Refined identification of structural defects in compact GIS based on X-ray digital imaging distortion compensation

The GIS equipment features a complex internal structure, where structural defects such as loose parts and cracks can significantly impair its normal operation. To address the distortion in radiographic images caused by the geometric relationship between the X-ray source and the detector, which reduces defect identification accuracy, this paper proposes a refined identification method for GIS structural defects based on X-ray digital imaging distortion compensation. A mathematical model for GIS structural defects is established through nonlinear dynamic analysis of the GIS equipment. An X-ray digital imaging system is employed to scan the GIS equipment and acquire digital images of its internal structure. Two-dimensional discrete stationary wavelet transform is applied to denoise the X-ray images. Subsequently, the digital phase mask method is utilized to perform distortion compensation on the denoised images. The fully distortion-compensated image is used as the input to an ensemble learning algorithm. In this algorithm, a support vector machine classifier is chosen as the base classifier, while the AdaBoost algorithm acts as the ensemble classifier to generate the final refined identification results for GIS structural defects. Experimental results demonstrate that the proposed method improves the accuracy of refined identification by utilizing distortion-compensated X-ray digital images of the GIS equipment.