The ‘Butterfly Effect’ in Injection Molding—A Connected Process
Curious about the “butterfly effect” in injection molding? Consider this.
According to Robert Gattshall with Plastics Technology:
“A recent column I wrote addressed whether there is a single process parameter that is more important than all the others. There isn’t—and that’s because in the injection molding process, all parameters are connected.
“A connected process is one in which multiple parameters can have an impact on multiple outputs. This makes it difficult to come up with a parameter or setpoint that won’t have some impact on multiple process outputs of an injection molding process. This is the reason behind Scientific Molding—it puts a priority on process outputs rather than the setpoints. Even the most minor, well-intentioned changes to a setpoint can have a ‘butterfly effect’ on an output or multiple outputs.
FIRST UP: SCREW-RECOVERY TIME
“Let’s say I’m reviewing a process that is currently running acceptable products, but I’ve noticed that the screw is recovering the full shot several seconds before the cooling time is completed and the mold starts to open. Thinking that I am optimizing the process to reduce the residence time of the material, I lower the screw RPM, which increases the recovery to what looks to be a more ideal time—just moments before my mold opens. I am not advocating for higher residence times; I am just using this as an example of what can happen to our outputs with a connected process. So, let’s continue this thought experiment and start pointing out what we just impacted in this process.
“First off, I definitely lowered the melt temperature of the material. Lower RPMs means less friction, resulting in less heat. In addition, less time in the barrel for the material also lowers the melt temperature. Lower melt temperature results in a higher injection pressure requirement to maintain the fill time. Lower melt temperature results in a higher viscosity, which means more force is required to fill the mold. If the process was already close to being pressure-limited, this screw RPM change might be enough to put us over the edge. Now, we have lost our velocity control during fill.
“Let’s keep going: Lower melt temperature means parts are going to cool faster—less heat in, less time required to pull it out. Our cooling time, however, wasn’t changed, so now our cooler plastic shrinks to the core and after demolding (ejection) the shrink is reduced, potentially resulting in larger part dimensions.
“Speaking of part dimensions, most molding machines run a process that fills the part under velocity control and packs the part with pressure control. A velocity-controlled process without cavity-pressure control has no real-time control of hold pressure. If my machine requires more injection pressure to fill the parts due to a higher viscosity of the colder material, then my hold pressure would need to go up to maintain the same shot weight and dimensions. Since hold pressure is a setpoint that doesn’t adjust itself, with a few exceptions, my parts are no longer packed out the same, which could result in part dimension issues.
“Another side effect of the lower RPMs could be less material mixing and a nonuniform melt. The higher RPMs could have been what was providing uniform color dispersion if I am using a salt-and-pepper blend, or was adding additional heat that ensured no unmelts in the materal flow.
“We can keep this experiment going, but I think the point is clear. I am not saying that we don’t want to optimize our process if something was missed during process development. We just need to do so knowing what can be affected and make sure that we challenge our changes to verify we didn’t take five steps back while trying to move forward. The above causes and effects do not have to be limited to just thought experiments; we know the relationships to be true based on predictions that were made and then confirmed through experiments we ran during our process-development protocols.
LIMIT PROCESS TWEAKING
“This is what I still find to be amazing about the injection molding process: What seems to be a minor change can have a significant impact on part quality and process robustness and repeatability. This idea of the connected process and the interactions of setpoints is the genesis of focusing on the outputs rather than setpoints.
“That concept is the foundation of Scientific Molding. What troubles me is the fact that many of us know the effects of process changes on outputs—even the unintentional effects they may have on peripheral outputs that are not as obvious—yet we still insist on process limits.
“Process setpoint limits that allow for technicians to make changes to the process throughout a run are counterproductive to a well-defined process. An injection molding process is a network of process interactions and it is nearly impossible to challenge each and every one of them.
“When I visit an injection molder that has well-established process development protocols but then has process limits for production, I can’t help but feel that time has been wasted. All the time spent establishing repeatable processes—running experiments to identify fill time, hold time, hold pressure, cooling time, melt temperature, and more (often with multiple experiments for each of these process outputs)—and then to allow them to change during the run boggles my mind.
“The key to Scientific Molding isn’t whether you run the correct experiments; it isn’t whether or not you have a rheology curve or a fill-balance study. It is about ensuring that outputs remain constant shot to shot, run to run. If outputs do not remain constant, then you can’t predict the end result of the ultimate process output: part quality.
“I have discussed process limits with many molders throughout the years, and the pushback that I have received is still astonishing. I spent a significant part of my career trying to eliminate a certain type of process limits at medical injection molders. The limits I am referring to are large setpoint or output limits that allow for the process to be altered, what I call defect threshold limits.
“I’ve often heard, ‘Well if I can’t adjust this setting during the run, what happens when my parts start showing _ (insert defect here).’ Once I was even told, ‘If I don’t have process limits, what do I do when I get wet material?’ I was basically speechless at that point. It was one of those moments in life when you don’t know exactly what to say, but you’re sure you shouldn’t say anything if you don’t want to be asked to leave the building.
“A well-established process isn’t going to eliminate issues or part defects—nothing can prevent these from occurring at some point. We do not live in a perfect world: steel wears out; material viscosity changes lot to lot; equipment starts to break down; valves no longer perform as well as they once did; molds get dirty; and, yes, material can get wet. But making process changes does not fix or resolve these issues long-term.
“Often, because our process is connected, the change we make to resolve one issue has that butterfly effect we mentioned above, and we have created more harm than good. By the way, going back to my being rendered speechless: I eventually told the molder that they should stop the machine and dry the material before continuing to run.
IDENTIFYING NORMAL VARIATION LEVELS
“Since we know the type of effects that changing just one setpoint on an injection molding machine can have on outputs, it is critical to ensure that your outputs are what we maintain. Once we have identified a robust process that provides output data with low variation, and outputs that the machine actually can achieve, we need to document exactly how much variation is normal for that machine, mold, and material within that process.
“Identifying what is normal variation allows us to establish control limits for the outputs. With these control limits, we can segregate parts made outside of them for further inspection or rejection. The point of process-control limits is not to establish limits that allow us to tweak a process—they should be established to ensure our process is running at the same outputs for which it was validated.
“There is so much about an injection molding process that is quite difficult, if not impossible, to control. That is why it is essential to monitor closely what we can control. I often think many of us confuse defect threshold limits with process-control limits, but there is a significant difference between them. One is the threshold at which a defect is created, the other is a limit that is established to prevent losing process control—thus the term ‘control limits.’
“Effective process monitoring is achieved through robust process development. Process-control limits should be large enough to compensate for normal process variation, but tight enough to prevent process changes and to segregate suspect parts.
WHAT CONSTITUTES A PROCESS CHANGE?
“There are several benefits to understanding the interactions within our connected molding process. We can even use it to our advantage for some of the more difficult-to-monitor outputs of our process. The one that jumps out at me the most is melt temperature. It is by far the most difficult output to measure—we can build smartphones with terabytes of storage, but we can’t measure melt temperature inside a barrel accurately.
“There are several different methods out there, a majority of which have zero repeatability. There have been some advances in ultrasonic technologies that may provide us an economically viable option in the future. Until then, we can use the machine’s outputs to help give us confidence that the melt temperature is similar to what it was during our validation runs.
“First off, it is critical that process changes have not been made. Verify that your fill time, shot size and transfer position have not been changed. Now all we need to do is make sure that our injection pressure is still within the normal variation that we established during process development.
“So, if I am filling the same amount of plastic into the mold in the same amount of time, and the force needed to do so is within acceptable levels, then the viscosity of the material must be similar. We can say with some confidence that our melt temperature must be within the same range as it was during our validation runs.
“Is it a perfect solution? No. It doesn’t give us a reading of what the actual melt temperature is, but by knowing that the other process outputs fall within our control limits, we can use data to support the assumption that the viscosity is equivalent.
“Of course, we also need to define what constitutes a process change. A modification to the hold-pressure setpoint during startup to bring the shot weight within range of the validated process shot weight is not a change to the process. An adjustment to the clamp open or closing settings that changes the output of mold open or close time is a process change. The former is an adjustment to reproduce one’s validated process and a typical part of startup after a changeover. The latter is a change that affects outputs—and not just the mold-open time.
“Let’s think about the interactions discussed earlier: Increased mold-open time causes my mold to cool more, meaning that my steel temperature will be different when plastic starts flowing into the mold. This can cause my parts to shrink more inside the mold and less after they have been ejected, which could result in larger parts.
“More likely is the effect the mold-open time has on the residence time of my material—the material stays in the barrel longer, causing higher melt temperatures. My point here is that we must define what our process is and completely understand what can happen as our setpoints and outputs interact within our connected process.
“Short story: While I was process engineering manager at a medical molder, we discovered an issue in the middle of a production run for one of our largest customer’s highest-priority parts. This job had a dedicated screw and barrel that were supposed to be changed along with the mold when it came time to run, but this time that step was missed. So, there we were with several hundred thousand parts molded and a customer who needed parts—but could they use the parts we had?
“Since we were output-driven from a process prospective, I decided to look at the data from the run and noticed that all the outputs were within the validated process limits. Even with the larger screw and barrel, the setpoints were set to achieve the process outputs that we had identified during process development. During the run, parts were measured and visually inspected and they were not only acceptable, the distribution of the dimensions hit the requirements of 1.33 and 1.67 for CpK an PpK respectively. Armed with all the data, a risk analysis was submitted and approved by the customer. We were able to ultimately prove that even with a different screw and barrel, the process had not been changed.
“I am sure that many of us are aware of what the four plastics variables are: time, temperature, pressure and velocity. When I discuss the interactions of an injection molding machine’s connected process, what I am ultimately referring to is the connection between time, temperature, pressure, and velocity. These four variables interact to form the injection molding process, and our control limits ensure that they influence one another consistently from shot to shot and run to run. Every parameter in an injection molding machine is a method of setting one of these four variables. On the surface, the injection molding process may seem not so complicated to the outside observer—’heat, squeeze, and squirt’ is a phrase that I sometimes hear.
“But our process is similar to the proverbial iceberg, where we see only 10% of the mass above the water, with the remaining 90% requiring us to look deeper to fully comprehend its size. There is so much going on below the surface of the injection molding process—thousands of interactions that influence our part quality. The deeper we dive into it, the more we start to understand why for so many years processing was considered more of an art than a science.”