A Neural Network Approach to Robotic Thermoplastic Tow Placement Process Control

摘要

An artificial neurla network (ANN) version of a model set for the dominant process physics of thermoplastic tow placement with on-line consolidation has been created. The ANN modle set is used in an optimization scheme for maximizing the fabricated material quality. Using ANN's a determinatio of optimal setpoints is possible in less than 1/2 second. Thus a fast responding control scheme can be implemented using the optimizer, with feedback from hardware and software snesors and communication to and from the robot system in about the same time scale. Several optimization procedures with different quality functions are tested and implemented in a GUI written in Lab VIEW. They show the ability to identify optimum set points for acceptable quality in minimal time.