Machine learning algorithms in production: A guideline for efficient data source selection

Abstract

Data acquisition, storage and processing becomes increasingly affordable and the use of machine learning algorithms feasible in the field of manufacturing. Even though state of the art machine tools are packed with sensors, the data’s benefits are difficult to assess in advance. Thus, this paper presents a management approach to select the most promising data sources regarding a defined objective. Quality Function Deployment matches the process specific objectives with preselected data sources. The preselection prevents the necessity to examine all possibilities while not restricting innovative solutions. This allows a targeted approach to fully exploit the advantages of machine learning. The approach is validated by a use case based on machine tool data.