You are on page 1of 8

This article was downloaded by:[University at Buffalo (SUNY)] On: 15 February 2008 Access Details: [subscription number 784375719]

Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Quality Engineering
Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t713597292

Improving Process Capability of Manufacturing Process by Application of Statistical Techniques


E. V. Gijo a a SQC and OR Unit, Indian Statistical Institute, Bangalore, India

Online Publication Date: 01 April 2005 To cite this Article: Gijo, E. V. (2005) 'Improving Process Capability of Manufacturing Process by Application of Statistical Techniques', Quality Engineering, 17:2, 309 315 To link to this article: DOI: 10.1081/QEN-200056494 URL: http://dx.doi.org/10.1081/QEN-200056494

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Downloaded By: [University at Buffalo (SUNY)] At: 11:28 15 February 2008

Quality Engineering, 17:309315, 2005 Copyright # Taylor & Francis Inc. ISSN: 0898-2112 print=1532-4222 online DOI: 10.1081/QEN-200056494

Improving Process Capability of Manufacturing Process by Application of Statistical Techniques


E. V. Gijo
SQC and OR Unit, Indian Statistical Institute, Bangalore, India

This article describes the application of statistical techniques in solving a problem of high rejection and rework due to variation in the machining process. The rejections in the shop floor of a company were studied through Pareto analysis for prioritizing the problems due to different sources. The grinding process was found to be the most contributing among all the processes towards rejection and rework. The process capability of this grinding process was evaluated and found to be very low. A step by step method was adopted by the application of several statistical techniques such as design of experiments for solving this problem. As a result of the study, the process capability has improved drastically. This has lead to reduction in the problems due to high rejection and rework in shop floor. Keywords Process capability; Pareto analysis; Gauge repeatability and reproducibility; Design of experiments; Orthogonal array; Analysis of variance, Signal to noise ratio; Optimum combination.

From the Pareto analysis, it is clear that almost 80 percent of the rejections are contributed by grinding and turning processes with grinding process alone contributing almost 60% of rejection. Based on this inference it was decided to undertake a study on grinding process capability.

INITIAL DATA COLLECTION To understand the current status of process capability of the grinding process, a sample of 40 components was machined in the grinding machine and the data were collected. The specification limits of this component were 18.98318.993 mm. The data thus collected are given in Table 1. The process performance indices, Pp and Ppk, were calculated from this data. The values of Pp and Ppk were found to be 0.62 and 0.49. This shows that the process is highly incapable of meeting the specified requirements. Hence, it was decided to conduct a detailed study for improving the process capability of the grinding process.

INTRODUCTION An engineering organization was faced with a problem of high rejection in one of its machining shops. The company management was not in a position to identify where exactly the rejections were originating and the root causes of the rejection related problems. In this context, a study was undertaken to identify the major causes related to rejection in the shop floor and taking remedial actions to eliminate the detected causes. The data regarding rejection were collected from the past records and a Pareto analysis was carried out for identifying the main causes of rejection. Figure 1 represents the Pareto diagram of the collected data.
Address correspondence to E. V. Gijo, SQC and OR Unit, Indian Statistical Institute, 8th Mile Mysore Road, Bangalore 560 059, India; E-mail: gijo@isibang.ac.in.

CAUSE AND EFFECT ANALYSIS A brainstorming session was carried out involving all the concerned people. The output of the brainstorming was put in the form of the cause and effect diagram given in Fig. 2. All the causes listed in the cause and effect diagram were physically verified and the causes, which were found not to have an effect on low process capability, were removed. For all other causes the following actions were initiated: . Gauge repeatability and reproducibility studies were conducted for the measuring instrument (dial comparator), and it was found that the gauge repeatability

309

Downloaded By: [University at Buffalo (SUNY)] At: 11:28 15 February 2008

310

Gijo

Figure 1.

Pareto chart for rejection.

and reproducibility percentage was less than 10. Therefore, it was decided to continue with the existing system of measurements. . The entire machine was dismantled and the tailstock and the headstock were perfectly brought on the center of alignment line. Previously, the different operating parameters of the machine were fixed by the trial and error method. No scientific method has yet been adopted to establish

the optimum level of operating process=machine parameters. Hence, it was decided to carry out a designed experiment for optimizing the process=machine parameters.

PLAN AND DESIGN OF THE EXPERIMENT After having a detailed study on the machine parameters, it was decided to select four parameters

Table 1 Data on grinding process Part no. 1 2 3 4 5 6 7 8 9 10 Dia. (mm) 18.989 18.992 18.991 18.986 18.989 18.989 18.991 18.99 18.985 18.989 Part no. 11 12 13 14 15 16 17 18 19 20 Dia. (mm) 18.987 18.99 18.992 18.984 18.982 18.993 18.993 18.986 18.987 18.987 Part no. 21 22 23 24 25 26 27 28 29 30 Dia. (mm) 18.992 18.989 18.988 18.989 18.988 18.993 18.995 18.989 18.988 18.99 Part no. 31 32 33 34 35 36 37 38 39 40 Dia. (mm) 18.994 18.989 18.987 18.988 18.989 18.989 18.99 18.989 18.99 18.99

Downloaded By: [University at Buffalo (SUNY)] At: 11:28 15 February 2008

Improving Process Capability of Manufacturing Process

311

Figure 2.

Cause and effect diagram.

Table 2 Factors and levels for experimentation Factors Feed rate Wheel speed Work speed Wheel grade

Code A B C D Div 1 (0.0008) 2200 250 A54

Levels Div 2 (0.0010) 2450 360 A60

Unit Mm=Rev RPM RPM

Existing levels.

(factors) for experimentation. The selected factors were Feed Rate, Wheel Speed, Work Speed, and Wheel Grade. Two levels each were identified for these selected factors, with the current operating level one among them. The different factors and their respective levels identified are given in Table 2. In addition to the main effects, interactions between Feed Rate and Wheel Speed (A B) and Feed Rate and Work Speed (A C) were also considered. Experimentation with four factors each at two levels and two interactions was conducted with the help of orthogonal array L8, (Phadke, 1989) in eight experiments. The factors and interactions are allocated to the orthogonal array as shown in the linear graph given in Fig. 3. In the linear graph, the nodes represent factors, and the line joining between two nodes represents the interaction between the factors.

Figure 3.

Linear graph.

The response variable of the experiment was selected to be variation of the outer diameter of the machined components. The physical layout for experimentation was prepared by taking the assigned columns and factors from the linear graph. The physical layout prepared is given in Table 3.

Downloaded By: [University at Buffalo (SUNY)] At: 11:28 15 February 2008

312

Gijo Table 3 Physical layout for experimentation Feed rate (A) (mm=Rev) (1) Div Div Div Div Div Div Div Div 1 1 1 1 2 2 2 2 Wheel speed (B) (RPM) (2) 2200 2200 2450 2450 2200 2200 2450 2450 Work speed (C) (RPM) (4) 250 360 250 360 250 360 250 360 Wheel grade (D) (6) A54 A60 A60 A54 A54 A60 A60 A54

Exp. no. 1 2 3 4 5 6 7 8

Table 4 Data of the experiment Feed rate (mm=Rev) Div Div Div Div Div Div Div Div 1 1 1 1 2 2 2 2 Wheel speed (RPM) 2200 2200 2450 2450 2200 2200 2450 2450 Work speed (RPM) 250 360 250 360 250 360 250 360 Response (mm) Wheel grade A54 A60 A60 A54 A54 A60 A60 A54 1 18.991 18.984 18.985 18.987 18.991 18.979 18.996 18.981 2 18.988 18.988 18.990 18.985 18.996 18.986 18.989 18.989

Trial no. 1 2 3 4 5 6 7 8

EXPERIMENTATION AND DATA COLLECTION The experiment was carried out for all the eight combinations given in Table 3 in a random order. The data were collected for all the eight combinations replicated two times. The data thus collected are given in Table 4.
Exp. no. 1 2 3 4 5 6 7 8

Table 5 10 log(s2) values S=N value 53.4692 50.9704 49.0298 56.9910 49.0298 46.1079 46.1079 44.9483

ANALYSIS The data were subjected to analysis of variance (ANOVA) to determine the significant factors, which have an effect on the variation of the outer diameter of the component. Since the variability was the main reason for low process capability, the response metric was taken to be 10 log(s2) (Montgomery, 1991). The response values were calculated for each experimental combination (see Table 5) and ANOVA was performed. The ANOVA table and average response tables [for 10 log(s2)], respectively, are given in Tables 6, 7, and 8.

The ANOVA table shows that factor A (Feed Rate) is significantly affecting the response. Similarly, factor D (Wheel Grade) and interaction AC (Feed Rate Work Speed) has significant effect on the response. From the average response curves of 10 log(s2) (see Fig. 4), it is evident that level one of both factors A and D give better results. Similarly, for interaction AC, level one for A and level two for

Downloaded By: [University at Buffalo (SUNY)] At: 11:28 15 February 2008

Improving Process Capability of Manufacturing Process Table 6 ANOVA table Source of variation A B AB C AC D Error (Pooled) Total DF 1 1 1 1 1 1 4 7 SS 73.60792 0.78144 4.00827 0.23836 11.38576 18.67308 14.48431 118.15107 MS 73.60792 0.78144 4.00827 0.23836 11.38576 18.67308 3.62108 F 20.32761 Rho% 59.24

313

3.14430 5.15677

6.57 12.74

Table 7 Average response table for main effects: [10 log (s2)] Factor Level 1 2 A 52.61510 46.54847 B 49.89433 49.26925 C 49.40917 49.75440 D 51.10958 48.05400

Thus the expected diameter at the optimum combination is 18:9873 18:9878 18:986 18:9873 18:9849 18:9878 18:9885 18:9878 18:9878 18:9896 mm By similar method, the expected value of response [ 10 log (s2)] at the optimum combination was found to be 55.33579. Therefore, the expected standard deviation at the optimum setting was estimated as 0.00171.

Table 8 Average response table for interaction AC C Level A 1 2 1 51.24950 47.56885 2 53.98070 45.52810

CONFIRMATION TRIAL A confirmatory trial was conducted by setting the operating parameters at their respective optimum levels. Forty components were machined under this optimum level. The average and standard deviation for these 40 values (given in Table 9) were found to be 18.9877 mm and 0.00125. Both of these values indicate that the process is now capable of meeting specifications. The process performance indices were calculated for this data and have shown substantial improvement. Comparison of process performance indices before and after the study is given in Table 10.

C were found to give better results. For the insignificant factor B, the optimum level was identified based on operational convenience. Thus the optimal factor level combination arrived at is A1 B1 C2 D1 .

PREDICTION OF THE EXPECTED PERFORMANCE The expected performance for the process average, under the optimal combination (Taguchi, 1988), was estimated as follows: ^opt A1 T A1 C2 A1 C2 T l D 1 T T i is the ith level  is the overall average and A where T average of factor A (i 1; 2).

CONCLUSION Thus, using a systematic approach, the optimal levels of the various factors of the grinding process had been obtained so as to improve the process capability and there by reducing the process rejections. The optimal values of the factor level combination thus arrived at are given in Table 11.

Downloaded By: [University at Buffalo (SUNY)] At: 11:28 15 February 2008

314

Gijo

Figure 4.

Average response curves.

Table 9 Data of confirmatory trial Part no. 1 2 3 4 5 6 7 8 9 10 Dia. (mm) 18.987 18.987 18.988 18.984 18.985 18.985 18.988 18.989 18.988 18.989 Part no. 11 12 13 14 15 16 17 18 19 20 Dia. (mm) 18.989 18.986 18.988 18.987 18.988 18.988 18.989 18.988 18.988 18.990 Part no. 21 22 23 24 25 26 27 28 29 30 Dia. (mm) 18.988 18.989 18.988 18.988 18.986 18.988 18.987 18.989 18.988 18.986 Part no. 31 32 33 34 35 36 37 38 39 40 Dia. (mm) 18.988 18.989 18.987 18.987 18.989 18.987 18.988 18.987 18.988 18.988

Downloaded By: [University at Buffalo (SUNY)] At: 11:28 15 February 2008

Improving Process Capability of Manufacturing Process Table 10 Process performance indices before and after study Process performance index Pp Ppk Before 0.62 0.49 After 1.33 1.24

315

ABOUT THE AUTHOR E. V. Gijo is a specialist in the Statistical Quality Control and Operations Research Unit of Indian Statistical Institute, Bangalore, India. He holds a Masters degree in Statistics from Mahatma Gandhi University, Kottayam, and a Masters degree in Quality, Reliability, and Operations Research from Indian Statistical Institute, Calcutta. He is an active consultant in the fields of Six Sigma, Quality Management, Reliability, Taguchi Methods, and allied topics in a variety of industries. He is a qualified assessor for ISO-9001 and ISO-14001 systems. He also teaches in the academic programs of the Institute.

Table 11 Optimum factor level combination Factors Feed Wheel speed Work speed Wheel grade Code A B C D Levels Div.1 (.0008) 2200 360 A54 Unit Mm=Rev RPM RPM

IMPLEMENTATION After observing the significant improvement in the process during confirmatory trial, the optimum combination was implemented in the process with immediate effect. With the introduction of this optimal operating level, the extent of rejection and rework has been reduced substantially.

REFERENCES
Montgomery, D. C. (1991). Design and Analysis of Experiments, 3rd ed. New York: John Wiley. Phadke, M. S. (1989). Quality Engineering Using Robust Design. Englewood Cliffs, NJ: Prentice Hall. Taguchi, G. (1988). Systems of Experimental Design, Volumes 1 and 2. New York: UNIPUB and American Supplier Institute.

You might also like