Abstract: Magnetic flux leakage detection is one of NDT methods for ferromagnetic materials. Quantitative identification is to identify the crack size through obtaining magnetic flux leakage signals. By combining principal component analysis (PCA) and neural network， a model was established to predict width and depth of the micro crack. The principal component analysis removed the data correlation and reduced the dimension of the input samples， so it can significantly simplify the network structure. BP neural network optimized by genetic algorithm (GA-BP neural network) can prevent the search process from running into the local optimal solution. Based on the theoretical calculation of magnetic dipole model and experiment on the artificial cracks， the algorithm applied in the quantitative recognition of microcracks was verified， which may lay the foundation for the early quantitative recognition technique of crack development stage.

Key words: magnetic flux leakage detection, principal component analysis (PCA), GA-BP neural network, microcrack, quantitative identification