CHAPTER-7 CASE STUDY

CHAPTER-7 CASE STUDY ... continuous improvement, leadership, ... Similarly the lower process capability index CPL is defined by the following...

0 downloads 78 Views 1MB Size
CHAPTER-7 CASE STUDY The study was conducted at a manufacturing facility specializing in plastic products. The particular plant is a part of the “parag group” of its parent company, which is made up of three divisions: filter, hygiene/medical, and plastics. This company‟s vision, which is to be innovative and quality-driven in its technically demanding applications, is determined by its guiding principles of continuous improvement, leadership, and commitment to anticipate and understand their customer‟s needs and expectations. Seventy percent of the plant‟s revenue comes from the plastic moulded parts, it produces for the automotive industry. The proposed DAURR (Diagnose, Analyze, Upgrade, Regulate and Review) principle along with various statistical tools of six sigma was applied to improve the quality of nylon-6 bush (KAMANI BUSH) produced by plastic injection moulding process. The production equipment employed in this study was a precision injection machine, model: PPU7690TV40G, over all dimensions 856×1500×2480 mm manufactured by the Targor Corporation. This machine was installed at Central Institute of Plastic Engineering and Technology (CIPET) Lucknow. 7.1. MODIFIED METHODOLOGY FOR SMALL ENTERPRISES

Having arisen in large corporations, Six Sigma is surely one of the most comprehensive approaches for company development and performance

improvement of products and processes. Nevertheless, it appears that the majority of small and medium sized enterprises (SMEs) either does not know the six sigma approach, or find its organization not suitable to meet their specific requirements. In the SME environment, there is little spare resource; every employee has a key role and usually several [Ryan‟s, 1995, cited by Macadam, 2000]. The challenges of smaller companies are “funding and logistics”, a “limited talent pool”, “multi-hat roles”, and “less exposure to management innovations in other industries”. However the other side of the coin is that it is easier to implement TQM in SMEs because the power of decision making does not depend on extensive hierarchies but lies within the owner managers. Since the SMEs have a much closer proximity to the customers and this proximity is coupled with a larger number of SME employees having direct customer contact and knowledge “therefore, the customer voice can be incorporated within SME operations without prolonged and formalized approaches” [Hale and Cragg, 1996, cited by McAdam, 2000]. Traditional loyalty to specific customers ,support improvement efforts, which are visible to the customers. Six Sigma depends on reliable ways of collecting the voices of the customer and translating these into critical-to-quality requirements of products and services. This close relationship and the high degree of communication with key customers appear to be significant advantages of SMEs in opposition to large corporations. The review of published literature on general QM and the cultural requirements that build the basis for any Six Sigma program, combined with the survey

responses, suggest that several factors have to be represented in a Six Sigma initiative within an SME context. These are (1) The Six Sigma roles should be restricted to the project leaders in the SME organisation (e.g. an “SME black belt”). The rest of the workforce and management staff should only participate in the awareness training. (2) A general six sigma concept for SMEs needs to be adjusted to the core requirements of ISO 9000 to enable a certification, which represents a major difference to Six Sigma programs in large corporations. (3) A training program has to be employed which is significantly shorter than in large corporations, but is still based on the well-proven methods and tools of QM adjusted to specific SME needs. (4) SMEs require consulting services which differ significantly from those usually found in the Market place working for larger corporations. SMEs require consultants and trainers offering modular services, which allow the addition or subtraction of elements without compromising the entirety of the concept and without risking the success for their target group. Keeping in mind the above problems and specific requirements of SMEs ,especially the plastic injection moulding industry, the DMAIC procedure of six sigma can be moulded into a cycle with slight modification.

Fig.7.1 Modified approach in a cycle The road map of this new proposed method is given in the Table 7.1. Based on this modified methodology a case study was carried in a small enterprise consisting of one expert (Black Belt) and few trained personnel. The results of this case study are discussed further in the article which are quite satisfactory. This modified approach will not put a large financial burden on small scale industries having, limited talent pool and less exposure to management innovations. The approach will help them to upgrade their existing system in a slow but steady manner. The experimental work was done at Central Institute of Plastic Engineering and Technology (CIPET), Lucknow.

Table-7.1 Road map for the proposed methodology

THE MODIFIED ROAD MAP FOR SMEs

Phases

Breakthrough

Phase Objectives

Strategy

DIAGNOSE

Define inputs to defining Define and Measure

customer expectations. Measure variability and current capability.

ANALYZE

Analyze

Modeling of the problem

UPGRADE

Improve

Optimize the process to attain “Six Sigma” ,Define capability and control process variation to maintain the desired capability level.

REGULATE/POKAYOKE Control

Transform corporate Culture.

REVIEW

Benchmarking for KAIZEN

Find out further improvement needed to achieve six sigma standard, sort out the areas of

improvement whether Man,Machine,Material or Method, Switch to diagnose stage.

Nylon-6 bush (KAMANI BUSH) produced by plastic injection moulding process has following specifications. Length- 44.4mm, internal diameter-16.1mm, Outer diameter-22mm.

Fig.7.2 Drawing of nylon-6 bush Application of modified methodology at manufacturing facility is described below 7.2 DIAGNOSE The first step in modified procedures is to define and measure the problems so that possible confusion in targets for improvement due to differences in cognition among project staff can be avoided. In this phase, the reasons for rejection and failure of nylon-6 bush were investigated by the management techniques like voice of customer and brain storming of production manager, quality engineer on the shop floor, as well as workers concerned with the production of the above product. After voice of customer and brain storming we

concluded that following four defects were playing important role in rejection and failure of the bush. 1. Sink marks 2. Stress cracking 3.Bulging defect (over shrinkage) 4.Low value of hardness After doing Pareto analysis we decided that bulging defect and low value of hardness are responsible for 80% of the failures and rejection. To measure the problem, the process capability index was ued to illustrate the most efficient way to circumvent the problems defined in the previous steps (bulging defect and hardness in this case). For the nylon-6 bush moulded in this study, the process capability index is measured based upon the data obtained from the bulging and hardness measurements (For the bulging measurement we used a dial gauge attached with a V shaped anvil placed over a surface plate while for the hardness measurement we used core hardness indenter). In general, the upper process capability index CPU is defined by the following equation.

CPU= USL- µ 3σ Where USL is the upper specification limit, µ is the mean of all measured values, and σ is the standard deviation. Similarly the lower process capability index CPL is defined by the following equation. CPL= µ-LSL 3σ

Where LSL is the lower specification limit. The higher value of CPU or CPL is known as Cpk. Before implementation of six sigma projects hundred random samples were taken from production line at ten different occasions in a week and over shrinkage in samples was measured, using dial gauge attached with a V shaped anvil, placed over a surface plate. SPC software was used to analyze the process.

Fig.7.3 Histogram for hundred random samples taken from production line at ten different occasions in a week (over shrinkage measurements shown on xaxis). Value of CPL calculated from hundred random samples for over shrinkage was 0.24 and production was of 2.38 sigma standard and the process mean was 0.1015.Analysis indicates that 19% defects are expected. This is obvious from the Fig.7.3 and Table-7.2.

Similarly SPC analysis for hardness was carried out with the help of hundred samples taken from production line at ten different occasions in a week. The histogram for these measurements is shown in Fig.7.4. Value of CPU calculated from hundred random samples for hardness was 0.56 and production is of 3.19 sigma standard and the process mean was 69.79. This is obvious from the Fig.7.4 and Table-7.3. Table-7.2 SPC software analysis for over shrinkage (bulging defect) in selected samples Cp values Decimal Points 2.00 Cpk 0.24 Number of Entries 100 CpU 0.24 Average 0.1015 Ppk 0.23 Stdev 0.03 Min Max Z Bench

0.05 0.15 0.71

Median Mode Minimum Value

0.1 0.12 0.05

% Defects PPM Expected Sigma

Maximum Value Range USL

0.15 0.1 0.12

PPM
19.0% 190000.00 238713.25 2.38 Observed *

Expected *

Z *

PPM>USL PPM

190000.0 190000.0

238713.2 238713.2

0.71

SPC analysis for hardness was carried out with the help of hundred samples taken from production line at ten different occasions in a week. The histogram for these measurements is shown in Fig. 7.4.

Hardness values

Fig.-7.4 Histogram for hardness of hundred samples (horizontal axis shows hardness, vertical axis shows number of samples) Table-7.3 SPC analysis for hardness in selected samples (for hardness measurements) Cp * Number of Entries 100 Cpk

0.56

Average

69.79

CpU

*

Stdev

2.77

CpL

0.56

Median

70

ZTarget/DZ

1.27

Mode

68

Pp

*

Minimum Value

65

Ppk

0.58

Maximum Value

76

Range

11

PpL

0.58

LSL

65

Skewness

0.19

USL

FALSE

Stdev

2.771773

Number of Bars

10.00

Min

65

Number of Classes

10.00

Max

76

Class Width

1.10

Z Bench

1.69

Beginning Point

63.9

% Defects

0.0%

Stdev Est

2.84

PPM

0.00

d2/c4

0.97

Expected

45568.98

Target

73.31532

Sigma

3.19 Observed

Expected

Z

PPM
0.0

45569.0

-1.69

PPM>USL

*

*

*

PPM

0.0

45569.0

% Defects

0.0

0.0

7.3 ANALYZE In the analysis step, the data collected from the process is analyzed in accordance with the CTQ (critical to quality) factors. The literature review suggests that to have better surface quality of moulded part, it is important to

correctly tune the settings of process parameters in injection moulding ,as well as precision machining of mould.

Fig. 7.5 Cause-and-effect diagram shown in fishbone schematics for the moulded bush. There are many factors that influence the quality of the moulded bush but eight factors are selected here as shown in bold italic type. According to the literature and with the cause-and-effect diagram shown in Fig.7.5, the most significant processing parameters selected were melt temperature(A), injection pressure(B) injection speed(C), mold temperature(D), packing pressure(E), packing time(F) , cooling time(G) and screw speed(H). Using Taguchi method L27 orthogonal experiment was performed setting above eight parameters at three different levels. Table-7.4 Different parameters and their levels Parameter

Symbol Values at different levels

Level-1

Level-2

Level-3

Melting temperature

A

2600C

2750C

2900C

Injection pressure

B

60 M Pa

75 M Pa

90 M Pa

Injection speed

C

40 Cm3/Sec

50 Cm3/Sec

60 Cm3/Sec

Mold temperature D

40 0C

600C

800C

Packing pressure

E

70 M Pa

75 M Pa

80 M Pa

Packing time

F

3 Sec

4 Sec

5 Sec

Cooling time

G

20 Sec

30 Sec

40 Sec

Screw speed

H

50 rpm

65 rpm

80 rpm

S/N (signal to noise) ratio in Taguchi method is helpful in the selection of process parameters at better levels. A large number of different S/N ratios have been defined for a variety of problems, though in this work we have used larger the better characteristic for hardness and smaller the better characteristic for over shrinkage (bulging defect).The method used to calculate S/N ratio for both the types is as follows: 7.3.1 Larger – the – better S/N ratio (η) =−10 log10(1/n∑1/yi2) [ i=1 to n] Where n = number of replications. This is applied for problems where maximization of the quality characteristic of interest is sought. This is referred to as the larger-the-better type problem.

7.3.2 Smaller – the – better S/N ratio (η) =−10 log10(1/n∑ yi2)

[ i=1 to n]

This is termed as smaller-the-better type problem where minimization of the characteristic is intended. Table-7.5 Average value of hardness (for three sample pieces) and S/N(signal to noise) ratio in 27 experiments performed at three different levels of parameters A, B, C, D, E, F, G, and H.

S.No.

A

B

C

D

E

F

G

H

Hardness S/N ratio

1

1

1

1

1

1

1

1

1

75

37.5

2

1

1

1

1

2

2

2

2

77

37.73

3

1

1

1

1

3

3

3

3

76

37.62

4

1

2

2

2

1

1

1

2

74.5

37.44

5

1

2

2

2

2

2

2

3

78

37.84

6

1

2

2

2

3

3

3

1

76

37.62

7

1

3

3

3

1

1

1

3

80

38.06

8

1

3

3

3

2

2

2

1

81.33

38.21

9

1

3

3

3

3

3

3

2

80.5

38.12

10

2

1

2

3

1

2

3

1

80.5

38.12

11

2

1

2

3

2

3

1

2

83.66

38.45

12

2

1

2

3

3

1

2

3

81.66

38.24

13

2

2

3

1

1

2

3

2

81.5

38.22

14

2

2

3

1

2

3

1

3

84.66

38.55

15

2

2

3

1

3

1

2

1

83.66

38.45

16

2

3

1

2

1

2

3

3

79.66

38.02

17

2

3

1

2

2

3

1

1

82.66

38.35

18

2

3

1

2

3

1

2

2

81.5

38.22

19

3

1

3

2

1

3

2

1

82.66

38.35

20

3

1

3

2

2

1

3

2

84

38.49

21

3

1

3

2

3

2

1

3

84.33

38.52

22

3

2

1

3

1

3

2

2

82

38.28

23

3

2

1

3

2

1

3

3

83.66

38.45

24

3

2

1

3

3

2

1

1

81

38.17

25

3

3

2

1

1

3

2

3

78

37.84

26

3

3

2

1

2

1

3

1

80

38.06

27

3

3

2

1

3

2

1

2

79

37.95

Using Taguchi method, L27 orthogonal experiment was performed setting the above eight parameters at three different levels for bulging (over shrinkage).The S/N ratio for the experiment is shown in Table-7.6. Table -7.6 Average value of bulging or over shrinkage (for three sample pieces) and S/N ratio in 27 experiments performed at three different levels of parameters A, B, C, D, E, F, G, and H Avg.

S/N

S. No.

A

B

C D

E

F

G

H

Bulging

1

1

1

1

1

1

1

1

1

0.116667

18.661

2

1

1

1

1

2

2

2

2

0.086667

21.243

3

1

1

1

1

3

3

3

3

0.135

17.39

4

1

2

2

2

1

1

1

2

0.142

16.95

5

1

2

2

2

2

2

2

3

0.083333

21.58

6

1

2

2

2

3

3

3

1

0.136667

17.29

7

1

3

3

3

1

1

1

3

0.1193

18.47

8

1

3

3

3

2

2

2

1

0.072

22.85

9

1

3

3

3

3

3

3

2

0.113333

18.91

10

2

1

2

3

1

2

3

1

0.106667

19.44

11

2

1

2

3

2

3

1

2

0.05

26.02

12

2

1

2

3

3

1

2

3

0.093333

20.59

13

2

2

3

1

1

2

3

2

0.18

14.89

14

2

2

3

1

2

3

1

3

0.106667

19.44

15

2

2

3

1

3

1

2

1

0.246

12.18

16

2

3

1

2

1

2

3

3

0.38

8.4

17

2

3

1

2

2

3

1

1

0.316667

9.9

18

2

3

1

2

3

1

2

2

0.38

8.4

19

3

1

3

2

1

3

2

1

0.362

8.82

20

3

1

3

2

2

1

3

2

0.303333

10.36

21

3

1

3

2

3

2

1

3

0.345

9.24

22

3

2

1

3

1

3

2

2

0.283333

10.95

23

3

2

1

3

2

1

3

3

0.234

12.61

24

3

2

1

3

3

2

1

1

0.28

11.05

25

3

3

2

1

1

3

2

3

0.182

14.79

26

3

3

2

1

2

1

3

1

0.13

17.72

27

3

3

2

1

3

2

1

2

0.175

15.14

Table-7.7 Average values of S/N ratio at different parameter levels S. NO.

Parameter Levels

Avg. S/N ratio

Parameter Levels

(Bulging)

Avg. S/N ratio (Hardness)

1

A1

19.26

A1

37.7933

2

A2

15.47

A2

38.2911

3

A3

12.3

A3

38.2344

4

B1

16.86

B1

38.11333

5

B2

16.66

B2

38.11333

6

B3

14.95

B3

38.09222

7

C1

13.18

C1

38.03778

8

C2

18.84

C2

37.95111

9

C3

15.02

C3

38.33

10

D1

16.83

D1

37.9911

11

D2

12.33

D2

38.0944

12

D3

17.88

D3

38.2333

13

E1

14.59

E1

37.9811

14

E2

17.96

E2

38.2366

15

E3

14.47

E3

38.10111

16

F1

15.11

F1

38.10111

17

F2

15.98

F2

38.08667

18

F3

15.94

F3

38.1311

19

G1

16.09

G1

38.11

20

G2

15.71

G2

38.12889

21

G3

15.22

G3

38.08

22

H1

15.32

H1

38.0922

23

H2

15.87

H2

38.1

24

H3

15.83

H3

38.12667

Fig.7.6 Variation in S/N ratio (vertical axis) because of change in parameter levels (for bulging)

Fig.7.7 Variation in S/N ratio (vertical axis) because of change in parameter levels (for hardness) From Fig.7.6 it is obvious that parameters affecting bulging in order of preference are A, C, D E, B, F, G, and H. The parameters A, C, D and E has

larger impact on the bulging defect therefore these parameters were selected to make regression and neural network models. While Fig.7.7 depicts that parameters affecting hardness in order of preference are A, C, E, D, G, F, H, B. The parameters A, C, E and D have larger impact on the hardness therefore these parameters were selected to make regression and neural network models. Both the properties hardness and bulging are affected the most by the parameters A, C, D and E therefore these will be considered for further analysis. The first order regression equation with the above variables did not give better results; therefore we opted for second order regression analysis. With the help of the above data we formed a second order regression equation for hardness as below HARD = 6.3732839507248 A +0.012305555556039 D -2.0035555555469 C +10.93255555366 E -0.011343209876806 A2 +0.00034861111110563 D2 +0.021377777777698C2 -0.072155555545211 E2 -1182.3028394404 + e ……………………. (1) Where „e‟ is an error term The regression statics for this model is depicted in Table-7.8

Table-7.8 Regression statistics for equation one Multiple Linear Regression - Regression Statistics Multiple R

0.985151

R-squared

0.970523

Adjusted R-squared

0.957422

F-TEST

74.079898

Observations

27

Degrees of Freedom

18

Multiple Linear Regression - Residual Statistics Standard Error

0.606962

Sum Squared Errors

6.631244

Table-7.9 Regression statistics (Analysis of Variance) for equation (1) Multiple Linear Regression - Analysis of Variance ANOVA

DF Sum of Squares

Regression

Mean Square

8

218.329741

27.291218

Residual

18

6.631244

0.368402

Total

26

224.960985

8.6523455840456

F-TEST

74.079898

Table-7.10 Regression statistics (Student Distribution Probability) for equation (1) Student (mathematical equation plotter)

Distribution

T-Test

10.5206

D.F.

18

Probability

Fig.7.8 Comparison between actual hardness values (dotted) and predicted hardness values (from equation one) for twenty seven experimental samples.

Fig.7.9 Residuals plot for twenty seven experimental samples.

Similarly the second order regression equation for bulging was formed as below BULGING = 0.061682428194518 A +0.0364504629528 D -0.10546870372806 C -0.33662925904979 E -0.0001034650209666 A2 -0.00030389351847241 D2 +0.001034425926114 C2 +0.0022465925911414 E2 +5.2475707471113 + e………….. (2) The Regression Statistics of this model is shown in table-7.11. Table-7.11 Regression statistics for bulging model shown in equation (2) Multiple Linear Regression - Regression Statistics Multiple R

0.984221

R-squared

0.968691

Adjusted R-squared

0.954776

F-TEST

69.614348

Observations

27

Degrees of Freedom

18

Multiple Linear Regression - Residual Statistics Standard Error

0.022217

Table-7.12 Analysis of Variance for bulging model shown by equation (2)

Multiple Linear Regression - Analysis of Variance ANOVA Regression

DF Sum of Squares Mean Square 8

0.274894

0.034362

Residual

18

0.008885

0.000494

Total

26

0.283779

0.010914588545104

F-TEST

69.614348

Table-7.13 Student Distribution Probability for bulging model shown by equation (2) Student (mathematical equation plotter)

Distribution

T-Test

2.7817

D.F.

18

Probability

Fig.7.10 Comparison between actual shrinkage values (dotted) and predicted shrinkage values for twenty seven experimental samples.

Fig.7.11 Residuals plot for twenty seven experimental samples. These second order regression models were further validated with the help of neural network models which have been prepared separately for both hardness and bulging predictions. The twenty seven results of Table-7.5 were used for training neural network model. We used neural net work model with following specifications for prediction of hardness value Minimum weight-0.0001 Limit of epochs-10,000 Initial weight-0.3 Learning rate -0.3 Momentum-0.6 Activation function –Log sigmoid function with no neurons in hidden layer

Fig.7.12 Comparison between actual hardness values (red) and predicted hardness values (green) for twenty seven experimental samples. Similarly the twenty seven results of Table-7.6 were used for training neural network model for prediction of bulging. The model had following specifications. Minimum weight-0.0001 Limit of epochs-10,000 Initial weight-0.3 Learning rate -0.3 Momentum-0.6 Activation function – log sigmoid function with no neurons in hidden layer

Fig.7.13 Comparison between actual shrinkage values (red) and predicted shrinkage values (green) for twenty seven experimental samples. From Fig.7.6 and Fig.7.7 it is obvious that factors A, D, C and E has the most significant affect on both the quality characteristics i.e. over shrinkage and hardness. Taguchi design of experiment (Table-7.5 and Table-7.6) indicates that hardness is maximum when parameters are at A2, C3, D1 and E2 levels while over shrinkage is minimum when parameters are at A2, C2, D3 and E2 levels. Since the parameters A and E give optimum value of both the quality characteristics at the same level a compromise was made between parameters C and D with the help of regression and neural net work models prepared with the help of experiment. The predicted values of hardness and over shrinkage (bulging) at different values of parameters A, C, D and E are shown in Table-7.14.After analyzing the predicted values of hardness and over shrinkage, obtained from both the models, we selected parameter A-275, D-80, C-54 and E-75, which shows best compromise between hardness and over shrinkage.

Table-7.14 Predictions made by proposed Regression and ANN models at different values of parameters D and C while parameters A and E are constant. Table-7.14 Prediction made by both regression and ANN models Parameters and their levels Neural network model Second order regression predictions model predictions S.NO.

A

D

C

E

Predicted Hardness

Predicted Predicted Predicted Bulging Hardness Bulging

1

275

45

52

75

83.86272195

2

275

50

54

75

84.28938385

3

275

55

56

75

84.47720709

0.198369 83.1596 0.198795

4

275

60

58

75

84.47592161

0.107416 84.2886 0.23122029

5

275

65

60

75

84.30075696

0.0587 85.60607 0.25672627

6

275

70

58

75

84.02592984

0.052041 84.86487 0.2006633

7

275

75

56

75

83.68894122

0.048561 84.3121 0.1376811

8

275

80

54

75

83.29563224

0.04709 83.9478 0.0677797

9

275

75

52

75

83.70390173

0.056641 83.09114 0.11268397

10

275

65

50

75

84.10593786

0.142654 82.12607 0.17354479

11

275

60

52

75

84.26794599

0.192166 82.22

12

275

55

54

75

84.37414015

0.228232 82.4636 0.182158

13

275

50

56

75

84.43304319

0.229088 82.915

14

275

45

58

75

84.45100007

0.184063 83.55497 0.16309569

15

275

40

60

75

84.43222966

0.106967 84.3833 0.143185

0.207976 81.4

0.11318675

0.2485 82.2190 0.1594505

0.18131140

0.176086

Confirmation experiment was done at the above setting of selected parameters i.e. melting temperature (A)-275oC, mold temperature (D)-800C, injection speed (C)-54 cm3/sec., packing pressure (E)-75 M Pa rest of the parameters like injection pressure (B), packing time (F), cooling time (G) and screw speed (H) were taken at moderate levels of 75 M Pa, 4 second, 30 second and 65 rpm respectively. Nine experimental runs were carried out based on the above parameter settings. The results of these runs are shown in Table 7.15. Table-7.15 Prediction made by both regression and ANN models vs. actual values obtained in experimental run. Parameters and their Neural network model Second order Confirmation levels predictions regression model experiment predictions results S.NO. A

D

C

E

Predicted Hardness

Predicted Predicted Predicted Bulging Hardness Bulging

Actu Actual al shrinkag Hard e ness

1

275

80

54

75

83.29563224

0.04709 83.9478 0.0677797 83

2

275

80

54

75

83.29563224

0.04709 83.9478 0.0677797 82.8 0.065

3

275

80

54

75

83.29563224

0.04709 83.9478 0.0677797 84

4

275

80

54

75

83.29563224

0.04709 83.9478 0.0677797 83.6 0.07

5

275

80

54

75

83.29563224

0.04709 83.9478 0.0677797 84.2 0.075

6

275

80

54

75

83.29563224

0.04709 83.9478 0.0677797 84.2 0.050

7

275

80

54

75

83.29563224

0.04709 83.9478 0.0677797 83.8 0.060

8

275

80

54

75

83.29563224

0.04709 83.9478 0.0677797 85.4 0.052

0.056

0.068

9

275

80

54

75

83.29563224

0.04709 83.9478 0.0677797 84.6 0.062

If we compare the predicted values of both the models as well as actual values obtained in confirmation run (Table-7.15), it becomes clear that both the models are predicting values of both the quality characteristics very close to actual value. To further consolidate results t-test was performed. 7.4 T test: For a normal population of size n with mean µ, variance σ2 and standard deviation s, student‟s t is defined as t = (x-µ)/s/n……………………………………. (3) While the confidence interval of µ is given by X ± t0.005s/n ………………………………………… (4) Where t0.005 is tabulated value of t at 99.5 percent confidence limit Table-7.16 T test for hardness with alternative hypothesis that true mean is greater than 83 at confidence level of 99.5 percent One Sample t-test H0

83

Alternative

greater

CI

0.99

Sample Mean

83.9555555555555

T-Test

3.61804719244034

DF

8

P-Value

0.00340091193327827

Since the p value is smaller than significance level and tabulated value of t (3.355) is less than calculated value (3.61804) null hypothesis that actual mean is 83 is rejected and alternate hypothesis that actual sample mean is more than 83 is accepted. Confidence interval for hardness in the above case can be given as 83.95555 ± 3.355×0.792324288/3 i.e. the hardness value lies between 83.06 and 84.8416 can be predicted at 99.5 percent confidence limit. Table-7.17 T test for over shrinkage with alternative hypothesis that true mean is less than 0.075 and at confidence level of 99.5 percent One Sample t-test H0

0.075

Alternative

less

CI

0.99

Sample Mean

0.062

T-Test

-4.65308989292077

DF

8

P-Value

0.000819043782834587

Since the p value is smaller than significance level and tabulated value of t (3.355) is less than calculated value (4.65308) null hypothesis that actual mean

is .075 is rejected and alternate hypothesis that actual sample mean is less than .075 is accepted. Confidence interval for over shrinkage in the above case can be given as 0.062 ± 3.355× 0.0083815/3 i.e. the over shrinkage value lies between 0.053 and 0.071 can be predicted at 99.5 percent confidence limit 7.5 UPGRADE PHASE First the optimal parameter setting decided in analyze phase such as melting temperature (A)-275oC, mold temperature (D)-800C, injection speed (C)-54 cm3/sec., packing pressure (E)-75 M Pa were employed and rest of the parameters like injection pressure (B), packing time (F), cooling time (G) and screw speed (H) were taken at moderate levels of 75 M Pa, 4 second, 30 second and 65 rpm respectively. The adequate number (nearly 100 to 150) of nylon-6 bushes were produced under the above setting of parameters and all the quality characteristics of bush were measured thoroughly. To measure the improvement in shrinkage values after process improvement, we selected hundred samples from the production line at different times in a week. The process capability index for these samples was calculated. The results of process capability analysis are shown in Table-7.18 and in Fig.7.14.

Fig.7.14 Histogram for hundred samples (over shrinkage measurement) If we compare the histograms for over shrinkage in diagnose phase and upgrade phase (Fig.7.3 and Fig.7.14) as well as process capability analysis in both the phases (Table-7.2 and Table-7.18) we can easily draw following conclusions. (1) Process capability index CPU has increased from 0.24 to 1.225 (2) Process mean has decreased from 0.1015 to 0.0615, which is very much desired. (3) Process has improved from 2.38σ standard to 5.18σ standard.

Table-7.18 Process capability analysis for hundred samples (for over shrinkage measurements) Cp

values

Number of Entries

100

Cpk

1.225

Average

0.06157

CpU

1.225

Stdev

0.011

CpL

*

Median

0.06

Ppk

1.193

Mode

0.06

PpU

1.193

Minimum Value

0.045

PpL

*

Maximum Value

0.09

Min

0.045

Range

0.045

Max

0.09

LSL

FALSE

Z Bench

3.674

USL

0.1

% Defects

0.0%

Number of Bars

10.000

PPM

0.000

Number of Classes

9.000

Expected

119.555

d2/c4

0.987

Sigma

5.170

Target

0.067776

Observed

Expected

Z

PPM
*

*

*

PPM>USL

0.0

119.6

3.674

PPM

0.0

119.6

To measure the improvement in hardness values after process improvement, we selected hundred samples from the production line at different times in a week. The process capability index for these samples was calculated. The results of process capability analysis are shown in Table-7.19 and in Fig.7.15.

Figure-7.15 Histogram for hardness (horizontal axis) If we compare the histograms for hardness in diagnose phase and upgrade phase (Fig.7.4 and Fig.7.15) as well as process capability analysis in both the phases (Table-7.3 and Table-7.19) we can easily draw the following conclusions. (1) Process capability index CPL has increased from 0.56 to 1.16. (2) Process mean has increased from 69.79 to 83.44, which is very much desired.

(3) Process has improved from 3.19σ standard to 4.99σ standard. Table-7.19 Process capability analysis for hundred samples (for hardness) Cp

*

Number Entries

of 100

Cpk

1.16

Average

83.44

CpU

*

Stdev

2.36

CpL

1.16

Median

83

ZTarget/DZ

0.57

Mode

83

PpU

*

Minimum Value

80

PpL

1.19

Maximum Value

90

Skewness

0.45

Range

10

Stdev

2.362673

LSL

75

Min

80

USL

FALSE

Max

90

Number of Bars

10.00

Z Bench

3.49

Number Classes

% Defects

0.0%

Class Width

1.00

PPM

0.00

Beginning Point

74

Expected

241.86

Stdev Est

2.42

Sigma

4.99

d2/c4

0.97

Target

82.08802

of 10.00

Observed

Expected

Z

PPM
0.0

241.9

-3.49

PPM>USL

*

*

*

PPM

0.0

241.9

% Defects

0.0

0.0

The above analysis shows improvement in process mean as well as process capability for both the quality characteristics therefore it was decided to control the process parameters at optimal levels as in the upgrade phase. 7.6 REGULATE PHASE (POKA YOKE) In regulate phase of the approach, the improvement reached in upgrade phase is standardized and adopted for production management of the process. The results must be clearly defined in the control plan in order to constantly monitor its process capabilities and retaining the fruitful improvements. The production equipment employed in this study is a precision injection moulding

machine,

model:

PPU7690TV40G,

over

all

dimensions

856×1500×2480 mm manufactured by the Targor Corporation. The machine is equipped with a built-in monitoring system together with a controller for the process parameters during injection moulding. Because of the built-in monitoring system it was not difficult to maintain the process parameters at the optimal levels decided in regulate phase, Poka Yoke was not needed. 7.7 REVIEW PHASE (KAIZEN)

In the review phase we compared the results obtained in upgrade phase with the six sigma standard so that further improvement (KAIZEN) can be done. As obvious in this study, process has been carried out up to 4.99 σ and 5.18σ standard for the two major quality characteristics bulging and hardness respectively, but there is still scope for the improvement. After brain storming with the shop floor workers, engineers and experts, out of 4Ms (Man, Machine, Material and Method) improvement is needed in mould design (Machine) because of following reasons. 1. In built monitoring and control system of the machine lefts no scope for operator (Man) intervention after the process parameters are set. 2. The material used for molding was tested and it was meeting the quality standards. 3. Method has already been improved (process parameters were already optimized), which leaves a little scope for improvement in method. With a vision for improvement in mould design we will switch to diagnose phase. The reasons for rejection and failure of nylon-6 bush will further be investigated. Keeping in mind the voice of customer, critical to quality factors which arise because of poor mould will be analyzed. This cycle (Diagnose, Analyse, Upgrade, Regulate and Review) will be carried out until six sigma standard is reached.