The Research And Application On Ultrasonic Inspection Of The Method Evaluaction Of The BIW Spot Welding

Spot welding has become the main connection technique in the manufacturing process of automobile body in white, and ensuring the quality of car body welding spot can guarantee the stiffness of overall architecture, the integrity and overall vehicle quality. Because of the complexity of the spot welding process, under the condition of guarantee welding complete, it is difficult to identify the quality of solder joint. And destructive after welding inspection and sampling observation is easy to cause the hysteresis and locality, and can't ensure the quality of solder joint meet the requirements of qualified. Ultrasonic nondestructive detection technology is to use the difference between parent metal and weld nuclear crystal structure,which can produce a reflection or penetrate the welding spot core at the detected part interface, which leads to the echo sequence containing solder joint quality characteristic information, then we can distinguish the welding spot quality according to the characteristic informationIn this paper, we analyzed the characteristics of ultrasonic echo and the dependency of welding spot quality, and how to choose welding spot characteristic value. We proposed a feature extraction method based on dynamic threshold according to character of welding spot ultrasonic echo signal to effectively extract these echo signals. On the basis of multivariate statistical analysis, we proposed BP neural network welding spot evaluation method and verified the effectiveness and accuracy of this method by metallographic test.The main work is as follows:1) The extraction of ultrasonic echo signal feature value. Using ultrasonic flaw detector to collect ultrasonic echo signal, and statistical analyzing ultrasonic echo signal characteristics and laws of different thickness welding spot, choosing attenuation rate, bottom echo number, middle echo number and bottom echo spacing as the criteria for welding spot quality evaluation. On this basis, we put forward multi-scale wavelet transform feature value extraction method based on dynamic threshold, which can intelligently identify spectrum peak and effectively extract feature value.2) Welding spot quality evaluation method based on BP neural network.We established a set of experimental system, including sample making,Ultrasonic echosignals collecting, characteristic information of ultrasonic echo signal extracting,characteristic value statistics, to get the data samples of welding spot quality characteristic value?Based on optimized BP neural network algorithm, we chose the eigenvalue and the thickness of the parameters as input data, different types of solder joint quality as the output, so as to realize the identification of the welding spot quality. At the same time, with the calculation of neural network identification,we got the recognition rate of the welding spot t quality.3) We developed welding spot quality evaluation software based on lab-view development platform. It can automatically positioning probe by controlling robots to collect high speed and high frequency real time supersonic echo signals.It can combine wavelet related filtering noise reduction method with Hilbert envelope spectrum analysis to deal with signals, then acquires the signals after filtering. We also can distinguish welding spot quality by optimized feature extraction method and intelligent identification method. In terms of microstructure observation of welding spot quality through metallographic test types, and compared with the welding spot quality evaluation results, we verified the validity of the evaluation method.