Prediction of Geometrical Accuracy in Wire EDM by Analyzing Process Data

In order to digitalize the wire EDM process, data-driven models for the evaluation of the process performance have to be developed. The challenge is the huge amount of data and the stochastically process behavior. Such a model could be used to predict the geometrical accuracy of machined workpieces as well as for development of new machining technologies. In this paper, the produced geometrical accuracy is predicted by continuously recorded process data. For this purpose, electrical parameters are measured and processed by using a FPGA (Field Programmable Gate Array) system to detect spatially resolved single discharges during the process characterized as normal or abnormal discharges. Afterwards, the recorded data are systematically reduced by averaging the process data over defined time intervals. For this purpose, unsupervised machine learning methods are used to ensure that no relevant information is lost. Finally, the processed data is analyzed and statistical parameters are correlated with the machined geometrical accuracy.