Estimation Method for Process Conditions of Cold Pressing Using Structural Analysis Data

Abstract

Using the design of cold press work as a case study, this research investigated a system that can determine the countermeasure corresponding to the difference data using machine learning by mapping the difference between the analysis data and the countermeasure taken for the difference data in multiple structural analysis data generated in the development cycle. The authors examined two types of classification: multi-label classification and multi-class classification using distance learning, and confirmed the prediction accuracy of each and the limitations of the system. In addition, based on the above results, we examined the type of analysis data that should be used for learning, and confirmed that the accuracy is better when the type of input data is narrowed down.