固体中裂缝型缺陷超声散射声场特性的研究

Crack type defects are volume-less and with sharp edges. They are very hazard. Study of scattering of ultrasound by crack is important for crack detection. This study is one of the basic problems and it is very challenging. The theoretical results lights the way of ultrasonic testing of materials.The penny-shaped crack which is one of the typical crack models was chosen as the basic model for this study. Both analytical and numerical methods were used to analyze the problem. Whereas in experiments two kinds of crack were made, the penny-shaped crack and the ribbon type one. The separation of variable method is applied in oblate coordinate system to obtain scattering far field by a longitudinal ultrasonic wave. The analytical solution expressed as a series of three kinds of oblate angle functions. This problem was also calculated by means of Boundarelemental method(BEM). Mathematically difficulties introduced by the essential high order singularity up to r-5 stemming from the sharp edges of crack. Some special techniques were used to overcome these difficulties. The results obtained from these two methods were compared to make confirm. The main features of the numerical results deduced from these two methods were coincided with each other. The characteristics of scattered ltrasonic field by cracks were examined. The directivities are complex and vary with the values of incident angles of ultrasound. We finded firstly that the scattered far fields show some interesting characteristics. When the incident angle varies from zero degree to a certain value, the strong peak of scattered field keeps around the z-axis roughly. Experimental results supported above conclusion approximately. This feature might be useful for crack direction determination. In order to find the features of the scattered fields by the cracks in general, a primary idea was presented on model recognition by means of neural network besides the BEM. A derivative arithmetic based on Chebyshev neural network was presented in order to determining the number of the hidden neurons justifiably and to speed up the convergence rate. The filter based on the sine basis function neural networks and the model structure based on Fourier transform were designed to try to form the echoes from the crack in order to extract the characteristics of the echoes as well. .Two systems of high speed waveform collection were formed. One of which is more advanced.

探索固体中裂缝型缺陷对脉冲超声波的散射规律，在超声无损检测应用中最常用的兆赫级频段分析超声散射回波的特性，利用神经网络、模式识别等数字信号分析技术，探讨裂缝型缺陷检测和识别的新方法和新技术。裂缝型缺陷是最具潜在危险性的缺陷，在许多情况下又是很难检测和辨认的缺陷，本研究结果对超声无损检测工业应用有重要指导意义。