Anomaly Detection Method for Hybrid Workpieces Using Dynamic Time Warping

Hybrid components consist of multiple materials, enabling the material distribution to be tailored to locally varying loads during the use phase. By selectively applying materials with high strength and density only to areas of a component that will be subjected to high local loads, the overall weight can be reduced. Hybrid components are manufactured through joining, forming, and subsequent machining. Material defects such as cavities or cracks, which can occur during joining and forming, significantly reduce the component's lifetime. These defects can be detected by monitoring the process signals of the machine tool. However, unavoidable deviations in the axial position of the material transition zone cause temporal shifts in the signals, impairing the performance of established monitoring methods. To monitor material defects in hybrid workpieces, this article proposes a new anomaly detection method based on dynamic time‐warping barycenter averaging that is robust against time shifts. For training, time series containing varying temporal shifts are used. The sensitivity and robustness of the new method when applied to hybrid workpieces are evaluated and compared to confidence limits that are common in industrial applications. Using the new method, over 97% of all material defects can be detected with no false alarms occurring.