Research of Tool State Recognition Based on CEEMD-WPT

Abstract

As a direct machining tool, the tool will inevitably wear out during production and processing. In order to grasp the wear state of cutting tools accurately and realize the accurate diagnosis in the cutting process, the CEMMD-WPT feature extraction method is proposed, which is based on Complementary Ensemble Empirical Mode Decomposition (CEEMD) and Wavelet Package Transform (WPT). Firstly, the CEEMD is used to decompose the Acoustic Emission (AE) signal that acquired by cutting tool. The AE signal is adaptively decomposed into several Intrinsic Mode Functions (IMFs) among with each IMF contains different time scale characteristic. Then, for less IMFs that still have mode mixing, is corrected with good local processing ability by WPT. The CEEMD-WPT combination algorithm not only can effectively solve the problem of the mode mixing after CEEMD, but also eliminate the influence of frequency mixing and illusive component after WPT treatment. Finally, this work select the first few IMFs component with large energy values, calculate the proportion of the total energy as feature vectors, and input them into the Support Vector Machine (SVM) for training and testing, to establish the tool state recognition system. Compared with CEEMD feature extraction method, the feature extracted by CEEMD-WPT method is more accurate and more representative, which lays a good foundation for later recognition.