+---------+---BP 115 0.5473 0.1905 (-*--)DBU 9 0.3363 0.0531 (-------*--------)J SRI 9 0.6053 0.1639 (-------*--------)

60

JB 84 0.5900 0.2083 (-*--)JSRI 17 0.4731 0.1914 (------*-----)KSR 67 0.5854 0.1954 (--*--)MLN 34 0.5529 0.1863 (----*---)RP 12 0.5793 0.0925 (-------*------)SAN 21 0.5186 0.2860 (-----*----) ------+---------+---------+---------+--- 0.30 0.45 0.60 0.75


The hypothesis cannot be rejected which implies that Operator has significant influence on the Log Standard deviation.The Box plot for different Operators is as shown in Fig.52.

Fig.52 Box Plot of Operators for Standard deviation for Machine E

From the box plot, Fig.52 it is evident that the operators have significant influence on the standard deviations. The operator DBU has low standard deviation compared to the other Operators.

One-way ANOVA: LOG STD versus Shift

Source DF SS MS F PShift 2 0.0071 0.0035 0.09 0.916Error 365 14.7235 0.0403

61

Total 367 14.7306 Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev --------+---------+---------+---------+-First 171 0.5608 0.1825 (---*---)Second 188 0.5525 0.2160 (---*---)Third 9 0.5668 0.2051 (------------------*------------------) --------+---------+---------+---------+- 0.490 0.560 0.630 0.700


The hypothesis is accepted which implies that shift doesn’t have any significant influence on the Log Standard deviation.

One-way ANOVA: LOG STD versus Collet Pressure


Individual 95% CIs For Mean Based on

Pooled StDevLevel N Mean StDev -------+---------+---------+---------+--3.0 207 0.5625 0.2121 (---*---)3.1 10 0.6455 0.1032 (-----------------*-----------------)3.4 151 0.5430 0.1872 (----*---) -------+---------+---------+---------+-- 0.560 0.630 0.700 0.770


The hypothesis is accepted which implies that Collet pressure doesn’t have any significant influence on the Log Standard deviation.

One-way ANOVA: LOG STD versus Preheat


Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev ---+---------+---------+---------+------5.4 82 0.5371 0.1810 (-----------------*----------------)5.5 286 0.5624 0.2055 (--------*--------) ---+---------+---------+---------+------ 0.500 0.525 0.550 0.575

62


The hypothesis is accepted which implies that pre heat doesn’t have any significant influence on the Log Standard deviation.

One-way ANOVA: LOG STD versus Weld heat


Individual 95% CIs For Mean Based on Pooled StDevLevel N Mean StDev ----+---------+---------+---------+-----14.0 64 0.5698 0.1864 (--*--)14.2 232 0.5552 0.2014 (-*-)14.3 57 0.5378 0.2243 (---*--)14.4 13 0.5861 0.1541 (------*------)14.5 2 0.6677 0.0524 (------------------*-----------------) ----+---------+---------+---------+----- 0.45 0.60 0.75 0.90


The hypothesis is accepted which implies that Weld heat doesn’t have any significant influence on the Log Standard deviation.

Histogram of UT Values of Machine E

63

Fig.53 Histogram of UT values for Machine E

Observations:

The histogram in Fig.53 looks right skewed with long tail.From the ANOVA it is clear that the influence of the operator, shift, and all parameters of the welding process exists on the UT values. The underlying distribution of the process can only be found out keeping this Operator, shift and all parameters of the process same, JSRI has consistent behavior the distribution is found using his data.

Fig.54 Probability Plot of Log normal distribution for JSRI

The data fits the log normal distribution as in Fig 54 the p-value is greater than 0.05, so log normal distribution can be assumed for process capability calculations.

64

Process Capability Analysis for Machine EExisting process:

Fig.55 Process capability for existing process

Process Capability after removing the out of control points:

From the above, Fig.56 it can be seen that sigma level of Machine C on UT is 5.55 with a PPM of 0.01 which is very good for any CTQ even in six sigma parlance.

65

Metallography data Analysis of Machine E

The MT data is discrete in nature, the possible values of are < 90, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, and 200. Since the data is discrete it is difficult to much analysis, so it recommended to improve the measuring system to infer more from the data.

The p chart is used to analyze the fraction of non conformities for one year (April’ 11- March ‘12), no non conformities are found so the process is in control.

P- Chart

Fig.57 P-chart MT values for Machine E

`

66

Fig.58 P-chart MT values for in control process Machine E

Process Capability of existing process:

Fig.59 Process capability of MT for Machine E

From the above,Fig.59 it can be seen that PPM of machine E on MT is 5470 which is not good considering the severity of lack of fusion on the weld quality. The causes for lack of fusion have been

67

identified in the FMEA more focus should be on these and necessary action should be taken to decrease the PPM.

Conclusion and Results:

Pareto Analysis: ECW (End cap welding) and DHT (Double Head Turning) stage contribute to around 65% of the defects and UT and upset variation contribute to 85.4 % of the total rejections at ECW.

UT Values: X bar - S chart: The Control limits are established for each machine.

X bar chart control limits are tabulated below:Machine UCL CL LCL

C 44.615 40.456 32.296D 46.47 41.07 36.68E 46.25 40.07 33.88

S chart control limits are tabulated below:Machine UCL CL LCL

C 6.868 3.784 0.700D 8.910 4.909 0.909E 8.615 3.802 0.00

These control limits of X bar and S chart can be used for monitoring the process.

ANOVA of UT Values:

The results of one way ANOVA are summarized in the table below.

Machine Operator ShiftSq.

PressureCollet

PressurePre Heat Weld Heat

C Significant Significant Significant SignificantNot

SignificantSignificant

DNot


Not applicable

Not applicable

Not Significant

Significant

E SignificantNot

SignificantNot

applicableNot

SignificantSignificant Significant

68

ANOVA on Log Standard deviation of UT Values:

The results of one way ANOVA are summarized in the table below.

Machine Operator ShiftSq.

PressureCollet

PressurePre Heat Weld Heat

CNot

SignificantNot


Not Significant

Not Significant

Not Significant

D SignificantNot

SignificantNot

applicableNot

applicableNot

SignificantNot

Significant

E SignificantNot

SignificantNot

applicableNot

SignificantNot

SignificantNot

Significant

The Underlying distribution for each machine is summarized in the table.

MachineDistribution followed

C Log NormalD Log NormalE Lognormal

Process Capability:

UT values

The process capability has been calculated using the Log Normal distribution and are in table below.

Machine Existing Process In Control ProcessZ bench Ppk PPM Z bench Ppk PPM

C 5.02 1.89 0.25 5.60 2.14 0.01D 4.28 1.56 9.32 4.46 1.64 4.08E 5.18 1.96 0.11 5.55 2.13 0.01

From the sigma value and PPM for each of the machine it is clear that process is very much capable of producing weld of customer satisfaction. Although Machine C has very good PPM individually but on comparison with other machines it has higher PPM, the aim should be achieve the six sigma quality for the machine.

69

MT Values

Machine ProcessAverage P PPM Z process

C 0.001322 1322 3.0064D 0.002700 2700 2.7822E 0.005470 5470 2.5466

The Machine C has less PPM while Machine E has the highest PPM.

Recommendations

The defects in ECW and DHT contribute to 65% of defects in the plant and in ECW process the rejections due to UT and Upset variation constitute 85% of total rejections. So primary focus should be on these two defects to reduce the defects in ECW process.

Currently no control charts are being used for the data collected on UT. As control chart help in effective process control, it is therefore recommended to use X bar- S chart to check if the process is in control or not and if it goes out of control then necessary actions be taken to bring it to find out the assignable causes for that.

After every day production using the UT values from the QC data it is recommended to do the process capability analysis using the MINITAB software to monitor the variation in sigma level and PPM for each machine if any major change in the value is noticed then corrective action should be taken to bring back the process to desired levels.

Similarly using the MT values of the process from the QC data it is recommended to do the process capability analysis and monitor the sigma level and PPM and if any undesirable changes are found corrective action should be taken to bring back to the desired levels.

In order to find out the high risk area involved in ECW stage Process FMEA is carried out. The details of the Process FMEA are described in the following section.

70

FMEA (Failure Mode and Effect Analysis)

Introduction:A failure modes and effects analysis (FMEA) is an inductive failure analysis used in product development, systems engineering / Reliability Engineering and operations management for analysis of potential failure modes within a system for classification by the severity and likelihood of the failures. A successful FMEA activity helps a team to identify potential failure modes based on past experience with similar products or processes or based on common sense logic, enabling the team to design those failures out of the system with the minimum of effort and resource expenditure, thereby reducing development time and costs. Because it forces a review of functions and functional requirements, it also serves as a form of design review to erase weakness (related to failure) out of the design. It is widely used in development and manufacturing industries in various phases of the product life cycle and is now increasingly finding use in the service industry. Failure modes are any errors or defects in a process, design, or item, especially those that affect the intended function of the product and or process, and can be potential or actual. Effects analysis refers to studying the consequences of those failures on different system levels.There are four types of FMEA:

1. System FMEA: Used to analyze systems and sub systems in the early concept and design stage. It focuses on potential failure modes between the functions of the system caused by system deficiencies.

2. Design FMEA: Used to analyze products before they are released to manufacturing. . It focuses on potential failure modes caused by design deficiencies.

3. Process FMEA: Used to analyze the manufacturing and assembly processes. It focuses on failure modes caused by process or assembly deficiencies.

4. Service FMEA: Used to analyze services before they reach the customer. It focuses on failure modes (tasks, errors, mistakes) caused by system or process deficiencies.

Process FMEA is done for the End Cap welding process to identify the major failure mode of the existing process.

Developing Process FMEA for End Cap Welding

In Order to have a proactive approach in preventing the potential failures occurring by understanding the potential causes and implementing the remedial measures that are appropriate to the risk is the essential feature of the methodology known as Process FMEA or popularly known as Process Failure Mode and Effect Analysis.

In this study, it is felt appropriate to make such a process FMEA to help the plant executives in adopting a proactive approach. The Process FMEA made for the End Cap Welding Process is given as a table below. As it is seen that the setting of the process parameters done once before the production (during the Setup only), this aspect is not considered as a sub stage of the process in this FMEA.

71

http://en.wikipedia.org/wiki/Design_review

http://en.wikipedia.org/wiki/Operations_management

http://en.wikipedia.org/wiki/Reliability_Engineering

http://en.wikipedia.org/wiki/Systems_engineering

http://en.wikipedia.org/wiki/Product_development

http://en.wikipedia.org/wiki/Product_development

It is essential to implement the required measures to prevent the occurrence of possible weld failures and update the Table periodically in due course of time based upon the RPN No (Risk Priority No).

PROCESS OF CONDUCTING FMEA:

1. Selecting Team and Brainstorming Sessions: A Team comprising of Officers, supervisors and Operators was made and Brainstorming sessions were conducted.

2. Prepare functional block diagram and/or process flow chart: Initially the whole ECW process was divided into many sub process and a process flow chart is prepared having inputs, process and output at each sub process.

3. Identification of Failure modes for each sub process and then possible Effects, possible causes and their current detection methods if any is done.

4. Fixing of scale of 1-10 to rate severity, occurrence and detectability for possible Effects, possible causes and their current detection methods.

5. Calculating RPN by RPN = SEV x OCC x DET

Process flow for End Cap welding Process:The End Cap Welding process is divided in to 14 sub processes as:

1. Tube Clamping

2. Tube Protrusion

3. Cap Holding

4. Head forward

5. Advance Helium

6. Helium purging

7. Chamber formation

8. Squeeze pressure application and butting

9. Pre heat

10. Cool time between Pre heat and Post heat

11. Post heat/ Weld heat

12. Hold time

13. Tube retraction.

The total process flow with Inputs, Process description, Outputs and also measuring opportunities at each sub process are identified and attached in the Annexure.

72

After finalizing the boundaries for each sub process, brainstorming sessions was conducted for each sub process to identify the failure modes, it’s possible causes and effects. The criteria for rating severity, occurrence, detection of possible effects, possible causes and current detection method respectively.

The criteria for Severity: Rating Criteria

10 Catastrophic failure and effects reactor function9 Leads to rejection of element8 Damage to the machine7 Cause Extreme Customer Dissatisfaction6 Delay in Production5 Cause A Loss Of Performance Which Is Likely To Result In A Complaint4 Cause Minor Performance Loss

3Cause A Minor Nuisance But Can Be Overcome With No Performance Loss

2 Be Unnoticed And Have Only Minor Effect On Performance1 Be Unnoticed And Not Affect The Performance

The rating criteria for Occurrence:

Rating Criteria10 More Than Once Per Day9 Once Every 3-4 Days8 Once Per Week7 Once Per Month6 Once for every 3 to 4 months5 Once Per Year4 Once Every 1-3 Years3 Once Every 3-6 Years2 Once Every 6-10 Years1 More than 10 years

The rating criteria for Detection:

Rating Criteria10 Defect Caused By Failure Is Not Detectable9 Very low

8,7 Low6,5 Moderate4,3 High2,1 very High

After giving rating to the Severity, Occurrence and Detection for all effects causes and current process control the RPN is calculated as:

73

RPN = SEV x OCC x DETThe FMEA table after giving rating and RPN is as annexed at the end of the report.

ANALYSIS:

Pareto Analysis

Pareto Analysis is done for each failure mode and is as below:The RPN of each failure mode is calculated and Pareto analysis is done to determine the vital few failure modes.

Process step Failure Mode RPN

Tube Clamping Insufficient Clamping Force 2502Excess clamping force 477Non uniform Clamping Force 1992Misaligned tube and collet axis 1440

Tube Protrusion Less protrusion 2133More protrusion 1296No Protrusion 702

Cap Holding Cap not Held properly 2793Cap not reaching small electrode 180Cap sided 324

Advance Helium Less or No Helium 1792Head Forward Head not moving 174

Head not reaching chamber 570Helium Purging Less or No Helium 1792Chamber formation Incomplete Chamber formation 928

Leak from the chamber 2128

Butting and Squeeze application Inadequate Squeeze pressure 1917

Excess Squeeze pressure 432No Squeeze Pressure 150No butting 204

Current Non uniform current flow 1410Less amount of current flow 3781More amount of current flow 2584No current flow through components

1088

Weld Head Retraction No Retraction 810

The Pareto chart, Fig.60 suggests that the following failure modes contribute nearly 80% of RPN more focus should be on these failure modes.

74

Less amount of current flow Less or No Helium Cap not Held properly More amount of current flow Insufficient Clamping Force Less protrusion Leak from the chamber Non uniform Clamping Force Inadequate Squeeze pressure Misaligned tube and collet axis Non uniform current flow More protrusion

Fig.60 Pareto Analysis of Failure modes

CONCLUSION:

The remedial action should be taken for failure modes with high RPN. The FMEA table can be used as a proactive process control tool for risk assessment. This table will be a learning document to the newly joined employees. It can also be used in trouble shooting. This table should be periodically looked after every modification in the process as the ratings of

severity, occurrence and detectablity change and the new RPN has to be calculated.

75

Annexure

76

Process Flow of ECW

77

FMEA table

78

Documents

Final Report