Development of a new machine learning-based informatics system for product health monitoring

Abstract

Manufacturing informatics aims to optimize productivity by extracting information from numerous data sources and making decisions based on that information about the process and the parts being produced. Manufacturing processes usually include a series of costly operations such as heat treatment, machining, and inspection to produce high-quality parts. However, performing costly operations when the product conformance to specifications cannot be achievable is not desirable. This paper develops a new machine learning-based informatics system capable of predicting the end product quality so that non-value-adding operations such as inspection can be minimized and the process can be stopped before completion when the part being manufactured fails to meet the design specifications.