A new look at check valves.

In the drive for injection molding quality, the influence of the non-return valve has been largely overlooked. Here are the first data on performance of a novel valve design, which could suggest changes in machines and molding techniques.

At a time when injection molded parts have grown more complicated than in years past, quality and consistency in molding those parts is becoming a paramount competitive factor in the marketplace. Statistical Process Control (SPC) is on its way to becoming a nearly universal necessity. Any deviation in product characteristics that can be eliminated or improved upon will certainly be required by the customer. Only the molder that can supply the lowest level of deviation will get the order.

It's no secret that one of the common, everyday obstacles to achieving the desired part-to-part consistency is the behavior of typical existing non-return valves--ring-and ball-type check valves. As any molder can testify, these valves do not always close immediately at the start of injection--or even close at all on some shots. These instances of late valve closing, or non-closing, are known as "flyers."

Despite its shortcomings, the same basic check-valve technology has been used for 30 years. Injection machine and controls manufacturers have compensated for erratic valve performance by developing microprocessor-based, closed-loop control of cushion size. They have been reasonably successful, but both molders and machine suppliers may have overlooked some disadvantages in relying on cushion to make up for inherent variability in valve operation.

In this article, we will present a totally new check-valve design (a patent is pending), which operates by a very different principle from that of other valves on the market. We will also present the first published data on its performance relative to other valves, both in laboratory tests at injection machine manufacturers, and in actual production use. Finally, we will point out that what has been learned through investigation of this new valve suggests some modifications of currently typical machine design and molding methods so as to eliminate other remaining sources of process variation.

NEW VALVE PRINCIPLE

As reported previously in PLASTICS TECHNOLOGY (Jan. '92, p. 17), the new valve, called the Dray Non-Return Valve (DNRV), invented by R.F. Dray, is commercially available from a new company, U.S. Valves, Inc. in Evansville, Ind. It differs from conventional ring and ball valves in one key respect. Ring and ball valves are closed by the action of melt flow through the valve. That is, when forward movement of the screw is initiated to start injection, some backflow of melt through the valve occurs, causing a pressure drop on the upstream side of the sliding or floating member (ring or ball). This causes the valve to close, assuming that the seats are aligned and clear of foreign matter and that valve wear does not cause leakage through sealing surfaces. With these types of valves, screw pullback is frequently used at the end of screw rotation (or sometimes at the start) to marginally increase the injection stroke and thus provide ample opportunity for the valve to close.

The DNRV, however, closes without either any melt flow through the valve or any screw movement. As shown in Fig. 1, this valve has a central piston with a larger surface area on the downstream end than on the upstream end. This piston moves freely under the influence of differential pressure on the two ends.

In other words, if the pressure on the larger (downstream) end (P1) multiplied by the piston area (A1) is different from the pressure on the smaller (upstream) end (P2) times the piston area (A2), then there will be a net force tending to move the piston in one direction or the other. Note that the valve will tend to stay closed except during screw rotation, when sufficient overpressure is generated upstream of the valve that:

TABLE 1-LAB TEST AT AUSTRIAN MACHINE MANUFACTURER
Test Conditions
Machine Size                     30 Ton, New, Micropro. Control
Screw Diam.                      35 mm
Material                         PP
Mold                             Single-Cav.
Samples Per Test                 50
Cushion                          None
Preclose (DNRV)                  None
Metering Stroke (Recovery)       52 mm
Inject Speed                     10 mm/sec
Backpressure                     73 psi
Pullback (Decompression)         5 mm
Valve Type              Ring         Repeater
Avg. Wt., g             35.085       34.117
Std. Dev.               1.088        0.103
Max. Wt., g             37.002       34.636
Min. Wt., g             33.416       33.970
Range, g                3.586        0.666
% Variation             10.221       1.952

(P2) x (A2) |is greater than~ (P1) x (A1).

After screw rotation ceases and backpressure is shut off (usually simultaneously), pressure will decay gradually on both ends of the piston. Since (A1) |is greater than~ (A2), when (P1) = (P2), then:

(P1) x (A1) |is greater than~ (P2) x (A2). Consequently, the valve will close.

Obviously, the greater the differential between (P1) x (A1) and (P2) x (A2), the more rapid the valve closure. Conversely, backward drift of the screw after rotation stops will slow the valve closure and allow small amounts of melt backflow. If screw drift is extreme, or excessive pullback is used, then valve closure may be delayed until the start of injection. However, with the DNRV, it is possible to "capture" the full, exact amount of melt metered by the screw. This is done by preclosing the valve prior to injection. All that is needed is to maintain hydraulic backpressure on the screw for one or two seconds after screw rotation ends. That will generate a higher pressure on the downstream end of the valve, immediately forcing it closed. Tests at both an American and a European machine manufacturer showed that control programs of existing equipment can readily be changed to incorporate delayed backpressure cutoff.

Some degree of screw pullback prior to injection may be used with the DNRV without causing the valve to reopen, provided that the pullback stroke does not exceed the overstroke of the valve piston.

TABULAR DATA OMITTED

Even if the valve is not preclosed prior to inject, it should be apparent that the valve will close very quickly upon initiating the forward stroke of injection. Since the length of piston travel necessary to close the valve is very small, the time required to close the valve, and thus the amount of backflow leakage, will be far less than with other valves. Most important, the amount of leakage will be consistent, as this type of valve is not vulnerable to the "flyer" phenomenon. The right-angle closing action of the piston eliminates problems of improper valve seating. And any slight wear that may occur on the upstream end of the piston will not affect its ability to close promptly, since there is ample overtravel of the piston.

NEW VALVE'S ADVANTAGES

The DNRV's closure-by-pressure feature is unique, as backflow is no longer an inherent and necessary aspect of valve closure. With preclosing of the DNRV prior to injection, backflow can be virtually eliminated. The only system that can perform comparably is a costly melt-accumulator system.

As noted above, wear is no longer a threat to consistent valve closing. The only moving part in the DNRV is the piston, and its sliding surfaces are not contacted by plastic flow. In other valves, sealing surfaces are constantly in contact with polymer flow. With abrasive filled materials, erosion of those surfaces does affect valve performance. Also, sealing areas in ring and ball valves are constantly exposed to any contaminants in the melt stream that may hinder effective valve seating.

Note also that the DNRV can be manufactured integral with the screw, therefore eliminating any chance of misalignment between screw and valve.

DETRIMENTAL EFFECTS OF CUSHION

Many, if not most, injection molders run their injection units with a cushion. The cushion is an inventory of melt between the screw tip and nozzle adapter, which allows for non-uniform performance of the non-return valve. But there is a downside in performance when using a cushion, which is often ignored or accepted as a necessary evil.

Polymer melts all tend to degrade with time. These changes in basic polymer or molecular architecture produce TABULAR DATA OMITTED inferior parts in critical applications. Since the presence of a cushion necessarily involves increased melt residence time, eliminating it will tend to improve molded part quality.

Molten polymer molecules are mobile random coils, which change their coil geometry during the application of shear. The greater the rate of relative motion between adjacent molecules, the greater the molecular orientation in the direction of flow. This phenomenon is responsible for shear-thinning viscosity behavior. Shear thinning is a function of the polymer's stress-relaxation time, which can vary from a few seconds to a minute or more, depending on melt temperature, molecular weight and molecular-weight distribution. Allowing the melt to sit in the cushion with little or no shear deformation allows molecular coils to return to the random state. On injection, this relaxed melt will appear more viscous, requiring much higher injection pressures.

TABLE 4-SIX-MONTH PRODUCTION TRIAL AT HOUSEWARES MOLDER
Test Conditions
Machine Size             700 Ton, 1978 Model, No Microprocessor
Screw Diam.              4 in.
Material                 LLDPE, 50 MI
Mold                     Laundry Basket, Single Cav.
Backpressure             None
Pullback
  End of Recovery        None
  Start of Recovery      0.5 in.
Preclose                 None
Cushion                  None
Test No.               1            2             3
Valve                  New                        Worn
Type                   Ring         DNRV          Ring
Recovery, in.          6.50         6.25          6.75
Avg. Wt., g            777.50       770.07        760.25
Max. Wt., g            779.00       770.30        785.00
Min. Wt., g            776.00       770.00        738.00
Range, g               3.00         0.30          47.00
% Variation            0.38585      0.03895       6.1822
Std. Dev.              1.323        0.105         16.49
6 Std. Dev.            7.930        0.630         98.94

It is a rare injection unit that has the melt zones accurately set to the melt temperature. With a short residence time, the effect of a temperature mismatch may not be significant, since heat transfer is time-dependent. Use of a cushion increases residence time and opportunity for heat transfer between the barrel and the melt. If you have a leaking or stick/slip non-return valve, the cushion melt-temperature profile usually varies from shot to shot. What's more, temperature stratification tends to occur with extended residence time in the cushion zone. It is not uncommon to find that the normal smooth, laminar flow in the cushion zone is shifted to tunnel flow, with the exterior melt moving more slowly--if at all--giving it yet a longer residence time than is desirable.

Finally, eliminating the cushion facilitates faster purging and color changes.

The conclusion is clear. Eliminate the cushion to make better parts. This will involve retraining the setup personnel to set proper limits on the forward and rear stops of the screw. But operating without a cushion also requires positive reliability of check-valve function.

LABORATORY TEST RESULTS

Laboratory test comparisons of the DNRV and conventional valves have been run at four different injection machine manufacturers in the U.S. and Europe. Results of one of these comparisons with a new, microprocessor-controlled Austrian machine are shown in Table 1. This test was intended to display a worst-case situation for a ring valve--i.e., slow injection with no cushion. Under these conditions, the ring valve performed far worse than it did with a cushion.

The DNRV was tested under the same conditions. With the DNRV, part-weight variation was much smaller, but could have been improved further by shortening or eliminating the screw pullback (decompression) stroke, as well as holding the screw position stable at end of recovery (no backward drift). Either pullback or screw drift can open the DNRV if the distance of travel is greater than the overtravel distance of the piston that closes off the melt-outlet channel.

The next series of tests (Table 2), run on a new press at a major European machine builder, shows the advantage of being able to preclose the DNRV before injection. This involved changing the injection machine's computer program to include holding of hydraulic backpressure after recovery for a given time. In some of these tests, we experimented with quickly raising the back-pressure after the screw stopped rotating to see what effect this had on valve closing. A nozzle shutoff valve was also added prior to test 5 so that pullback could be eliminated without causing mold drool.

In all tests, the DNRV gave much more consistent results than the ring valve. Eliminating pullback greatly reduced part-weight variation with the DNRV (compare tests 2-4 with test 5). Tests with the longer (4-mm) pullback stroke after preclosing exhibited the greatest variation in DNRV performance (tests 3 and 4). The one exception was test 6, which may have been influenced by too short a preclose time (0.5 sec), not allowing for full preclosing. Note that in test 7, longer preclose time (1.0 sec) reduced part-weight variation. Longer preclose time also appears to yield better results in comparing tests 3 and 4.

In tests 7 and 8, results improved when the preclosing time stayed the same but the pressure was reduced. That may result from the unusual experimental procedure of suddenly increasing the backpressure after screw rotation stopped. The sudden compression of the shot may have caused a backflow of melt and consequent part-weight variation.

Table 3 shows tests run on another new press at a major U.S. machine builder. They are interesting because, in three of the four comparisons, tests with the DNRV actually showed greater part-weight variation than with a ring valve. We believe the evidence points toward poor screw performance (inappropriate design for the material being tested). In tests 3A and 3B, the DNRV gave dramatically better part uniformity. In this case, the metering stroke was deliberately reduced so as to produce short shots. This emphasized the differences in valve performance, since they were not "hidden" by the presence of a cushion.

An important clue to the poor results with the DNRV in the other tests in this series is the variability in recovery time, which was significantly greater for the DNRV tests (1B, 2B, 3B) than the ring-valve tests (1A, 2A, 3A). That recovery-time variation was greater with the DNRV even in tests where the part-weight variability was much smaller than with the ring valve (tests 3A and 3B).

We believe the screw was surging as a result of inadequate plasticating performance. Surging would produce temperature variations in the melt, and therefore differing degrees of melt density and melt pressure in the accumulated shot beyond the valve. The DNRV closes at the end of recovery when the pressure equalizes on both sides of the valve, and remains closed prior to inject. So surging and melt-temperature variations would mean that the DNRV would "trap" different shot weights on each cycle.

A ring valve would not exhibit this "trapping" effect and would permit backflow through the open valve to even out variations in the accumulated melt pool caused by surging--yet another example of how current machine and valve design accommodate and help disguise each other's limitations.

What's more, it was evident that the screw actually surged more with the DNRV than with the ring valve. The reason is that the 1.75:1 diameter ratio between the two ends of the DNRV piston required 75% higher melt pressure upstream of the valve in order to keep it open and melt flowing through it during screw rotation. Thus the 125 psi hydraulic pressure downstream of the valve required 219 psi upstream of it. This translates into a difference in melt pressure between 1040 and 1820 psi.

The marginal conveying ability of this screw with PP material was spotlighted by doubling the backpressure on the ring valve from 125 psi in test 1A to 250 psi in test 4A. This increased the average recovery time by 17.5% and the recovery-time standard deviation by 27%. With the DNRV, the screw could not even recover at a 250-psi backpressure setting, so the setting was dropped to 50 psi. Compared with test 1B at 125 psi, the average recovery time dropped 45% and the standard deviation (or "sigma") fell 60%. The range of part weights also was slightly lower.

PRODUCTION PLANT RESULTS

Table 4 shows typical results from a normal production run in a large housewares molding plant. The DNRV was run in production here for six months on a 1978-model, 700-ton, U.S.-built "workhorse" machine without microprocessor controls. Note that the molder operated without a cushion, providing a good comparison of valve performance.

Even without a cushion, this machine with a brand-new ring valve held weight variation within 0.1 oz on a 1.7-lb part. But variation was reduced 90% with the substitution of the DNRV. After six months of 24-hr, seven-day production, no measurable wear was seen on this valve.

In striking contrast, test 3 shows a large degradation in performance with a worn ring valve before it was replaced by the new valve in test 1.

Table 5 shows results of production molding of a small (1.3-oz), polycarbonate housing on a sophisticated, microprocessor-controlled Netstal machine at Nypro, Inc. in Clinton, Mass. Test 3 was run a week or two before the other test and under slightly different conditions, but it provided the only part-weight data available at press time.

To evaluate those weight data, first note that the 55-mm screw displaces 2.376 cc of volume with 1 mm of stroke. Position readout accuracy in these tests was 0.1 mm (0.004 in.), yielding a volumetric measuring precision of 0.2376 cc. Estimated specific gravity of black polycarbonate melt is 0.7 g/cc. Therefore, rounding off the recovery and cushion positions to the nearest 0.1 mm can alone account for 0.214 g of weight variation. A three-sigma range of weight variation totaling 0.043 g in test 3 indicates highly uniform performance of the DNRV--particularly in view of the decline in average weight over the 50-shot sample by about 0.018 g as the process stabilized after startup.

The injection stroke on the DNRV with preclose (tests 1A and 2A) was about 2.05 mm shorter than it was without preclose (tests 3A and 4A). After the DNRV piston closes, it overstrokes to avoid reopening if pullback is used after preclose. During this overstroke, a small amount of the shot is drawn into the end of the DNRV. On the specific DNRV used in this test, the screw moves forward 1.33 mm to displace plastic into the space vacated by the piston during overstroke. The remaining 0.7 mm difference in screw stroke is due to melt compression.

The DNRV closes faster than a ring valve, as shown by comparing injection strokes and recovery times of tests 1A through 4A with those of tests 1B through 4B (disregarding test 3). Less shot loss from backflow during valve closing results in less shot volume to recover and hence faster recovery.

That a DNRV closes more consistently than a ring valve is shown by the lower three-sigma range of recovery TABULAR DATA OMITTED time (tests 1A through 4A) than for the ring valve (tests 1B through 4B). More consistent valve closing produces a more consistent shot size to be recovered.

IMPLICATIONS FOR PROCESS CONTROL

Scientific testing of one machine variable or component, such as a check valve, requires eliminating or minimizing the effects of all other relevant variables. We have not come close enough to accomplishing this in testing the DNRV.

Obviously, there are many major sources of process variations in injection molding, including the plastic material itself, part and mold design, mold-temperature control, and mold clamping force. However, our primary interest is in the injection unit. The foregoing discussion has highlighted some of the most important variables there, including screw design, recovery speed and backpressure, melt residence time, and last but not least, shot volume.

Obtaining precisely consistent shot volume is one of the most complex aspects of the process, and itself can be broken down into a number of factors. One is the necessity of obtaining a precisely consistent position and pressure at start of injection. Unlike water or hydraulic oil, plastic melts have widely varying degrees of compressibility. This compressibility is evidenced by the screw "bounce-back" effect often seen when hold pressure is released, as well as the smaller initial bounce-back seen when screw backpressure is released at the end of recovery. Thus, to inject a precise amount into a mold, it is important to start injection at a precise position and pressure and with no voids in the accumulated melt.

For best control one must:

1. Stop screw rotation at a precise position. Avoid variations due to the response time of the machine control.

2. Avoid screw pullback. Avoid mold designs that require pullback to prevent mold drool. Pullback through an open non-return valve introduces voids in the accumulated shot and allows inconsistent flow through the valve. Avoid non-return valves that require pullback to improve their repeatability.

3. Compensate for position drift after stopping screw rotation. In addition to bounce-back when recovery backpressure is released, the screw often drifts backward. This is probably due to continuing decay of the pressure profile along the screw established during rotation. As the pressure decays, flow occurs both toward the feed throat and forward into the accumulated shot as pressure equalizes. Maintaining recovery backpressure--or better, maintaining position control after "Stop Rotate"--will avoid this drift. However, many (probably most) machine hydraulic systems and control systems require modification to accomplish this.

4. Obtain a repeatable pressure at the injection starting position. As noted above, plastic compressibility makes this necessary for precise shot-size control. This is not easy to achieve with current technology. Valves used to control recovery hydraulic backpressure are often inaccurate and inconsistent in operation at these low pressures. Further, variations in viscous resistance to movement between the plastic-filled screw and the barrel, along with seal and piston-ring friction in the screw-drive system, mean that low hydraulic pressures do not translate efficiently into control of plastic pressure. Therefore, closed-loop control of melt pressure in the shot chamber will probably be required. Once a repeatable injection starting position and melt pressure have been achieved, the DNRV's preclosing feature will effectively "trap" that precise amount of melt so that none escapes back through the valve prior to or during injection.

A second major priority, as we have seen, is to close the non-return valve reliably and without variation. All conventional non-return valves have significant leakage flow out of the accumulated shot during valve closing, at the same time that flow is starting into the mold. Ring and ball valves are usually--but not always--quite consistent in the amount of accumulated shot weight that "escapes" before the valve closes. Both lab and production data show that the DNRV is much more consistent, due to the small plunger motion required to close, and its dependency on pressure only (no escape flow is required) for closing action. No test of the DNRV has ever shown a "flyer" such as has been seen even under laboratory conditions with ring valves.

Just as important as a repeatable injection starting position and pressure is to stop injection at a precise forward position and pressure. Bottoming the hydraulic cylinder at a controlled pressure accomplishes this. Most modern machines can provide the needed pressure control.

We strongly believe that preclosing the non-return valve will be the way of the future to improve control of the injection molding process. The preclosing must be done at a precisely controlled position and plastic pressure. Closed-loop control of pressure at valve closing will probably be required. Cushion, the crutch for valve performance and lack of position and pressure control, can then be eliminated. Screw design, mold-temperature control, and plastic material consistency will then become the major variables in the molding process.

The View From a Controls Supplier

Most process-control systems have a short-correction or cushion-correction feature intended to compensate for variations in material viscosity. This feature is extremely important to the processor, as it can effectively compensate for lot-to-lot variations in raw material and differences in regrind ratio.

A common problem with using process controls in this manner is associated with leakage of the check valve. The electronic control has a tendency to interpret changes in mold filling caused by a leaking check valve as a reflection of changes in melt viscosity, rendering the control feature ineffective.

The DNRV will evidently eliminate this problem and will make the short-correction feature of process controls a more profitable and effective tool than ever before.