Portable analyzers find what ails your process.
When your injection molding process comes down with some puzzling ailment and starts making bad parts, what are you going to do? You could hope processing experience or just plain intuition turns up a cure. You could wait for a house call from your materials or machinery supplier. Or you could solve the problem yourself with some diagnostic help from a portable machine analyzer, or PMA.
These suitcase-sized data-acquisition and monitoring devices hook up to molding machines and auxiliary equipment, gathering the kind of information that helps identify the root causes of processing problems. They display process information, often in real time, as graphical profiles of shot variables. The information can also be viewed as statistical trend charts on a personal computer, which may be incorporated into the analyzer itself. "Portable machine analyzers are like having an electro-cardiogram of your molding process," explains Curt Norby, v.p. of T.G. Branden, maker of the Shotscope PMA.
PMAs can also be employed preventively to check the health of your process. Some sophisticated molders use PMAs during process development and start-up to quickly find the optimal processing conditions for new molds. Other molders have used this powerful new tool to gain the upper hand in dealing with machine vendors. Some of these shrewd molders now require PMA tests to validate the repeatability of new machine performance before signing the check. With one exception, PMAs cost less than $9000, putting the benefits of analysis well within reach of many small molding shops.
Current PMA offerings fall into two camps. Three lower-cost units support a minimalist approach to data acquisition. The other camp is occupied by a more expensive unit with more robust capabilities. Some PMA users consider this to be "overkill" while others claim they couldn't live without the extra channels and higher data-acquisition speeds.
When evaluating a PMA, smart shoppers will consider two important issues hidden beneath the hardware specs - the supplier's monitoring philosophy and the PMA's user-friendliness.
These two issues drew the most attention from GE Plastics' Polymer Processing Development Center in Pittsfield, Mass., during a recent pre-purchase evaluation of three analyzers - the PMA from Hunkar Laboratories, the A5000 Shotscope from T.G. Branden, and the PM 4000 from RJG Technologies. The fourth injection molding PMA, from Nicollet Process Engineering, came on the market too late for this evaluation.
Because all three candidates did a good job of taking the data from the molding machine. GE's evaluation barely touched on the hardware specs. "In my opinion, the performance is equal on these units from a data-collection standpoint," says Greg Tremblay, a GE process-development engineer specializing in data acquisition. "Hardware is hardware as long as it works. As long as it captures the whole molding cycle, the real choice between these units mostly comes down to software." Tremblay goes on to cite the user interface and data-analysis capabilities as the two key software considerations.
After looking at the three units, GE decided to buy several RIG PM 4000s for its traveling technical-service representatives, though the Shotscope came in a "close second." As for the Hunkar unit, Tremblay says the model he tested could not retain data for more than 400-500 hundred cycles, had problems connecting to some machine signals, and exhibited some user-hostile quirks. Tremblay is the first to note however, that Hunkar has since updated its PMA with a new model that addresses all his complaints.
HOW MUCH DATA DO YOU NEED?
Every PMA creates its profile of a shot from a collection of machine signals. Typical analog inputs from transducers include clamp and injection pressure, screw position and speed, and a sampling of process temperatures. Digital signals - taken from the machine controller or, more rarely, from temporary limit switches - show events such as injection forward or mold open.
All the PMA units offer the user some degree of flexibility in assigning inputs to the data-acquisition channels. The list of possible events to capture extends even to auxiliary equipment. For example, Tremblay reports using RJG's unit to measure everything from cooling-tower activity to plant humidity and electrical-power fluctuations. Despite this kind of user-defined monitoring flexibility, PMAs still tend to embody the monitoring philosophy of their manufacturer.
RJG, to take one example, promotes a minimalist strategy, based primarily on cavity-pressure and fill-time monitoring. "Ninety five percent of the time, fill rate and cavity pressure build extremely effective SPC," explains company president Brad Watkins. "We don't recommend doing 32-variable studies. Sometimes you can have too much data and not enough information. Our approach is to look at a maximum of four variables to decide whether a part is good or bad - even if you haven't solved the problem yet." For that reason, RJG's system comes with four to six channels. "More than four variables at a time is kind of overwhelming," says Watkins.
Besides fill rate and cavity pressure, the choice of an additional variable or two to monitor depends on the specific process - for example, cooling-water temperature and flow rate are useful to monitor in PVC molding applications.
Two other units also stick to four channels - not counting temperature inputs - in their standard configurations. T.G. Branden's Shotscope offers four "dynamic" channels for real-time graphical display of 1000 points per waveform. According to Norby, monitoring just these four variables can paint a picture of up to 26 processing parameters. "One variable can tell you about more than one parameter," he says. Likewise, Hunkar's new Smartbox, which first appeared at NPE '94, needs only four channels to categorize machine repeatability for 19 different process parameters, according to marketing director Michael Mueller.
Tremblay, who has performed nearly 20 machine evaluation studies over the past six months, comes down squarely in this minimalist camp. "I started out thinking that I'd monitor the world. But having too much data is not necessarily a good thing," he says. "In almost every study where we helped a customer, we just looked at injection pressure, cavity pressure, screw velocity, and screw position." On the digital side, he adds times for injection forward, holding, plastication, and mold open to his list of key variables.
THE TECHNOPHILE'S CHOICE
Not all PMA vendors and users agree that simpler is better. Molders who want to perform complex studies may want a more elaborate monitoring system. One unit stands out on this score, the PMA 360 from Nicollet Process Engineering. Introduced at NPE '94, it samples up to 100,000 points/sec rather than the 100 points/sec typical of other units. At 32 channels (16 analog and 16 digital), the PMA 360 has the largest standard input capacity of all the machines on the market. It's also priced at $11,500 - which is more than the other standard units.
Whether such muscular hardware is overkill depends on whom you ask. Tremblay, for one, doesn't want the extra information; 3M Co. in St. Paul, Minn., does. In fact, the development of Nicollet's PMA began with a request from 3M molding technology specialist Rob Peterson, who conducts studies he deems too complex for the other three units. "The niche we were looking for was a high-end, multi-channel, high-speed data-acquisition system. We couldn't find something that fit that niche, so we went to Nicollet," he says.
3M uses its Nicollet units for advanced process research and sensor development on molded products ranging from Scotch tape dispensers to eye implants and magneto-optical discs. Even Peterson has yet to use all 32 channels on his PMA 360 at one time. But he doesn't rule out the possibility, given the number of potential variables on every molding machine. "The general-purpose molder can get by with four to six channels and solve most of his problems," says Peterson. "The more sophisticated molder with tight dimensions and downstream automation, who is looking to optimize yields, probably can't get by with less than eight channels and probably needs more."
Molex Corp., one of the world's largest molders of electrical components, which operates 500 injection machines worldwide, values its 16-channel Nicollet unit primarily for its high speed. "Unless you scan at 100,000 cycles/sec, you're not getting the kind of signal you can depend on," says Greg Ford, international molding development manager. "Some units I've seen scan about every 2 sec. The whole world can go by in 2 sec."
As an example of the fleeting events Molex studies, Ford cites the ability to check servo-valve repeatability with the PMA 360, even though the PID program controlling servo operation typically runs in less than 80 millisec. "If you don't collect the data fast enough, you can't see how well the valve is programmed. If you are fast enough, you can back-calculate the constants and see if the PID is well tuned." This kind of sophisticated detective work has allowed Molex to solve a number of difficult problems, such as tracing short-lived injection-pressure anomalies back to servo-valve repositioning errors.
EVALUATING EASE OF USE
After making sure a unit's hardware meshes with your monitoring philosophy, PMA shoppers should ask, "Is it easy for me to use?" Since the answer depends on individual preferences, the best bet is to try before you buy - as Tremblay did for the GE evaluation.
Every PMA vendor reports recent enhancements in user-friendliness. Hunkar, for instance, has switched from the plain-vanilla LCD readout on its older models to a touchscreen interface on its new Smartbox system.
And T.G. Branden has fine-tuned the Shotscope's graphical interface. While users have always been able to access numerical readouts of processing parameters by positioning a cursor on a shot's waveform, the company has now added that ability to the trend screens showing historical SPC graphs. The Shotscope also has another handy feature, a "macro" language that lets users automate common tasks by stringing together actions from lists that appear on the screen. These macros can automatically execute a common calculation, or they can contain different monitoring setups for specific machines - e.g., presses with one-stage or three-stage holding phases.
RJG's Dartnet software now has an enhanced on-line "help" system that wins high marks from Tremblay, as does its menu-driven interface. In Tremblay's view, another helpful feature is the ability to overlay displays of waveforms for different variables - a feature offered to some degree by all the vendors.
Nicollet, meanwhile, is the only vendor to use a Microsoft Windows-based interface - a move that won praise from every user surveyed for this article. Nicollet's industry-standard computing environment makes it easy to transfer data into other software packages, such as spreadsheets or databases - a task Tremblay found difficult with the other units. "Ultimately, I think all the vendors will have to go to Windows in order to compete." he predicts. Like the Shotscope unit, Nicollet's interface also lets users automate common calculations and setups.
LOOKING FOR TROUBLE
A monitoring philosophy may dictate which data to collect, but it doesn't begin to tell you what the data means. To help you track down the root causes of processing snags, all the PMA systems provide a suite of SPC/SQC analysis tools. Each PMA system offers a variety of statistical calculations and related displays, including histograms, x-bar and R-charts, Cpk, and limit determination. Nicollet's SPC software adds the analysis of non-normal data sets, according to marketing director Dan Whealon.
Beyond basic SPC/SQC, the packages all contain additional analysis functions designed to give a simplified view of the process. For example, packages from RJG, Branden, and Nicollet can integrate the area under the waveform. "Knowing the area under the curve can simplify finding what's gone wrong with the process by giving you one number to represent what the machine did," Watkins says. For example, he explains that the area under the injection-pressure curve is a sensitive indicator of any process changes.
Hunkar's "class factor" system like, wise boils down the complex task of comparing and evaluating waveforms into a single meaningful number. As reported previously (PT, April '93, p. 15), the class-factor system derives from the company's studies of 27,000 machines in actual production. Statistical analysis of these machines' performance enabled company president Denes Hunkar to classify them into nine categories of repeatability for 19 molding parameters. When loaded with Hunkar's SPC-Pro software, the Smartbox PMA can analyze a designated sample of production cycles from a particular machine and assign class factors to individual parameters and to the machine's overall performance. With a single number, molders can see how the machine stacks up against others in the field and against its own past performance.
Molex's Ford says the class-factor approach lacks detail, but others appreciate its convenience. Dr. Mark Polczynski, manager of Eaton Corp.'s Plastics Initiative program in Milwaukee, finds that class factors provide a comprehensible language for describing machine performance to the non-technical managers who often control capital budgets. This aid to communication is especially important at Eaton because many of its 30 different molding sites were acquired over the past 50 years. "The machines are 20-30 years old," says Polczynski. "And when we asked ourselves why have such old machines, we found it was because we couldn't justify new ones to management without the data. For me to go to management and show them that of 27,000 machines in the field, some of ours are performing in the bottom 10% is a powerful statement."
RELATED ARTICLE: How to Get Hooked Up
For quick and easy attachment of an analyzer to a machine, Tremblay and other experienced users recommend "connection kits." These consist of quick-disconnect hydraulic fittings for pressure transducers and a wiring harness that remains on the machine controller to provide immediate access to digital signals. "When I get to the machine, I just plug in and start collecting data," Tremblay says.
For the analog signals, most PMA vendors and users recommend installing a separate set of transducers that works in parallel with those serving the machine controller. Without an independent set of sensors, the PMA will be unable to help detect cases when a processing problem actually stems from faulty machine transducers. Using the same transducers for machine control and PMA is "like auditing your own books," says GE's Tremblay. A second set of sensors also avoids problems that could arise if the PMA interrupted control signals.
Nicollet is the only vendor that prefers signals from the machine's existing transducers. Molex's Ford notes that this approach works only because the PMA 360 does a good job of buffering the machine signals and is fast enough to pull an accurate portrait of the original analog signal from the machine controller. "You have to be able to get to that root signal," he says. To uncover problems that may lurk in the machine transducers themselves, Ford does dimensional studies on parts using a special mold designed to yield process information. "We're really running two separate analyses," Ford says. One relies on the PMA; the other validates that first analysis with part-to-process correlations. "If there's a problem with the machine transducers, it will show up in the samples."
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||includes related article|
|Date:||Feb 1, 1995|
|Previous Article:||How to extrude plastomer blown film.|
|Next Article:||How molders put PMAs to work.|