Declarative Programming Approaches for Robust Anomaly Detection in HPDC Process Data

The increasing demand for lightweight components with complex geometries and high performance has significantly driven the use of High Pressure Diecasting technologies in the transportation sector. However, the production of high-quality HPDC components necessitates precise control of process parameters. Within this scope, the Data and Metadata for Advanced Digitalization of Manufacturing Industrial Lines (metaFacturing) project, funded by the EU Horizon program, is trying to solve some of the challenges. The focus is on a digitalized toolchain that will optimize the use of raw materials, incorporating recycled ones, reduce operator costs and effort, as well as waste caused by out-of-specification production results. In this paper, a solution for the challenge of the structured data characteristics is presented using two differently trained classifiers for anomalous and non-anomalous data in order to improve classification performance. To additionally increase efficiency and utilize process data characteristics, different declarative programming methods are investigated, such as ILP, ASP, CLP, and CP.