Fundamental Study
For Supplier Quality Improvement
at AP&T Presses AB Sweden
Fritz Alum Yah & Jian Che
The Thesis comprises 15 credits and is a compulsory part in the Master of Science in Industrial Engineering with a major in Quality & Environmental Management,
Fritz Alum Yah
Jian Che
Master thesis
Subject Category: Technology
Series Number: 3/2008
University College of Borås
School of Engineering
SE-501 90 Borås
Telephone +46 033 435 4640
Examiner: Henrik Eriksson, Chalmers University of Technology and
Sahlgrenska University Hospital, Sweden
Supervisors: Bertil Hansson, AP & T Presses AB, Tranemo, Sweden
Roy Andersson, University College of Borås, Sweden
Client: AP&T Presses AB, Tranemo, Sweden
Date: June 5th, 2008
Keywords: supplier quality, histogram, pareto chart, stratification and process
capability study
Abstract
This thesis aims at securing the quality of components from AP&T’s suppliers and to get a picture of different suppliers’ processes. In this thesis, samples supplied by various AP&T suppliers where measured at AP&T to gather data on how well the parts met design
specifications. The collected data were analyzed using histogram and stratification. Further analysis was done using pareto chart and capability indices. Only the critical dimensions identified from the pareto chart together with discussions with experts on the shop floor were further analyzed using capability indices. This was done to focus more on the critical
dimensions which needed more attention. The findings were critically studied and suggestions were made which served as a fundamental base for supplier product quality improvement plan at AP&T. However, there was no direct contact with the supplier process to ascertain its being in control; therefore, product characterization was carried out. In this case, all the results obtained from the analyses only gave a momentary picture of the process. Therefore, the results can not be used for future predictions.
In the analysis, special attention was laid on the shape, spread and centering of the histogram. Studies were also done on the significant few; got from the pareto chart. Use was made of capability indices to get the momentary fall-outs of ppm and the percentage of the
Abstract ... 3
1. Introduction ... 5
1.1 Background ... 5
1.2 The Company ... 5
1.3 Objective ... 5
2. Theoretical Frame of Reference... 6
2.1 Data Collection... 6
2.2 Histogram ... 6
2.3 Normal Distribution ... 6
2.4 Pareto Chart... 7
2.5 Stratification ... 8
2.6 Process Capability Study... 9
2.6.1 Product Characterization ... 9
2.6.2 Capability Indices... 9
2.6.3 Confidence Interval ... 11
2.6.4 Tolerancing... 11
3. Methods ... 12
3.1 Choice of Research Approach... 12
3.2 Data collection Technique... 12
3.3 Application of Process Capability Study ... 12
3.3.1 Application of Histogram... 12
3.3.2 Application of Pareto Chart... 13
3.3.3 Application of Process Capability Indices ... 13
3.4 Validity and Reliability ... 14
4. Empirical Work... 15
4.1 Situation ... 15
4.2 Measurement and Analysis ... 15
4.2.1 Histogram Analysis ... 15
4.2.2 Stratification Analysis ... 37
5. Further Analysis... 41
5.1 Pareto Chart Analysis... 41
5.2 Process Capability Index Analysis... 43
6. Conclusions and Suggestions... 49
Supplier Quality Improvement Chapter 1: Introduction
1. Introduction
1.1 Background
With the development of economy as well as other subsequent aspects, people have come up with higher demands on the product quality from needs to satisfaction. As a matter of fact, quality has been paid much attention all over the world in the past 20 years and it is quite important to implement quality management program in all types of companies. However, supplier quality management is becoming more and more significant in all fields of work, especially in the manufacturing industry. Components or parts from suppliers are some of the inputs of manufacturing processes and they have much influence on the whole fabrication process as well as the output of the process.
Without good quality, the corporation will lose customers, which means there is no business for the company. Gradually, the corporation can not exist in the competitive market for a long time. In order to ensure the quality of products, there is need to analyse and improve supplier quality which includes both supplier process and its output.
1.2 The Company
AP&T is a world-wide company, which designs and manufactures presses, tools and automation machine. Meanwhile, the company has complete production lines for the sheet metal and tube forming industry. They are specialized in forming technology and automated processes. As a supplier, AP&T aims to be a One Responsible Partner.
AP&T has wide customers, including subcontractors in the automotive, domestic appliances and indoor climate businesses, and also an extensive range of products that are useful for many other metal forming companies. At the moment, AP&T has subsidiaries and
representatives in over 20 countries world wide. The company’s strategy is “We will be where our customers are” [1].
1.3 Objective
2. Theoretical Frame of Reference
2.1 Data Collection
One of the most important steps in quality improvement program is data collection. From the very start, in order to ensure sound data collection, the purpose of the data collection must be clear. It is important to know what the quality problem is and what facts are required to elucidate the problem. Until these are known, it is not possible to move on to collecting data [2]. Note that unauthentic data collection not only causes a lack of proper information but also produces problems such as, making erroneous decisions and increasing project costs [3]. A data collection sheet (or check sheet) is designed. In doing this, emphasis is laid on ease of data entry and data analysis. This can be done easily with the use of spread sheets like
Microsoft Excel. It is worth noting that the application of a computerized system connected to the internet has various advantages. For example, data can be easily transferred/accessed to/from various locations, off-site analysis can be easily done and benefits from just-in-time intervention can be reaped. Also, observations from different origin can be easily recorded using different colours [3].
2.2 Histogram
Histogram is made of bars very similar to that of bar charts but for the fact that there are no gaps between the bars. They are only used for quantitative variables that are continuous in nature [4]. It is a graphical form of frequency distribution and is very good in handling large amounts of data. To construct a histogram, the range of the data must be divided into intervals called class intervals, cells or bins. The horizontal axis is used to represent the measurement scale for the data and the vertical axis to represent the counts or frequencies. They are easily interpreted when they class intervals are of equal width. Information deduced from a
histogram is greatly affected by its display; therefore, care must be taken in selecting the number of bins. When too few or too many bins are used, the histogram does not become informative. In practice, what works well is choosing the number of bins approximately equal to the square root of the number of observations. However, computer packages have default setting for number of bins. In any case, small data sets change dramatically in appearance if the number and/or width of bin changes [2] [5].
Note that histogram gives a visual impression of the shape of the distribution of the measurements together with information about the scatter or dispersion of the data [6].
2.3 Normal Distribution
Probability distributions can be divided into 2 general classes: discrete and continuous distributions. Discrete distributions are most often used in situations where the data refer to some sort of count. For example we may want to know the number of items that have a certain characteristic. Continuous distributions (often called density functions), on the other hand, deals with measurement type data. For example measurement of weight and time which could be considered to be on a continuous scale [7].
Supplier Quality Improvement Chapter 2: Theoretical Frame of Reference
Its density function can be written as
( )
(
−)
−∞〈 〈∞ −=
x xe
x
f
2/2 2,2
1
μ σπσ
……… (2-1)To get the standard normal distribution, we transform the normal distribution by setting μ = 0 and σ2 =1 Then we get
( )
/2 22
1
xe
x
f
=
−πσ
[7] ………. (2-2)The shape of this distribution is illustrated below in fig 2.1.
Fig 2-1 the normal distribution (diagram taken from the book Introduction to statistical quality control) [5]
As mention above, the shape of a histogram can be used to determine if a distribution is normal or not. Normality of a data can also be determined from a normal probability plot, or a stem-and-leaf diagram [5] [6].
However in process capability studies normal or non-normal distribution may be encountered. If the distribution is normal, analysis of the data is done directly. In the case of non-normal distribution, the data will have to be transformed to normality using mathematical functions or the quantile based approach proposed by Clements [8]. The objective is to transform the data to normality in order to be able to employ the conventional process capability indices which are based on the normality theory.
2.4 Pareto Chart
In a quality improvement program, several problems are encountered. Generally, only one problem can be solved at a time. When deciding in which order the problem should be addressed, the Pareto chart is of great help [2].
Fig 2-2 Diagram of pareto chart taken from an example in the book Quality from customer needs to customer satisfaction [2].
Each defect type is represented by a rectangle whose height represent the number of defectives on the left-hand scale. Some times, there is a right hand scale which shows the accumulated percentage of defective. It is often drawn with the defect type with the highest frequency placed furthest to the left and the least to the right. Note that it is also possible to draw a pareto chart based on the experienced consequence costs of the different types of defects. Therefore it is not only the total number of defects of complains that determine what step to take [2] [7].
2.5 Stratification
Stratification is a technique used in combination with other data analysis tools. When data from a variety of sources or categories have been lumped together, the meaning of the data can be impossible to see. This technique separates the data so that patterns can be seen [9]. Stratification can be used when data come from several sources, such as shifts, equipment, materials, suppliers or various population groups. Additionally, it is also applied when data analysis may require separating different sources.
2.6 Process Capability Study
“Process capability study is an improvement methodology where a product characteristic is measured and analyzed in order to determine the ability of the process to meet the
specification for the characteristic studied.” (After Deleryd, 1998b).
Wide use of capability studies is partly a result of demands on suppliers from customer
companies and increase use of international standards such as ISO 9000, QS 9000 and ISO/TS 16949, where capability studies are required [2].
Customers usually lay various complains and therefore, in-order to carry out process capability study, it is important that, the most important characteristic be identified. This is followed by, planning the study to know what is to be measured and how to measure what. Data is then collected. Prior to data collection, the process must be in statistical control if the study is to fulfil its purpose. The capability of the process is assessed using tools such as: capability index, histogram, box plots and normal probability plots. Capability study can be seen to comprise of three steps clearly linked to the first three phases in the continuous improvement cycle as presented by Deming [2] [10]. The fourth step, ‘initiate improvement efforts’, is the result of the study. This is illustrated in fig 2-3.
Fig 2-3 The basic steps of process capability studies, Deleryd [10]
2.6.1 Product Characterization
Process capability study usually measures critical-to-quality characteristics on the product and not the process itself. It is a true process capability study when the analyst can directly
observe the process and can control or monitor data collection activity. In this case, inferences can be made about the stability of the process over time. When only sample units of the product are available, say from the supplier, the study is more properly called product
characterization. In such a case where process stability can not be established, result from the study gives only a momentary picture of the process and we can say nothing about the
dynamic behaviour of the process or its state of statistical control [2][5]. 2.6.2 Capability Indices
The process capability ratio Cp simply measures the spread of the specification relative to the
six-sigma spread in the process. It does not take into account where the process mean is located relative to the specifications.
σ 6 LSL USL Cp − = ……… (2-3) Where: USL = upper specification limit
LSL = lower specification limit σ = process standard deviation
Equation 2-3 assumes process has both upper and lower specification limits.
The percentage of the specification band, P, used up the process can also be calculated using the Cp value. 1 ⎟⎟100 ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ = p C P ……… (2-4)
Cpk on the other hand is a more relevant index as it takes process centring into consideration.
It is simply the one-sided PCR for the specification limit nearest to the process average. To characterize process centring satisfactorily, Cpk must be compared to Cp.
⎟ ⎠ ⎞ ⎜ ⎝ ⎛ = − = − = σ μ σ μ 3 , 3 LSL C USL C Min Cpk Pu pl ………. (2-5)
Where μ = process mean
Cpu and Cpl are upper and lower one-sided PCRs respectively for one-sided specifications.
If: Cp = Cpk, the process is centred at the mid-point of the specifications.
Cpk < Cp the process is off centre.
Cpk = 0 the process mean is exactly equal to one of the specification limits.
Cpk < 0 the process mean lies outside the specifications.
Cpk < -1 the entire process lies outside the specification limits.
The value of Cpk relative to Cp is a direct measure of how off-centre the process is operating.
However, a high positive value of PCR is always desired. Cp is usually said to measure
potential capability while Cpk measures actual capability [5] [11].
For any fixed μ in the interval from LSL to USL, Cpk depends inversely on σ and becomes
larger as σ approaches zero. This property can make Cpk unsuitable as a measure of
centring. That is a large Cpk does not tell us any thing about the location of the mean. A much
better indicator of centring is Cpm [5]. Cpm is an index that accounts for characteristics with
pre-established target value [10].
(
)
2 2 6 T LSL USL Cpm − + − = μ σ ………..(2-6)Boyles (1991) noted that both Cpk and Cpm coincide with Cp when μ = T and decreases as μ
moves away from T. Cpmk like Cpk which is developed from Cp, is developed from Cpm
[5][10]. It is defined as
(
)
(
)
2 2 2 3 , 1 T LSL USL Min T C Cpmk pk − + − − = ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ − + = μ σ μ μ σ μ ……….(2-7)The major condition required in process capability study is that the process should be in statistical control.
2.6.3 Confidence Interval
A confidence interval provides a range in which one can claim, with a specified degree of confidence to contain an unknown value. It is usually needed since only estimates ofμ and σ are often available. It expresses only the statistical uncertainty, which is uncertainty resulting from sampling variability. It does not represent uncertainty that results from possible failures in the assumptions upon which the method is based on. Selection of an appropriate confidence level depends on the specific application and the importance of any decision that might be made based on the calculated confidence interval. For example financial decisions on a new project require a high degree of confidence. Calculation of confidence interval for a normal population can be got from Handbook of Statistical Methods for Engineers and Scientist by Harrison M. W. [7].
2.6.4 Tolerancing
Variation can be found in every process. There are no identical parts in reality. Thereby, each dimension must have a tolerance. Tolerance demonstrates how much variation is acceptable in a dimension between minimum and maximum limits. There are various types of tolerances, for instance, bilateral tolerance, unilateral tolerance, limit tolerance, single limit tolerance and title block tolerance [12].
Bilateral Tolerance means that variation is accepted in both directions from a specified dimension.
Unilateral Tolerance means that variation is permitted in only one direction from a specified dimension.
Limit Tolerance shows the maximum and minimum limits of a dimension. Single Limit Tolerance specifies a maximum or minimum limit only.
Title Block Tolerance is interpreted using the title block when dimensions are shown without any tolerances [12].
3. Methods
3.1 Choice of Research Approach
There are four types of research approaches, description, explanation, exploration and
prediction. Description is used to show the facts and characteristics of a specific problem in a well-defined area. Explanation is used to establish causal relationships among a number of variables and show the connections and influences in between. Exploration is used to address issues when the knowledge has some limitations. Prediction is used to make a prognosis of the future development of a phenomenon [13].
In this thesis, explorative research approach has been mainly used; since we are always dealing with data from the measurements and trying to find out (explore) some properties behind these data by means of quality control tools. Meanwhile, we also explain the application of different tools as well as their inter-relationship.
3.2 Data collection Technique
As is known to us all, there are two data collection techniques, qualitative and quantitative. Regarding our task, we did not have history data for parts measurements, all the data were primary data. Quantitative methods were used mostly. Detailed measures were to make use of excel tables for all entries, and to further transfer the data from Excel sheet to a new sheet were analyses were performed using Minitab.
Data were collected from the measurement results directly. A number of numerical data were handled; hence the quantitative methods were applicable in our case. Excel tables were used as is one of the easiest methods to handle large volumes of data.
3.3 Application of Process Capability Study
As mentioned in chapter one, process capability study is an improvement methodology which entails measurement and analysis of data with the ultimate objective of quality improvement applying: histogram, pareto chart and process capability indices just to name some of the tools. Due to the fact that samples were got from the supplier and there was no direct contact with the supplier process to ascertain its being in control, product characterization was performed in this thesis. However, the procedures were essentially that of process capability analysis. The only difference was in the interpretation of results.
3.3.1 Application of Histogram
- The centering of the histogram: was used to analyse the aim of the process.
- The width of the histogram: was used to analyse the process variability about the aim. - The shape of the histogram: to decipher if it is a normal (bell-shaped) distribution or not. In case of non-normal distribution, the pattern was used to shed more light on the variability in the process. For example histograms with two or more peaks reveal that multiple
populations were mixed together. These populations could be different heads on a machine, different spindles, different suppliers different mould cavities and so forth.
Advantage was taken of the histogram as it gives an immediate visual impression of the process performance. Other inferences were also made from it as it some times immediately shows the reason for poor process performance. The illustrations below are good examples by Douglas C. Montgomery. Fig 3-1a shows a process with adequate potential capability but the process target is poorly located. Fig 3-1b shows a process with poor capability resulting from excess variability.
Fig. 3-1 (a) Poor process centering. (b) Excess process variability.
(Diagram from the book Introduction to statistical quality control by Douglas Montgomery)
3.3.2 Application of Pareto Chart
Supplier improvement programs, which consist of broad attempts to tighten-up all procedures rather then identifying and attacking the vital few problems, can fail (J. M. Juran quality control handbook, 1988) [14]. With this in mind, the pareto chart of nonconformities was used to identify the vital few dimensions that contributed most to the nonconformities in a sample. Also, experts in the shop floor were consulted to identify those pertinent dimensions, that though did not bear much nonconformities were critical in the quality of the final assembly. For such dimensions, attention was emphasized.
3.3.3 Application of Process Capability Indices
As mentioned in chapter 2, process capability indices give a quantitative measure of the process capability. However, only the process capability indices of the critical part
dimensions were analyzed. The index Cp was calculated and used to calculate the percentage
of the specification band used up by the process. This gave a picture of the process spread. The index Cpk was also calculated and the results analysed as presented in section 2.5.2.
≤ − − − 1 1 , 2 / 1 2 n X Cp α n Cp 1 1 , 2 / 2 − ≤ − n X Cp α n ………3.1
Where and are the lower α/2 and upper α/2 percentage points of the chi-square distribution with n-1 degrees of freedom [5]. α is the sum of the two areas in the tails of the normal distribution.
1 , 2 / 1 2 − − n X α X2α/2,n−1
3.4 Validity and Reliability
Validity considers the correctness of measurement results which is supposed to be measured. If the results are correct enough, they can be regarded as validate. Reliability is concerned with variability of the measurement results, in other words, the results can not be reliable if they are not stable. In order to increase validity and reliability, we usually use several data sources to answer a question and see if the findings are similar.
In this thesis, data analysis based on a large number of measurement results was carried out. Thus data validity and reliability were considered so that the analyses can be used as a base for decision-making. Measurement results from different data sources were looked at. Stratification, one of the seven quality control tools, was employed to compare the variation from different operators and the variation from different measuring equipments.
4. Empirical Work
4.1 Situation
As mentioned earlier, AP&T deals with short series production whereby in a year about 20 hydraulic presses are produced. AP&T does the designing of the various machine components and contracts out some of the parts to some suppliers. It is the suppliers’ obligation to
manufacture the contracted components to desired specifications. Different suppliers manufacture different components and there is need for an evaluation to be done in order to have an idea of the quality situation of the contracted parts. Samples from 3 different suppliers were analyzed: one from Noros, 3 from Mekator and one from Buttorps Bruk.
4.2 Measurement and Analysis
The empirical work began with application of histogram to perform general analyses of all measured dimensions of parts from all suppliers. The measured dimensions were identified by encircled letters on the drawing part.
4.2.1 Histogram Analysis
The measured dimensions were: a = 8±0.1, b=20±0.2, c=40±0.3, d=Ø40F8, e= Ø90JS8, g=-0.02 as can be seen in the drawing above. Data got can be found in appendix I. Drawing 1b: Plan of Block
The measured dimensions were f´ and f´´. (f´and f´´ = 81.3±0.3, same value but measured from different locations) Data found in appendix I.
2. Histogram Analysis of Measured Dimensions
C 1 Fr e q u e n cy 8 ,1 0 8 ,0 7 8 ,0 4 8 ,0 1 7 ,9 8 7 ,9 5 7 ,9 2 2 0 1 5 1 0 5 0 7 ,9 8 8 ,1 M e a n 8,005 S tD e v 0,01882 N 7 H i s to g r a m o f a N o r m a l 7
Fig 4-1 Histogram of a (part 1)
C2 Fr e q u e n cy 20,2 20,1 20,0 19,9 19,8 19,7 19,6 18 16 14 12 10 8 6 4 2 0 19,8 20 20,2 M ean 19,89 S tD ev 0,1219 N 77 Norm a l H is togr a m of b
Fig 4-2 Histogram of b(part 1)
There are quite some non-conformities in the dimension measurement of ‘b’, but most of the observations deviated from the target value.
C3 Fr e q ue nc y 40,24 40,16 40,08 40,00 39,92 39,84 39,76 35 30 25 20 15 10 5 0 39,7 40 40,3 M ean 40,07 S tD ev 0,04563 N 7 Histogr am of C ´ Norm al 7
Fig 4-3 Histogram of c’ (part 1)
C4 Fr e q u e n cy 40,8 40,6 40,4 40,2 40,0 39,8 50 40 30 20 10 0 39,7 40 40,3 M ean 40,08 S tDev 0,09718 N 7 Histogram of C´´ Normal 7
Fig 4-4 Histogram of c’’ (part 1)
For the dimension measurement of ’c´´’, the results seem much better since all the observations were around the target value except one non-conformity.
Note the mark difference in the shapes of histograms c´ and c´´ which were the same dimension measured from different locations on the block.
C5 Fr e q u e n cy 4 0,0 6 4 0,0 5 40 ,04 4 0,0 3 40 ,02 40 ,01 25 20 15 10 5 0 4 0,0 25 4 0,0 64 M ean 40,03 S tD ev 0,008898 N 7 Norm a l
H is togr a m o f d ( inte r na l dia me te r )
7
Fig 4-5 Histogram of d (part 1)
C6 Fr e q ue nc y 90,03 90,02 90,01 90,00 89,99 89,98 20 15 10 5 0 89,973 90 90,027 Mean 90,00 StDev 0,01114 N 7 Normal
Histogram of e(external diameter)
7
Fig 4-6 Histogram of e (part 1)
For the external diameter, the completed results seemed good because most observations fall within specification limits except several defects.
C7 Fr e q u e n cy 81,60 81,52 81,44 81,36 81,28 81,20 81,12 81,04 25 20 15 10 5 0 81 81,3 81,6 M ean 81,39 S tDev 0,01706 N 7 Normal Histogram of f' (opposite) 7
Fig 4-7 Histogram of f’ (part 1)
C8 Fr e q ue nc y 8 1,6 0 81 ,52 81 ,44 81 ,36 8 1,2 8 8 1,2 0 8 1,1 2 81 ,04 30 25 20 15 10 5 0 81 81 ,3 8 1,6 M ean 81,40 S tD ev 0,01667 N 7 Norm a l H is to gr a m o f f''(a dja c e nt) 7
Fig 4-8 Histogram of f’’ (part 1)
For the dimension measurement of ’f´´’, the results seemed quite similar to the results from ‘f´’. This indicated the same dimension measured from different positions gave similar results. C9 Fr e q u e n cy 0,0225 0,0150 0,0075 0,0000 -0,0075 -0,0150 25 20 15 10 5 0 -0,02 0 0,02 M ean 0,005260 S tD ev 0,006275 N 7 Norm al Histogram of g (tolerance) 7
Fig 4-9 Histogram of g (part 1)
Part 2: Piston (KOLV VL-CYL) Name of Supplier: Mekator Drawing 2: End View of Piston
Dimension measured were a= Ø90 0.1 2 . 0 −
− , b= Ø84e9, c= Ø40h8, d=28±0.1, e=79.5±0.1,
2. Histogram Analysis of Measured Dimensions C1 Fr e q u e n cy 8 9 ,9 6 8 9 ,9 2 8 9 ,8 8 8 9 ,8 4 8 9 ,8 0 8 9 ,7 6 1 4 1 2 1 0 8 6 4 2 0 8 9 ,8 8 9 ,9 M ean 89,85 S tD ev 0,03946 N 4 No rm a l H is to gr a m o f a 2
Fig 4-10 Histogram of a (part 2)
There were quite some non-conformities in the dimension of ‘a’.
C2 Fr e q ue nc y 83,94 83,92 83,90 83,88 83,86 83,84 14 12 10 8 6 4 2 0 83,841 83,928 M ean 83,91 StDev 0,01497 N 4 Normal Histogram of b 2
Fig 4-11 Histogram of b (part 2)
C3 Fr e q ue nc y 40,08 40,06 40,04 40,02 40,00 39,98 39,96 20 15 10 5 0 39,961 40 M ean 39,99 StDev 0,01761 N 4 Normal Histogram of c 2
Fig 4-12 Histogram of c (part 2)
There were several non-conformities in the dimension of ’c’. C4 Fr e q ue nc y 28,60 28,56 28,52 28,48 28,44 28,40 14 12 10 8 6 4 2 0 28,4 28,5 28,6 Mean 28,52 StDev 0,03862 N 4 Normal Histogram of d 2
Fig 4-13 Histogram of d (part 2)
C5 Fr e q ue nc y 79,59 79,56 79,53 79,50 79,47 79,44 79,41 12 10 8 6 4 2 0 79,4 79,5 79,6 Mean 79,46 StDev 0,009468 N 4 Normal Histogram of e 2
Fig 4-14 Histogram of e (part 2)
For the dimension measurement of ’e’, all the observations were acceptable, although they were not so centred on the target value.
C6 Fr e q u e n cy 136,20 136,14 136,08 136,02 135,96 135,90 135,84 12 10 8 6 4 2 0 135,8 136 136,2 Mean 136,0 StDev 0,02212 N 4 Normal Histogram of f 2
Fig 4-15 Histogram of f (part 2)
C7 Fr e q u e n cy 0,03 0,02 0,01 0,00 -0,01 -0,02 -0,03 -0,04 12 10 8 6 4 2 0 0,03 M ean -0,001762 StDev 0,01447 N 4 Normal Histogram of g 2
Fig 4-16 Histogram of g (part 2)
For the dimension measurement of ’g’, the results indicated the degree of concentricity between the two circles.
C8 Fr e q ue nc y 40,3 40,2 40,1 40,0 39,9 39,8 39,7 25 20 15 10 5 0 39,7 40 40,3 Mean 40,00 StDev 0,03020 N 4 Normal Histogram of h 2
Fig 4-17 Histogram of h (part 2)
C9 Fr e q ue nc y 80,6 80,5 80,4 80,3 80,2 80,1 80,0 79,9 12 10 8 6 4 2 0 79,913 80,6 M ean 80,57 StDev 0,01515 N 4 Normal Histogram of i 2
Fig 4-18 Histogram of i (part 2)
Part 3: Internally threaded ring (Mutter med fas Sugventil) Name of Supplier: Mekator Drawing 3: Section View and Plan of Ring
2. Histogram Analysis of Measured Dimensions C1 Fr e q u e n cy 20,20 20,15 20,10 20,05 20,00 19,95 4 3 2 1 0 20,05 20,2 M ean 20,09 S tD ev 0,06282 N 6 Norm al Histogram of a
Fig 4-19 Histogram of a (part 3)
For the dimension measurement of ‘a’, the results did not look good since one non-conformity was found out of five observations.
C2 Fr e q ue nc y 1 6 5 4 3 2 1 0 1 M ean * S tD ev * N 6 Norm a l H is to gr a m o f b
Fig 4-20 Histogram of b (part 3)
Part 4: Piston Suction Valve Name of Supplier: Mekator Drawing 4: End View of Piston Suction Valve
Dimensions measured: a= Ø59.9±0.05, b= Ø55h9, c= Ø50h11, d=6.3 , e=7.5 , f=760 ,
g= Ø27f8, h= Ø36g7. Data found in appendix IV.
2.
Histogram Analysis of Measured Dimensions C 1 Fr e q ue nc y 5 9 ,9 4 5 9 ,9 2 5 9 ,9 0 5 9 ,8 8 5 9 ,8 6 9 8 7 6 5 4 3 2 1 0 5 9 ,8 5 5 9 ,9 5 9 ,9 5 M ea n 59,92 S tD ev 0,009333 N 2 N o r m a l H i s to g r a m o f a 0Fig 4-21 Histogram of a (part 4)
For the dimension measurement of ‘a’, all the observations were within tolerance limit, but not centred enough.
C2 Fr e q ue nc y 5 5,0 6 5 5,0 4 55 ,02 55 ,00 5 4,9 8 5 4,9 6 54 ,94 9 8 7 6 5 4 3 2 1 0 5 4,9 26 55 M ean 55,01 S tD ev 0,01959 N 2 Norm a l H is to gr a m o f b 0
Fig 4-22 Histogram of b (part 4)
C3 Fr e q ue nc y 50,08 50,04 50,00 49,96 49,92 49,88 49,84 12 10 8 6 4 2 0 49,81 50 Mean 50,02 StDev 0,02819 N 2 Normal Histogram of c 0
Fig 4-23 Histogram of c (part 4)
For the dimension measurement of’ c’, almost all the observations were bigger than upper tolerance limits. C4 Fr e q ue nc y 6,48 6,45 6,42 6,39 6,36 6,33 6,30 5 4 3 2 1 0 6,3 6,5 Mean 6,332 StDev 0,01785 N 2 Normal Histogram of d 0
Fig 4-24 Histogram of d (part 4)
C5 Fr e q u e n cy 7,68 7,64 7,60 7,56 7,52 7,48 6 5 4 3 2 1 0 7,5 7,7 Mean 7,506 StDev 0,02114 N 2 Normal Histogram of e 0
Fig 4-25 Histogram of e (part 4)
For the dimension measurement of ‘e’, there are several non-conformities which were lower than minimal limit.
C6 Fr e q u e n cy 76,08 76,04 76,00 75,96 75,92 5 4 3 2 1 0 75,9 76 Mean 75,99 StDev 0,04290 N 2 Normal Histogram of f 0
Fig 4-26 Histogram of f (part 4)
C7 Fr e q ue nc y 26,9850 26,9775 26,9700 26,9625 26,9550 26,9475 4 3 2 1 0 26,947 26,98 Mean 26,98 StDev 0,006126 N 2 Normal Histogram of g 0
Fig 4-27 Histogram of g (part 4)
For the dimension measurement of ‘g’, four observations out of twenty fell out of tolerance limit. C8 Fr e q u e n cy 36,08 36,04 36,00 35,96 9 8 7 6 5 4 3 2 1 0 35,966 35,991 M ean 36,01 S tD ev 0,03460 N 2 Norm a l H is togr a m of h 0
Fig 4-28 Histogram of h (part 4)
Part 5: Valve (Föröppnare Sugventil) Name of Supplier: Buttorps Bruk Drawing 5: Plan of Valve
Dimension measured: a=Ø38±0.05, b= Ø22g7, c=182±0.5, d=320 , e= Ø21.80 Data found
in appendix V.
2 . 0
− −0.2
2. Histogram Analysis of Measured Dimensions
C 1 Fr e q u e n cy 3 8 , 0 4 3 8 , 0 2 3 8 , 0 0 3 7 , 9 8 3 7 , 9 6 1 2 1 0 8 6 4 2 0 3 7 , 9 5 3 8 3 8 , 0 5 M e a n 3 8 ,0 1 S tD e v 0 ,0 1 2 3 4 N 3 N o r m a l H i s t o g r a m o f a 0
Fig 4-29 Histogram of a (part 5)
Supplier Quality Improvement Chapter 4: Empirical Work
C2 Fr e q ue nc y 21,990 21,985 21,980 21,975 21,970 21,965 21,960 14 12 10 8 6 4 2 0 21,972 21,993 M ean 21,98 S tD ev 0,005958 N 3 Norm al Histogram of b 0Fig 4-30 Histogram of b (part 5)
There was just one non-conforming in the dimension measurement ‘b’.
C3 Fr e q ue nc y 182,42 182,28 182,14 182,00 181,86 181,72 181,58 12 10 8 6 4 2 0 181,5 182 182,5 M ean 182,0 S tD ev 0,03558 N 3 Norm al Histogram of c 0
Fig 4-31 Histogram of c (part 5)
Fig 4-32 Histogram of d (part 5) C4 Fr e q u e n cy 32,00 31,96 31,92 31,88 31,84 31,80 12 10 8 6 4 2 0 31,8 32 Mean 31,93 StDev 0,03234 N 3 Normal Histogram of d 0
There was no non-conformity in the dimension measurement’d’, thus the results were good.
C5 Fr e q ue nc y 2 1 ,7 8 2 1 ,7 5 2 1 ,7 2 2 1 ,6 9 2 1 ,6 6 2 1 ,6 3 2 1 ,6 0 7 6 5 4 3 2 1 0 2 1 ,6 2 1 ,8 M ean 21,73 S tD ev 0,02684 N 3 No r m a l H i s to g r a m o f e 0
Fig 4-33 Histogram of e (part 5)
For the dimension measurement of ‘e’, there was no non-conformity.
Supplier Quality Improvement Chapter 4: Empirical Work
4.2.2 Stratification Analysis
1. Stratification of Data Collected By Two Different Operators
In order to have an idea of the variation of the of the measurements taken by 2 different operators, both operators measured dimensions b, d and e of part 5 and the stratification plots were done as can be found below. Collected data is found in appendix VI.
Dimension b M e a s u r e m e n t ( m m ) Fr e q ue nc y 2 1 ,9 9 0 2 1 ,9 8 5 2 1 ,9 8 0 2 1 ,9 7 5 2 1 ,9 7 0 2 1 ,9 6 5 2 1 ,9 6 0 1 4 1 2 1 0 8 6 4 2 0 O p e r a to r I
Fig 4-34 Stratification of operator I (dimension b, part 5)
Dimension d Measurement (mm) Fr e q u e n cy 32,00 31,98 31,96 31,94 31,92 31,90 31,88 31,86 14 12 10 8 6 4 2 0 Operator I
Fig 4-36 Stratification of operator I (dimension d, part 5)
Measurement (mm) Fr e q ue nc y 32,00 31,98 31,96 31,94 31,92 31,90 31,88 31,86 14 12 10 8 6 4 2 0 Operator II
Supplier Quality Improvement Chapter 4: Empirical Work
Dimension e Me a sure me nt (mm) Fr e q ue nc y 21,795 21,780 21,765 21,750 21,735 21,720 21,705 21,690 20 15 10 5 0 O perator IFig 4-38 Stratification of operator I (dimension e, part 5)
Me a s ur e me nt (mm) Fr e q u e n cy 21,795 21,780 21,765 21,750 21,735 21,720 21,705 21,690 20 15 10 5 0 O pe rator II
Fig 4-39 Stratification of operator II (dimension e, part 5)
2. Stratification of Data Collected By Two Different Machines
Likewise, dimension f from part 4 was also measured using different measuring equipments (digital height gauge and digital calliper). Data can be found in appendix VII and the plots below.
Note: machine I = digital calliper; machine II = digital height gauge. Dimension f M e a s u r e m e n t ( m m ) Fr e q u e n cy 7 6 ,0 8 7 6 ,0 4 7 6 ,0 0 7 5 ,9 6 7 5 ,9 2 5 4 3 2 1 0 M a c h i n e I
Fig 4-40 Stratification of machine I (dimension f, part 4)
M e a s u r e m e n t ( m m ) Fr e q u e n cy 7 6 ,0 2 7 6 ,0 0 7 5 ,9 8 7 5 ,9 6 7 5 ,9 4 7 5 ,9 2 7 5 ,9 0 6 5 4 3 2 1 0 M a c h i n e I I
Fig 4-41 Stratification of machine II (dimension f, part 4)
Supplier Quality Improvement Chapter 5: Further Analysis
5. Further Analysis
5.1 Pareto Chart Analysis
(1) Pareto Chart of Square Block
num b e r o f d e fe ct s ac cu m u la te d % o f d e fe ct s C 2 C o u n t 6 , 3 6 , 3 0 , 0 C u m % 6 5 , 6 8 7 , 5 9 3 , 8 1 0 0 , 0 1 0 0 , 0 2 1 7 2 2 0 P e r c e n t 6 5 , 6 2 1 , 9 O th e r g e d b 3 5 3 0 2 5 2 0 1 5 1 0 5 0 1 0 0 8 0 6 0 4 0 2 0 0 P a r e t o C h a r t o f F y r k a n t s t å n g ( s q u a r e b l o c k )
Fig 5-1 Pareto chart of Square Block (part 1)
This chart shows which dimension contributed more to the defects. The results indicated dimension ‘b’ had much more non-conformities, and dimensions ‘a’, ‘c´’, ‘c´´’, ‘f´’ and ‘f´´’ did not have any defect. However, the severity of any defect could only be identified with the help of experts working with this part.
(2) Pareto Chart of Piston
nu m b e r o f de fe ct s A ccu m u la te d % o f d e fe ct s C 2 C o u n t 3 5 ,7 1 4 ,3 1 4 , 3 1 4 , 3 1 4 ,3 7 ,1 0 , 0 C u m % 3 5 ,7 5 0 ,0 5 6 4 , 3 7 8 , 6 9 2 ,9 1 0 0 ,0 1 0 0 , 0 2 2 2 2 1 0 P e r c e n t O th e r d i g c b a 1 4 1 2 1 0 8 6 4 2 0 1 0 0 8 0 6 0 4 0 2 0 0 P a r e t o C h a r t o f P i s t o n
This chart shows which dimensions contributed more to the defects. The results indicated dimension ‘a’ had much more non-conformities, dimensions ‘b’, ‘c´’, ‘g’, ‘i´’ and ‘d´’ came after that. And ‘e’, ‘f’ and ‘h’ did not have any defect. However, the severity of any defect could only be identified with the help of experts working with this part.
(3) Pareto Chart of Internally threaded ring
nu m b e r o f D e fe ct s accu m u la te d % o f d e fe ct s C 1 C o u n t 1 0 P e r c e n t 1 0 0 , 0 0 , 0 C u m % 1 0 0 , 0 1 0 0 , 0 b a 1 , 0 0 , 8 0 , 6 0 , 4 0 , 2 0 , 0 1 0 0 8 0 6 0 4 0 2 0 0 P a r e t o C h a r t o f M u t t e r
Fig 5-3 Pareto chart of Ring (part 3)
This chart shows which dimensions contributed more to the defects. The results indicated dimension ‘a’ contributed all; since we measured two dimensions, and dimension ‘b’ did not have any defect. However, the severity of any defect could only be identified with the help of experts working with this part.
(4) Pareto Chart of Piston Suction Valve
nu m b e r o f de fe ct s A ccu m u la te d % o f d e fe ct s C 2 C o u n t 3 2 , 7 2 1 , 2 2 1 , 2 9 , 6 7 , 7 7 , 7 0 , 0 C u m % 3 2 , 7 5 3 , 8 1 7 7 5 , 0 8 4 , 6 9 2 , 3 1 0 0 , 0 1 0 0 , 0 1 1 1 1 5 4 4 0 P e r c e n t O th e r g e f h b c 5 0 4 0 3 0 2 0 1 0 0 1 0 0 8 0 6 0 4 0 2 0 0 P a r e t o C h a r t o f P i s t o n S u c t i o n V a l v e
Supplier Quality Improvement Chapter 5: Further Analysis
This chart shows which dimensions contributed more to the defects. The results indicated dimension ‘c’ had much more non-conformities, and dimensions ‘a’ and ´d’ did not have any defect. However, the severity of any defect could only be identified with the help of experts working with this part.
(5) Pareto Chart of Valve
nu m b e r o f de fe ct s A ccu m u la te d o f % d e fe ct s C2 Count 1 0 Percent 100,0 0,0 Cum % 100,0 100,0 Other b 1,0 0,8 0,6 0,4 0,2 0,0 100 80 60 40 20 0
Pareto Chart of Föröppnare
Fig 5-5 Pareto chart of Valve (part 5)
This chart shows which dimensions contributed more to the defects. The results indicated dimension ‘b’ had much more non-conformities, and other dimensions did not have any defect. However, the severity of any defect could only be identified with the help of experts working with this part.
5.2 Process Capability Index Analysis
Since the process capability studies were performed on a manufacturing process with many characteristics to be examined, process capability indices were chosen for the further analysis. Capability indices are unitless and take into consideration process location and variance with one-sided or two-sided specifications, with or without a target value for the process mean. Sample mean and sample variance were not used because they could be cumbersome as they are not unitless [15]. Cp, Cpk,Cpl and Cpu were used.
NOROS
Dimensions a, e and g were identified as most critical from the square block supplied by Noros. See drawing 1a. Below is the analysis using Minitab.
8,10 8,07 8,04 8,01 7,98 7,95 7,92 LSL Targ et U SL P rocess D ata S am ple N 77 S tD ev (Within) 0,0188841 S tD ev (O v erall) 0,0188841 LS L 7,9 T arget 8 U S L 8,1 S am ple M ean 8,00506 W ithin O v erall Process Capability of a (using 95,0% confidence)
P otential (Within) C apability
C P U 1,68 C pk 1,68 Low er C L 1,40 U pper C L 1,95 C C pk 1,77 C p O v erall C apability P p 1,77 Low er C L 1,48 U pper C L 2,04 P P L 1,85 P P U 1,77 1,68 P pk 1,68 Low er C L 1,40 U pper C L 1,95 C pm 1,71 Low er C L Low er C L 1,44 1,48 U pper C L 2,04 C P L 1,85
O bserv ed P erform ance P P M < LS L 0,00 P P M > U S L 0,00 P P M T otal 0,00
E xp. Within P erform ance P P M < LS L 0,01 P P M > U S L 0,25 P P M T otal 0,26
E xp. O v erall P erform ance P P M < LS L 0,01 P P M > U S L 0,25 P P M T otal 0,26
Fig 5-6 Process Capability of a (Square Block, part 1)
Note
the values of Cp and Cpk. Also note the confidence interval for Cp is 1, 48≤ Cp≤ 2, 04and that for Cpk is 1, 4≤ Cpk≤ 1, 95. However, the momentary image of this process is that it
produces 0, 26 non-conforming parts in a million. Specific calculations for Cp and Cpk can be
found below, Cp= σ 6 LSL USL− = ) 018822 . 0 ( 6 9 . 7 1 . 8 − =1.77 Cpk=Min ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ − − σ μ σ μ 3 , 3 LSL USL =Min ⎟⎟⎠ ⎞ ⎜⎜ ⎝ ⎛ − − ) 018822 . 0 ( 3 9 . 7 0050649 . 8 , ) 018822 . 0 ( 3 0050649 . 8 1 . 8 =Min ⎟⎟⎠ ⎞ ⎜⎜ ⎝ ⎛ ) 018822 . 0 ( 3 1050649 . 0 , ) 018822 . 0 ( 3 0949351 . 0 =Min (1,681279, 1,8606755)=1.68
9 0 ,0 3 9 0 ,0 2 9 0 ,0 1 9 0 ,0 0 8 9 ,9 9 8 9 ,9 8 LS L Targ et U S L P rocess D a ta S a m ple N 77 S tD e v (W ithin) 0,0111776 S tD e v (O v e ra ll) 0,0111776 LS L 89,973 T a rget 90 U S L 90,027 S a m ple M e an 90,0015 W ith in O v erall Process Capability of e (using 95,0% confidence)
P otentia l (W ithin) C apability
C P U 0,76 C pk 0,76 Low e r C L 0,62 U ppe r C L 0,90 C C pk 0,81 C p O v erall C a pa bility P p 0,81 Low e r C L 0,68 U ppe r C L 0,93 P P L 0,85 P P U 0,81 0,76 P pk 0,76 Low e r C L 0,62 U ppe r C L 0,90 C pm 0,80 Low e r C L Low e r C L 0,67 0,68 U ppe r C L 0,93 C P L 0,85
O bserv ed P e rform ance P P M < LS L 0,00 P P M > U S L 25974,03 P P M T ota l 25974,03
E xp. W ithin P e rform ance P P M < LS L 5327,60 P P M > U S L 11384,73 P P M T ota l 16712,34 E xp. O v e ra ll P e rform ance P P M < LS L 5327,60 P P M > U S L 11384,73 P P M T otal 16712,34
Fig 5-7 Process Capability of e (Square Block, part 1)
Note
the values of Cp and Cpk. Also note the confidence interval for Cp is 0, 68≤Cp≤ 0, 93and that for Cpk is 0, 62≤ Cpk≤ 0, 9. However, although the process looks fairly centered, the
spread of the histogram is too large and there are some few non-conformities. This gives a momentary image of a process that produces 16 712, 34 non-conforming parts in a million. The specification band used by the process is: P= 100 123,46
81 , 0 1 100 1 = ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ = ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ p C 0,025 0,020 0,015 0,010 0,005 0,000 -0,005 Target USL P rocess D ata S ample N 77 S tD ev (Within) 0,0053587 S tD ev (O v erall) 0,0053587 LS L * T arget 0 U S L 0,02 S ample M ean 0,00619481 W ithin O v erall Process Capability of g (using 95,0% confidence)
P otential (Within) C apability
Note
the value of Cpk. Also note the confidence interval for Cpk is 0, 7≤ Cpk≤ 0, 86. In thiscase, there is no lower specification limit as we are trying to get the quality of concentricity of two circles. We however have a momentary image of a process that produces 4994,32 non-conforming parts in a million.
MEKATOR
Dimensions a, c and g were identified as most critical from the piston suction valve supplied by Mekator. See drawing 4. Below is the analysis using Minitab.
59,94 59,92 59,90 59,88 59,86 LSL Target USL P rocess Data Sample N 20 StDev (Within) 0,00945658 StDev (O v erall) 0,00945658 LSL 59,85 Target 59,9 U SL 59,95 Sample Mean 59,9165 Within Overall Process Capability of a (using 95,0% confidence)
Potential (Within) C apability
C P U 1,18 C pk 1,18 Low er C L 0,78 Upper C L 1,58 C C pk 1,76 C p O v erall C apability Pp 1,76 Low er C L 1,21 Upper C L 2,32 PPL 2,34 PPU 1,76 1,18 Ppk 1,18 Low er C L 0,78 Upper C L 1,58 C pm 0,86 Low er C L Low er C L 0,69 1,21 Upper C L 2,32 C P L 2,34 O bserv ed Performance P PM < LSL 0,00 P PM > U SL 0,00 P PM Total 0,00
Exp. Within Performance P PM < LSL 0,00 P PM > U SL 198,17 P PM Total 198,17
Exp. O v erall Performance P PM < LSL 0,00 P PM > U SL 198,17 P PM Total 198,17
Fig 5-9 Process Capability of a (Piston Suction Valve, part 4)
Note
the values of Cp and Cpk. Also note the confidence interval for Cp is 1, 21≤ Cp≤ 2, 32and that for Cpk is 0, 78≤ Cpk≤ 1, 58. However, the momentary image of this process is that it
is producing 198, 17 non-conforming parts in a million.
50,08 50,04 50,00 49,96 49,92 49,88 49,84 LSL Target USL P rocess D ata S am ple N 20 S tD ev (Within) 0,0285596 S tD ev (O v erall) 0,0285596 LS L 49,81 T arget 49,905 U S L 50 S am ple M ean 50,0195 W ithin O v erall Process Capability of c (using 95,0% confidence)
P otential (Within) C apability
C P U -0,23 C pk -0,23 Low er C L * U pper C L * C C pk 1,11 C p O v erall C apability P p 1,11 Low er C L 0,76 U pper C L 1,46 P P L 2,45 P P U 1,11 -0,23 P pk -0,23 Low er C L * U pper C L * C pm 0,26 Low er C L Low er C L 0,23 0,76 U pper C L 1,46 C P L 2,45
O bserv ed P erform ance P P M < LS L 0,00 P P M > U S L 850000,00 P P M T otal 850000,00
E xp. Within P erform ance P P M < LS L 0,00 P P M > U S L 752628,20 P P M T otal 752628,20
E xp. O v erall P erform ance P P M < LS L 0,00 P P M > U S L 752628,20 P P M T otal 752628,20
Fig 5-10 Process Capability of c (Piston Suction Valve, part 4)
In this process, -1<Cpk< 0 which implies the process mean lies outside the specification limits.
The process has a momentary production of nonconformities of 752 628,2ppm. The specification band used by the process is: P= 100 90,09
11 , 1 1 100 1 = ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ = ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ p C 26,9850 26,9775 26,9700 26,9625 26,9550 26,9475 LSL Target U SL P rocess D ata S am ple N 20 S tD ev (Within) 0,00620675 S tD ev (O v erall) 0,00620675 LS L 26,947 T arget 26,9635 U S L 26,98 S am ple M ean 26,9766 W ithin O v erall Process Capability of g (using 95,0% confidence)
P otential (Within) C apability
C P U 0,19 C pk 0,19 Low er C L 0,03 U pper C L 0,34 C C pk 0,89 C p O v erall C apability P p 0,89 Low er C L 0,61 U pper C L 1,17 P P L 1,59 P P U 0,89 0,19 P pk 0,19 Low er C L 0,03 U pper C L 0,34 C pm 0,37 Low er C L Low er C L 0,31 0,61 U pper C L 1,17 C P L 1,59
O bserv ed P erform ance P P M < LS L 0,00 P P M > U S L 200000,00 P P M T otal 200000,00
E xp. Within P erform ance P P M < LS L 0,96 P P M > U S L 289158,02 P P M T otal 289158,99
E xp. O v erall P erform ance P P M < LS L 0,96 P P M > U S L 289158,02 P P M T otal 289158,99
Note
the values of Cp and Cpk. Also note the confidence interval for Cp is 0, 61≤ Cp≤1, 17and that for Cpk is 0, 03≤ Cpk≤ 0, 34. However, the momentary image of this process is that it
is producing 289158, 02 non-conforming parts in a million.
The specification band used by the process is: P= 100 112,359 89 , 0 1 100 1 = ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ = ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ p C BUTTORPS BRUK
Dimension b was identified as most critical from the valve supplied by Buttorps Bruk. See drawing 5. Below is the analysis using Minitab.
21,99 0 21,98 5 21,98 0 21,97 5 21,97 0 21,96 5 21,96 0 LSL Target USL Process Data Sample N 30 StDev (Within) 0,00600967 StDev (O v erall) 0,00600967 LSL 21,972 Target 21,9825 U SL 21,993 Sample Mean 21,9795 Within Overall Process Capability of b (using 95,0% confidence)
Potential (Within) C apability
C PU 0,75 C pk 0,42 Low er C L 0,26 Upper C L 0,58 C C pk 0,58 C p O v erall C apability Pp 0,58 Low er C L 0,43 Upper C L 0,73 PPL 0,42 PPU 0,58 0,75 Ppk 0,42 Low er C L 0,26 Upper C L 0,58 C pm 0,52 Low er C L Low er C L 0,39 0,43 Upper C L 0,73 C PL 0,42 O bserv ed Performance PPM < LSL 33333,33 PPM > U SL 0,00 PPM Total 33333,33
Exp. Within Performance PPM < LSL 105005,33 PPM > U SL 12518,40 PPM Total 117523,72
Exp. O v erall Performance PPM < LSL 105005,33 PPM > U SL 12518,40 PPM Total 117523,72
Fig 5-12 Process Capability of b (Valve, part 5)
Note
the values of Cp and Cpk. Also note the confidence interval for Cp is 0, 43≤ Cp≤0, 73and that for Cpk is 0, 26≤ Cpk≤ 0, 58. However, the momentary image of this process is that it
is producing 117 523, 72 non-conforming parts in a million.
The specification band used by the process is: P= 100 172,414 58 , 0 1 100 1 = ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ = ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ p C
As can be seen from the further analyses, the processes for Noros and Mekator in fig 5-6 and fig 5-9 respectively have quite low process variability as indicated by the percentage of the specification band used by the process. They both have good Cp values. However both
6. Conclusions and Suggestions
6.1 Conclusions
In the manufacturing industry, supplier quality has a direct influence on the fabrication and the quality of final products. Stable supplier process will result in conformable products which provide good materials for customers, like AP&T. Therefore, more and more
companies have started valuing supplier quality improvement by means of looking at process capability indices, such as Cp, Cpk,Cpl and Cpu.
Three important suppliers for AP&T Company: NOROS, MEKATOR and BUTTORPS BRUK were evaluated. From Minitab, a number of data with different parameters were got. A simple comparison by looking at some of these parameters was done. For example, if we check the parameter ‘ppm (parts per million)’, their magnitudes are intolerable and will contribute a substantial percentage of the final cost. This cost usually ends up as a price to pay by the customers.
Normally, supplier capability evaluation consists of two facets: a) qualifying the supplier’s design through the evaluation of production samples. b) qualifying the supplier’s capability to meet quality requirements on production lots [14]. Here, we have focused on the former facet since product characterization has been employed, and a momentary picture of the specific product can be given in the specific time. Thus the results could not be as supporting a base for future prediction.
6.2 Suggestions
After parts measurement and data analysis, we come up with several suggestions for the company we worked with. These suggestions are centred on supplier quality, meanwhile considering all the stakeholders’ needs. From the perspective of R&D departments, it is better for designers to set up wider tolerance limits, provided that the parts function in a good way. Under this circumstance, the purchasing department can have extensive options, and hence wise decision. From the perspective of purchasing department, documented procedures in detail are needed to monitor how suppliers deal with the products before delivery. From the perspective of quality department, the related personnel should work together with purchaser and make an improvement plan to stimulate supplier to take some measures aiming at higher quality. For instance, suppliers can be compelled to conduct process capability study in order to eliminate special causes, achieve a stable process and thereby consolidate high quality in the long run.
List of Reference
[1] AP&T Presses AB, (2007). AP&T is One Responsible Partner [online]. Available: http://www.apt.se/ [accessed May, 2008]
[2] Bo Bergman, Bengt Klefsjö (2003). Quality from Customer Needs to Customer Satisfaction. Second Edition. Studentlitteratur, Lund.
[3] Ming-Kuan Tsai, Jyh-Bin Yan, Chang-Yu Lin (2006). Synchronization-based model for improving on-site data collection performance. Automation in Construction 16 (2007) 323-335.
[4] Michael C. Fleming and Joseph G. Nellis (2000).Principles of Applied Statistics. Second Edition. Thomson Learning, UK.
[5] Douglas C. Montgomery (2005). Introduction to Statistical Quality Control. Fifth Edition. John Wiley & Sons Inc., United States of America.
[6] Douglas C. M., George C. Runger and Norma F. Hubele (2001). Engineering Statistics. Second Edition. John Wiley and Sons Inc., United States of America.
[7] Harrison M. Wadsworth (1998). Hand Book of Statistical Methods for Engineers and Scientist. Second Edition. McGraw-Hill, United States of America.
[8] S.Ahmad, M.Abdollahian, P.Zeephongsekul(2007). Process Capability Analysis for Non-Normal Quality Characteristics Using Gamma Distribution. Proceedings of the International Conference on Information Technology (2007) 425-430
[9] American Society for Quality, (2004).Data Collection and Analysis Tools, Stratification [online]. Available:
http://www.asq.org/learn-about-quality/data-collection-analysis-tools/overview/stratification.html [accessed May, 2008]
[10] Mats Deleryd (1998). A pragmatic view on process capability studies. Int. J. Production Economics 58 (1999)139-330.
[11] W. L. Pearn, P. C. (2004), Testing process performance based on capability index Cpk with criteria values, computers and industrial engineering 47 (2004), 351 - 369
[12] Gary K. GRIFFITH (2000). The Quality Technician’s Handbook. Fourth Edition. Prentice-Hall, United States of America.
[13] University of Gothenburg, (2003). ERP System Effects –A Comparison of Theory and Practice [online]. Available: http://www.handels.gu.se/epc/archive/00002854/01/02-03-58D.pdf (2003) (accessed May, 2008)
[14] J. M. Juran, Frank M. Gryna (1988). Juran’s Quality Control Handbook. Fourth Edition. McGraw-Hill Book Company, USA.
Benämning
Supplier
Ritningsnr.: HV000150
Rev.
Nr. 01-06
Ordernr. N/A
Benämning : Mutter med
fas Sugventil 200-250T
Uppmätt av : Fritz/Carl
Datum:2008-04-16 unit: mm
Nr.
Nominellt värde (a) Uppmätt värde(a) Nominellt värde (b) Uppmätt värde(b)
Machine I 1 75,9-76 76,02 Digital Caliper 2 75,9-76 75,93 3 75,9-76 75,98 4 75,9-76 75,9 5 75,9-76 76,01 6 75,9-76 75,97 7 75,9-76 76,03 8 75,9-76 76 9 75,9-76 75,95 10 75,9-76 76 11 75,9-76 75,99 12 75,9-76 75,98 13 75,9-76 75,98 14 75,9-76 76,08 15 75,9-76 75,99 16 75,9-76 75,96 17 75,9-76 75,94 18 75,9-76 75,99 19 75,9-76 75,98 20 75,9-76 76,07 Machine II 1 75,9-76 76
Digital Height Gage 2 75,9-76 75,926