Study On Multi Coupled Nonlinear Vibration Signals De-noising Method Of Mechanical System

Generally, measured vibration signals of mechanical equipment in modern industrial production are often annihilated by noise due to the complexity of the working environment, which has brought great inconvenience to the subsequent fault diagnosis process. Therefore, It is necessary to deal with the noise for collected vibration signals of mechanical system. Usually, the vibration of different position points in the mechanical system is related to each other, so there is usually a coupled relationship between the signals collected at multiple locations. The coupled of multi-point signals is more significant especially for the flexible body. If the commonly used single point de-noising method is used to remove the noise from these signals, the coupled information is usually will be treated as noise filtering.Therefore, it is very necessary to de-noising the multi-point coupled signals collected in the mechanical system simultaneously and synchronously.This paper set the rolling bearing system multi-point coupled vibration signals as the research object, carry out multi-point signals noise reduction research based on Kernel Principal Component Analysis(KPCA). In the same mechanical system,the vibration signals of different position points often have the characteristics of coupled relation, for this characteristic, an improved KPCA is used to de-noising the multi-point signals to keep the coupled information between these signals. For the problem that the effect of KPCA on the signal de-noising is greatly influenced by the kernel function and its parameter, and there is no method to determine the optimal parameters efficiently and accurately. So, the paper put forward a Parallel Analysis method(PA) to choose kernel parameter excellently in part and a kind of Particle Swarm algorithm(PSO) to choose kernel parameter excellently in overall situation.The main content of the research are as follows:1. The principle of KPCA denoising is studied, and this main theme will be analyzed from the following several aspects: the vibration signal of outer ring rolling bearing, the characteristics of Multi-point signals in mechanical system and the de-noising effect of kernel width parameter of KPCA. Denoising method based on phase space reconstruction and KPCA is proposed,The Multidimensional Scaling(MDS) is used to reconstruct the signal after noise reduction, So the KPCA is introduced into the field of mechanical vibration signals de-noising from the image de-noising field.2. For the multi-point coupled nonlinear vibration signals de-noising, a de-noising method based on KPCA of kernel parameter local optimization is proposed. A parallel analysis method is proposed for the joint optimization of the kernel width parameters of the radial basis function(RBF) and the maximum principal component, It can select the optimal kernel width parameter in a given local interval quickly and accurately, so as to improve the de-noising effect and efficiency of KPCA. And this method is applied to the multi-point coupled nonlinear signal denoising, and the effect is good.3. For the multi-point coupled nonlinear vibration signals de-noising, for the defect of optimization of kernel parameters by PA. Put forward a de-noising method based on KPCA of kernel parameter global optimization.Improve the original particle swarm optimization(PSO), set the mean square error between the signal and the noise signal as the fitness function of the PSO, so as to optimize the kernel width parameter of RBF within the range of real numbers. The multi-point coupled signals are adopt to verify the de-noising performance of this method, and the effect is excellent.