Wire Electrical Discharge Machining (WEDM) is widely used for the precision manufacturing of complex components, particularly those made from difficult-to-cut alloys. One of the main challenges affecting process stability and accuracy is the lack of information about the actual discharge location along the wire during machining. Conventional methods face difficulties to monitor or control this spatial discharge behavior in real time, leading to energy loss and reduced efficiency, especially when the workpiece thickness varies. This study proposes a data-driven methodology for discharge localization in WEDM based on process signals and machine learning modeling. A novel electrical indicator, defined as the voltage-scaled difference of the reciprocals of the upper and lower discharge currents, was introduced to quantify the spark position along the wire. Calibration and validation experiments confirmed a monotonic and physically consistent relationship between this indicator and the actual discharge position. Machine learning regression techniques, particularly isotonic regression, were applied to establish a robust mapping model capable of achieving over 96% localization accuracy with spatial errors within 2 mm. Furthermore, the developed discharge localization framework can be extended to workpiece thickness estimation and adaptive adjustment of machining parameters. Experimental results demonstrate that, when integrated into adaptive control, the system improves discharge stability and energy utilization efficiency, leading to a 25.5% increase in machining productivity, and a 28.4% decrease in total energy consumption. This physically interpretable, data-driven approach provides a solid foundation for intelligent monitoring and process optimization in WEDM.
Wang et al. (2026) studied this question.