摘要：AbstractShear cutting processes are characterised by physical quantities that change in time and space because of influencing process effects. These measurable properties can be recorded with the use of sensors and assigned to the causal process effects through evaluation with analytical models. By making conclusions about these process effects, the product quality of the processed sheet can be monitored based on tool wear effects and continuous part quality can be ensured by preventive tool change. Here, analytical models are often not available or show a quantifiability that needs improvement for the respective process effects. The main reasons for this are cumulative signal interference in the sensor signals caused by the drive concept of the tool movement, the machine dynamics and the control and signal processing cycle of the machine tool to be automated. Previous methods for such prediction purposes are either based on large and complex neural networks or have an unsatisfactory performance. To solve these problems, this paper presents a novel method based on automatic feature extraction to quantify tool wear effects. By linking the recorded process signals, the inherent information content about the tool wear condition can be recovered, and signal interference is significantly reduced. The linking approach pursued describes a two-stage procedure that takes into account the physical origin of the individually recorded process signals from the multivariate time series data and combines them into a new physical quantity through smart pairing. The effectiveness of this method is evaluated based on real industrial data and shows excellent results for mapping the tool wear effects using a K nearest neighbour (KNN) algorithm.
关键词：KeywordsFeature extractiontime series analysistool wearclassificationshear cutting
