Breakout detection for fast EDM drilling by classification of machining state graphs

Abstract

Due to its capability of machining hard-to-cut materials as well as its high machining efficiency as compared with conventional electrical discharge machining (EDM) processes, fast electrical discharge drilling (fast EDM drilling) is widely applied in industries such as mold and die as well as aerospace component manufacturing. The breakout detection is an essential technique for hole completion judgment and back-strike prevention. This paper presents a novel method, called classification of machining state graphs (CMSG), for online detection of breakout events. A machining state graph (MSG) is formed by the recent changing patterns of feature signals, which would change abruptly when breakout happens. Then, the detection problem is solved by classification of real-time MSGs. In this paper, the feature signals were selected to be normal discharge ratio, short circuit ratio, and servo feedrate of the tool electrode. The signals were preprocessed in order to improve the detection accuracy and reduce the decision lag. A classification model was built to classify MSGs. To simplify the modeling process and improve the generalization ability of the detector, a pattern recognition (PR) algorithm was adopted as the core algorithm for classification. The classification model of the detector was acquired through offline training and loaded on the start-up of the control system for online detection. Performance judgment criteria were proposed and experimental results proved the high performance of the proposed method.