LAUSR

Abstract Process anomalies and fluctuations in product quality are widespread problems in discrete manufacturing. There have been various control approaches to tackle the challenge. This paper presents a cross-process control approach that combines engineering knowledge and data analytics techniques. An initial rule basis is generated by experts using simulation models. To achieve a data driven enhancement concerning process and product quality, a PLC-based connector is developed to record and unify real process data from heterogeneous data sources. The data is processed in the cloud and inferred using online modeling techniques. Neural networks with autoencoder structure are applied to extract unknown features, to iteratively refine the knowledge base and thus to optimize quality control.