Iterative Monitoring of Wire-EDM Stability through Pulses Energy Distribution and A K-Means Regression Model

This study proposes an iterative model to capture the pulse train energy distribution generated by WEDM (Wire electrical discharge machining) for real-time stability monitoring. It aims to analyze the correlation between machining parameters and EDM energy features and to predict the risk of wire breakage through statistics of the big data from pulse trains. First, quantitative investigations analyze EDM features, material removal rate (MR), and energy distribution to explore the corresponding logic of each feature. Secondly, the data feature sets are captured to train a regression clustering model as a WEDM model for material removal rate. A logic expression clustering algorithm with a K-Means clustering model is established to predict the machining stability. It is found that one typical symptom in advance of wire breakage is an energy distribution leak from EDM pulse trains. Therefore, an evaluating algorithm is adopted to formulate the corresponding EDM stability index vectors. Finally, based on the stability index vector, the monitoring system issues an early warning before the WEDM deterioration. It is confirmed that this monitoring system can more comprehensively evaluate WEDM stability than the traditional stability index based on the EDM’s abnormal pulse ratio. The experiment results confirm that this monitor strategy can maintain a high processing efficiency in real-time, and an earlier warning of breakage before 0.77 seconds is possible with this model.