Spatial Annotation of Time Series for Data Driven Quality Assurance in Additive Manufacturing

Abstract

One of the biggest challenges for artificial intelligence in industry is the lack of labeled application data. Particularly for time series data, labeling requires a large amount of time for data preparation and expert knowledge both in data analysis and in the application domain. In this work, we propose a methodology for labeling time series solving the two barriers identified above in an additive manufacturing use case. Our approach correlates spatial and temporal features of process defects by means of a spatial sensor. By applying our method, we were able to achieve shorter labeling time while obtaining high-quality labels.