Application And Research Of Empirical Mode Decomposition In The Analysis And Monitoring Of Cutting Chatter

With the rapid development of high speed and high precision numerical controlprocessing technology as well as the wide application of difficult-to-machine materials inheavy equipment, cutting chatter has become one of main obstacles in the improvement ofproduction efficiency and processing precision. Classical time-frequency signal analysismethods such as wavelet analysis are based on linear Fourier transform and thus have intrinsicdeficiencies in accurately extracting symptom features of nonlinear and nonstationary cuttingchatter signals. In this thesis, based on the theory of empirical mode decomposition, anadaptive decomposition method is used for the fast and effective recognition of cutting chattersignals.Based on the two key concepts of empirical mode decomposition, instantaneousfrequency and intrinsic mode function, the theory and steps of empirical mode decompositionare analyzed. The decomposition process of this method is adaptively based on time scalefeatures of signals. The advantages of empirical mode decomposition are discussed.With impeller milling process as an example, empirical mode decomposition is used toadaptively decompose spindle vibration signals. Two values, mean square frequency andone-step autocorrelation function, are extracted from intrinsic mode function as sensitivefeatures of cutting chatter. The change of the two features is earlier than that of signals’amplitude in time domain and they can be regarded as possible symptoms of cutting chatterrecognition. The least-squares support vector machines is used as multi-classification methodto accurately recognize the status of cutting vibration in impeller milling process.Aiming at solving the problem of the real-time monitoring of cutting chatter in numericalcontrol machining, an empirical mode decomposition-based monitoring module is developedand successfully integrated in a machine tool monitoring system. This system, which canintegrate with various numerical control systems, has good cross-platform ability andexpansibility. In hardware design aspect, the general structure of monitoring system isproposed and hardware selection such as sensors is discussed. In software design aspect, Qt isused for interface development and C language is adopted for the realization of empiricalmode decomposition-based signal analysis module. With the experiments of300M steelmilling machining and impeller milling process, the validity and practicability of proposedcutting chatter monitoring system are verified.