Anomaly Detection Strategies for NVH based Production Quality Testing of high volume Automotive Electric Drive Units and related Rotating Machinery (en)
* Presenting author
Abstract:
The integration of machine learning methods into production quality testing of high volume (automotive) rotating machinery became increasingly immanent over the past months and years. In this follow-up paper we want to delve into more recent progress in the application of NVH based anomaly detection techniques, specifically focusing on feature selection, performance and ranking techniques. Also tool chains and necessary pre-processing steps like dataset handling, consolidation and balancing will be identified. Finally strategies for clustering and validation of training results are discussed, concluded by a glimpse on chances and challenges of integrating the discussed NVH methods into global corporate IT big data and cloud processing infrastructures.