Machining Center Spindle System Time Series Analysis

With the progress of science and technology, the modernized machine tool is required high speed, high precision and good machining capacity and efficiency. In order to be satisfied with this development, the research of the machine tool's dynamic characteristics becomes very necessary. Its dynamic capabilities will effect on the whole machine tool's dynamic traits. It is very important to study particularly the spindle system's characteristics.In the dissertation, TH6350 bed machining center produced by the limited Corporation of Xi'an Transportation University and Kunming Machine Tool is the research object. Our study focuses on the dynamic characteristics of spindle system at idle and the modal parameters extraction using time series analysis method.The works have been done include:1.The problems of signal traits are addressed, including Gaussianity Test and the detection of Nonlinearities by using Higher Order Spectra Analysis (HOSA) and Hilbert Transform, respectively. HOSA is a useful tool for dealing with the system's Non-linearities, specially the Bispectrum. This method not only detects the signal nonlinearities but also affords some causal information. Then, Hilbert Transform is based on the assumption of the excitation to be a white noise process. It's main use is to detect the degree of Nonlinearity, strong or feeble.2.Time domain approaches are available in which modal parameters are estimated from output signal only, making assumption as to be the nature of the input. By modeling the time domain response waves on spindle system under working condition and the knowledge of these estimates, natural frequencies of the system are identified directly. Therefore, the system's dynamic characteristics are derived. Tow methods of time series analysis, multivariate autoregressive process and optimum AR modeling, are discussed. Analysis of a multivariate autoregressive process deals with the estimation of the parameters by stepwise least squares method and the order of the AR process is determined by a new criterion function SBC. Optimum AR modeling for every-small segment a more accurate autoregressive model can be built and FPE criteria are introduced to get optimum model order. The analysis of the vibration signal under working condition shows that these methods are valid.3.Detecting nonlinearities in time series by surrogate-data method. This method identified the nonlinearity in the time series is based on the surrogate-data that come into being by the null hypothesis of linear characters such as average and variance of the original data. The statistical discriminating quantum of treatment data and surrogate-data is calculated and compared. If the treatment data accords with null-hypothesis, this is considered as an indication of nonlinear behavior. This research is in order to detect the validity of a linear time series analysis.Some conclusions can be obtained from work above-mentioned:1.After identification and estimation problems of time series modeling under working condition, it is shown that time series analysis can describe the system's dynamic characteristics to some extent. Under some conditions such as in longer time signals, a time series model of an optimum order can be received. It can be instead of the original data. The modeling of spindle system of TH6350 bed machining center illuminates this. On the other hand, the relationship between coefficients of time series model and system's modal parameters is considered. The time series analysis, a time domain approach, is a useful method for research of mechanical system's dynamic characteristics.2.From signal traits analysis and detection of nonlinearities, it is evident that nonlinear components and periodic frequencies are in time series of spindle system of TH6350 bed machining center. Furthermore, the main vibration frequencies of system focus on low frequency domain from 2HZ to 1200HZ and output signal identity is non-Gaussianitys non-symmetry and feeble nonlinearity of quadratic phase coupling. This conclusion can offer valuable reference for the model modification and the choice of methods for dealing with dynamic data.