A new way to evaluate injection machine performance.

A New Way to Evaluate Injection Machine Performance

Computerized SPC monitoring can make for more informed buying decisions and tell when machines need maintenance or replacement.

What's the most important factor that will affect plastics processing in the 1990s? Information! More and more processors will learn to take advantage of new computer technologies that can make them better informed about the performance of their machines and processes. This will lead to more sophisticated, unemotional decisions on which machines to buy for a particular task. This will give them unequivocal and up-to-the-minute information on the performance capabilities of their machines in place, so they can better judge when they need maintenance or deserve to be replaced. And the end result of better information will be increased productivity and better product quality.

This article illustrates the practical value and importance of the information that is now electronically available to processors. It is based on real-world production data from 498 injection machines of all makes and vintages, at 38 molding plants that are using commercially available hardware and software to continuously monitor their machines' performance. Over more than three years, millions of cycles' worth of data from these machines and plants have been collected and analyzed by standard statistical process control (SPC) methods.

What does all this show? As we shall see in the detailed discussion below, the popular wisdom about the relative performance of U.S.-made machines relative to those imported from Europe or the Far East requires some adjustment to reality.

Second, overall observations indicate that if all injection molding machines performed as well as the best, we could see an 18% to 25% increase in productivity in the average plant. Contrary to some recent surveys that would have us believe that approximately 60% of processing problems come from human interruptions, this multi-plant study shows that the vast majority of interruptions come from the process drifting out of specification owing to misbehavior of the molding press.

Third, the quality of today's injection machines--judged by statistical repeatability of performance--is not up to the standard of some other industrial machinery, such as for metal cutting or semiconductor manufacturing. An acceptance criterion for customers of the latter machines is a process capability index, or Cpk, of 1.3 or greater for every process parameter. Today's injection molding machines might fulfill that criterion in one or two parameters, but will almost certainly fall short in some other critical areas.

However, I believe that a standard of repeatability comparable to the most stable and precise industrial manufacturing equipment is not an unattainable goal for an injection press. I believe that as the processor starts demanding this level of quality, he can get it, as the technology is here. He must, however, be willing to pay an extra $3000 to $5000 per machine to get it. When we consider the long-term efficiency that will result from this investment, over the lifetime of the machine, it is indeed a small price to pay.

USING A COMMON `YARDSTICK'

The data disclosed in this article represent the performance of over 300 new machines, of 1988 or '89 vintage, except for a few 1979 models, whose manufacturer has not changed its technology since that date. All these data were obtained during operation on automatic cycle, so as to provide a clear understanding of the capability of the machine itself. For the same reason, no data have been included where manual process adjustments were made. The data were all recorded automatically from machines linked to Hunkar Laboratories CIM-1 networks, to eliminate human intervention in the recordkeeping process.

To evaluate the performance of all these machines, the following five parameters were used, as these are the ones most critical to quality and productivity in "general" molding applications:

* Cycle time: This parameter can be of some importance to product quality when processing certain materials. More generally, cycle repeatability has a major impact on productivity. We have seen cycle times drift considerably over the performance standard, with big consequences on the real cost of machine utilization.

* Injection time: This parameter is important in critical jobs where filing conditions affect quality of the molded part. Large, thin-wall parts, and large, multicavity molds are two examples.

* Plasticating time: This indicates the performance of the screw, its drive mechanism, and check ring. It is critical to processors who must have consistent melt conditions; it may also be significant to those whose overall cycle time is affected by variations in plastication time.

* Hold pressure: This parameter is supremely important, even though apparently not well understood, as it affects the degree of packing, which has a direct bearing on dimensional tolerance and part weight.

* Backpressure: Some experts indicate that 80-95% of the heat imparted to the melt comes from conversion of mechanical shear energy of the screw, which is a function of rpm and backpressure. Thus, backpressure has a major impact on melt quality and consistency.

Table 1 shows the performance standards for each of the above parameters that was used to evaluate the machines studied. The process capability indexes (Cpk's) reported here are based on comparing the machines' actual performance to these standards. While this standard may seem strict, it is based on what the study shows modern machines should be able to do. Starting with the average repeatability measured for each parameter on the 300 + machines studied, the tolerance standard was set wide enough that two-thirds of the machines in the study could meet that standard.

CPK AS A MEASURE OF QUALITY

The process capability index, or Cpk, has long been used by quality-control people and statisticians to evaluate manufacturing processes. Cpk is handy because it reduces a chart full of data points to a single number, which can readily be compared with another Cpk to determine which is the better, or more capable, process. As explained in the accompanying box, Cpk is also important because it correlates with reject rates. This means that any process parameter that has a low Cpk value may be responsible for producing reject parts. Obviously, depending on the part molded, certain parameters are more critical than others.

I believe that by the middle of this decade, no molder will survive unless each process parameter in his molding machine is providing a Cpk of 1.0 or greater. Some processors are, in fact, operating at this standard of quality most of the time, and they are in a far better position to provide high-quality molded parts with higher productivity than their competitors who are operating at a lower standard of quality.

COMPARISONS OF MACHINE PERFORMANCE

Table 2 reports statistical analysis of the production performance of 280 machines in the 150-310 range, grouped by whether they were manufactured in North America, Europe or Asia. Each row of data reports the average performance of groups of up to 15 presses of like tonnage (indicated by the last three digits) and vintage (first two digits), and built by the same OEM. The only exceptions are lines 3 and 4, which represent single machines, for reasons discussed below. While it was not possible to obtain a sample of every injection machine nameplate, most of the well known makes are represented here.

Data shown are average Cpk values for the five critical injection parameters noted above, as well as an overall average for the machine category. Cpk's were calculated by comparing the spread of actual data (no fewer than 100 shots from each machine in the study) with USL's and LSL's based on the "performance yardstick" shown in Table 1 and discussed above.

Looking at the average performance line for all three groups, it is apparent that the North American manufacturers (from which 167 machines are represented in the sample) have made the greatest effort to provide consistent cycle times, as well as good performance for injection, plastication and backpressure control. The logical conclusion is that North American machine builders perceive that their market puts heavy emphasis on cycle-time consistency and on all-around performance. Note, however, that hold-pressure control is not performing too well.

It's interesting to note how different the performance was on the individual machines represented in lines 3 and 4, even though these two machines were made by the same manufacturer, having consecutive serial numbers, running identical materials and molds, and installed side by side in the same plant. This suggests that either variations in manufacture of the machines or in their "tuning" by the manufacturer made the difference. To the processor, this means he cannot blindly assume identical performance of all machines coming from the same source, even if of the same model. Rather, he would be wiser to conduct acceptance testing of each machine (the methods used in this study could serve as a model).

Furthermore, we have found that brand-new machinery will change in performance after two months' running time, necessitating recalibration or possibly the extension of the machine supplier's performance "guarantee" until the machine's performance is requalified sometime after initial installation.

Turning now to the Asian-built machines, the results from the 81 presses in the sample tend to dispel the myth that machines built in that part of the world necessarily perform better than machinery made in North America. To be sure, there are some excellent machines built in that part of the world, which provide healthy competition for machinery OEM's on this continent.

From the average performance of this machine group, it is evident that the Asian producers seem to concentrate their engineering effort on consistent plasticating performance--i.e., producing good melt quality--because this is where the Cpk's are highest. Plasticating time and backpressure seem to be the focus of their attention, though sacrifies possibly have been made in design aspects that relate to cycle-time consistency.

As for European-made machinery (32 presses in the sample), the concentration of engineering efforts seem to be on plastication and packing. Note that the hold-pressure control capability of the European machines seems to be the highest of all three groups.

Line number 10 shows the average production performance of a group of identical brand-new machines after only two month's use. Their performance had also been analyzed while they were still in the machine builder's plant. In that pre-shipment test, we sensed the possibility that the machines had been "super-tuned" in order to make an especially good impression. After two months in production at the customer's site, those same machines were retested. As shown in Table 3, the Cpk values showed significant degradation of injection-time consistency, hold-pressure control, and backpressure control. The overall average Cpk decreased from 2.03 to 0.93.

We have seen this phenomenon to be characteristic of most proportional control designs. As proportional hydraulic valves "wear in" during production, their consistency and repeatability will change. Thus, we would advise the processor to keep an eye on changes in his machine performance and re-analyze it periodically after it has been in production for some time.

Although there are no brand names attached to the data in this article, this study showed unquestionably that certain European machines that command a higher price do not necessarily justify it with superior performance. In our population sample, these machines did not produce more consistently than lower-reputation European or even non-European models. To the processor, this should suggest the value of investigating and verifying manufacturers' claims himself, to obtain the sort of objective data discussed here.

PERFORMANCE BY TONNAGE RANGE

Table 4 groups machine data in three tonnage ranges. Each row represents an average of a group of similar machines of the same make. The analysis indicates that certain traits seem to predominate according to machine size.

* Under 100 tons: The concentration of design effort was obviously placed on optimizing the cycle repeatability of the machine. Hold-pressure and backpressure consistency is dismal in the small-machine class. Our opinion is that the builder simply has not found the cost of good hydraulic control to be justifiable in this area, in view of the overall price of the machine.

Some machine marketers and some molders advocate the idea of using several small machines with smaller molds in place of one large machine with a high-cavitation mold. We feel that the evidence shows that only if small-machine builders strengthen the pressure-control performance of their presses will they be truly very competitive with large machines.

* 150 to 250 tons: We find that the manufacturers in this tonnage range have endeavored to provide good all-around machine performance.

* 300 to 610 tons: Suppliers of large machines evidently have concentrated their design effort on injection performance and on maintaining melt quality. This seems to be appropriate to the large parts for which big machines are typically used.

However, it is too bad that cycle-time repeatability in large machines is just not there. Possibly this was overlooked as an important factor, since most large machines operate in a semi-automatic mode--hence, the processor has always attributed cycle-time variations to the operator who opens and closes the gate. But we forecast that in the 1990s, a significant proportion of large machines will be operating in automatic mode. At that time, it will be supremely important to achieve good machine cycle consistency.

All hydraulic designers agree that it is more difficult to make large-machine systems less susceptible to hydraulic viscosity changes than on small machines. Yet the technology exists to compensate for oil viscosity. So it should be possible to build a highly repeatable machine even in a large tonnage range.

Table 3 shows that the overall average Cpk figure for each size range decreases with increasing tonnage. This means that the larger the machine, the higher the probability of producing reject parts. That again indicates that extra investment is worthwhile to make large machines perform consistently, particularly in view of the cost associated with running these size ranges.

NEW CONTROLS ON OLD PRESSES

Table 5 shows what performance can be achieved when older machines are retrofitted with a "true" closed-loop control system. Item 25 is a single 1982, 75-ton machine; item 26 shows averages for 14 machines, all 1987 375-tonners; and item 27 is a single 1971, 476-ton model.

Processors should note that most machinery suppliers today claim to have closed-loop controls. For the vast majority, this means that the loop is closed between the moving member of the control valve and the command signal going to the control valve itself. If the command signal calls for the valve to be 50% open, for example, the control loop makes sure only that the valve is open 50%. Generally, these controls do not measure the actual injection speed or hold pressure or backpressure and provide direct closed-loop control from these.

True closed-loop control, using what is termed a master control loop, means that the hydraulic valve is commanded automatically to perform the final parameter of interest. This means that injection speed, for example, is measured and the valve will be opened to whatever degree is needed to provide the injection speed required.

Table 4 shows that retrofit closed-loop controls of the latter type can make older machines compete very well with brand-new machines. In some respects, they are even superior.

However, note that in some cases cycle-time consistency cannot be improved, since retrofit controls generally do not control the clamp end of the machine. To the processor this means that he should consider investing in retrofit controls only for a machine that is basically in good shape on the clamp end.

This argument is borne out by the performance of Item 27 in Table 4, a 1971 machine whose clamp just did not operate consistently and whose extruder rpm deviated significantly because of a worn extruder drive. We have a very large database that indicates that if the clamp end of a machine is rebuilt or repaired, and the extruder drive is in good condition, the investment in retrofit master closed-loop control will make the older machine competitive in a world-class sense.

QUALITATIVE ANALYSIS OF PERFORMANCE

Up to this point, we have discussed quality in terms of single numbers (Cpk). This has the advantage of removing the subjectivity that results when data are analyzed in a qualitative sense. Still, it is easier for most people to visualize machine performance in terms of a graph of actual data, rather than a statistical abstraction from those data. And graphic analysis certainly adds another dimension to the understanding of machine performance.

There are two main types of charts used for qualitative analysis. One of these is the "Z-chart," shown in Figs. 1A to 5A. This is a plot of individual data points for each cycle along the horizontal time axis, taken from a typical machine of the group shown in row number 2 in Table 2. This is a 1988 model of a very popular make, having all the latest technology.

Superimposed on the data in the Z-charts shown here are the nominal setpoints and USL and LSL values around those setpoints, based on the standard "yardstick" criteria in Table 1. Thus, all parameter data that fall outside those limits very likely indicate reject parts. (This suggests how an automatic quality-control system, with 100% good/bad parts discrimination, can be implemented from the same computer monitoring system that produces the charts.)

The second predominant type of chart for qualitative machine analysis is the "X-bar and R chart." Figures 1B to 5B are X-bar and R charts corresponding to the data in Figs. 1A to 5A. The upper portion is the X-bar chart, which plots average values from sample subgroups of a certain size, as well as the average of those averages, and the [+ or -] 3-sigma UCL and LCL around that overall average. The lower chart is the "R" or Range chart, showing the difference between the highest and lowest values in the sample subgroups used to plot X-bar values above. Again the average range and [+ or -] 3-sigma values are also shown.

Figures 6-10 show an example of a "good" machine, in terms of repeatability performance. These data are for one typical machine in the group shown in row 5 of Table 2, a 1979 model. This shows that indeed, certain machines can perform consistently, even after many years of production, according to the "yardstick" standard in Table 1. A processor would not be unreasonable to demand this quality of performance from all machine suppliers, since it can be provided by at least one. We have analyzed several machines of this particular brand and have found them to be fairly represented by these data. (Please do not call for the name of the supplier.)

Figures 11-15 show the performance of a 1988, 610-ton machine. It is a rare occasion to see this size machine operating in an automatic mode; therefore, our data provide an unusual insight into the capabilities of a large hydraulic system.

This machine was started up around 7:00 a.m. It is clearly evident that an hour and a half later the machine was still drifting as it was warming up. As oil viscosity changed, cycle time increased, plasticating time decreased, and hold pressure decreased as well.

How is it possible that hold pressure was decreasing while backpressure was maintained within very tight limits? The probable answer, as noted above, lies in the concentration of the machine builder's effort on controlling backpressure while not doing as well on other parameters. (Or, is it due to the use of low-cost hydraulic control valves?)

Considering the cycle-time drift from 40.65 sec to 41.25 sec, we should ask how much time was wasted due to the instability of this parameter. A 0.6-sec variation may not seem very significant, but at 88 cycles/hr, this adds up to 53 sec lost per hour. This particular processor valued his overall machine-time cost at $125/hr, which puts a value of $1.84 on the 53 sec/hr loss. Based on a 6000-hr annual operating time for the machine, this lost-time cost totals $11,040. That this rather significant conclusion was reached from considering only one parameter on only one machine in the plant suggests the potential value to processors of machine performance analysis. [Table 1 to 5 Omitted] [Figure 1A to 15 Omitted]