Most of us in plastics molding accept the fact that mold development is an iterative process—make it, try it, tune it, and repeat that till it’s making good parts. Normal, right? Yes, but maybe normal needs an engineering review.
Developing and validating a mold that makes Cpk-capable parts is an inefficient, time-consuming, expensive, and frustrating process for many moldmakers. Increased costs, schedule delays, lower quality, and frustration travel right back up the supply chain to the OEM. Revenue, market share, and profits suffer up and down the entire supply chain. Product performance is degraded. Assembly operations can grind to a halt. Tight tolerances, high-cavitation molds, a large number of critical design dimensions, and complex part designs make a tough job even tougher.
If you haven’t heard of it, Algoryx has developed software that conducts mold characterization studies (MCS) intended to demystify moldmaking. Steve Tuszynski, president of Algoryx Inc. (Los Angeles, CA) says his system provides the following benefits:
• Cost: Reduces development costs for the moldmaker by eliminating iterative mold and press tuning cycles.
• Time: Accelerates delivery schedule for the moldmaker and shortens time-to-market for the customer.
• Quality: Provides the customer a mold of the highest possible quality (Cpks).
• Cash flow: Enables the moldmaker to get paid sooner by delivering data-based proof (the MCS) that the mold meets customer Cpk requirements.
• Product performance: Identifies and eliminates unnecessary tolerance relaxations that reduce product performance and can result in assembly line shutdowns.
Figure 1. Changing any one of the five elements of the molding system affects all other elements, resulting in 10 project interactions.
Figure 2. The science behind Algoryx: five press settings, five different levels, 17 setups. This “bead on a wire” graph shows that you can slide the bead along the wire by changing press settings but you can’t take the bead off the wire.
Figure 3. The predictor dimension gives you the optimum tooling adjustment required to make in-spec parts.
Why is mold development such a mystery? Tuszynski says the first and most obvious culprit is that what you get out of the mold depends on which press settings you use to make the parts. This simple fact can have devastating consequences. One combination of press settings can tell you that you need to increase a mold steel dimension to bring the part to target. A different combination can tell you that you need to decrease that same mold dimension to bring the part to target! Some moldmakers will tune a mold four to five times during development and end up right back where they started.
Dimensional sensitivity to changes in press settings also causes moldmakers to ask for tolerance relief—a request is made for a first “problem” dimension, later followed by a second request for a second “problem” dimension, and so on. Many tolerance relaxations are unnecessary, degrade performance, and cause assembly-line shutdowns. Tuszynski says one of the most common complaints he hears from design engineers is, “Those moldmakers keep coming back to me half a dozen times and each time they want to relax a different tolerance! Don’t they know what they’re doing?”
The second culprit is the fact that injection molding is loaded with, to use DOE lingo, simple and complex interactions. These interactions result in complex response surfaces that confound decision-making. DOE methods work for only the most simple and trivial cases (two to three critical dimensions or cavities, for instance). Algoryx is not a DOE method.
The third culprit is what Tuszynski calls project interactions (Figure 1). Any time you change one of the five elements of the molding system, the change affects the remaining four elements. The resulting 10 project interactions add to the complexity of mold development.
The fourth culprit is inertia. Both management and operational staff are reluctant to re-engineer and rethink their technological and quality processes. This inertia, coupled with a fear of statistical methods, prevents them from adopting more efficient methods.
What do OEMs and molders do to work around these problems? Some have the moldmaker deliver a steel-safe mold. The molder then incrementally tunes the mold and requests tolerance relaxations until the molder is able to make “good enough” parts. Some ship the production press to the moldmaker for mold tuning and then ship the mold and press combination back to the molder for validation. This illustrates the expensive lengths some OEMs will go to in their attempt to solve the moldmaking mystery.
Some OEMs buy a mold that meets a mold (steel) specification instead of buying a mold that is capable of making good parts. Tuszynski notes that there is a world of difference between the two approaches. OEMs can order a mold in three ways:
1. Apply a constant shrink factor to all part dimensions and say, “Build me a mold to meet these steel dimensions.” This is the worst way, because while the mold may meet the mold drawing, there’s no guarantee the mold will make good parts.
2. Take the previous step and say, “Because we’re unsure, we’ll build some mold dimensions steel-safe” so the mold can then be shaved several microns at a time to get to where it needs to be without overshooting. Steel-safe, says Tuszynski, means you’re building a mold that’s virtually guaranteed to make bad parts.
3. Give the part drawing to the moldmaker and say, “Build me a mold that makes parts that meet this print.” The moldmaker gets good parts on his tryout press, but when the mold is shipped to the molder’s press, the molder may be unable to make good parts.
All of these techniques are inefficient, costly, and can result in molds that produce marginally acceptable parts. In some cases, molders give up and simply block off the cavities that produce bad or low-Cpk parts. Production costs increase.
How does Algoryx solve these complex problems? The answer, says Tuszynski, begins with the fact that all part dimensions produced by a mold are linearly related. Furthermore, one dimension in one cavity is the statistically best predictor of all other dimensions in all other cavities, the predictor dimension (PD).
Figure 2 shows an example, based on real part data, of the linear relationship (regression line) between two dimensions. When press settings are changed, that changes the location of where the part dimensions are located along the regression line. Think of this as a bead on a wire. You can slide the bead along the wire by changing press settings but you can’t take the bead off the wire. The regression line in Figure 2 represents the relationship between these two dimensions for all possible combinations of press settings in the process window.
Figure 3 shows how to perform optimum mold tuning, whereby the change is the offset between the regression line and the design target intersection. Algoryx uses the regression lines in conjunction with spec limits to eliminate unnecessary tolerance relaxations. By tuning the mold and tuning tolerances, molders are able to define changes required to meet Cpk requirements, to simulate and test changes prior to issuing ECOs, and to eliminate iterative development.
Algoryx will generate the biggest cost and schedule savings, Tuszynski says, when the MCS is done immediately after the mold is made functionally operational. He recommends using scientific/decoupled molding to define the center of and limits to the process window.
During production, Tuszynski says you can expect Algoryx to help in these areas:
• Eliminates one shift per week for 24/7 operations.
• Reduces energy costs by 4%-5%.
• Measures only the PD in one cavity instead of all in-process dimensions in all cavities.
• Monitors only the PD to accept good shots and reject bad shots to allow for a less costly and complex vision system.
• Uses the difference between the PD and the operating target to drive press control settings to ensure that only good parts are produced.