Head Defect Detection And Recognition Based On Wavelet Theory Of Friction Welding

Friction welding, which has been using in many high technology areas such as aviation, spaceflight, nuclear energy and ocean exploitation, is a high-quality, efficient and cost saving solid state joint technology. It is possible to produce some kinds of defects in practically welding process, for example lack of bonding, weak bonding etc. Defects in friction-welded joint can make its mechanics characteristic descend sharply so the efficiently nondestructive testing to the defects is becoming very important.In this paper using the ultrasonic to detect the friction-welded joint and get the ultrasonic testing signals. Two denoising methods, which are discrete binary system wavelet transform denoising and lifting scheme denoising, are used in order to efficiently denoise and properly test defects. Following using the ultrasonic testing signals as researching objects, respectively optimize the parameters of two denoising methods and analyze their effects to the denoising, which include decomposition level, wavelet function, threshold value and lifting step, scheme styles, threshold value. Finally by compared it can be found that the better denoising method to this research is the discrete binary system wavelet transform denoising. The successful denoise to the researching signals is the base of the following defects' identification.In this paper putting forward a new feature extraction method, which is using wavelet transform to the signal's power spectrum. By validated two conclusions are acquired, one is that this method can describe the original signals by using few feature dimensions so it decreases the calculation and another is that signals' power spectrum is easier identified than original signals. A kind of BP neural network is established to identify the defects in the friction-welded joint basing on the extracting features. The identifying results are consistent with the testing exactly. This research also shows that the approaching coefficients produced by wavelet transform can represent the original signal's information so they can be signal's features directly.Because the fractal theory is a efficient tool to research into irregular signals in the nature, in this paper using it to analyze the ultrasonic testing signals of the friction-welded joint. Researches show that the ultrasonic testing signals have the fractal characters and can be using the fractal dimension to describe their complexities. So calculate the box dimensions of three kinds of ultrasonic testing signals and get their relations. The researches of this part provide features for the identify of defects in friction-welded joint from the point of view of fractal theory, which is contributed to go deep into researching into the defect's test and identification.