Automation system of polymer press molding by robotic manipulation and bayesian optimization based parameter exploration

Abstract

In this paper, we described an automation system using a robot arm and machine learning for the press process of the polymer materials development. In the system, the press machine is operated through robot manipulation and control signal from the system. As an evaluation of a molded polymer, we constructed a method to recognize the polymer by image processing and calculate its thickness. Evaluation functions were proposed to evaluate the experimental process, which considers the thickness of molded polymer and press time. We implemented experimental parameter exploration by Bayesian optimization for the next experiment using those evaluation values. Through verification experiments, it was confirmed that the system could proceed the experiment sequentially by proposing press parameters that satisfies the experiment conditions.