MANAGEMENT CONCERNS TO CONTROL DEFECTS THE FIRST STEP This book is about solving defects, and it is impossible to be effective in solving defect problems if there is no historical record about defect prob- lems. In addition, it is impossible to keep track of costs without good de- fect records; thus these records will be important enough to impact the growth and perhaps the survival of the company. In most plants keeping track of the scrap is not always the first priority, maintaining production is always more important. Only when a customer calls or perhaps when scrap costs go sky-high does something happen to trigger some extra effort on a particular defect problem. Then the first step is to examine the scrap records. Hopefully, there are some worthwhile records to examine, BUT ALL TOO OFTEN THERE ARENT ANY RE- ALLY GOOD RECORDS - and the net result is a wild trial and error effort to find the solution. This is an extremely expensive way to operate, and it cannot compete with the those who use technical approaches and document them properly. A. SOME BASIC RECORD KEEPING Good records require some simple basics. These should be incorporated in the plant scrap reporting system and used at all times; it should not be used only during a crisis. Certainly the record keeping system should contain the information that a die casters customer will request, and it is para- mount that it contain good information for tracking costs; but these re- quirements often dont supply the information required by the process manager or engineer; who, after all, are the ones who have to solve the problem. 1 2 NADCA DIE CASTING DEFECTS Permanently solving the defect problems will absolutely require the inclusion of whoever is responsible for the final solution of defect prob- lems, and structuring the records so they can get the information they need. This information should include the following basics: Description of the defect Reason code, or assigned cause for the scrap. Listing of the number of defects (scrap percentage) by: die, by cavity number, by machine, by shift, by operator and possibly by inspector, metal heat number, and set up sheet number. Probably the description of the defect and the reason code is the most im- portant. This will vary some from plant-to-plant, and it should be tailored to the specialties and needs of a particular production operation. For example, some operations are mostly concerned with the visual appearance and may have several types of visual defects listed; while an- other plant may have several types of porosity listed with only one type of visual defect. B. REASON CODES (or description of defects) A typical list of reason codes would be as follows; but each plant should make their own list to match their special needs. Certainly others can and should be added as needed. 1. Visual appearance a. Nonfills b. Cold flow c. Discolored d. Laminations 2. Porosity a. Shrink b. Gas c. Flow 3. Leakers 4. Blisters 5. Sinks 6. Cracks 7. Drags 8. Solder 9. Bending 10. Distortion 11. Trim damage 3 Record Keeping and Management Concerns SHIFT REPORT FOR SCRAP THIS REPORT PREPARED BY: SHIFT NUMBER DATE PART No. CAVITY No. MACHINE No OPERATOR LOT OR MARKINGS: MACHINE SHUT DOWN BECAUSE OF SCRAP? HOW LONG? COMMENTS: Note: All scrap must be listed, including startup or other scrap thrown away at the machine by anyone. No. GOOD PIECES No. BAD PIECES REASON CODE (SEE LIST) C. TYPICAL REPORTING FORM A form that would include these items should be developed for each plant. A typical form is shown below. Having a good form is important, but assigning the responsibility for filling out the form accurately and completely is even more important. Just who maintains the form will vary from plant-to-plant; it can be the operator, the inspector, the foreman, or someone else; but most important is the discipline to do it correctly during the production shift when the best information is available. Note that it is important that ALL scrap be reported, including start up or warm up shots. These can be included in a special category if de- sired, but it must be reported. After all, these shots are not free. Every shot on a $100,000 die that will live for 200,000 shots is worth $0.50, regardless of whether it is a start up shot or not. There is probably an- other $.025 spent for expendables (mostly sleeves and tips) for every shot. This amounts to a good deal of money, and managers should insist on having the data. Management must take the responsibility to see that the reporting gets done; and that whoever does it has the training and the time and freedom to do it properly. The lack of management emphasis is usually the biggest cause for failure of these systems, and it can be very costly. For example, if the scrap is not defined and reported properly, then correcting the problem will usually be based on the observations of the past 2 hours. This will always result in a correction that isnt adequate and must be done over and over until finally done right. This should not be- come a way of life. Table 1.1. Typical form to gather scrap information. 4 NADCA DIE CASTING DEFECTS D. COMMUNICATING THE DEFINITIONS OF DEFECTS In all the communicating that must occur between people in any given plant; between the plant and the customer, between the plant and the ven- dors, and between the plant and other plants, there can be many big prob- lems defining and understanding defect problems. There are many instances where an operator or quality control person defines something as a defect when it really is not; and once the communi- cation is started, it may result in a series of actions that are always expen- sive and frequently unnecessary. This can be avoided by making sure ev- eryone understands the proper nomenclature and description of casting defects. Often the most serious problems are between the customer and the die caster. The customer may change their minds, or the definition of a par- ticular defect may vary when new people become involved, or when some- thing brings focused attention on a casting problem. Frequently it is mostly a problem of being able to define the defect, and these problems can often be avoided if the die caster establishes good definitions of defects early in the relationship. For all these reasons (and more), it is very advantageous for the die caster to have good definitions of defects available to everyone. A simple but critical factor is just having the people in the die cast plant talk the same language and understand each other. Thus when a night shift fore- man says we had a real problem with sinks and cold flow last night, then everyone knows what he means. The best way to solve this problem is to build a defect board with example castings and with a name of a particular defect tagged to each casting. Along with this display should be a book of pictures that show each defect. These should be 8 X 10 and in color (if possible), and placed in a three ring binder with the pictures in clear plastic holders. Several of these books should be made up; a minimum would be one book for each shift. When these books are first made up, the common reaction will be we dont need that, everyone here knows what we are talking about; but typi- cally within 6 months there will be several examples of how this system saved much more than it cost. The intent of the board and book of pictures is to define the names of defects for this plant. This will not change with different customers level of acceptance. These acceptance levels certainly may change, but the first step MUST be to be sure everyone uses the same name in defining defects, and that everyone understands what each type of defect looks like. A sepa- 5 Record Keeping and Management Concerns rate effort should be made to define the different levels of acceptability for each customer or product as needed; and this may change, but the defect definition board should not change. As an example, one big factor will be to define porosity defects as either gas or shrink (this will be addressed later in this book). The two types of porosity require entirely different corrective action, so defining the type of porosity is very important if quality improvements are expected. Certainly there are instances where even metallurgists will not agree on the appearance of the two kinds of porosity; but for most practical situ- ations, a system will be beneficial if it defines that porosity can be sepa- rated visually into either gas or shrinkage; or into a category that would include both gas and shrinkage. Even if it is not 100% accurate, it is criti- cal that this separation be made and reported if corrective action is to be effective. To do this, a definition of the two types of porosity must be established so people can communicate accurately. E. ADDITIONAL DEFECT REPORTING FACTORS An absolute essential for any defect reporting system is that it be available at the beginning of the day shift for all activity for the previous day. This should always be the case, and the report should be provided for everyone who could possibly want one (and even to those who dont think they need one) because defect correction is one of the most important activities in any die casting plant. It should be important enough that it is on everyones desk every morning. Operators or anyone else who needs it should get it, or know exactly where they can get a copy. The report should include, perhaps as a separate page, the information managers need for decisions. Probably the most important would be a daily Pareto chart showing the scrap dollars associated with defects by part number and by reason code. This will allow the managers to focus on the few problems that cost the most money. This particular chart can pay huge dividends for good managers, and is worth the cost to prepare. Another important portion of a good reporting system is an estimate of the production time or dollars wasted because of defects that were run the previous day. This would include machine time devoted to making scrap, down time for die casting machine caused defect problems, repair time for dies, approximate time and costs for engineering, process technicians, and production labor, etc. This is intended to bring out costs associated with defective castings. Routine repair items should not be included, of course; but this daily 6 NADCA DIE CASTING DEFECTS estimate of the cost of scrap is extremely valuable in managing the scrap reduction effort. Among other things, this helps make better management decisions about the effectiveness of the engineering effort and the effec- tiveness of new equipment. A real problem with most reporting systems is how to handle those defects that are discovered downstream in the process. Defects uncovered in machining are often not included in the daily scrap reporting system. It is almost always worth the cost to develop a system that will provide immediate feedback to the operator or process technician; not only of those defects detected in the production area, but those uncovered downstream. If the costs of running a half shift of poor quality is reported correctly, the cost of developing a method of getting immediate feedback will be quickly apparent. A system to machine current production samples during the produc- tion machining operation so as to provide quick feedback for porosity would be the best and usual procedure, but if this cannot be done; then some kind of special set up must be developed. Using X-ray or flouriscope or density measurements are the type of solution that provides the best chance for good statistical data, but even using an old lathe or milling machine that has lost its accuracy can be a good and inexpensive solution. F. PRE-ENGINEERING Some mention must be made of the requirement for pre-engineering; this is the first step in preventing defects. Unlike other casting processes, most die casting defects are process dependent, which means that the pre-engi- neering absolutely must include process engineering. The process decisions should not be based entirely on experience or guesses. The proper methods for this engineering design effort are taught in the NADCA courses on gating, PQ 2 , thermal analysis, and process con- trol. Defect solution is no longer a black art, and proper engineering will go a long way to eliminating what used to be big defect problems. This pre-operational engineering work, if properly done, will put the process into the known optimum operating windows; and will give the most robust design possible. This should all be done prior to sampling and production. Trial and error is fun for some, but it is too expensive and slow for what is needed today. This engineering effort must include a formal metal flow prediction and gating design, PQ 2 calculations, thermal analysis, and a review and a report on the process conditions required to meet the quality requirements 7 Record Keeping and Management Concerns for this particular part. All of this must be documented and available for others to use in later improvements. The days when verbal directions and estimates were good enough are long gone. Still, a great many (perhaps even the majority) of plants will attribute defect solution to a required trial and error program, usually emphasiz- ing metal flow and gating changes. In fact, in some plants, the gate changes never stop until the die is worn out. Doing the proper engineering up front (it does take extra time and effort) should give a robust process and usually means putting the die on and having it run easily and well from the first shot. The next step in reducing defects (once the process is running and stable) is to identify, analyze, and reduce any variation in the process. In fact, a consistent effort in this area is the only possible way to achieve the ppm defect levels now being required. G. STATISTICAL AND OTHER ANALYSIS METHODS FOR REDUCING VARIATION AND FOR DEFECT CORRECTION There are well proven statistical techniques available to determine which are the main influences for process variations. SPC charting is valuable and should used where appropriate as a monitoring tool for the process after using analysis tools. Analysis tools that can be used to define the important variables and forecast the areas where effort should be focused include the FEMA and various Design of Experiments methodologies (Shanin, Taguchi, etc.), FEMA will try to predict the problems in the end use of the casting or in the downstream processes from different casting defects. This then in- dicates the relative importance of various defects, which is especially valu- able for products that have large production volumes, and for those that have a safety impact. This information is then used to modify the part design and/or the process procedures and equipment so the more critical types of defects will be less common. A FEMA should be done as early as possible in the design process so the information will be available to the part designers, the process engi- neers and die designers, who should use it to minimize the impact from different types of defects. Training is essential for those involved; also the effort (and cost) involved makes this technique more appropriate for very high volume work or for products that have a potential for safety related failure. The design of experiments techniques (DOE) will tell the process en- 8 NADCA DIE CASTING DEFECTS gineer which variable(s) are the most important for the defect of concern. They are required if the process is to develop into one that will reduce scrap to the parts per million (ppm) level. For example, this book will describe a sink defect as one that is nor- mally reduced by temperature control or by pressure adjustment. For any given part and machine configuration, however, it may be difficult to de- termine just which pressure or temperature adjustment would be the most critical, especially to reduce the scrap from say 1% to .5%. A properly run DOE would provide this kind of answer. The most common DOE techniques are Shanin, Taguchi, and the clas- sical factorial experimental methods. Each of these requires an understand- ing of statistics, and some training in how to apply these methods in an industrial environment; but the training is well worth the effort because these techniques are the only way to determine the exact cause and effect relationship for some defect situations. Some of the most useful and basic data needed for correcting defects comes from the more simple procedures. The usual method of determining the quality standards for those cast- ings that require judgmental standards is to ask the customer to define some bad pieces. The customer may or may not respond with a good standard. If this is the case, it is imperative for good defect control to establish in plant standards purely for the purpose of internal process management, and not to depend on the customers response. Good process management requires setting rating standards, and this must be done; even if it has to be done independently of the customer. Adjustments can then be made if the customer changes, but the internal standards can stay constant. Setting standards on judgmental defects should be done by establish- ing a rating system. It takes time and effort, but it will pay dividends. It is invaluable for avoiding disputes with the customer; but even more impor- tantly, it will allow the use of DOE and other statistical techniques that ensure the best quality for the least cost. Also, it will make available pro- cess control techniques that are essential if the die caster is to approach the ppm defect levels. Without a rating system, the operator or process engineer only has at- tribute data. This is somewhat like asking a machinist to machine a shaft to a 1 diameter, but the machinist is not told what the actual output diam- eter really is; the only information provided is whether the output is good or bad. As in die casting, this guarantees a lot of very expensive trial and error work. Without feedback or a rating system, the operator is forced to make 9 Record Keeping and Management Concerns process adjustments by judgments based on some indicator of the output quality. Usually the surface appearance is used because nothing else is available. However, surface appearance is not a good indication of many quality characteristics (gas porosity, for example). This makes it impos- sible to make improvements in the process. Much better is prompt feed- back from a system with a good rating procedure. A rating system is based on establishing quality levels. These are usually set to be from 1 to 10, with 10 being the best, but it can be from 1 to 5 if necessary; although this scale is rather course and does not de- fine the process as well. The dividing line between good and bad pieces could be at a rating of 4 on a scale of 1 to 10, with 4 being just barely good enough to pass, and a rating of 3 being just barely a reject. If the 1 to 5 scale is used, then the dividing line could be at 2. Other methods can be developed. This can then be used to determine where the process is running; trends can be determined, and process corrections made before the first bad piece is made. Determining how much variation is in the process can be done with a short capability study. This gives a quick picture of process condi- tions, and is a required portion of the process development if the scrap is to be reduced to the ppm levels. H. AN EXAMPLE OF A DEFECT CORRECTION PROCESS FOR VERY LOW SCRAP RATES The following steps illustrate one way to approach a defect problem where the scrap rate is low, for example to go from a scrap rate of 1% 3% to something approaching .1% (1000 ppm). Assume that surface finish is the problem, and that a rating system had been developed going from 1 to 10 (10 best). A short capability can be run with at least 30 parts taken in 6 samples of 5 sequential parts. This will give a quick snap shot of where the process is operating, and show if defects are being made even if no defects appear in the samples. An approximate standard deviation can be calculated by dividing the aver- age range of each of the 6 groups of 5 samples by d 2 , which is 2.326 for subgroups of 5. This can be shown in the table that follows. The estimated sigma () will provide an estimate of the variation as shown by this sample; in this case = 1.182, which is larger than one would like. This puts the lower limit of the distribution curve (as defined by the 3 limit) below 4, and since anything below a rating of 4 was not an 10 NADCA DIE CASTING DEFECTS SUBGROUP 1 2 3 4 5 6 7 8 SAMPLE 1 8 6 8 8 8 9 9 6 SAMPLE 2 7 8 8 8 9 7 5 8 SAMPLE 3 8 8 8 7 8 6 7 8 SAMPLE 4 7 5 6 6 8 6 8 7 SAMPLE 5 6 7 5 8 6 7 8 8 TOTALS 36 34 35 37 39 35 37 37 AVERAGES 7.2 6.8 7 7.4 7.8 7 7.4 7.4 RANGES 2 3 3 2 3 3 4 2 AVE OF GROUP AVERAGES, = X = 7.25 AVE OF RANGES, = R bar = 2.75 EST. SIGMA, OR R bar /2.326 = 1.182 note: VALUE OF CALCULATED SIGMA = 1.125 VALUE OF X MINUS 3 X ESTIMATED SIGMA = 3.828 VALUE OF X MINUS 3 X CALCULATED SIGMA = 3.999 acceptable casting; then there will be a number of castings that will be rejects. Note that no castings were found in the samples taken that were re- jects, but the location of the overall average, and the width of the distribu- tion curve as defined by the estimated variation shown in this sample, makes it certain that there will be some defects. The exact percentage of defects can be calculated at about .31%, but this an approximation, and for these purposes, the exact calculation is not as important as knowing that the process is where it needs to be, and more work needs to be done to reach the goal. Note that an accurate calculation of sigma is also shown, and it is close to the estimated value for these numbers. While the estimated value using this method will not always be as accurate as in this example, it will be close enough for the process engineer to have a good idea of where the process stands for the sample taken. This example assumes all the surface defects were combined into one rating number, but if there is more than one quality issue, it may often be best to use a separate rating system for each of the judgmental type defects. Table 1.2. Sample calculation of an estimated process capability. This table represents 8 samples of 5 sequential castings taken for the purpose of establishing a quick picture of the process capability. The value will provide an indication of where the process is performing on the average; in this case it is at 7.25 on a scale of 1 to 10, where 10 is the best. 11 Record Keeping and Management Concerns If the short capability study comes out as shown in the example, then the process technician should work on two things: one is to increase the average rating ( X ), the other is to reduce the variation. The first step to increase the average quality level is to define the im- portant process variables for this casting. This is typically done with a DOE, but first the factors to use in the DOE have to be determined. Hope- fully, this book will be of some help in narrowing down the possible cause factors; the process engineer should also do some brainstorming with quali- fied people. Also, some simple techniques can be very powerful to narrow the vari- ables to the correct ones. One is to determine if the problem happens from time-to-time or from shot-to-shot. These are Shannins techniques, and are described in Keki Bhotes book World Class Quality. Mr. Bhote uses the Shanin methods to suggest some simple but effective methods for pro- cess control that are intended to reduce defects where the quality rating is judgmental. Usually die casting will have high shot-to-shot variation, which would include plunger variations (fill time, pressure, gate velocity, biscuit thick- ness, spray variations, etc,). Die or metal temperature would usually be examples of typical time-to-time variations. Once the probable variables are selected, a DOE would probably be the best next step because there are usually a number of interactive vari- ables in die casting. Other studies, such as a correlation analysis could be done, but they will miss the possible interactions. The old fashioned full factorial is always a good method. Once these important variables are defined, the appropriate settings can be determined with calculations and some follow-up on confirming experiments. Variations in these factors should then be studied and reduced as much as possible. This is discussed in more detail in the next section. I. MAINTAINING THE SETTINGS, AND REDUCING VARIATION Generally there are only a few really important variables, and limits need to be set on these based on the information from the DOE. Once these limits are set, then monitoring the variables should be done (at least for a short time) to make sure they can be controlled within limits, and to reduce the variation even further if possible. Monitoring can be done with SPC charts, run charts, or by using pre- 12 NADCA DIE CASTING DEFECTS control. Pre-control charts are often overlooked, but they are accurate and they provide better information for process control, and they are much easier for the operator to use. As an example, if pre-control charts were to be used with the system discussed earlier, where 4 is the worst acceptable rating; then the pre-con- trol charts should have a line drawn at 5.5. The area between 1 and 4 is colored red or called the red area. The area between 4 and 5.5 is colored yellow or called the yellow area, and the area between 5.5 and 10 is the green area. Five parts in a row are checked on start-up to verify that the process is in control and within limits, and they must all be within the green area to be comfortable that the process is ready to run. After this check, then two parts in a row are checked at intervals; these intervals are found by taking the time between stoppages (reds) and dividing by 6. The two parts are used as follows: a combination of yellow and green is OK to keep running, two yellows means the process needs adjusting (but keep running), and one red means the process should be stopped and corrected. Using some kind of charting, whether it be pre-control, SPC or just run charts, will be important; but once the process is stable, it is desirable to discontinue charts that are not needed. The important thing is to do some charting until it can be demonstrated that the process is stable. However, it is also extremely desirable to monitor all process variables as much as possible and to make every effort to reduce variation wherever it is found. These ideas and comments are intended to convince the reader that there is a lot more to correcting a defect than just getting a good idea of what may be the cause. Since most die casting defects are controlled by the process, it also requires good process discipline and control. A good system to reduce defects requires considerable management skill and attention if it is to get good results. The system must have the capability of doing the proper engineering up front, then analyzing defect problems to find the proper variables for process control. Only then can the proper process discipline to correct the defect be imposed.