Application Study Of RBF In Crack Testing Of Steel

With the rapid development of the scale of China's industry, the steel industry as a traditional heavy manufacturing plays an important role in it. The quality of steel has a decisive action to the quality of steel product, in addition the products'useful life,reliability and so on. In the field of the steel product testing, the electromagnetic nondestructive testing was regularly used; the existing steel quality testing instruments have got great results in performance of the steel testing. However, in some steel defects, for instance, the aspect of the crack detection, they still have some problems of low accuracy and stability. The reason is that the chip is relatively backward, and the algorithms is unsatisfactory, therefore, to improve the existing instrument, and design a new steel crack nondestructive testing system is necessary, in this way the precision and speed of crack detection testing could be improved.To develop the rapid electromagnetic nondestructive testing technology, this paper combine the RBF algorithm and SOPC into FPGA,studying on the applied research of new crack nondestructive testing system.This paper extracted steel performance index characteristic signal by using the initial magnetic conductance method, removed the noise through steel testing instrument, got the useful information, then send the parameter into the RBFNN model. According to the rules of judgment, get the correct output. In the process of building neural network model, this paper analysised the neural network design,parameter selection,training sample processing,and the selection of the transfer function and the weight.Through the testing of the cracking steel product sample, the precision rate of the new steel crack testing system that designed in this paper was ninety percent, comparing with the regular testing model, it has great stability and high efficiency. The thought in this paper was worth to make deeper research in the steel defects testing field.