Ultrasonic Testing And Single Processing Of TiAl/40Cr Diffusion Bonding

Since the following features exist, such as kissing-bond, small defects of dispersion distribution and so on, the quality testing of diffusion bonding is always the research puzzle in the field of nondestructive testing. With the rapid development of computer technology, the fusion of advanced signal processing method and ultrasonic testing technology becomes the research hot-spot, this can provide a new way for the detection of diffusion bonding. The wavelet transformation method was adopted in this paper, and features analysis and extraction were proceeded on the detection signals of diffusion bonding. In addition, signal identification was done based on the method of Support Vector Machines.Diffusion bondings which contain feature defects were made by using the following two methods controlling the surface treatment states before welding and changing the welding technology. The optimal ultrasonic scan parameters were used to detect the samples, and the ultrasonic signal array which correlate with the joint states were obtained.The following two factors were chosen as the judgement criterion, one can achieve the precision feature time-frequency locating towards the interface echo wave signal, and the other can disclosure the implied information connation. And the complex Gaussian wavelet continuous transformation was chosen as the signal analysis method, at the same time the parameters optimization was proceeded.The inner relation between the signal features of ultrasonic testing and interface states of diffusion bonding was analyzed profoundly, and feature extraction was carried out on the signals which were processed by complex continuous transformation. Dimension standard energy, dimension response coefficient and time-frequency phase-shift were extracted from the following three aspects-wavelet domain energy, dimension response, time-frequency phase-shift, and the joint interface states were attributed by using the above feature parameter. At the same time the advantages and applicability of feature parameter were studied profoundly. Training model of feature parameter was proceeded by using the method of Support Vector Machines, and the optimal combination threshold distribution of feature parameter was determined, then the automated identification of ultrasonic signal was achieved, and it can be used to attribute the interface states of diffusion bonding. The experimental results of destructive testing show that this method is satisfactory.