Abstract: Improved RBF neural network is applied on eddy current nondestructive quantitative detecting. Owning to the disadvantages of OLS in network structure optimization, authors put forward using Fisher ratio method to optimize the RBF centers, orthogonal transform and forward selection search method are used to optimize structure. The results show that the neural structure is simplified strongly, the converge precision and class separability is improved, and this method is likely to be used for detecting online.