Feed-forward control: how it improves injection molding.

A new generation of injection molding machine controllers "looks ahead" to see where the process is going not simply where it's been.

What if an injection molding controller could predict the future and then react to process events at the very moment they happened? "The controller becomes ideal," answers Tom Bulgrin, senior software engineer at Van Dorn Demag. No longer would controls merely wait for the process to deviate from setpoint before responding. With a truly predictive control, molders could say goodbye to clamp-braking mishaps, sloppy injection profiles, temperature overshoots, and just about any other problem rooted in controller response time. Far from being some imaginary form of high-tech fortune telling, predictive control is commercially available thanks to "feed-forward" control techniques.

At least four makes of injection molding controllers now incorporate some type of feed-forward control, a well-known form of digital signal processing. "Feed-forward has been around for a while in the continuous processing industries, but now we've applied it to a non-linear, seemingly unpredictable system like injection molding," says Bulgrin, one of the designers of Van Dorn's Pathfinder controllers. Cincinnati Milacron traces the lineage of feed-forward controls on its Roboshot servo-electric machines back to machine-tool applications.

Van Dorn Demag uses feed-forward to control motion and temperature; Milacron uses it for melt pressure and velocity. Ube Industries  uses feed-forward to predict and control the transition between injection speed and pressure control. And Toshiba Machine Co. applies feed-forward in a more limited way - to control temperature only.

LOOKING OVER THE HORIZON

How does digital feed-forward predict the future? As Bulgrin explains it, feed-forward is a type of "digital filtering technique." Deep within the controller, the digital filter, which consists of various mathematical operations, transforms incoming digital signals representing the state of the process. With a properly designed filter, the rate of change of a known signal can be used to calculate the signal's predicted value in the near future - say, a few milliseconds from now. Bulgrin says one of Van Dorn's feed-forward filters looks ahead by "simple linear extrapolation."

The four digital controllers described here all apply feed-forward to a similar task - cutting controller response time and variability to improve machine performance. But the similarity ends when you compare how each supplier works feed-forward techniques into an overall control strategy.

* Van Dorn Demag: In the new PLC-based Pathfinder controllers, digital feed-forward applies to all the machine's motion-control functions, including clamp braking, ejector speed cut-off, and injection speed. Van Dorn also uses some feed-forward elements as part of a much more complex temperature-control model. "In temperature control, feed-forward is only a small part of a more advanced theory," says Bulgrin.

In all cases, though, digital feed-forward removes or reduces response delays- time lags created as the PLC sequentially scans inputs, executes programming, and issues some output to trigger a new process operation. Bulgrin points out that by the time traditional controls finish all these tasks, whatever you were trying to control may have overshot its setpoint.

Digital feed-forward also takes aim at "scan-time uncertainty," or the variation inherent in response times. Tom Richards, Van Dorn Demag's manager of software development, explains that this variation comes from the fact that events occur at any point within the scan. PLCs, however, wait until the end of a scan to begin a new operation. In other words, some events have been waiting longer than others by the time the scan is completed. "By responding instantaneously to predicted events, you eliminate that scan-time uncertainty," Richards says.

Van Dorn Demag believes that digital feed-forward control necessarily goes hand in hand with a unique PLC programming method. Called "state-transition programming," it conceives of the injection molding process as a sequence of discrete states - such as mold closing, injection, or screw recovery - joined by critical transitional events. According to Bulgrin, state-transition logic addresses the fact that feed-forward prediction is a linear technique used in a process with sudden nonlinear changes - like solenoid valve shifting. "With every shift of the solenoid, the entire problem changes suddenly," he says. Van Dorn's digital filter anticipates these transitional events at least one PLC scan in advance and than mounts a response that coincides with the actual event. "It pre-arms the controller for the next state," as Richards puts it.

Richards cites clamp travel as one of the classic problems addressed by digital feed-forward. "Let's say we're going from 'clamp-close-fast' to 'clamp-close-slow' within one inch of slamming the molds together. In the past, you dialed in one inch and 'go slow,' but there would be tweaking - that is, moving back - until you were sure the clamp could react in time. But say you want a faster cycle time" so you increase clamp speed and bang, the clamp slams."

By contrast, with the anticipatory abilities of feed-forward, this clamp-braking transition is based on time not position. "If you dial in one inch, you will get one inch, regardless of speed and pressure," says Richards.

* Cincinnati Milacron: On the Roboshot all-electric machines built by Fanuc in Japan, a 32-bit RISC-type microprocessor controller combines digital feed-forward techniques with adaptive control. The heart of this Series 16 controller is Real-Time Melt Pressure Control, or RTM.

When using the Series 16, an operator typically starts off by running the machine in a traditional closed-loop mode - with the RTM turned off. The melt-pressure profiles developed in this closed-loop mode can than be stored in the controller's memory. Once RTM is on, it "learns" how to duplicate that melt-pressure trace, making velocity-profile alterations automatically and in real time. "It will change the injection process to achieve the pressure curve," says product manager Kent Royer. "We're able to have process repeatability even with external influences such as changes in temperature, regrind, viscosity, or screw and barrel wear." Additional process alarms can be set around the pressure curve and extruder torque requirements to detect material changes.

In a separate but related RTM mode, users can edit existing melt-pressure traces - or create new ones from scratch. In this case, the system adapts to the edited trace, again by making appropriate velocity-profile changes. "The operator doesn't have to calculate velocity changes; he just tells the machine what sort of pressure curve he wants," says Royer. Whether adapting to a stored or edited curve, the RTM typically brings the process to the requested profile in three cycles or less, Royer adds.

The RTM system uses feed-forward to predict where the pressure will be one or more micro-processor scans into the future. With that data, the system's "learning engine" can continuously adjust velocity to reach pressure-profile targets.

The Series 16 controller also makes use of digital feed-forward techniques for motion-control tasks, such as ejector and clamp speed and position. In these critical functions, the Roboshot machines boast position accuracy and repeatability of [+ or -] 0.0005 in. regardless of speed changes during different phases of the motions. "It's basically a machine-tool controller applied to an injection molding machine and having the same type of repeatability," says Royer.

* Ube Machinery: For the past six years, Ube has used feed-forward elements in its "Software Servo" system, which features injection-speed control based on time and screw position. According to sales manager Dan O'Keefe, the controller's feed-forward elements define where the screw should be positioned at 1-millisec intervals in order to meet a given velocity set-point. In this case, the feed-forward feature seeks to eliminate the effects of response time on velocity-to-pressure transition. "It allows high-speed injection until the moment before the cavity is filled, resulting in superior fill capacity," says O'Keefe. Ube says feed-forward helps it control velocity to 0.1 mm/sec.

* Toshiba: The S10 and V10 controllers both apply feed-forward techniques to temperature control, according to special projects manager George Hinkfus. The controller adjusts heater output continuously to adapt to look-ahead predictions of when the temperature will cross its setpoint. "Feed-forward allows temperature to become uniform more quickly. You don't get the peaks and valleys of traditional temperature control," says Hinkfus.