Parts I & II
Ernest Griesser
Armen Sahagen
Abstract
Precision injection molders are plagued with variations in shot weight control which are beyond the control of the machine and their operators. Often the plastic resin uniformity is blamed, when the shot-to-shot control of the machine is really at fault. Symtomatic of these problems are machines that require pressure adjustment several times per shift to maintain a steady fill/flash ratio, and machines that lose control from cold drafts, etc.
These random variations are due to a combination of machine and shop floor ambient conditions acting upon the repeatability of the pressure transducers used to control the process, as well as the basic repeatability characteristics of the pressure sensors themselves. After all, any process is only as good as the accuracy and repeatability of the pressure sensor used to control it! This is why molders are bypassing standard machine pressure sensors to use in-mold and in-nozzle pressure transducers in place of hydraulic and linear position pressure transducers. Sometimes they achieve improved shot weight control and sometimes not, depending upon the sensor used to by-pass the hydraulic control system. All sensors are not created equal!
Part I, follows on a study reported at ANTEC '96 , in which the author continues to examine the performance of various types of pressure transducers and load cells, and their locations, on the shot weight repeatability of the injection molding process. Basic injection machine process control is examined, together with studies made on various auxillary pressure transducers and load cells compared to standard control with hydraulic and linear positioning pressure transducers.
Part II discusses the advantages of open and closed-loop pressure control with Silicon on Sapphire melt pressure and temperature nozzle transducers and gives examples of ultra-precision molding as achieved on standard molding machines modified with these new melt pressure/temperature transducers. Actual results show molders have achieved a standard deviation in shot weights of 0.0007%, compared to 0.035% to 0.065% for standard electric and hydraulic machines with a minor investment in a new type of pressure transducer controlling from the nozzle melt, hot runner, or mold areas.
Part I- Analyzing Injection Machine and Sensor Repeatability
Introduction - The Problems
Figure 1 shows a typical mold fill pressure profile for a single cavity mold. As the mold is filled, the filling pressure increases to a selected peak pressure (Ppeak), which controls the initial fill of molten plastic. As the plastic cools, it shrinks, and the packing pressure (Ppack) forces more molten plastic into the mold to make up for shrinkage, until the sprue or gate is frozen. At that exact point, providing the gates do not freeze prematurely, the mold is filled to full capacity, subject to variations in Ppeak and Ppack, mold temperature, injection speed, etc., from shot-to-shot.
Increasing Ppeak above the optimum value will increase shot weight above optimum and cause flash on the parts. Decreasing Ppeak will cause a lesser filled part with dull finish, blemishes and possibly sink marks, unless Ppack can make up for the deficiency. By the same token, fluctuations in Ppack will also affect part weight, but seldom will affect flash, as this is more related to Ppeak. Obviously then, shot weight is directly controlled by Ppeak and Ppack, which, in turn, need to be controlled within a narrow range to produce consistent weight parts with best appearance and with consistent properties. This is especially important for precision molding applications.
The problem with injection molding machines is that Ppeak and Ppack generally vary too much for precision molding applications . Figure 2 shows the typical variation that can be seen in Ppeak from shot-to-shot for 62 consecutive shots in a typical hydraulically controlled press. A process analyzer using nozzle pressure and hydraulic transducer was employed to monitor the process. The peak pressure variation measured in this example was 1.78%, indicating a total, six sigma deviation of about 2%. Similar variations can be seen in electrically operated presses.
Table I shows the process capability of a 50 ton standard hydraulic machine used to mold polycarbonate CD's over a 24 hour period. Note that Ppeak varies 338 psi or 2.0% from shot to shot. This explains why bi-refringence, an important property of CD's is difficult to control.
Table II shows the process capability of a 250 ton Klockner hydraulic machine molding 32 cap closures in a hot runner mold.
This machine shows Ppeak variations of 430 psi or 3% at the nozzle, monitored with a Sensonetics Silicon on Sapphire nozzle pressure using an NPE process analyzer data acquisition system. From the above examples, it is evident that a 2-3% variation in Ppeak is common in standard injection machines.
Figure 3 shows the results of a study by the Plastics Institute of Aachen, Germany , in which five different, recent vintage, injection molding machines, including three all-electric machines were evaluated with five different plastics resins in two molds. Fig 4 shows the results of their part weight study plotted in part weight distribution format on a 75 gram part weight basis. The one sigma standard deviation in shot weight averaged 0.05%, with a range of 0.035% to 0.065%. These one sigma figures can be improved 50-90 fold, merely via controlling the machine injection pressure with a Silicon on Sapphire melt pressure sensor, as shown in the narrow distribution curve in the center of the normal spread. (To be discussed further in Part II.)
The question arises; if this broad distribution is the norm for shot weight control on standard machines, where are these variations coming from, and how can we eliminate them to accomplish real precision molding? Referring again to Fig 2, we see an excellent correlation coefficient of 0.97 between hydraulic fluid pressure and nozzle melt pressure, indicating cause and effect. In other words, the fluctuations present in hydraulic system shot-to-shot variations cause the variations in the pressure at the nozzle, which, in turn, affect shot weights proportionally.
Figure 5 shows a plot of shot weight variation versus nozzle melt pressure variations from the previous examples, empiracally correlating the low and high nozzle pressure variations with the low and high range shot weight variations described in Fig 3. Note that high precision requires very low nozzle pressure fluctuations.
Figure 6 diagrams the typical hydraulic machine operating system as follows:
1. The drive at A rotates the screw, which deposits its accumulating shot behind the nozzle and in front of the screw, which backs out of the barrel as the shot build-up progresses. As the screw backs out it becomes shorter, and generates less adiabatic heat in the polymer. This means the back portion of the shot is slightly cooler than the front portion, where the screw was longest initially. Excessive variation in melt temperature can also cause quality problems.
2. When the shot and mold are ready, the extrusion unit advances to kiss the mold and inject the molten plastic. In this step, a check valve near the screw tip seals off backflow to turn the screw into a ram, while a pressure intensifying piston behind the screw multiplies the hydraulic pressure by ten to twenty times to assure sufficient injection pressure. Without pressure amplification, the plastic would ooze, rather than flow into the mold. Note that excessive and variable backflow through the check valve can also contribute to shot weight repeatability problems.
3. The hydraulic system fluid transducer is located at B and is typically rated at 0-2,500 psi, with a repeatability error of 20 psi, and a response time of <5 milli-sec.
4. As the multiplying piston intensifies the pressure on the plastic melt, the repeatability of 20 psi is also multiplied, typically by 10, 15 or 18 times. This means that the best possible repeatability of the melt pressure at the nozzle can be no less than 200 psi for a 20,000 psi (700 Bar) rated machine. Adding the additional variations caused by hydraulic fluid temperature changes, room temperature variations, hysteresis of the transducer, hysteresis caused by hydraulic fluid compression, and friction from the screw and intensifying pressure piston, obviously raises this repeatability error from 200 psi to the 300 to 450 psi range, as measured, or 2-3% of actual injection pressure from shot-to-shot for a 20,000 psi machine, and higher for a higher pressure machine.
5. The friction of the piston, and especially of the screw in the barrel, can consume from 10-30% of the overall pressure theoretically applied to the melt. This reduces the pressure at the nozzle by that amount and contributes to a portion of the variability of Ppeak and Ppack. This explains why measuring and controlling Ppeak and Ppack at the nozzle, mold or hot runner is superior to hydraulic control as a means of process control for precision molding applications. The next step becomes one of selecting the most suitable pressure sensor to use in the nozzle or mold in order to maximize the benefit.
6. For machines operating from linear positioning transducers, the repeatability of the shot position and the cushion position determines accuracy. While linear positioning transducers have capabilities to 0.001" resolution, the microprocessor I/O capacity is almost always compromised by bit limitations with resultant programming position stops in the range of 0.004 to 0.010" apart. The effect this has on repeatability of shot volume is illustrated below for machine controls having 0.010" position spacings, with a resultant calculated repeatability of 0.014" using the square root of the sum of squares method:
Repeatability of PLC Controlled
Screw Position with 0.10" Resolution Between stops
| Screw Diameter | Volume | PS Weight | PP Pellets |
| Inches | cc | grams | count @ 60/gm |
| 1.0 | 0.18 | 0.19 | 10 @ 60/cc |
| 1.25 | 0.30 | 0.31 | 16 |
| 1.5 | 0.40 | 0.43 | 22 |
| 2.0 | 0.72 | 0.75 | 39 |
| 2.5 | 1.18 | 1.23 | 64 |
| 3.0 | 1.62 | 1.70 | 88 |
| 4.0 | 2.88 | 3.02 | 156 |
| 5.0 | 4.70 | 4.93 | 254 |
| 6.0 | 6.49 | 6.81 | 350 |
This table provides a comparison of variability in volume, weight of PS, and pellets count of PP, that can be expected from various size screws that are controlled within 0.010" in the shot and cushion positions. For screws controlled to within 0.005", repeatability will be half of the amounts shown. These variations are caused by the microprocessor bit capacity insufficient for 0.001".
Discussion - Sensor Selection
Because any process control can only
be as good as the sensors used, it is important for molders to understand
the requirements of the process and the performance of available pressure
sensors sold for this purpose. Table III lists the generic load cells and
pressure transducers commercially available for this purpose, by type,
along with specification features and calculated results.
Fundamentally, to catch a fast moving
Ppeak pressure, it requires a very fast responding pressure transducer, in the range
of 1-5 milli-sec in response time or faster. Thus, any pressure transducer slower
than 5 milli-sec is too slow to catch peak pressure in the same place all
the time and therefore can never function properly for this purpose. Accordingly,
this eliminates all liquid-filled and push-rod type transducers that are
available (or could possibly be used) for use in a nozzle or mold. The
pressure sensors remaining are made up of indirect cavity load cells, and direct
cavity or nozzle quartz and Silicon on Sapphire pressure transducers.
The next requirement is for basic accuracy
and repeatability, with emphasis on repeatability, to be as good as possible.
After all, a 20,000 psi (700 Bar) sensor with a 1% repeatability can basically
control to only 200 psi, only slightly better than the hydraulic
transducer it replaces, while a .1% repeatability silicon on sapphire pressure transducer can
basically control to 20 psi. Therefore, transducer specifications
are of utmost importance in the sensor selection process. Accordingly,
the sensors remaining are excellent repeatability load cells and silicon on sapphire
pressure and pressure/temperature transducers.
The next step in the selection process
is to determine the effect of the press and plant ambient conditions on
the performance of the pressure sensors. This can readily be calculated from the
square root of the sum of the squares of all the errors involved(1). Sources
of error magnification include; ambient temperature variation effects,
hysteresis effects, and friction effects of ejector pins for load cells.
In fact, only the load cells are affected by all three of the foregoing
error magnifying factors, leaving the Silicon on Sapphire direct mold cavity, nozzle
or hot runner pressure transducer as the most repeatable sensor.
Table IV shows the results of a study
of the accuracy and repeatability of all the foregoing sensors, taking
into account modest ambient conditions most often found in injection molding
presses and plants. These results were calculated from the aforementioned
square root of the sum of the squares method (1), using typically expected
ambient effects and specification data from the manufacturers. Fig 7 shows
the calculated repeatability, in bar graph form, of the sensors listed
in Table IV. Note that the Silicon on Sapphire pressure transducer is significantly better
in calculated repeatability than any of the other types available.
In Part II, the author will examine
the results of a number of beta site trials that confirmed the superiority
of silicon on sapphire melt pressure transducers for precision control of injection
molding.
Part II- Results of improving shot weight precision
As discussed in Part I, the pressure sensor
selection process is crucial to improving accuracy and repeatability for
precision molding. Not all pressure sensors were found to provide the same degree
of improvement over the standard hydraulic transducer. The main reasons
are; lack of sufficient response time to catch peak and pack pressure in
the same place from shot-to-shot, insufficient accuracy and repeatability,
and/or excessive machine and room temperature ambient effects on pressure
readings. Remember, any process control system can basically be only as
good as the accuracies and repeatabilities of the sensors used in the system.
Fig. 7 shows the results of calculated
repeatabilities of the transducers and load cells most often used in injection
molding. Note the typical repeatability in the range of 300-350
psi
calculated for these transducers and
compare this to the repeatability of 300 - 450 psi actually measured
on several standard hydraulic machines. This indicates that these calculations
are providing slightly conservative values in the proper ballpark.
Excellent repeatability is the single
most requirement for precision molding shot weight control. Note that specification
values do not tell the complete story unless the fine print is read and
calculations of the temperature, hysteresis, etc. effects are made on repeatability.
To obtain better repeatability of pressure
control, a Silicon on Sapphire pressure transducer was installed in a nozzle adaptor, Fig 8,
of a 170 ton Van Dorn HT Hydro-Toggle Machine in Van Dorn's molding lab
in Strongsville, OH. Table V shows the results obtained on this machine
with a Pathfinder 2500 Controller. Using the Silicon on Sapphire nozzle pressure transducer
to control the cycle by bypassing the hydraulic pressure transducer, a pair of gears
were molded in a pinpoint gated two cavity gear mold having a shot weight
of 75 grams.
Using the melt temperature capabilities
of the pressure transducer, Fig 9, it was observed that melt temperature declined
by 1.5%F at the end of each shot as the melt was ejected past the nozzle
pressure transducer position and recovered at the start of each shot. This drop-off
was due to the progressive shortening of the screw as it backed out of
the barrel during the recovery cycle. Additionally, there was a random
variation of 1%F from the heaters cycling on and off.
Fig 4 shows the part weight distribution
after proper conditioning and weighing by Standard Laboratories, Inc. an
independent laboratory. The parts had a average weight of 74.8900 grams,
with a total six sigma spread of 30 milligrams. This means, that the one
sigma spread was 0.0007% compared to 0.05% (.035-.065%) for a regular machine,
for an incredible improvement of 50-90 times.
Table VI, in another case, this time
with large parts, Delphi Division of GM, in Anderson, IN was molding a 10 Kg
Cadillac PP bumper structure with 14% rejects, when they had a silicon on sapphire
nozzle pressure transducer installed along with a new screw on an old Cincinnati
Milacron 1000 ton machine with relay controls. Without changing the controls,
they controlled the machine from the nozzle melt to produce zero reject
parts having a weight variation of only 3 grams for virgin resin
and 5.5 grams for 100% regrind. The virgin resin runs provided a
standard deviation of 0.01%, or 5 times better than a recent vintage machine
closed-loop control.
Table VII, in still another case, this
time molding EPDM rubber, Elastimold in Albuquerque, NM purchased a dozen
nozzle pressure transducers and installed them on all their Desma 4-station rotary
molding machines and some of their Lewis vertical presses. The results
reduced rejects to zero and eliminated operator adjustments of fill pressure,
which were taking place between 3-6 times per shift. As a result of these
changes, Elastimold reported savings of $30,000/year per machine, with
a payback of less than two months and will proceed to install these silicon on sapphire
pressure transducers on all of their 100+ machines in three plants.
Maximum shot weight is a very important
property, because it controls surface uniformity, appearance, and gloss
as well as physical properties. In the case of pin-point gated parts, one
has to be careful that the gates do not freeze before a totally full shot
can be accomplished. In a recent experiment with such a four cavity mold,
parts with dull surface finish and blemishes, thought to be from the tool,
were actually caused by incomplete fill, as shown in Fig 10. When the melt
temperature or mold temperature was raised, the fill improved, and surface
appearance improved remarkably. The difference in shot weights between
the two sets of 22.5 gram parts was 0.15 grams, or 0.67%, an actual difference
of nine more pellets (60/gram) in the best appearance parts compared to
the poor appearance parts. All this was caused by the gates freezing before
the part was completely filled!
Another problem in achieving consistant
shot weights concerns PLC microprocessor programming. Because typical microprocessors
have finite programming possibilities due to their microprocessor size
and capacity, the range of positions or pressure points are often compromised.
For example; a linear transducer is typically repeatable to 0.001
inches, yet for programming purposes, the number of positions available
for selection by the PLC is reduced to less than 0.001 inches apart to
perhaps 0.003 to 0.010 inches apart. Accordingly, a machine inherently
capable of controlling linear position of the shot to 0.001 inches
may actually be limited by the PLC program to 0.003 to 0.010". For
a 50mm, or 2 inch screw, this means that repeatability will range between
0.24 to 0.72cc, or in case of polystyrene, a variation of 0.25-.75
grams per shot. For a 100 gram shot, this variation amounts to 0.25-.75%,
while for a 50 gram shot it amounts to 0.50-1.50%, and for a 25 gram shot,
the variaton amounts to 1.0-3.0%. This is another cause of precision variation
that can be eliminated with Silicon on Sapphire melt pressure control.
There are a number of good reasons
to monitor and control the plastic melt in the nozzle area, Table VIII.
As in all plastics processes, the plastic melt pressure and temperature
define the quality of the plastic at point of injection. Maintaining uniform
melt pressure and temperature is important for quality repeatability from
shot-to-shot. Also, shot weight is controlled by pressure, both peak and
pack, to fill the mold.
Hydraulic pressure is not a true indication
of nozzle pressure due to frictional losses from the multiplying piston,
screw and molten plastic in the barrel. Anyone who has removed a screw
from a machine with molten plastic in it can testify that the material
friction represents a very large frictional pressure drop. Typical pressure
drops range from 10 to 30%, depending on the viscosity of the melt. Glass-filled
engineering plastics represent the highest pressure drops. While we have
seen fairly high correlation coefficients between hydraulic pressure and
nozzle melt pressure areas, the constant poor repeatability of 300-450 psi,
versus the best possible repeatability of 20-40psi with a silicon on sapphire
transducer precludes perfect correlation. If it is important for shot weight
precision to control in the nozzle, mold or hot runner manifold to an accuracy
of 20-40 psi, then that is what must be done, and only Silicon on Sapphire transducers
can accomplish it.
Additionally, continuously monitoring
nozzle melt pressure and temperature alerts operators to the key dependent
variables in the process. Barrel temperatures do not indicate real time
plastic melt temperatures, just as hydraulic pressures do not indicate
real time nozzle melt pressures. Different types of plastics develop different
amounts of adiabatic frictional heat in the barrel. Molding ten or twenty
different plastics materials a month on a single, general purpose screw
in a 20/1 L/D extruder, will not generate the same results for all the
plastics and the screw will probably be close to ideal for no more than
20-30% of them. Any molder interested in ISO-9002 should monitor and control
the nozzle melt conditions. These dual pressure/temperature Silicon on Sapphire transducers
are the best and most accurate sensor for the job.
Conclusions
The author has shown in a number of
experiments and examples (Fig 5), that melt pressures at the nozzle of
standard injection machines can typically fluctuate 2-3%, and is caused
by the repeatability of the hydraulic and the linear position transducers
employed for control.
As a result, typical shot weight reproducibility
accuracies of injection molding machines are in the range of six sigma
= 0.3%, an amount that is too large for precision molding requirements,
and is fundamentally inherent in machine operating systems due to sensor
type and ambient effects on repeatability and hysteresis.
A new type of plastics melt pressure/temperature
transducer (Fig 9) has been developed which has self-contained, in-diaphragm,
solid-state sensing of both pressure and temperature, within a highly durable
sapphire diaphragm. This new, leading edge technology sensor, offers opportunities
for 10 times better hysteresis-free pressure control, and 50-90 times better
shot weight repeatability over currently available pressure transducers
used in machine hydraulic systems or in external sensing points at the
nozzle or in the mold.
Tested in numerous types of machines,
both in mold and nozzle locations, with a wide variety of molded products
and materials ranging from thermoplastics to thermosets, this new sapphire
pressure/temperature transducer is destined to become the new paradigm
for precision control in the injection molding industry. It eliminates
hysteresis and temperature effects on pressure readings, and has a true
repeatability of 0.1% of full scale. It is ideal for ISO-9000 molders
wanting to monitor and control the plastics melt quality in their injection
molding operations and is the only sensing method suitable for unmanned
after-dark molding.#
References:
1) E. Griesser & K. Pinkham, "Analysis
of Transducer Accuracy on the Repeatability of the Injection Molding Process"
SPE ANTEC, 1996.
2) RG Speight, EP Yazbak, et al, "In-Line
Process Measurements for Integrated Injection Molding. SPE ANTEC, 1996
3) N. Kudlic & W. Michaeli, "Reproducibility
of Injection Molding Machines"
SPE ANTEC, 1996