A Fast Haar Transform and Concurrent Learning - based Monitoring Approach and Its Application to Tool Breakage Detection

This paper describes an effective monitoring approach for manufacturing processing by combining the recursive in-place growing FIR-median hybrid (RIPG-FMH) filters, the in-place fast Haar transform (IP_FHT) and the concurrent learning (CL). Meanwhile, the approach is applied to detect tool flute breakage during end milling by analyzing the feedmotor current signatures. RIPG_FMH can preserve the shape of signals and provide good noise attenuation in real time; here it will be applied to preprocess the signals of feed-motor current during end milling. IP_FHT can analyze the feed-motor current signals in real time, and a wavelet coefficient is selected for further monitoring. An advantage of the method is that can save the memory place of computer and reduce the computation time by comparison with a traditional wavelet transform. The determination of the suitable thresholds based on concurrent learning is another very important issue of the described method, which can effectively detect the abnormal state of a system. The application procedure and the effectiveness of the proposed method have been delineated by a case study; the result indicated that the proposed approach possessed an excellent potential application to tool condition monitoring in manufacturing. Index Terms FIR-median hybrid filters; Fast Haar transform; Concurrent learning; Recursive; Tool flute breakage; End milling. IEEE Trans. Mechatronics, MT01-068R (Revised)