Putting Blender Accuracy To the Test

Blender accuracy is very much in the eye of the beholder. Hard data are rare, as is any firm definition of what accuracy means. "About accuracy, we all pretty much say what Maguire says," admits the marketing head of a competing blender manufacturer.

Maguire Products wasn't the first to market a weigh blender, but it was first to capitalize on a huge injection molding appetite for inexpensive batch blenders that could tally resin ingredients to calculate part cost. The market evolved with other suppliers following Maguire's lead, either by private-labeling or imitating its product. But in the past few years, as increasing numbers of extrusion processors also bought batch blenders, new competitors came in, brashly claiming better quality--particularly mixing quality. One of those, Process Control Corp., last year introduced the first test protocol to verify mixing quality.

Maguire and its competitors typically claim their batch weigh blenders are accurate to ±0.1% to 0.25%. "When pressed for an explanation, it is rare that any blender manufacturer knows the basis for its accuracy claim," writes Steve Maguire, president of Maguire Products. (He declined to be interviewed for this article but did submit documents on the subject.) "It's a confusing subject because there is no accepted standard in the plastics industry to express blender accuracy. I personally believe that marketing pressure requires that we all make claims at least equal to our competitors without really knowing how they are derived."

A claim of ±1% accuracy, for example, could mean that a blender set to dispense 4% color concentrate actually delivers from 3% to 5% colorant, which would be ±1% of the total batch weight. Or, it could mean 3.96% to 4.04% colorant, which would be ±1% of the 4% target ratio. "We do better than the latter, reaching 3.99% to 4.01% delivered, or 0.25% of target," says Maguire. "If asked to deliver 10%, we reach 9.99% to 10.01%. This is ±0.1% of requested target."

AEC/HydReclaim reports metering accuracy of up to 0.1% per component for free-flowing ingredients and "total machine accuracy up to 0.25%."

Maguire adds, "It would be more meaningful to discuss average error in grams per batch and then compare this to the batch size. These are more meaningful numbers. But customers do not ask for that. They ask for 'percent accuracy.'"

Another measure of accuracy concerns not weighing precision, but mixing uniformity. Accurate weighing may ensure that a batch as a whole contains within 0.1% of the target amount of additive. But mixing quality could lead to differing amounts of additive being present in each increment that enters the feed throat of a molding machine or extruder. That's what Process Control found in its testing of a "leading model" of blender (which Maguire identifies as a Maguire WSB-44).

The test involved taking sequential plug-flow samples of a batch and counting colorant pellets. Blend homogeneity was reported as two standard deviations divided by the mean of all measurements. Process Control found that the leading model "benchmark commercial blender" gave 50% variation in blend homogeneity. By the same test, Process Control's new Guardian Blender, engineered for more thorough mixing, reportedly gave 6.3% variation in blend homogeneity.

Maguire argues that for Process Control to have obtained such poor results on the Maguire, it must have bypassed the mix cycle altogether and just let dispensed material pour through the blender. Process Control insists that it ran all equipment following Maguire's instruction manual.

Counting pellets

Last summer, Process Control submitted its Guardian blender to outside evaluation by Plastics Product Review magazine (published in Traverse City, Mich.), which performs independent product testing and does not accept advertising. Plastics Product Review used Process Control's plug-flow-tube test protocol. The magazine plotted a flow chart of the percentages of colorant observed in a series of shots and found it hovered between 3.5 and 4.5% when 4% was set as the target. That amounts to a 14% variation.

Blenders from Mould-tek, Conair, and Maguire were to have been evaluated next. All three companies agreed to submit blenders for testing but so far have not done so, says Plastics Product Review editor Adam Schultz.

Conair explains that it isn't opposed to giving the magazine a blender for comparative testing. It just wants to be sure of the test criteria, test materials, and its blender performance. Conair's AutoWeigh name is on some 4000 blenders built around Maguire components and controls. Conair has started to conduct its own accuracy tests on the AutoWeigh blender, says product manager Gary Hovis. Conair is using a plug-flow test similar to that used by Process Control. Testing isn't yet completed.

No more law of averages?

The heart of the testing controversy is averaging. For years, blender makers totaled gravimetric weights over 20 to 30 or more batches and averaged them. Averaging works fine to calculate part cost, but it doesn't ensure part-to-part uniformity, critics say. "If you put your hand in freezing water and then in boiling water, the average would be quite comfortable," sniffs William Orozco, international sales manager of Mould-tek Industries. Opponents argue that averaging makes any blender look good if the time line is long enough.

Process Control CEO Joe Robertson says his batch weigh blender doesn't need averaging to make it look good because it is accurate within one batch. Meanwhile, Maguire states that all makes of blenders "achieve 0.1% accuracy only over time, only with pellets that are fairly uniform in shape and weight, only if the hopper material is level and held steady (for uniformity) and internal parameters are set correctly."

Customer tests with real-world materials typically show a "variation of ±3% within any given portion of the blended batch," Maguire claims. Its dispensing software compensates for errors in one batch by metering more or less in the next batch. According to Maguire, variation isn't significant because a blender's mixing hopper holds two to four batches, and an extruder screw can hold three to four times the volume of the mixing unit, or roughly 12 batches.

Plastics Product Review had previously rated Maguire's Micro Blender in January 1998, giving it high marks for quality, but using Maguire's averaging approach. The magazine took three lots of 30 batches each, a total of 90 batches, and recorded 0.4% accuracy. But after the magazine subsequently reviewed Process Control's blender, it now uses that company's plug-flow method to test blending accuracy.

Until recently, blender suppliers tended to avoid testing. John Hotchkiss, v.p. of engineering at Process Control and chairman of the blender subcommittee of the Society of the Plastics Industry's Machinery Division, says he tried four years ago to get the subcommittee to draft accuracy standards, but the rest of the committee was not interested in pursuing it.

Factors in accuracy

Up to now, blender-accuracy claims have tended to focus on dispensing accuracy. That depends on load cells, batching software, and dispensing mechanisms such as air-actuated slide gates, pinch valves, auger feeders, and vibratory feeders.

Load cells themselves are pretty accurate: Maguire says its 3-kg load cells are accurate to 0.025 g, or one pellet. Its controller reportedly has "software resolution" of four pellets. On the other hand, Maguire warns, "Vibration is always an issue, and accurate readings when vibration is present require more time."

Mould-tek claims superior metering accuracy based on its pinch-valve dispensing and precise computer control. A bar closes off a rubber flange, like fingers dispensing a pinch of salt. The valve is fully open until 80-90% of target weight is dispensed, then it feathers open and shut rapidly until the target is reached. Each dose is weighed by the quick-acting load cells of Mould-tek's RLTS (real-time live scale) system.

A number of factors in the process can affect blender accuracy: resin type, pellet size, density, flow characteristics, regrind type and level, batch size, and throughput rate. Oversized, heavy, or irregular particles can cause difficulty in mixing in the blender. Color masterbatch pellets tend to be heavy and often of a different size from virgin, which can hinder uniform mixing. Easy-flowing particles can cause errors by rushing too fast through slide gates, while sticky resins clump and bridge. AEC/HydReclaim says auger feeders are most accurate for both slippery and sticky resins. But augers pulsate: The first half turn delivers two-thirds of the material, and the second half turn delivers the rest.

Regrind affects accuracy in several ways. When regrind content varies, software calculates changing amounts of virgin and adjusts the level of colors and additives. But as the color fraction gets smaller, it's hard to measure accurately. Geometry of the regrind particles affects accuracy, too.

"Film regrind is impossible in a batch blender," notes Conair's Hovis. "Sheet and bottle regrind can be used, but only with adaptations for extra mixing."

Blender size and throughput affect accuracy both directly and indirectly. Blenders range from 10 lb/hr for lab models up to 10,000 lb/hr for central blending In general, bigger blenders are more accurate. Slide gates may be accurate to ±2.5 grams with an optimal resin, but percentage accuracy depends on batch size--2.5 g is 0.25% of a 1-kg batch or 0.014% of an 18-kg batch. Central blenders have the disadvantage that material must be conveyed to processing machines by airvey system or gaylords. This transport can cause demixing. Blender vibration also demixes and can upset load cells momentarily, causing metering errors.

Blenders aren't equally accurate at all throughputs. HydReclaim's biggest batch blender, model OS-100, for example, has a throughput range of 50 to 5000 lb/hr. Blenders are generally more accurate at dispensing larger amounts. Yet some blenders double their output by following a gravimetric batch with several rapid volumetric batches, which aren't weighed, so accuracy decreases.

The number of ingredients affects batching cycle time and accuracy. More components take more time to dispense and raise the risk of error. Maguire has software that manages up to 12 components. HydReclaim doesn't build a batch weigh system for more than six components. Beyond that, HydReclaim recommends using a continuous weigh blender instead.

The accuracy debate has focused attention on blender mixing effectiveness. Some suppliers are already addressing the issue. For example, AEC/HydReclaim introduced a new agitator design, called the Opti-Mixer, which is a whorl of interwoven paddles with holes in them to improve blend homogeneity. Meanwhile, Conair uses a ribbon-style mixer for hard-to-mix materials whose particles have different sizes or geometries, like mixing thermoformed sheet regrind with virgin pellets.

This year, Conair also introduced the XB Series extrusion blender, which has a reverse-flighted auger mixer and components from TSM Control Systems, an Irish blender maker that has an office in Atlanta. It pulls material from the edge of the hopper to the center and positively feeds material to the outlet. "This is a superior mixing method for all materials and especially for ones that tend to demix," Conair's Hovis asserts.

What accuracy do you need?

How should you approach suppliers' claims and counter-claims? Look carefully at what's really being measured--metering accuracy, mixing, or weighing--and how important that is to your process. Not everyone may find the accuracy dispute relevant. For example, Steve Maguire remarks, "With over 13,000 units sold, if we had a mixing problem, our customers would tell us, not just our competitors." He adds that "extruders do an excellent job of mixing. One color pellet can disperse across 20,000 natural pellets." So differences in blenders may or may not produce perceptible differences in product color or quality. On the other hand, Process Control points out that an accurately dispensed and homogeneously mixed blend could save you money on costly additives.

Different processes require different levels of accuracy in metering and mixing. "Generally the people doing injection molding aren't as picky as the people doing extrusion, and not nearly as concerned with [blender] accuracy," says Fred Eichhorn, electrical engineering manager at Novatec Inc. Automotive molders are exceptions, he says. "They'll quite often bring their own materials and go to different blender manufacturers and run tests. Engineers will sit around and literally count pellets from a 5-lb batch to see which blender is more accurate."