Stainless Steel Resistance Spot Welding Ultrasonic Inspection Defect Identification

As a pressure welding technology with high production efficiency,low heat input and easy automation,resistance spot welding has become a reliable manufacturing technology in the field of automobiles and rail vehicles.Because of the high current,short processing and invisible nugget in the process of stainless steel resistance spot welding,the defects such as splashes,pores,and unfused are easily generated,which affect the reliability of the product during service.Therefore,there has a great significance to carry out efficient and accurate ultrasonic testing and defect identification to stainless steel resistance spot welded joints.Firstly,spot welding samples with defects such as pores,spatter,and lack of fusion were prepared,and ultrasonic A-scanning was used to detect the joints to obtain echo signals.At the same time,the defects were verified by metallographic inspection.The wavelet packet transform and empirical mode decomposition were used to denoise the ultrasonic detection signals respectively,and the denoising effects of the two were compared and analyzed.The experimental results showed that the wavelet packet noise reduction had good localization characteristics,and could obtain a higher signal-to-noise ratio(42.48 d B)than the empirical mode decomposition for the noise reduction of the ultrasonic echo signals of pores,splashes,defects and no fusion.Next,feature extraction was performed on the noise-reduced echo information from the time domain,frequency domain,and time-frequency domain,and a total of 18 features were extracted.The correlation between eigenvalues and defects was studied also.The redundant features were eliminated by principal component analysis method,and 11 dimensionless features with a cumulative contribution rate of more than 90% were selected.Each feature could accurately characterize various defects.Taking the obtained features as the input value of the defect identification classifier,a defect identification model based on BP neural network and probabilistic neural network were established.And the BP neural network was optimized by the additional momentum method,At the same the accuracy of defect recognition by different models was analyzed.The research results showed that the correct rates of defect recognition by BP neural network and probabilistic neural network were 98% and 99.7%,respectively.However,the recognition accuracy rate of the BP neural network optimized by the additional momentum method is increased by 1.8%,and both neural networks could obtain better results in the defect recognition problem.Finally,combined with ultrasonic testing technology and EBSD technology,the relationship between ultrasonic parameters and the internal structure of welded joints was explored.By establishing the relationship between the ultrasonic characteristic parameters and the microstructure of different regions of the welded joint,the preliminary exploration of the evaluation of tissue defects by ultrasonic testing wasrealized.The test results showed that the grain size and crystal orientation strength of different regions of the stainless steel spot welded joint had obvious effects on the ultrasonic attenuation coefficient.The ultrasonic attenuation coefficient of each region increases with the increase of grain size and crystal orientation strength,and the mutation occured at the equiaxed crystal.The local orientation difference of the solder joint reflects the internal defect density,and the attenuation coefficient of each region is proportional to the KAM value.