Research On Machine Tool Cutting Excitation Techniques And Dynamic Parameters Estimation Based On Responses

The performance of machine tools is affected greatly by the dynamics of the structure. Experimental modal analysis (EMA) is one of the main methods to identify the dynamics of structures in static state. However, due to the great differences of dynamics between machining operation and static state, the results of EMA may not characterize the dynamics of machine tool accurately. Operational modal analysis (OMA) is a new identification process which derives the dynamic information only from structural responses (outputs) while the structure is in operation. Owing to its practical and economical advantages and great potential for applications, operational modal analysis attracts large increase of research recently. However, the theory of OMA isn’t yet perfect. Because of the absence of input information, it is hard to estimate the frequency response function (FRF) which is of great significance to many applications. Also, many new problems occur when OMA comes to Mechanical Engineering. First, it is hard to meet the assumption of white noise excitation of OMA for mechanical structures like machine tools. Second, the responses of rotation equipment often contain periodic signals which will cause the severe problem of distinguishing between natural frequencies and harmonic frequencies. The goal of this dissertation is to deal with the above problems, to contribute some valuable studies and trials on excitation methods, signal processing, operational modal analysis theories and automatic modal parameters identification algorithms, and to propose a complete methodology for OMA carried out in mechanical structures like machine tools.First, the dissertation presents lots of cutting excitation methods to generate random impulse excitations and to uncouple the dynamics of the machine tool structure and the cutting effects. A mathematical model of the random impulse signal is built and simulated to analyze the relation among the frequency range and power of the exaction, the impulse parameters and the cutting parameters. The patterns of the excitation signal is summarized to provide theory foundation for control and adjustment of the excitation. Random rotation cutting method, random steps cutting method, random single (curve or polyline) step cutting method, gradual changing step cutting method and step point cutting method are proposed to generate the excitation and uncouple the dynamics of machine tool structure and the dynamic effects of cutting process.Second, frequency response function is estimated based on the idea of using the calculated cutting force excitation instead of measuring. Valuable information, such as the cutting parameters, the tool geometry, the workpiece material, and the cutting coefficients which are all known is made use of to predict the cutting forces. Then the calculated cutting forces are employed to replace the missing input measurement to estimate the FRF together with measured responses. A function is introduced into the cutting force prediction model to evaluate the influence of spindle speed variation on the cutting forces.Third, identification methods are presented to eliminate possible spurious modes during operatonal modal analysis. The spurious modes are first categorized according to their origins and characters, and then identification methods are proposed according to their different characters. Two novel tools, the spectrum abruptness and the frequency fence, are proposed to identify and filter out the harmonic modes even when the harmonic frequencies are not known.Finally, the dissertation discusses the key problems of automatic operational modal analysis, the poles automatic grouping and automated estimation of modal parameters from the grouped poles. A so-called poles distribution criteria is presented to efficiently clear the numeric poles based on the regulation that physical modes locate closely while numeric modes spread randomly. An automatic modal parameters estimation algorithm is proposed based on the idea of minimizing the standard deviation of the estimated parameters. Combined with the former harmonic modes elimination method, an automatic OMA modal parameters estimation algorithm Auto-LSCE is realized by Matlab.In general, the dissertation presents a cutting excitation modal analysis method based on output responses only. The work deals with the problems of the OMA theory, such as the estimation of frequency responses function and automatic modal parameters extraction, and new counterparts when OMA comes to machine tool structure like excitation, spurious modes identification. The results are helpful to promote the development of OMA theory and its spread in Mechanical Engineering.