Wire EDM failure prediction and process control based on sensor fusion and pulse train analysis

The study aims to develop a neural network classification model to predict machining failures during wire electric discharge machining. Also, a process control algorithm retunes the process parameters based on the remaining useful time before failure. In the proposed methodology, an artificial neural network (ANN) classifier receives four in-process discharge characteristics as input. These extracted features are discharge energy, spark frequency, open spark ratio, and short circuit ratio. Output classes are labeled normal machining, wire breakage, and spark absence. One hundred eight experiments were conducted according to a full factorial design to train the classifier model, with 90% classification accuracy.Parallelly, another trained ANN model predicts the remaining useful time before failure, based on which process parameters are retuned to restore the machining stability. The algorithm was successful in ensuring continuous failure-free machining.