Online detection of welding pore defects in steel bridge decks based on acoustic emission

To achieve online monitoring of defects in the robot intelligent welding process of orthogonal anisotropic steel bridge decks,a pore defect acoustic emission detection method was proposed based on fast Fourier transform (FFT)and support vector machine (SVM). The acoustic emission characteristics of the welding and defect generation processes in steel bridge decks were explored by conducting robotic welding experiments. The parameters of acoustic emission signals,such as amplitude,counts,peak frequency,and center frequency,in the non-damage and pore defect cases behave with significant overlaps and low correlations. However,the Fourier spectrums of signals from the pore defect case exhibit more high-frequency energy distributions. Therefore,taking spectrums as the input,a radial basis kernel SVM model was established for classifying the two cases. Experimental results demonstrate that the proposed method outperforms other machine learning models,including naive Bayes,random forest,and linear kernel SVM models,in terms of accuracy (95. 4%)and recall (94. 3%). It can be used for online detection of pore defects in the welding process,exhibiting strong robustness and practicality. © 2024 Southeast University. All rights reserved.