Parameters Identification, Contact Analysis And Prediction Of Fractal Rough Surface

The passive intermodulation(PIM) of microwave devices have generated a certain impact on normal operation of many communication systems. However, the main causes for PIM are material nonlinearity and contact nonlinearity of electrical components. Contact nonlinearity is the nonlinear effect, such as electron tunneling effect, caused by inadequate contact between the surfaces of electrical compone nts which are actually rough from a micro perspective. An essential section in the study of microwave devices' contact nonlinearity is the analysis on contact status of microscopic rough surface. Hence, this paper studies a modeling approach of microscopic rough surface based on fractal geometry and its contact analysis method. Then, the variation of contact stress and real contact area is analyzed.1. The modeling approach of rough surface based on fractal geometry and the relevant parameter identification method have been studied. Firstly, the modeling approach and self-affine properties of Weierstrass-Mandelbrot fractal surface function is discussed. By adjusting the fractal parameters, their impact on the morphology of rough surface model is revealed. Then, a fractal parameter identification method based on power spectral density is studied. At last, a fractal parameter identification method based on wavelet transform is proposed. In virtue of the filter property and the limit of energy, wavelet transform can extract the waveform information of different frequencies from contour curve. This method is able to extract both fractal dimension and fractal roughness parameter, and is more precise in contrast with the method based on power spectral density.2. The analytical approach of contact between fractal rough surfaces has been discussed. Firstly, a finite element model of waveguide flange is built to analyze the macroscopic distribution of contact stress on interface of flange. Then, a finite element model of microscopic rough surface is built for analysis on the variation of contact stress and real area of contact on micro-view. Furthermore, the effect of plastic deformation on rough surfaces is also discussed. Finally, estimation on the real area of contact on the interface of flange is calculated by considering both macroscopic and microscopic contact status.3. The prediction method of real area of contact between rough surfaces has been studied. Firstly, several typical support vector regression algorithms and kernel functions have been discussed. And Least squares support vector machine and radial basis function kernel are chosen to contribute the prediction model. Then, the support vector machine is trained, with fractal dimension, fractal roughness parameter and contact load being its input data and real area of contact being its output data. The hyper-parameters of the support vector machine are determined by optimization algorithm, whose objective function is generalization error estimated by K-fold cross validation. The globally optimal solution of hyper-parameters is obtained via a two-stage optimization algorithm combined with coupling simulation annealing algorithm and grid search algorithm. Finally, the real area of contact is able to be predicted through the support vector regression algorithm.