Automatic Tuning and Regulation of Injection Molding by the Virtual Search Method

Abstract

Methodical specification of process inputs for injection molding is hindered by the absence of accurate analytical models. For these processes, the input variables are assigned either by trial and error, based on heuristic knowledge of an experienced operator, or by statistical Design of Experiments (DOE) methods which construct a comprehensive empirical model between the inputs and part quality attributes. In this paper, an iterative method of input selection (tuning) referred to as the Virtual Search Method (VSM) is introduced that conducts most of the search for appropriate machine inputs in a ‘virtual’ environment provided by an approximate input-output (I-O) model. VSM applies the inputs to the process only when it has exhausted the search based on the current I-O model. It evaluates the quality of inputs from the search and updates the I-O model for the next round of search based on measurements of part quality attributes (e.g., size tolerances and surface integrity) after each process iteration. According to this strategy, VSM updates the model only when needed, and thus selectively develops the model as required for tuning the process. This approach has been shown to lead to shorter tuning sessions than required by DOE methods.