Abstract:

In the process of ultrasonic testing of vehicle welding spot, ultrasonic echo is unsteady signal and it is not easy to extract the characters, at the same time many types of defects make automatic determination and recognition of welding spot quality become more difficult. Therefore, an intelligent method to recognize the quality of welding spot based on ultrasonic parameterized and J-divergence entropy model is proposed. By establishing parameterized model of ultrasonic echo signal and based on EM algorithm, a parameters estimation algorithm of ultrasonic echo signal is proposed. According to the time-frequency characteristics of ultrasonic echo and supervising the effectiveness of these characteristics by the J-divergence entropy to get the optimal characteristics subset. Use the method to achieve the recognition of welding spot quality. The actual test results of welds have verified the validity of this method and the accuracy of recognition.

Key words: J-divergence entropy, parameterized model, ultrasonic testing, welding spot quality