Application of improved BP algorithm based on information entropy of signal in recognizing defects of seamless tube

Flaw signals of seamless tube are investigated in frequency domain. These flaws are internal flaws, outside flaws, and holes. Two-dimension spectrum entropy indexes, which are regarded as character indexes, are practical in feature abstraction of defects, and the seamless tube flaws are recognized by the improved BP neural networks algorithm according to analyzing deeply the forward neural network dynamics model and its Jacobian matrix. The experiments have demonstrated that the recognition algorithm is of a perfect precision which could reaches as higher as 100 percent, and it is suitable to real time monitoring. It will have a wide application prospects.