Characteristics Analysis Of Ultrasonic Interface Signals And Defects Intelligence Recognitions Of TiAl/40Cr Diffusion Bonding

Diffusion bonding has been widely used in aerospace, electronic and nuclear industry, especially in jointing of new materials. Diffusion bonding defects can significantly degrade integrity of interface, strength and service performance of bonding structures. Thus it is significant to evaluate type, size and location of the defects to guarantee the performance and life of the bonding structures. In general, the defects are only a few micrometers in size and very difficult to detect by conventional ultrasonic testing, especially in dissimilar bonding structures. Some ultrasonic energy is still reflected from perfectly bonded interface due to the effect of impedance mismatch between materials to be bonded.Diffusion bonding samples were prepared at various welding temperatures and press. Ultrasonic tests were performed by C-scan method to obtain C-scan images and signals of samples. Characteristics of unbonded, kissing bond, micropore and perfectly bonded area were represented on the points of view of microstructure, shear strength, fractography and ultrasonic interface signals.Characteristics of ultrasonic wave reflected from air gap between TiAl and 40Cr were analyzed. Relationships between reflection coefficient, ultrasonic frequency and air thickness were demonstrated. Amplitude-frequency and phase-frequency characteristics of three kinds of defects and the perfectly bonded interface of TiAl and 40Cr diffusion bonding were studied based on Fourier Transformation. Effects of gain and central frequency of transducers on the above characteristics were analyzed. Significant differences of the amplitude-frequency and phase-frequency characteristics are found between three kinds of defects and the perfectly bonded interface. Three kinds of defects and the perfectly bonded interface can be distinguished by the characteristics. The amplitude-frequency and phase-frequency characteristics aren't influenced by the gain when the increasing of gain doesn't result in the signal saturation. As the central frequency of the transducer increases, the amplitude-frequency and phase-frequency characteristics are still maintained.Time-frequency characteristics were studied based on Continuous Wavelets Transform. Effects of time-frequency characteristics of four complex wavelets, parameters of wavelets, scale and step were analyzed. The unbonded, the kissing bond, the micropore and the perfectly bonded interface can be characterized by time-frequency energy image and time-frequency phase image. Time-frequency analysis demonstrates that the signals have the same reflective characteristics at every second.Defects recognition models were established. Effects of numbers of train samples, kernel functions, parameter optimization methods and characteristics on the model performance were systematically analyzed. Defects of the unbonded, the kissing bond and the micropore are recognized successfully. Empirical formula of relationship between shear strength and bonding ratio is established to predict strength of samples. There is a good agreement between predicting strength and testing shear strength.