Running In State Recognition And Quality Optimization Design Based On Multiple Analysis And Attractor Shape

Friction wear occurring on the interface of relative motion is one of the important factors causing component failure and material loss.It not only results in significant economic losses due to material consumption,but also significantly reduces the operational reliability and service life of key components.Therefore,conducting research on the wear state identification and quality evaluation of mechanical equipment is of great significance for improving the stability of mechanical equipment operation and prolonging its service life.It has been proven to have strong system dependence and time-varying characteristics.Extracting features closely related to wear status and wear performance from friction signal time series using nonlinear dynamic methods has become a new research direction.This paper uses multifractal and chaos theory to extract dynamic features of friction signals generated during the wear process,analyzes their evolution laws during the wear process,and uses this to construct an optimization design of wear parameters based on dynamic analysis results.The main work and achievements of this paper are as follows:(1)Considering that the measured friction signal contains interference information such as inherent background noise((quasi-)periodic components)of the experimental device and environmental random noise(random components),the empirical mode decomposition(EMD)method is applied to extract nonlinear components.According to the power spectrum analysis results,the dynamic characteristics of each-order intrinsic mode function(IMF)component obtained by EMD decomposition are identified.IMF components with quasi-periodic and random features are removed.The remaining IMF and residual components are reconstructed,and the mean squared error(MSE)and normalized cross-correlation(NCC)between the reconstructed signal and the original sequence are calculated to verify the reconstruction accuracy.The experimental results show that EMD decomposition can effectively realize the extraction of nonlinear components of the measured friction coefficient signal.(2)The multifractal detrended fluctuation analysis(MF-DFA)algorithm is used to characterize the multiple analysis properties of the friction coefficient:Firstly,the influence trends of the fitting polynomial order,weight factor,and sliding window scale on the multifractal spectrum and spectrum parameters are analyzed based on the binomial multifractal theory model.The results show that the weight factor |q|max=10 causes small errors at the end of the multifractal spectrum and the range of sample sizes from 8000 to 3000 can save calculation time while ensuring calculation accuracy,and the influence of the fitting polynomial order on the analysis result is minimal.On this basis,the multifractal spectra and spectrum parameters of the friction signal in different wear stages are extracted using the sliding window method.The results show that the multifractal spectrum of the friction coefficient signal exhibits different positions and shapes in different wear time but overall shows an inverted bell-shaped curve,indicating the multifractality of the frictional coefficient signal.However,there is no obvious variation rule for the multifractal spectrum parameters,which is due to the fact that the variation of the friction coefficient signal during the wear process mainly reflects the time dependence of the amplitude,and the degree of signal fluctuation does not change significantly.(3)In order to overcome the difficulty of the MF-DFA algorithm in characterizing the time-varying friction coefficient signal with almost identical oscillation amplitude,the attractor morphology analysis method is introduced to improve the wear state identification accuracy.The friction signal is reconstructed into a high-dimensional phase space,and the principal component analysis method is applied to project the high-dimensional data onto a three-dimensional space/two-dimensional plane to construct the wear attractor.The trajectory of the wear attractor gradually converges during the wear process and eventually forms a stable"line bundle," and the correlation dimension shows a trend of "initial oscillation gradually increasing and final stability." The attractor projection matrix two-dimensional distribution matrix parameters are used to construct the attractor morphology,and the effectiveness of the new characterization parameters is verified by combining the three-dimensional histogram of the phase point.During the wear process,the two-dimensional distribution matrix parameters of the attractor projection matrix follow the evolution law of "decrease-stability," which realizes the effective characterization of the chaotic attractor morphology of the friction signal and can be used for wear state recognition research.Compared with the correlation dimension,this group of parameters can effectively describe the variation law of the amplitude of the friction signal sequence during the wear process.(4)By changing the contact load,sliding speed,and initial surface roughness of the lower specimen,a multi-factor and multi-level orthogonal test of the wear process is carried out.The stable friction coefficient signals are extracted to extract feature parameters,and the range analysis and one-way analysis of variance reveal the system dependence of the multifractal spectrum width and the two-dimensional distribution matrix parameters of the attractor projection matrix on the contact load,sliding speed,and initial surface roughness,respectively.A multi-objective optimization design of wear parameters is performed using the nondominated sorting genetic algorithm with elite strategy,and the Pareto optimal solution set and corresponding optimal wear parameter set are obtained.The fitting results show that to obtain stable friction signals and form better wear surfaces,combinations of larger roughness and larger load or speed should be avoided.