Abstract: In the process of charging and discharging, the explosion-proof battery carried by the mining rubber wheel vehicle will produce continuous vibration, which is transmitted to the boom through the vehicle structure, making the welding seam of the welding plate gradually loose and difficult to withstand the instantaneous impact force, resulting in local small deformation of the thin plate. However, this deformation is only presented by slight changes in the signal spectrum, which makes it difficult for traditional vision-based detection methods to detect small shifts in the thin plate, resulting in low detection accuracy. Therefore, the deformation detection method of welded sheet part of the boom of explosion-proof storage battery rubber wheel car for mining is put forward. The ultrasonic signal of the welded thin plate of the boom is obtained by Top-Hat transformation, and its time-frequency domain characteristics are analyzed to extract the characteristics of the high and low frequency components during the deformation process of the thin plate. The extracted features of high and low frequency components are taken as the input of Support Vector Machine (SVM). The convex quadratic optimization problem is used to extract the small changes of frequency components in the time-frequency domain features. The dung beetle optimization algorithm is introduced to optimize the penalty factor parameters of SVM, and the optimal decision function is established combining the small changes of frequency components. It can detect the deformation of thin plate accurately and solve the problem that it is difficult to detect the small change of frequency component accurately. It is proved that the proposed method can accurately extract the time-frequency characteristics of the ultrasonic signal after the bending of the thin plate, and the R-square index and Pearson correlation coefficient are close to 1, indicating that the detection precision of small deformation is high.