700 readers say: mold analysis makes the grade.
How well is computer-aided mold analysis living up to its potential? That's the question we set out to answer with the first broad-based survey of user experience ever published--and, as far as we know, ever attempted--since this type of computer-aided engineering (CAE) software became commercially available in 1978. With advice from CAE software suppliers and users, PLASTICS TECHNOLOGY designed a four-page questionnaire and mailed it together with our January 1992 issue to all 20,000 PT readers involved in injection molding and/or tool design.
Their responses leave no doubt that CAE mold analysis is providing significant benefits to the vast majority of firms using it. They said it is saving them time, money, and raw material, cutting their scrap and reject rates, improving their product quality, and helping them get new products to market faster. Those "user" firms said 10% of their tools are being designed with mold analysis today, but they expect 40-50% of their molds to incorporate it in the future. (Those are median percentages for the group.) Even today's non-users of CAE predict that they'll be using it on 30% of their molds in the future.
But like any new technology--and anything having to do with computers, in particular--CAE mold analysis comes in for some criticism. It seems to perform better in some circumstances than others, and users would like to see some improvements. Both sides of user experience are covered in this report.
By the end of February we had received 696 responses, 403 of them (58%) from companies that had some experience, either direct or indirect, with computerized mold analysis; and 368 of them (53%) claiming to be fairly knowledgeable about it. A total of 314 respondents said both that they were knowledgeable about mold analysis and that their firms had experience with it. This is the "core group" for much of the survey analysis that follows. We telephoned a number of them for further clarification of important issues.
That 58% of respondents come from firms that are CAE "users"--whether they perform mold analysis themselves or have it done by an outside firm--no doubt exaggerates the true extent of CAE penetration in the injection molding field. Just a year ago, PLASTICS TECHNOLOGY surveyed 380 custom injection molders and found that only 16% of them made use of mold analysis. At best, 50% of the top-flight, "world-class" custom molders took advantage of CAE (see PT, April '91, p. 96). If one assumes similar proportions among PT readers and plants, this earlier research suggests that the current survey may have obtained a 13% response rate from "user" firms, far higher than the overall response rate of 3.5%. Although even "non-users" were asked to return the questionnaires, it is likely that persons uninvolved with the technology were less inclined to respond.
HOW DO THEY USE CAE?
Respondents obtained access to mold analysis by a number of means. Just under half of them actually have performed the analysis in-house. The majority have sought outside help--from expert consultants who specialize in CAE services, resin suppliers, and moldmakers (in that order of frequency), and sometimes from the respondent's customer.
When all respondents, users and non-users alike, were asked how they would prefer to obtain mold analysis in the future, around half said they'd prefer to do it in-house. They said that would allow them fastest access and unlimited freedom to test out various solutions and "what-if" experiments. Respondents also cited the importance of gaining in-house experience with CAE (which would come only from doing it themselves), the desire to custom-tailor the analysis to their own and their customers' needs, and concern over maintaining confidentiality of results. Quite a few also considered in-house CAE to be most cost-effective in the long run.
On the other hand, many respondents said they could not justify the capital and personnel resources needed for in-house mold analysis. They considered it more cost-effective to seek outside services on the occasions that they needed it. Some hoped or expected to get mold analysis for free as a service from resin suppliers, whom they thought would be best qualified anyway, since they know their materials best. (The same was said by some about moldmakers and tooling.)
Perhaps not surprisingly, those who have used mold analysis in the past are much more eager than inexperienced persons to do it themselves in the future (57% vs. 31%). Experienced CAE users, if seeking outside help, would go to a consultant first, resin supplier second, and a moldmaker as a distant third. Non-users would pick moldmakers over resin suppliers, however. Among those who would pursue in-house CAE, outright lease or purchase of software is preferred by a large margin over modem access to a time-sharing network.
Among the CAE users, a surprising 60% have experience (directly or indirectly) with more than one brand of mold-analysis software; 27% have experience with three or more. Table 1 shows which software suppliers they are familiar with. (Among the "others" cited are Cisigraph, Techanalysis, Computer Resources, ICAM Technologies, Injection Molding Industries, and the University of Wisconsin-Madison.) Note that the relative frequencies of mentions in the table may not accurately reflect the relative number of copies of different CAE software in use. Many respondents were merely citing the type of software used by their moldmaker, resin supplier or consultant.
Figure 2 shows that engineering workstations are by far the most popular type of computer hardware used for mold analysis today. PCs are a distant number two.
Figure 3 shows how often the various types of CAE software features are used (only actual users' answers are counted here). Respondents were asked to answer Most of the Time, Sometimes, Rarely or Never for each feature, and their answers were converted to a numerical scale from 3 to 0 respectively, in order to arrive at an average frequency. Clearly, mold-filling simulation is used significantly more often than cooling simulation. Three specific aspects of filling simulation--locating gates, sizing and balancing runners, and predicting or optimizing weld lines and gas traps--are also fairly popular uses. Understandably, the newer and/or less well known CAE software functions of cycle and cost optimization, shrinkage and tolerance analysis, warpage and stress analysis, and material selection for processability scored far lower on the usage scale.
Figure 4 addresses an extremely important, but often overlooked, question of when to perform mold analysis. All but a small fraction of CAE users know that it should be done while the tool is being designed and before any steel is cut. Unfortunately, in the real world, mold analysis is often not performed until mold construction is under way, when the flexibility to make changes may be limited. And just as often, companies wait until they encounter a molding difficulty or part-quality problem before they seek help from CAE (see, for example, PT, Oct. '91, p. 23).
HOW WELL DOES IT WORK?
Figure 5 shows the answers to the question, "Do you believe that computerized mold analysis helps design molds that will make parts 'right the first time'?" Users overwhelmingly answered Yes. As the figure shows, respondents from firms that have not yet tried mold analysis are also convinced it works.
Even the negative answers were qualified: Several users commented that mold-analysis is only a tool that helps get you closer to a right answer, but does not guarantee a bullseye on the first shot. Respondents noted that success depends a lot on the knowledge and skill of the CAE user, especially because the software programs must make simplifying assumptions that cannot account for the wide range of variables in real-life applications.
"Even though you have an analysis done, by whatever software you want to use, it still takes a skillful interpreter to see what is there," says Timothy Erwin, senior tooling engineer at Pitney Bowes, Inc., Danbury, Conn. "Unless you go back through the data, you can make a mistake. You could easily design in unreasonable machine parameters, because the computer doesn't tell you anything about molding machines."
In the right hands, mold analysis can indeed achieve stunning results. Witness the crash program to develop the new IBM laptop computer (see PT, May '91, p. 210). The Plastics Technology Center of Lexmark International, Inc. in Lexington, Ky., and Leap Technologies of Otsego, Mich., jointly developed 25 major plastic components and 85 customized buttons without making a single prototype mold. With the aid of mold analysis, they went directly from "prototyping on the computer" to hard tooling, molding production parts within six months of initiating the project.
Kuss Filtration of Findlay, Ohio, had a similar success on its first experience with mold analysis. The company molded a TP polyester fuel filter for Robert Bosch Corp., to go on the new GM Saturn cars. The complex part is round with a flexible lip on the outer diameter and a stack on one side and is insert molded to precise dimensional requirements. After building a prototype mold, Kuss had mold analysis performed to determine shrinkage and optimum gate location. According to project engineer Edward J. Fitzpatrick, the analysis helped Kuss to come up with "a pretty complex gate design" for a four-cavity production mold. The parts came out right the first time, Fitzpatrick reports. In his experience, most tools must be reworked up to three times before they're ready for production. "With mold analysis, we bypassed a lot of the problems that we could have had. We were the first one to supply approved production parts to Bosch for their part of the Saturn."
IS IT ALWAYS A 'MUST'?
The next question we asked was in what circumstances is mold analysis a cost-effective tool. The answers of the "knowledgeable user" group appear in Fig. 6. Although 37% of respondents recommend it for most or all molds, the largest area of agreement is that CAE is cost-effective mainly with more complex part shapes. Some respondents commented that it helps with unbalanced cavity layouts and family molds.
"Mold filling analysis doesn't have to be used in every case, just applications in which experience cannot predict the results," says Harvy Lahe, supervisor of R&D at Husky Injection Molding Systems Ltd., Bolton, Ontario. "People should look at the design and estimate whether it is a risky situation or could be a potential problem."
CAN YOU BELIEVE THE COMPUTER?
Respondents were then asked to rate the accuracy of mold-analysis predictions on a four-point scale from Very Poor to Excellent. The average ratings of the knowledgeable users appear in Fig. 7. None of the categories achieves a perfect 4.0 (Excellent) rating, as that would require nearly unanimous agreement among all respondents. However, it is evident that respondents think mold analysis does a very good job on simple parts with loose tolerances in single-cavity molds, particularly if using amorphous resins. However, with complex shapes, tight tolerances, very thin or very thick walls, multiple cavities, and crystalline resins, accuracy of the results drops down into the "pretty good" range. The lowest accuracy rating was for predicting behavior filled and reinforced materials.
TABLE 1--FLOW-ANALYSIS SOFTWARE WITH WHICH RESPONDENTS HAVE EXPERIENCE(a) Software Percent of Supplier Respondents Moldflow 66% Advanced CAE Technology 31% CAE Services 19% SDRC 16% Graftek 9% Plastics & Computer Inc. 8% Others 5% Don't Know 8% a Experience may be first-hand or via a consultant, moldmaker or resin supplier.
"One thing flow analysis has not come to grips with yet is orientation effects of glass. It's important when you get into some of the more structural applications, where you want to make sure you get glass in the right areas of the part to ensure structural integrity," notes Clive Copsey, design manager at BFGoodrich's Applications Engineering Design Laboratory in Avon Lake, Ohio. Some brand-new software is beginning to deal with this phenomenon. However, as noted by David Kazmer, design engineer at Advanced Design Engineering, GE Plastics, Pittsfield, Mass., "Some fiber effects can only be modeled in three dimensions," whereas today's CAE software really works in a sequence of 2-D layers. In response, software suppliers foresee future versions providing a limited number of "hybrid" flow elements to help model 3-D fiber or melt orientation effects in critical areas of the part.
Reports Husky's Lahe, "In extreme thin-wall applications, we find that the programs can give some unrealistic results. To ensure accurate results for thin-wall applications, it is important that the software is capable of correctly predicting shear thinning and solidified layer build-up."
Some resin suppliers see limitations in the way today's software models the behavior of crystalline thermoplastics. Says Timothy Brasel, design engineering manager at Himont Inc. in Wilmington, Del., "With semicrystalline resins such as polypropylene, the commercial software does not do a very good job of handling the way they crystallize. It does a very simplistic job of estimating when the material is going to solidify. That becomes very important in thin walls."
TABLE 2--JOB FUNCTIONS OF RESPONDENTS Primary Mold Analysis Job Function Users Non-Users Production or Mfg. Eng. 46% 42% Gen'l & Corp. Management 24% 38% Res. & Devel. 19% 10% QC/QA 2% 3% Purchasing 1% 1% Other (Incl. Design) 7% 6%
Adds GE's Kazmer, "I would question the accuracy of modeling solidification, crystallization, fiber orientation and final part deformation. In a crystalline polymer, you get a complex micro-structure forming. Warpage is a function of pressure, shear rate, cooling rate, and the crystallization kinetics of the resin. Similarly, you utilize the velocity of the material to try to get estimates of the orientation of the material. Nobody models all that." Software suppliers reply that some of those effects are accounted for in current software, and more will be modeled in new versions now at the precommercial development stage. Software and resin suppliers are even working together to develop suitably accurate flow analysis for the exotic behavior of liquid-crystal polymers.
Another material supplier's critique comes from Robert Goodman, injection molding supervisor at Phillips 66 Co., Bartlesville, Okla. "Most of my experience has been in PPS material and thin-wall parts like connectors. In those applications, where the flow path is changing constantly around a lot of holes, the software has a tendency to overrate the pressure required to fill very intricate components."
Figure 8 shows how well satisfied the knowledgeable users are with the accuracy and reliability of specific software functions. Overall mold-filling prediction gets a high rating, as do weld-line and gas-trap prediction. Users cite only middling satisfaction with cooling predictions, and are even less satisfied with predictions of cycle time, injection pressure or clamp tonnage, shrinkage dimensions and tolerances, and warpage or stress. To be fair, the last two are relatively new additions to CAE software packages and among the least well known. "Warpage is a new application and they've got quite a way to go," says Goodman of Phillips 66. Copsey of BFGoodrich agrees: "Warpage prediction is not there yet. While it can indicate the trends, the absolute numbers are way off."
TABLE 3--PRIMARY ACTIVITY AT THIS LOCATION Mold Analysis Users Non-Users Plastics Processing 56% 61% Moldmaking 8% 6% Other Mfg. 16% 22% Design & Engineering 15% 6% Res. & Devel. 6% 2% Corp. HQ 2% 3%
Adds Brasel of Himont, "Shrinkage and warpage analysis needs more material data than is generally available from resin suppliers. PvT (pressure-volume-temperature) databases are very limited. We only have a few of our resins characterized for pvT. It's a fairly specialized test, not something that is commonly done, it's expensive, and we can't do it in house."
Figure 9 looks at other aspects of user satisfaction or dissatisfaction. The ratings indicate that users are, on average, relatively neutral in their assessment of software speed and ease of use, the breadth of materials databases, and customer training and support. They appear to have stronger reservations about software compatibility with other CAD or finite-element analysis (FEA) programs--and, of course, about software cost.
Lower cost was, in fact, the most common answer to the question of how mold analysis could be made more useful or valuable. Less expensive CAE software, respondents said, would be accessible to more potential users. Making software simpler and easier to use was the next most common recommendation, followed by faster-running programs (which is also a function of the computer hardware used), more seamless integration with other CAD/CAE programs, and improved material data.
TABLE 4--PERSONAL FAMILIARITY WITH MOLD ANALYSIS At Firms That Use It Do Not Use Personally Experienced 41% 2% Familiar in Some Detail 36% 16% Generally Aware 22% 56% Totally Unfamiliar 1% 25%
As for CAD integration, the biggest complaint of some of the more technically sophisticated users is that they cannot "import" existing solid models of a part design into mold-analysis software, which treats the flow essentially as if it were over a 2-D surface. (Several surfaces can be stacked up in layers to give a picture of what happens through the thickness of a part.) "It would be nice if they could iron out some of the interfaces between CAD and the analysis package, so you could build a solid model and use it for your analysis without changing the model," comments Michael Brazee, CAE engineer at DTM Products, Inc., a custom molder in Boulder, Colo. "What's needed next is solid modeling in the flow-analysis software," agrees Goodman of Phillips 66. "Right now, there's no way that you can take every part and translate it into a flat surface. It's not accurate and really doesn't represent what's going on in the part."
Some CAE software suppliers reply that this would add "an order of magnitude" to computer processing time or to the computer power required. But at least two commercial CAD software packages do allow automatic importation of solid models into finite-element meshing programs in preparation for mold analysis. And even users of other programs, suppliers say, can create their CAD solid models in such a way as to make them much easier to transport into CAE analysis.
Another commonly mentioned criticism was the need for more and better materials data, which are critical to accurate mold-analysis predictions. "Materials databases aren't updated frequently enough, and we are continually having to add things that are not there," is a typical complaint. CAE software suppliers do continually expand and update their software databases, though maybe not fast enough or completely enough to satisfy everyone. Explains Robert McIntyre, development leader at the Materials Engineering Center of Dow Plastics in Midland, Mich., "We have several hundred commercial thermoplastic product designations. It is impractical and unnecessary to completely characterize each and every grade for injection molding simulation. Frequently, a mold-filling analysis is to be performed using a specific grade for which data do not exist. Experienced CAE users with materials knowledge often use available data for another grade in the same product family after comparing rheology to ensure that its flow behavior is representative of the specific grade."
Jerry Austin, president of Glacier Design, Inc., a CAE consulting firm in Baraboo, Wis., agrees that "families of materials tend to behave very similarly, and it depends on what you are trying to establish in a tool as to how critical the rheology numbers are." He explains that it may suffice to use a fairly similar material for the purposes of balancing a multicavity tool. But in attempting to determine how thin a wall section can be made and still fill adequately, "it is very important to have good rheology data on the exact material you want to run."
Says Husky's Lahe, "To get accurate results, you need accurate data on the exact grade of resin you intend to mold. In comparisons we have made to actual results, we have found our accuracy to be within 10%. It's hard to say how much of that 10% is due to inaccurate material data, but we believe it plays a role."
A half-dozen CAE users said they'd like to see comparative tests of the accuracy of different analysis programs. An example is Himont's Brasel: "There needs to be more published benchmark validations showing the accuracy of various software predictions. It would be nice to see test data on more than just a plaque, where they have run the filling analysis, have molded parts, and then correlated the two and shown which packages do the best job in certain geometries or resins."
HOW BIG ARE THE BENEFITS?
Knowledgeable CAE users are firmly agreed that mold analysis saves time and money in getting a mold built correctly. The median savings, they say, is 20%. Users also report that the median number of times they have to get a new tool reworked before it runs right in production is two times with mold analysis and three times without it. But, as one user comments, it's not just the number of reworks that counts, but also how extensive the reworking must be. With mold analysis, he said, the adjustments are more minor.
As was noted above, such time savings are of no small importance in today's competitive environment. Says BFGoodrich's Copsey, "The benefit is in decreased time to market. We're seeing a lot of people trying to compress three- to four-year time-frames down to two to three years or less."
A substantial majority of knowledgeable users also agrees that mold analysis helps save time in getting a new mold to run correctly once it's built, helps determine the correct molding conditions or "processing window" for a mold, and helps maximize product quality. At DTM Products, mold analysis "is integrated into our quality management system," says president Robert J. Grubb. "It's done on every single mold that we build."
Using mold analysis to help define the process conditions for running a mold may become an important trend, in the view of some software suppliers and other industry experts. But resin suppliers warn of the potential pitfalls of designing a mold to run under a set of conditions that may not be duplicated in the molding plant. GE's Kazmer says, "You can say the melt temperature is 570 F and the mold temperature is constant at 120 F, but when you get to the actual molding machine, how well can you control your process? When you understand those shortcomings, then you are in a good position to utilize CAE technology."
As with every aspect of mold analysis, good communications are key. John Bozzelli, Dow Plastics development specialist, warns that mold fill time is a critically important parameter in mold analysis. But the information on what fill time was used for the analysis often gets "lost" by the time the mold is built and put into production. "The people who operate the mold never hear what that 'ideal' fill time was supposed to be."
Around half the CAE users find that mold analysis also helps reduce production of scrap parts and helps reduce wasted material consumption in runners and parts. Median savings in scrap and rejects has been 10% with mold analysis, users say. Copsey of Goodrich says the computer has come in handy for a lot of customers that are trying to eliminate "overdesigned" parts and reduce wall thickness to save material while maintaining moldability.
Survey respondents report a median saving of 10% in cycle times with the aid of mold analysis. But in Fig. 10, relatively few users acknowledge that mold analysis helps either to predict cycle time or to achieve the shortest cycle.
When users were asked which benefits from mold analysis were most important to their company, their priorities were clear:
1. Saving time and money in tool development.
2. Maximizing product quality.
3. Saving time in getting a mold running right once it's in the press.
WHAT MAKES IT WORK
Figure 11 shows how the knowledgeable CAE users rate the importance of various factors in getting the most value out of mold analysis. On a three-point scale (Little or No Importance, Somewhat Important, or Very Important), the two that come out on top are making sure that mold analysis is done before the tool is built, and good communications between CAE analyst, part and mold designer, and molding engineers. CAE consultant George Markus of Advanced Plastics Design, Inc., Fairfield, Ohio, says these two factors cannot be overemphasized, and ignoring them is most often responsible for unsatisfactory results.
A number of the negative comments on mold analysis received in this survey came from molders who were discouraged by one bad experience. Typically, they had worked with a consultant or other outside source, but evidently did not work closely enough with that source to effectively communicate their needs. Consequently, the consultant came up with a proposed solution that was unworkable but could easily have been avoided if there had been more discussion at the outset.
As Fig. 11 indicates, it's vitally important that someone with intimate knowledge of the injection molding process be brought into the mold-analysis effort. Says DTM's Grubb of his own CAE team, "The people that operate it and utilize the result are molders--not some ivory-tower engineers who never had the gut-wrenching experience of having to answer to a manager for a shift's worth of defective parts."
Peter Fukuyama, director of product development for O'Sullivan Corp.'s Gulfstream Trim Div. in Winchester, Va., cautions that CAE users "have to have a good understanding of what the molding machine can and cannot do, and include that within their model when they are going through the iterations. Most people, when they incorporate an injection profile into their filling simulation, don't mirror the injection profile of a traditional hydraulic injection molding machine. They might put in a flat injection profile, or they might put in a straight line with a very firm slope. That's not the real world. The closer your process parameters are to reality, the closer you'll get to accuracy."
Another example he cites of the need to input "real-world" conditions when doing mold analysis concerns runnerless hardware. "If you are using a hydraulic shutoff gate valve, you have a secondary restriction within it that creates a larger pressure drop than you would typically have with a straight orifice or conventional hot-runner system. Consequently, the CAE prediction is not as close to the real world as it potentially could be. If you modify your input conditions based on the orifice area that you really have, you get very close."
Andrew Doll, design engineer at General Industries in Elyria, Ohio, offers another example: "Flow packages, in their current stage of development, won't predict core shift of large parts with long, narrow cores. The flow analysis will show you the unequal pressures around the core, but then you've got to manually calculate the deflection of the steel."
Doll says the best solution is a team approach. "Once we have built a computer model, we involve process engineers, who contribute their on-the-job molding knowledge of typical fill times and mold temperature expectations. That information gives us what we need to rerun the model. Frequently we run 10, 15 mold-analysis iterations per part. Our ultimate determinations are based on sharing those iterations with everyone on the loop from design through manufacturing."
Communication with the moldmaker is just as important. Says Austin of Glacier Design, "If you've got a moldmaker who keeps the wall thickness within careful parameters of what it is supposed to be, then the end result will be much closer to what you predicted with the computer."
As Fig. 11 suggests, good understanding of the computer software itself is equally essential. O'Sullivan's Fukuyama points out that some early-generation software may have an insufficient number of finite-element "nodes" available to adequately "mesh" a very complex part, "If you have a very complex part and have too few data points, you are going to get a very poor prediction."
Who Answered the Survey
Responses came in from 42 states, Puerto Rico, Canada and Mexico (only 21 responses, or 3% of the total, came from the last two). The pattern of responses was quite representative of the overall geographic distribution of our injection molding and mold-design readers. The distribution of job functions among respondents was also highly representative (see Table 2). Note that the "user" group included about twice as many R&D personnel as the "non-user" respondents, balanced by about 37% fewer persons in general and corporate management.
Table 3 shows the primary activity at the respondent's location. It probably stands to reason that the CAE "user" plants included a greater percentage of design/engineering and R&D locations than the "non-user" group.
Although plastics processing was cited by only 50-60% of respondents in Table 2 as the primary activity at their location, 83% of respondents said that some injection molding was performed where they work, and 10% said it was performed at another location of the same firm. That leaves only 7% of respondents from firms that do no injection molding at all.
What's more, 58% of respondents' firms do some moldmaking--51% for their own use and 18% for others' use; some do both. (It's interesting to note that, in this sample, 59% of the injection molders also make molds, and 84% of the custom moldmakers also do some injection molding.)
The custom/captive ratio among injection molding firms was almost exactly 50/50 for the survey sample--with no difference between "user" and "non-user" groups--whereas 60/40 is closer to the true percentage. This suggests that mold analysis may have penetrated deeper into captive molding, which includes many large manufacturers in the appliance, automotive, medical, packaging, and other fields that were represented in the survey. These large manufacturers obviously have more financial clout to invest in CAE than most custom molders.
Supporting the hypothesis that larger firms are more likely to invest in mold analysis, both the median and average number of injection machines among responding "users" (22 and 35.8 machines, respectively) are about double those for "non-users." They are also significantly larger than the average plant in PLASTICS TECHNOLOGY's Injection Molding Census, which has 11.4 machines. (The median in the Census is somewhere in the 1-10 machine range.)
As Table 4 shows, respondents from firms using mold analysis have much greater personal familiarity with the technology than do respondents from "non-user" firms. Among the core group of 314 "knowledgeable user" respondents, 53% have individual hands-on experience with mold analysis.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||includes related article; computer-aided mold analysis|
|Author:||De Gaspari, John|
|Article Type:||Cover Story|
|Date:||Apr 1, 1992|
|Previous Article:||New styrenics toughen & compatibilize.|
|Next Article:||Low-tech/no-tech parts-handling solutions.|