Unsupervised Product Quality Monitoring in Hydraulic Metal Powder Presses using a Minimal Sample of Sensor and Actuator Data

Machine learning and artificial intelligence provide methods and algorithms to take advantage of sensor and actuator data in automated production systems. Product quality monitoring is one promising application for data-driven modeling, in particular in cases where the quality parameters cannot be measured with reasonable effort. This is the case, e.g., for defects such as cracks of workpieces in hydraulic metal powder presses. However, challenges such as an enormous variety of products with different product and process parameters set high requirements for data-driven product quality monitoring. In particular, the monitoring models have to be adjusted to these changes in every single batch. Feasible concepts further have to tackle problems such as unlabeled data and minimum sample sizes of only a few observations. Latter is required in order to receive an accurate monitoring model already after a few products of the current batch. Therefore, this paper proposes an unsupervised product quality monitoring approach for hydraulic metal powder presses based on dynamic time warping and non-linear regression to detect anomalies in sensor and actuator data of minimal sample size. Furthermore, a preprocessing step, isolating only relevant intervals of the metal powder compaction process, is introduced, facilitating efficient product quality monitoring. The evaluation on an industrial data set with 37 samples, generated in test runs, shows a true-positive rate for detected product quality defects of 100% while preserving an acceptable accuracy. Moreover, the approach achieves the output within less than ten seconds, assuring that the result is available before the next workpiece is processed. In this way, efficient product quality management is possible, reducing time- and cost-intensive quality inspections.      «

Machine learning and artificial intelligence provide methods and algorithms to take advantage of sensor and actuator data in automated production systems. Product quality monitoring is one promising application for data-driven modeling, in particular in cases where the quality parameters cannot be measured with reasonable effort. This is the case, e.g., for defects such as cracks of workpieces in hydraulic metal powder presses. However, challenges such as an enormous variety of products with di...     »