Machine learning for quality prediction in abrasion-resistant material manufacturing process

Abstract

Quality monitoring and prediction plays a key role in improving product quality and achieving automated quality control in manufacturing processes such as the abrasion-resistant material manufacturing process. Traditional methods that rely on the use of first-principle models are difficult to formulate due to the increasing complexity and high dimensionality of manufacturing processes. Data-driven machine learning methods offer an efficient way to learn models for quality prediction, in which the meaningful process information can be learned directly from large amounts of measured process data at different stages. In this paper, based on data collected throughout an abrasion-resistant material manufacturing process, product quality prediction of burned balls is achieved with the use of the Support Vector Machine classification algorithm.