Printer Friendly

SPC for small to medium foundry's sand system.

SPC for Small to Medium Foundry's Sand System: Part 3

Now that you have set up your testing procedures, trained your operators, supervisors and technicians, and are taking data, it's time to learn how to use your SPC program.

I hear a lot about control charts. What are they?

Control charts enable you to display your data over time so that process variations are identified and you can take corrective action. They are the keys to continued process improvement.

(At this point you should begin to realize that the purpose of SPC isn't just to keep your operation humming along the way it is today, although it will help you do that. SPC's major benefits come from using the data to improve your opeations, to decrease the amount of variability in your sand system. As you go through the rest of this discussion, keep in mind the concept that SPC is forever, and that properly used, it will lead to ongoing improvements in your operations and profitability.)

The most commonly used charts--and the ones you should start with--are the X (X-bar) and R charts.

The X chart is a plot of the average of the values of your readings for a particular variable against time. That's the number which you calculate by averaging all of the data for a single batch of sand or time period.

The R chart is a plot of the range, or the difference between the highest and lowest values you found when you measured each of your samples--again, against time. Usually X and R charts are plotted on the same piece of paper, with the X chart at the top of the page and the R chart at the bottom.

Examples of X and R charts are shown in Fig. 1. The data from which the chart was made is included, and shows the kind of data sheet to use.

In other words, the X chart measures the batch-to-batch variation, while the R chart measures the variation within each batch.

At the next interval (that is, the time you measure the samples), plot the values again. Connect the dots with a line. Do this for each of the tests you make throughout the day.

Now, look at the lines that you have drawn. Are they pretty much horizontal or do they tend to crawl up or down the chart? Are they primarily consistent or do they move up and down so much that they look like a profile of the Rocky Mountains? Do they seem to vary in a random manner or do they have a pattern?

Depending on the appearance of these charts, you will be able to tell whether your process is in control or out of control.

Control Limits

What are control limits?

Control limits are values of X or R that are set to tell you when you are out of control or when your process should be looked at because it is about to be out of control. In Fig. 1, they are marked [UCL.sub.X], [LCL.sub.X] for upper control limit and lower control limit for X and [UCL.sub.R], for upper control limit for R. Note that R has only an upper control limit.

Here we should make a distinction between process specifications and statistical control limits. Your process (or engineering) specifications are the limits you want to achieve because you know from experience that sand that falls within these limits produces satisfactory castings. Statistical control limits are calculated from statistical rules and based on the performances of the system.

Why are they different? Shouldn't they be the same? Why can't I just use the process specifications? After all, I know what values I must have to make good sand.

The reason that you can't use just the engineering process specifications goes back to the reason for an SPC system. You see, just keeping track of when you exceed your limits doesn't do much for you--by that time you've already produced sand that isn't within specification. What you want is a system that will warn you in time so that you don't exceed your limits, and therefore, don't produce bad sand.

There's another reason for having two sets of limits. The difference between your engineering specification limits and the statistical control limit tells you how good your process capability is.

To see why, let's consider what these statistical control limits are. You can find them after you've completed the first day of charting.

You begin by finding X (X double-bar). This is merely the average of all the X's you recorded during the day. It tells you what the overall average performance of your system is.

Then you calculate the R (R-bar). This is simply the average of the ranges you observed in your measurements during the day.

Now you set upper and lower control limits. These depend on the number of samples you take for each point on the chart (more detail will be foundin books and manuals on SPC), but for three samples per point, a good approximation is:

UCL on X = X + R; LCL on X = X - R; UCL or R = 2.5 R.

Draw the engineering specification lines on the chart and then draw the statistical control limits on it. Now you are ready to use the chart to learn something about your sand system.

Look at the statistical control limits that you calculated. Do they lie inside the engineering specification limits? (In other words, could you group all of your data between the limits, as shown in the upper left of Fig. 2?) If so, that's a good start.

But if not, and the data looks like the upper rightof Fig. 2, you have a problem. Your sand system isn't under control. Before you go any further, go over your sand system--the equipment, materials and operators' training--to find out what's wrong. You can't start an SPC program with an out-of-control system.

Figure 2 shows one other possibility. At the bottom you see a chart which shows that the data falls within a narrow range (narrow enough that you can be pretty sure that you have the process under control), but in the wrong place. Here the process is under control but out of specification.

In other words, you're pretty good at consistently making bad sand. That's not what you want. Either your specifications are incorrect or you should change your sand practice to bring the data back into the specification limits.

Go back to Fig. 1 now and take a look at the specification limits (given in the box on the data sheet in the upper right hand corner), and the upper and lower control limits (given on the X chart). Notice that the lower specification limit (28.0) is outside the lower control limit (28.11). That's good because a reading can reach the lower control limit--which says that it's time to take action to correct the situation--before the specification is violated and bad sand is made.

Note, however, that the upper control limit (36.57) is outside the specification limit (34.0). This says that there is a serious problem in controlling LOI in this foundry and that management action is necessary to correct the situation.

Warning Signs

But, assuming that your control limits are within the specification limits, look at your points. An out-of-control condition exists when any of the following is present:

* one or more points fall outside the upper or lower control limits;

* there are more than six points in a row on the same side of the center line;

* the data points trend up or down;

* more than half of the points are above or below the center line;

* an obviously nonrandom pattern is present.

Any of these circumstances tells you that you should look into your sand system. That means going back over the materials, people, procedures and equipment performance to see that they're all doing what they should be.

Note that the system can exceed the statistical control limits without exceeding the engineering specification limit. When that happens, you are getting a warning that something in your sand system isn't right before you make bad molds and while you still have time to correct the situation. That's the purpose of SPC--to let you know of problems before they affect the quality of your product.

While we tend to think that just the X chart is important, the R chart also gives good information. It will indicate (frequently faster than X dta will) when your raw material is varying and when the equipment (muller) is malfunctioning.

Based on your performance, update the statistical control limits each day. Your goal is to get them where you want them within the engineering specification limit, and hopefully well within these limits. Of course, when you find that you can operate within tighter limits, you adjust them so that your molds become more uniform and your castings better.

As you can see, SPC is a continuing process that grades your performance every business day of the year. Used properly, it will help you avoid scrap and rework, and improve your profit.

Reducing Scrap

When you get to this point in your SPC program, it's time to consider assigning one of your staff the job of learning advanced SPC concepts. What you have up to now is just an introduction to what SPC is capable of doing.

You mean there's more than just control charts?

Indeed there is. The example below will show you how you can use your SPC system to cut costs.

For instance, if you have been analyzing your causes of scrap, you can make up another type of chart, called a Pareto chart. (Pareto is the man who discovered that, in most cases, there are a few significant problems, which if solved, will have maximum impact on the total situation.)

The Pareto chart ranks causes of scrap (starting with the most important) and shows you where the largest contributor to the scrap is to be found (see Fig. 3). Instead of wasting valuable time chasing down a small scrap problem, you can concentrate your efforts and energy on the large contributors. (Note that instead of looking at the frequency of occurrence, you may concentrate on the scrap's contribution to cost and choose to work on the most costly items.)

The situation shown in Fig. 3 is an example of what might be the case for a foundry. In this case, overall foundry scrap has been plotted. However, if there are a few high-running jobs, you would want to prepare similar charts for each of them. This chart shows that the foundry's most important problem is burn-in.

Now that you know what defect you want to tackle first (in this example it's burn-in because it's the largest cause of scrap), determine what the possible causes of this scrap are. To do this you can use another tool called the Cause and Effect Diagram, also known as the "Fishbone" Diagram. This diagram shows the relationship between possible causes of a final outcome (in this case, the possible causes of burn-in).

A good way to develop this chart is to "brainstorm." Assemble a group of individuals who are somewhat familiar with the process. Causes for the effect are usually separated into five groups: people, material, methods, equipment and environment.

For burn-in, we can think of at least ten possible variables or causes. Now it's time to go back to the X and R charts on our sand system. Take a good look at the methylene blue clay, moisture and sand temperature charts. If they are showing that they are predominantly on the wrong side of the center line of the chart, you have identified the problem and can take action to correct it.

Of course, when you have corrected the burn-in problem you'll find that blows are the biggest cause of scrap. So you will set to work to get rid of blows using the same procedure: Gather your staff and prepare a Fishbone diagram of the possible causes of blows, look at your control charts for clues and fix the problems.

Then go on to scabs, and then to sand inclusions, until you've fixed every problem in the foundry. Of course, that day won't ever quite be reached, but your goal is to get as close to it as possible.

After a basic understanding of the foundry system has been achieved, the use of more advanced techniques such as Designed Experiments (also called "Taguchi Methods") can be employed effectively.

Committe (4-M) members who contributed to this series are: Van Champion, Talladega Foundry & Machine Co, Inc; Bill Edison, George Fischer Foundry Systems, Inc; Steve Neltner, The Hill & Griffith Co; Pat O'Meara, Intermet Corp/Lynchburg Foundry Co; Tom Piwonka, Univ of Alabama/Tuscaloosa; and Jack Vincent, Mueller Co.
COPYRIGHT 1989 American Foundry Society, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1989, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Part 3; statistical process control
Publication:Modern Casting
Date:Nov 1, 1989
Previous Article:Hunter marks 25 years of molding innovations.
Next Article:Upgrading castings by hot isostatic processing.

Related Articles
SPC for small to medium foundry's sand system: part II.
Failure mode and effect analysis: a step toward total quality assurance.
Quality and computer modeling for the next decade.
Total quality control: a 'top down, bottom up' approach.
Sand SQC cuts foundry defects.
Effectively using Gage R&R and measurement systems variability.
Process control tips for a phenolic urethane nobake foundry.
Gain a competitive edge with TQM.
More than maintaining: maintenance of advances in process control at Intat.
In Pursuit of high tech: what foundries must consider to become State-of the-Art: three engineering firms detail their vision of what it takes to be...

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters