Cost-effective process modeling and optimization methodology assisted by robust migration techniques

Abstract

While data-based models are frequently used as an effective means of investigating a variety of complex processes in engineering, they are typically restricted to a specific process being modeled. This paper demonstrates the application of migration-based approximation to relate similar yet nonidentical processes, using an extension to the model calibration method. A robust method is also used to identify the plausible migration function most consistent with experimental data. In addition, we apply the migration-based modeling method to process optimization wherein a specified target of the process is to be explored; this is achieved by using an infill criterion that assesses the conditional likelihood that a process response will attain the target. Two examples, namely, a pair of test functions and real-world injection molding processes, are used to demonstrate the method.