======================== Tutorial for 'Variation' ======================== This is a tutorial to use in conjunction with "Variation". It has been written assuming that the user has no previous experience of Control Charts. This tutorial does, however, assume that you are familiar enough with the Windows environment to use menus, command buttons and dialogue boxes. There is guidance on using the windows environment in the file INSTRUCT.TXT. If you want to read this document on-screen and carry out the instructions at the same time, make the Notepad or word processor screen displaying this document wide enough to see whole lines of text, and adjust the height of the screen so that it is just enough to see a few lines of text. When 'Variation' is running, change the size and positions of the main windows so that most of this document remains visible. If you have a very small screen, it might be easier to print this document and work from a paper copy. To print this document, open it in Windows Write, or another word processing application, select the entire document and format the text in 10 point Courier before printing. Throughout this document, indented texts are step-by-step instructions which you should carry out on your computer. Run 'Variation' now: double click on the icon in the Program Manager or double click on VARI.EXE in the File Manager read the message on the opening screen then press the Enter key or mouse click on the Start button BEFORE YOU START THE TUTORIAL: If you have not already done so, read the first 4 pages of the on-line instructions carefully. This will take about 5 minutes. Read pages 5 - 7 later: select Instructions on the menu bar or press F1. Use the Next button or press Alt-n to change pages. When you have read page 4, click on the Close button or press Esc. NEXT, RUN THE FUNNEL EXPERIMENT: This will take about 7 minutes. In Funnel Experiment mode, the scales are changed to have zero in the centre. choose 'New' from the 'File' menu click on 'Funnel Experiment Simulation' in the dialogue box press o.k. now follow the on-screen instructions One of the lessons to be learnt from the funnel experiment is that tinkering with a stable system (in our case, moving the launcher) will create more variation on the output than just leaving it alone. Now you are ready to learn how to control a process using statistical methods. During the following exercise, if you want explanations of various section of the programme, you may press the 'Explain' buttons as they appear in windows or dialogue boxes. This will slow you down so you may prefer to leave the explanations until after you have completed the tutorial. This programme automatically creates charts and calculates control lines for these charts from the results of the bouncing ball 'process'. The calculations for the control lines involve the use of published tables, but they are not complex. So for real processes, it is not difficult to draw these charts on paper. It is very important for the process operator to be actively involved with the charts. He or she should plot results without too much delay. Also, notes should be written on the chart when anything occurs which might affect the process (e.g. changes of operators / sources of raw materials / machinery maintenance etc.). These notes may be needed later to track down the source of a special cause of variation. START OF TUTORIAL ----------------- The tutorial will take about 20 minutes. With 'Variation' running, do the following: choose 'Open' from the 'File' menu Select "TUT1.VAR' in the dialogue box and press o.k. (if there are any previous results visible on the screen, remove them by choosing 'New' from the 'File' menu, click on 'New File, retain settings' then press O.K.) Imagine that you are the operator of a machine - the launcher. Your job is to fire balls at the target and get them to land as close as possible to the ideal value of 500 (we are no longer using centre zero scales). First we need to 'centre' the process. We will fire a number of balls, find the average, then move the launcher. Fire off 50 shots WITHOUT MOVING THE LAUNCHER click on '50' in the 'Shots' frame in the main window press 'Start' When all the balls have been fired, calculate the mean or average landing position choose Mean from the Calculate menu use all the results in the calculations (i.e. accept the default values in the From and To boxes) click on Calculate. note the Mean value - integer part only, ignore what comes after the decimal point press Cancel or Close Now move the launcher so that the spread of future shots will centre on the 500 mark. subtract the Mean value (average) from 500. You have to do this manually move the launcher by this amount (click on the Help button in the Launcher Position frame if you do not know how to do this) The process should now be centred (do not worry if you got the calculations wrong, it is not important for the rest of the exercise). Next, we need to find out if the process is "statistically stable". Fire off another 100 shots then create a control chart. with 50 selected in the Shots frame, press Start twice choose Control Charts from the Charts menu accept the default chart type of X-bar and Range with subgroups of 5 so press O.K. The results will be arranged into subgroups of 5 shots. For each subgroup, the average will be plotted (X-bar) along with the largest value in the subgroup minus the smallest value (Range). The point at which the launcher was moved is shown on the chart as a vertical line with the letter 'P' above. This is the equivalent of writing a note on a paper chart and is the sort of information an operator should record somewhere on the chart. Use the mouse to click on the vertical line to see the note. mouse click on the vertical dotted line on the control chart. Hold down the mouse button to read the note Next we calculate the control lines. The purpose of these lines is to show when we should suspect that something has changed which affects the process (in other words, a special cause of variation has occurred). Of course we know that the launcher has been moved and moving the launcher is a special cause of variation, so we should calculate the lines with results which come after the launcher move. press Calculate Lines above the X-bar chart in the From box, enter 51, this shot number was just after the launcher was moved (if you clicked on the note line on the control chart before opening this dialogue box, you can enter the number simply by clicking on the words under the From box) make sure that the last subgroup is in the To box (you can do this quickly by clicking on the words under the box) press O.K. If all the results after moving the launcher are within the control lines, then the process is probably stable or "in statistical control". This means that all the variation comes from common causes. Common cause variation is just the normal random variation which is inherent in the process. If some of the results after moving the launcher are outside the lines, then this has probably been due to a special cause of variation. The process is described as "unstable" and because there is every reason to believe that this type of change will happen again, life is going to be pretty chaotic. If we want to be in full control of a process we must use the charts to identify when special cause variation occurs, determine if things were better before or after the change, then make one of these situations permanent. So let us carry on producing. Fire off another 100 shots return to the main window, but do not close down the control chart. Minimise it or leave it in the background with 50 selected in the Shots frame, press Start twice Look at the control chart and do not recalculate the control lines. All we need to know now is whether there has been any change in the process since the lines were calculated. Is the output stable? bring the control chart to the foreground look to see if any of the points are outside the control lines (ignore the points before the control lines were calculated) It looks as if something unusual happened around shot 200. Subgroup average drops below the control line so a special cause of variation has occurred. As an operator, your job is to produce results as close as possible to 500 but the average landing position has suddenly changed. You could, of course re-centre the process (move the launcher). This might help, but you have no idea whether things might suddenly change back to normal. The only really satisfactory solution is to investigate and find the source of the special cause of variation, learn from what happened, and make sure that the change does not occur again. A word now about tolerance limits. In most industrial processes, the operator is given conformance or tolerance limits as well as the target value. However, these tolerance limits should always be looked on as representing the MINIMUM acceptable quality from the process. World class quality does not come not from treating everything within the tolerance limits as equally acceptable. We must try to produce as close as we can to the target value. This is what the customer really wants. In our bouncing ball process, an unknown special cause of variation made the subgroup average fall at around shot 200. It might be that the individual results are still within the specified tolerance limits, but our customer would prefer the results to be 500. So we must make efforts to produce with the average output at 500 and the minimum variation that our process is capable of. So we must investigate and remove special causes of variation even though we are still producing within tolerance limits. At this point you should be at shot 250. I can tell you that investigations show that that one batch of balls has slightly less bounce than normal. We discard this batch and demand from our supplier that they supply us with statistically stable product (they can only be sure of doing this by using control charts). I have removed this special cause of variation so things should return to normal from shot 251. Fire off another 50 shots to see this. So this time, there is no need to change the position of the launcher. return to the main window but do not close the control chart window fire off another 50 shots The control chart should show clearly that a change occurred around shot 200 and things returned to normal around shot 250. bring the control chart to the foreground look at the plots on the X-bar chart In this tutorial, the computer calculates and plots large quantities of results at once. This means that we might not detect special causes of variation for some time after they occur. If an operator plots results manually, then he or she might know quickly that a change had occurred. In this case, the process was stable in the initial stages. The control charts clearly indicated that a change had occurred compared with the time when the lines were calculated. What happens if the process is unstable while the control lines are being calculated? Open the file TUT2.VAR. Now, there will be special causes of variation in the early stages. choose 'Open' from the 'File' menu in the main window Select "TUT2.VAR' in the dialogue box and press o.k. (if there are any results visible at this stage, remove them by choosing 'New' from the 'File' menu, click on 'New File, retain settings' then press O.K.) Fire off 100 shots and then create a control chart. We will not worry about centring the process: with 50 selected in the Shots frame, press Start twice choose Control Charts from the Charts menu accept the default chart type of X-bar and Range with subgroups of 5 so press O.K. Now calculate the lines. We have not moved the launcher so we can use all results in the calculations. press Calculate Lines above the X-bar chart make sure that 1 is in the From box make sure that the last subgroup is in the To box press O.K. You should now see that the process is indicating instability even though the data used to calculate the lines contains instability. We will need to investigate the special causes of variation and remove them to bring the process under control. After that, we can be confident that when we centre the process by moving the launcher, the position will not need to be changed again (unless, of course, another special cause of variation comes along). This ability of Shewhart control charts to detect special causes of variation, even when these special causes of variation are present in the data used to calculate the control lines is very important. Most commercial processes are not naturally in a state of statistical control. The control lines are set at 3 times sigma from the average. Sigma is similar to standard deviation but uses calculations based on the spread of results within subgroups. It is a mistake to calculate the control lines from individual results. See for yourself: in the main screen choose Mean from the Calculate menu click on Show 3 x S. Dev. lines on Minichart click the Close button note the warning and click O.K. Lines at 3 times standard deviation from the average will now be drawn on the minichart on the main window. Notice that this chart does NOT detect the special variation whereas the x-bar chart does detect it (by 'detecting' we mean that there are points plotting outside the control lines). You can spool back the minichart to see previous shots by clicking the scroll bar to the right of the chart. You have now come to the end of the tutorial. =============== END OF TUTORIAL =============== I hope that you are beginning to understand how control charts are used. By distinguishing between special cause variation and common cause variation, control charts can help operators and managers to run processes which produce on-target with minimum variation. Note that process capability indexes (like Cpk) work on the assumption that the future will behave like the past. This assumption is only valid if a process is stable. If special cause variation is present, we must find the root cause and stop this from occurring again in the future. If no special causes are present and the average output is on target, then if we are still getting unacceptable output, we must take steps to reduce the common cause variation. For example we might need better machinery, more maintenance or less common cause variation within raw materials. The simulation used in this programme is based on an industrial process where the output is measured on a continuous scale. There are similar techniques using different types of charts for processes where the important features are counted (such as the number of flaws or errors). Distinguishing between common cause variation and special cause variation is just as important in non-manufacturing processes. For example, if the latest quarter's sales figures in an area are up compared with the previous period, we need to know if this is simply part of the natural ups and downs inherent in the selling process (common cause variation), or due to something unusual which has significantly altered the success rate - such as an improved advertising campaign or some extra-special effort from a salesperson (special cause variation). Imagine what happens if we think that the advertising campaign or the extra efforts were responsible for the improved figures but, in fact, they were simply due to the normal common cause variation. We will probably declare the campaign a success and adopt it across the country, and we might give the salesperson a prize. Next quarter, we are astonished to find that some areas get reduced sales with the new advertising campaign and the 'top' salesperson is well down the league of achievers. We do not understand why we did so badly and disillusion sets in. In fact we did not do badly, this is simply common cause variation at work again. ======= Summary ======= Let us go over the main points that you should now understand about processes, and the use of this type of control chart: 1. All processes contain variation. 2. World class quality comes from producing with the average on target and with the minimum of variation about the target. 3. To achieve the above, we must distinguish between special cause variation and common cause variation. We need to know this difference because the things we will have to do to remove or reduce the two types of variation are very different. 4. The way to distinguish between common cause variation and special cause variation is to use a control chart. 5. The calculations for the distance of the control lines from the average are based on an average dispersion measurement. 6. We must use our knowledge of the process when deciding how to sample results and arrange them into subgroups. We should do this in a way which we know will reduce the chances of special cause variation occurring within subgroups. The range of each subgroup is a convenient dispersion measurement. 7. If we do not know much about the process and we cannot be confident that little special cause variation will be present within subgroups, then we should use a Moving Range chart where the dispersion measurement is based on the difference between subsequent individual results. 8. Operators should understand the charts. They should be involved in producing them and plotting the results. 9. Before we can consider a process to be "under control", efforts must be made to remove special causes of variation. We must also learn from each incident of special variation and take action to make sure that these types of changes do not happen again - this may require close co-operation with suppliers (some companies are not capable of this type of management - they will always exist in a state of chaos even if they have control charts). 10. If the process is statistically stable, any adjustments that we make to the process average will create more variation on the output. =============== WHAT TO DO NEXT =============== Return to the default settings: choose New in the FILE menu, select New File, Default Settings click O.K.. Fire off some shots, then create special causes of variation either by moving the launcher or by changing the variation settings (Choose Variation from the Control menu). Read all the messages available from the Explain buttons as you find them. Create control charts and experiment with control lines calculated from different sections of the data. Read about 'Taguchi Loss' (Choose Loss from the Calculate menu and press the Explain button). The VPROGx .VAR files --------------------- As well as the tutorial files, 9 other variation files have been provided. Try to detect the unknown special variation created by these files. With this exercise, do not attempt to keep the balls landing close to 500. The only purpose here is to give more experience in detecting special cause variation using control charts. After Variation is running, open one of the files. choose OPEN in the FILE menu, select one of the VPROG files. Do not use the control menu to look at the pre-programmed variation. Instead, use control charts to try to figure out what is happening to the centring and spread of the results. Fire off one or two hundred shots before looking at the results (don't change the Launcher position, this will add another source of variation). You may want to calculate the control lines for the X-bar and Range chart after about 100 shots. Do not go beyond 500 shots, there will be no special causes of variation after that. In most cases it should be possible to determine what sort of change has taken place by looking at the X-bar and Range charts with subgroups of 5; you should also be able to tell approximately when the change happened. (Note: It is not possible to determine from the charts if the changes were caused by the velocity of the ball or the bounce factor). With some of the programmes, however, it may be difficult or even impossible. One of the programmes explores how little change the charts will detect (you will have look for 8 consecutive plots on one side of the centre line to find a signal with this file). One of the files has been deliberately produced to cause variation which gives misleading results with subgroup size 5. (The lesson here is that it is vital to subgroup intelligently. Using your knowledge of the process, you must choose a subgroup size where there is little chance of special variation being present.) One of the files will cause no special variation at all. This has been included so that you keep an open mind when interpreting the results. When you think you have determined the special causes of variation, look to see if you are right, select Variation from the Control menu in the main window look at the Programmed Variation Changes section. This will show the shot number where special variation occurred, and whether it affected the Centring (average) or Spread (range). Game Mode --------- When you are happy with the concept of what control charts are trying to achieve, try the game. Read the game instructions carefully before you start, perhaps more than once. It is important that you understand the process and all the factors which may influence it before you try to play the game. choose New from the File menu click on New game click on O.K. click on Game Instructions on the menu bar or press F1 Good luck! FINALLY ------- I recommend that you read Donald J. Wheeler's book "Understanding Statistical Process Control" before applying SPC to real processes. See the final page of the on-line instructions in File mode for details of the book. I hope that you enjoy using 'Variation'. I would be very grateful for any feedback from users. Steve Horn, 21 Benjamin Drive, Bo'ness, West Lothian EH51 0QS, United Kingdom CompuServe 100116,3151 Internet steve@horn.demon.co.uk 8 January 1995