Metal Corrosion Assessment Method of Power Equipment Based on Multi-Feature Fusion and PCA Dimensionality Reduction Analysis

The "exposure method" was used to conduct field test on Q
235b
steel, a metal material for power transmission and transformation equipment in the substations, and the corrosion images with different corrosion degree were obtained. On this basis, firstly, the relevant characteristic parameter values of images with different corrosion levels were calculated by color moment, information entropy, singular value, wavelet and fractal analysis methods. Then, the principal component analysis(PCA) method was used to reduce the dimension of the multi-dimensional image features. Finally, the particle swarm optimization least squares support vector machine (PSO-LSSVM) algorithm was used to establish the metal corrosion evaluation model of power equipment. The results show that the colour moment feature, information entropy, singular value, wavelet coefficient and fractal feature corresponding to the metal corrosion image can reflect the evolution law of the metal corrosion morphology well. The corrosion evaluation model established by PCA and PSO-LSSVM can accurately evaluate the corrosion degree of metals, and the evaluation accuracy rate is 94.6%.