Research Of Eddy Current Nondestructive-Testing System Based On DSP

Eddy current nondestructive detecting is a nondestructive inspecting method. It's widely used in metal material and equipment detection online. But there are too many factors affecting eddy current detecting. So qualitative detecting, especially quantitative detecting, is always a key question. Recently, there are many signal processing methods used in instrument, such as phase analysis, frequency analysis and amplitude identification method, but defect identification depends on operating personnel and there is extremely large safety incipient fault. In this paper, neural network combined with DSP applies to eddy current nondestructive detecting. It solves the difficulty of defect identification.Introduce the theory of eddy current nondestructive detecting system based on DSP and complete the design of hardware and software. In the hardware design, the basic structure of DSP is given, and DSP as the kernel chip applies to generate the sine waveforms frequency range of 0~2 MHz. In the software design, the methods of eddy current nondestructive detecting based on neural network are adopted in this paper. Because of the defects of BP network, this paper adopts RBF neural network having better performance than BP. However, though many algorithms can decide RBF centers, there are some disadvantages, such as they can't optimize in global and have slow converge speed. Therefore, Fisher ratio algorithm is used, but it doesn't belong to the most optimization in class separability, so improved Fisher ratio algorithm is presented. Both of them are programmed by MATLAB. The results show that the neural structure is simplified strongly, the converge precision and class separability is improved using Fisher ratio and improved Fisher ratio algorithms.