Vous êtes sur la page 1sur 30

Manufacturing Technology (ME461)

Instructor: Shantanu Bhattacharya

Review of previous lecture


1. 2. 3. 4. 5. 6. 7. Failure mode effect analysis Seven tools of QC Histogram Check sheet Pareto chart Cause and effect diagram Scatter daigram Defect concentration diagram Control chart

Control Chart
A typical control chart is a graphical display of a quality characteristic measured on a sample versus the sample number. There are 3 essential elements of a control chart:

1. Center line representing the average value of the quality characteristic corresponding to the in control state. 2. Upper control limit (UCL) represented by the upper horizontal line of the chart. 3. Lower control limit (LCL) represented by the line below the central line.
The selection of the upper and the lower points also depend on the notion that all the sample points should fall between them if the process is in control. Points lying outside these limits signal the presence of assignable causes. The process should then be investigated to eliminate assignable causes.

A general model of a control chart


Suppose y is a sample statistic that measures some quality characteristic of interest, and let y be the mean y be the standard deviation. The CL, UCL and LCL are given by: UCL = y + ky CL = y LCL= y ky Where k is the distance of the control limits from the center line. If K =3, the cotrol limits are typically called three sigma control limits. Control Charts are categorized into variable control chart and attribute control chart.

Variable Control Chart


If the quality characteristic can be measured and expressed as a continuous variable, then it can be conveniently characterized by measures of the central tendency and variability. Variable control charts can explain the process data in terms of both location (in terms of average) and spread (in terms of piece to piece variability). For this reason the control charts for variables are always developed and analyzed in pairs. The X chart is widely used for controlling central tendency. Charts based on either the sample range (R chart) or the sample standard deviation (S chart) are used to control the process variability. The range chart is relatively efficient for smaller subgroup sample sizes, especially below 8. The S chart is used with larger sample sizes.

The attribute control chart


Many quality characteristics cannot be measured in a continuous scale. This happens when the quality of an item is judged as either conforming or non conforming to a specification.

For example, items which have cracks, missing components, appearance defects, or other visual imperfections may be rendered as rejects, defective or non conforming items.
Control charts to control such quality characteristics are called attribute control charts. The different attribute control charts are the following: 1. The p chart for proportion of units non conforming (from samples not necessarily of constant size). 2. The np chart for number of units non conforming (for samples of constant size). 3. The c chart for number of non conformities or defects (from samples of constant size). 4. The u chart for number of non conformities per unit ( from samples not necessarily of constant size).

The benefits of control charts


Control Charts provide a large number of benefits. The compound effect of all these benefits is an overall improvement in the quality.

Some of the distinct benefits are as follows:


1. Control charts are effective means of monitoring statistical control. 2. Control charts help predict the performance of a process when the process is in a state of statistical control. 3. They provide a common language about the process, such as between 2 or 3 shifts, between production and maintenance supervisors, or between suppliers and producers. 4. Control charts help direct corrective measures in a logical manner by identifying the occurrence of assignable causes. Some of the assignable causes may require special resources and involvement of the management. This helps in avoiding confusion, frustration and high cost of possible misdirected efforts to solve the problem.

Characteristics of Data ( )
The notations are for the average and the standard deviation for a sample of data.

We think as a typical value of the Xs in a sample, or as a point around which the numbers tend to cluster.
The standard deviation s is a typical deviation of an observation x from the average It is a deviation descriptive of variation within the sample. From a well organized frequency distribution as we have already seen above we can say the following: 1. Between

Standard deviation

This formulae simplifies calculations in a frequency table.

Mean and Variance in a Frequency Table

Calculation of Mean and Variance using the frequency method

Checklist necessary for X and R charts


It is helpful to visualize the decisions and calculations that must be made and the actions that need to be taken for plotting a X and a R chart. They are the following: 1. Decisions preparatory to the control charts Some possible objectives of the charts. Choice of variables. Decisions on the basis of subgroups. Decisions on size and frequency of subgroups. Setting up the forms for recording the data. Determining the method of measurement. 2. Starting the Control Charts Making and recording measurements and recording other relevant data. Calculating the average X and range R of each subgroup. Plotting the X and R charts. 3. Determining the trial control limits Decision on required number of subgroups before control limits are calculated. Calculation of trial control limits. Plotting the central lines and limits on the charts. 4. Drawing preliminary conclusions from the charts Indication of control or lack of control. Interpretation of processes in control. Relationships between processes out of control and specification limits.

Decisions preparatory to control charts


Some possible objectives of the charts are the following: 1. To analyze a process with a view to (a) secure information to be used in changing specifications or in determining whether a given process can meet specifications. (b)To secure information to be used in establishing or changing production procedures. (c)To secure information to be used in establishing or changing the inspection procedures or acceptance procedures. 2. To provide a basis for current decisions during production as to when to hunt for causes of variation and take action intended to correct them, and when to leave a process alone. 3. To provide a basis for current decisions on acceptance or rejection of manufactured or purchased product.

Choice of Variables
The variable chosen for control charts for X and R must be something that can be measured and expressed in numbers, such as dimension, hardness number, tensile strength, weight etc. From a standpoint of the possibility of reducing production costs, a candidate for a control chart is any quality characteristics that is causing rejections or rework involving substantial costs. From an inspection standpoint destructive testing always suggests an opportunity to use the control chart to reduce costs.

Decision on the basis of sub-grouping


The key idea in Shewhart method is the division of observations into what Shewhart called rational subgroups. The subgroups should be selected in a way that makes each subgroup as homogeneous as possible and that gives the maximum opportunity for variation from one subgroup to another. As applied to control charts on production, this means that it is of vital importance not to lose track of the order of production. Particularly, if the purpose of the control chart is to keep detecting shifts in the process average, one subgroup should consist of items produced as nearly as possible at one time; the next subgroup should consist of items all produced at a single time later; and so forth. Decision on the size and frequency of subgroups Shewhart suggested 4 as the ideal sub group size. In industrial uses 5 seems to be a better alternative because of ease of calculations. The essential idea of the control chart is to select subgroups in a way that gives minimum opportunity for variation within a subgroup. It is therefore desirable that the size be as small as possible. Subgroups of two or three may often be used to good advantage, particularly where the cost of measurements is so high as to veto the use of larger subgroups. Larger subgroups of 10 or 20 are sometimes appropriate if it is desired to make the control chart insensitive to small changes.

Setting up the forms for recording the data

Layout of data should be as per convenience of calculation and analysis. The forms should have a recording space for item of measurement, unit of measurement, and operator remarks about tool change, operator change, machine change etc.

Starting the Control charts


Making and recording the measurements and recording other relevant data:
The actual work of the control charts start with the first measurements. Any method of measurement will have its own inherent variability; it is important that this is not increased by mistakes in reading measurement instruments or errors in recording data. Calculating the average and range for each subgroup:

Plotting both the charts:


Determining the trial control limits: 1. Decision is needed on the required number of subgroups before control limits are calculated. 2. The fewer the subgroups used the sooner the information thus obtained will provide a basis for action but the less the assurance that the action will be sound.

Calculation of trial control limits:

Drawing Preliminary Conclusions from the chart


Indication of control or lack of control: Lack of control is indicated by points falling outside the control limits on either of the charts. When the points fall outside the control limits, we say that a process is out of control, this is equivalent to saying that assignable causes of variation are present and this is not a constant cause system. In contrast if none of the points fall outside the control limits then No assignable causes are present.
Interpretation of processes in control: With the evidence from the control chart that a process is in control, we are in a position to judge what is necessary to permit the manufacture of product that meets the specifications for the quality characteristic charted. The control chart data gives us estimate of : 1. The centering of the process. 2. The dispersion of the process. Actions based on the relationship between the specifications and the centering and dispersion of a controlled process depend somewhat on whether there are two specification limits, a maximum or upper limit or a minimum or lower limit. This would again depend heavily on the parameter that we choose to plot.

Possible relationships of a process in control to upper and lower specification limits

Case I: The spread of the process 6 is appreciably less than the difference between the specification limits. Case II: The spread is approximately equal to the difference between the specification limits. Case III: The spread of the process is appreciably greater than the difference between the specification limits.

Case I: When the process 6 is appreciably less than difference between specification limits
Frequency curves A,B,C, D and E indicate various positions in which the process can be centered. With any of the position A, B and C practically all the products manufactured will meet the specifications as long as the process stays in control. In general when conditions A, B and C come it represents the ideal manufacturing situation. When the control chart shows that one of this control chart exists, many different possible actions may be considered depending on the relative economy.

For example it may be considered economically advisable to permit X to go out of control if it does not go too far, i.e., the distributions may be allowed to move between positions B and C. This may avoid the cost of frequent machine setup and of delays due to hunting of assignable causes of variation that will not be responsible for unsatisfactory product. Or where acceptance has been based on 100% inspection, it may be economical to substitute acceptance based on control charts. Or where there is an economic advantage to be gained by tightening the specification limits, such action may be considered. With the process in the position D some points will fall above the upper specification limit. Similarly with the process in position E some products will fall below the lower limit. In either case the obvious action is to bring the centering of the process towards that of A.

Case II: When the process 6 is equal to the difference between specification limits In this situation only the process
exactly centered between the specification limits, as in position A, will practically produce everything conforming to the specifications. If the distribution shifts away this exact centering as in B or C, it is apparent that some of the products will fall outside the specification limits.

Here the obvious action is to take all steps possible to maintain the centering of the process. This usually calls for continuous use of the control charts for X and R with subgroups at frequent intervals and immediate attention to points out of control. If fundamental changes can be made that reduce dispersion that eases the pressure. Consideration should also be given to changing of the tolerances.

Case III: When the process 6 is less than the difference between specification limits
The third type of situation arises when the specification limits are so tight that even with the process in control and perfectly centered some non conforming parts still get produced as in position A.

This primarily calls for a review of tolerances. It also calls for a fundamental change in the process that will reduce the process dispersion. It is still very important to maintain the centering of the process; the curves in position B and C show how a shift in process average will increase the non conformity.

Possible relationships of a process in control to a single specification

I: Low value of distribution ( ) is appreciably above LSL. II: Low value of distribution is at LSL. III: Low value of the distribution is appreciably below LSL.

Possible relationships of a process in control to a single specification


The first situation is one in which there is a margin of safety. The second is a one in which the specification is just barely met as long as the process stays in control. The third is one in which some non conforming products are produced unless there is a fundamental change in the process spread or increase in process average. All three distributions A,B and C have the same lower value. However, distribution B with a greater dispersion must have a greater process average than A for the low points to be at the same level. Similarly distribution C should have a greater process average than B. It is evident that the greater the distribution the higher the average must be for the entire distribution to fall above the lower specification limit. The relationship between average and cost is vital in many instances. For example in the filling of containers a reduction in dispersion may reduce cost by reducing the average overfill. On the other hand, the less the dispersion, the more important it is that the process average does not go out of control. This is illustrated by comparing distributions A and C in Case III. In Case III both process averages have shifted an equal amount below their position. However, the proportion of bad product, as indicated by the area of the distribution below the specification limit, is much greater in A than in C.

Use on control charts by a purchaser to help suppliers improve their processes.


Facts of the case: A manufacturer of electronic devices had trouble with the cracking of a certain small cross shaped ceramic insulator used in the device. The cracking generally took place after the manufacturing operations were nearly completed and did so in a way that made it impossible to salvage the unit. Hence the costs resulting from each cracked insulators during manufacturing operations suggested that others might be likely to crack under service conditions. In an effort to improve the situation, all incoming insulators of this type were given 100% inspection. This 100% inspection failed to decrease the percentage defective units. A simple testing was then constructed to measure the actual strength in flexure by testing insulators to destruction. From each incoming lot of insulators 25 were tested. As the insulators came from 2 suppliers, control charts (X and R) from both suppliers were separately maintained. The tests showed that both suppliers had approximately the same % defectives but the explanations for defectives were defectives. Supplier A had higher average strength but complete lack of anything resembling statistical control. Supplier B on the other hand had excellent statistical control but at a level such that an appreciable part of frequency distribution was below the required minimum strength. This diagnosis of the situation was brought to the attention of both suppliers and they were encouraged to exchange information about production methods. Finally product control could be achieved.

Milling a slot in an aircraft terminal block.


Decisions preparatory to the control chart: High percentages of rejections for many of the parts made in the machine shop of an aircraft company indicated the need for examination of the reasons for trouble. As most of the rejections were for failure to meet dimensional tolerances, it was decided to try to find the causes of trouble by the use of X and R charts. These charts which off-course required the measurements of dimensions, were to be used only for those dimensions that were causing numerous rejections. Among the many dimensions, the ones selected for control charts were those having high costs of spoilage and rework and those on which rejections were responsible for delays in assembly operations. This example deals with one such critical dimension, the width of a slot on the duralumin forging used as a terminal block at the end of the rear tail or an airplane. The final matching of the slot width was a milling operation. The width of the slot was specified as 0.8750 . The design engineers had specified this dimension with an unilateral tolerance because of the fit requirements of the terminal block; it was essential that the slot width be at least .8750 in and desirable that it is very close to this value.

Milling example continued


Due to the small no. of available inspection personnel it was decided that for each chart the sample would be inspected would be approximately 5% of the total production of the parts in question. It was decided that all measurements would be made at the place of production and it was decided that a part be measured every 20 parts produced. The subgroup size was thus fixed as 5. The figure on left shows all observations. The method to inspection to secure data for each of the charts was to measure two portions of the slot width using a micrometer screw gauge and average these measurements.

Milling example continued


Starting the control charts: The actual measurements of the first 16 subgroups are shown in the table. These numbers of subgroups correspond to a production order for 1600 of these terminal blocks. At the time of the 12th subgroup, before the completion of this production order and before the calculation of the central line or control limits, the quality control inspector noticed that the machine operator was occasionally checking performance on a terminal block that had just come off the machine and was still hot. After the 12th subgroup the operator was asked to do measurements after some time of cooling is allowed.

Determining the trial control limits: Calculations of the trial control limits was made after the first 16 subgroups which completed the production order. As shown in the figure on the right these were done by finding out the A2 and D4 factors for a subgroup size 5. Drawing preliminary conclusions from the graphs: Subgroup 1 is above the upper control limit on the R chart. Subgroup 10 is below the lower control limit on the X chart. Moreover the last 10- of the 6 points fall below the central line.

Milling Example Continued


It is obvious that the measurements made are not the result of a constant system of chance causes. If subgroup 1 is eliminated from consideration, R for the remaining 15 subgroups is 536/ 15 = 36. This gives us a revised upper control limit D4 (R) = 2.11 (36) = 76. Subgroup 5 falls exactly on the upper control limit in the X chart. Hence a second revision of R with subgroup 5 eliminated seems reasonable. Therefore R = 460/14 = 33. . An estimate may be made of from the R values. With a d2 value of 2.362 corresponding to the subgroup size 5 we get a = R/ d2 = 0.0014 in. The value of the natural tolerance or spread of the process is therefore 6= 6 (.0014) = .0084 in. This spread may be compared with tolerance spread: U-L = .0050in

It is evident from the process that the natural tolerance of the spread is substantially greater than the specified tolerance. Therefore, unless the process dispersion is reduced the process will keep producing non conforming components.

Vous aimerez peut-être aussi