Abstract: A numerical method of quantitative nondestructive testing(NDT) of metallic foam was developed on the basis of the direct current potential drop(DCPD). A feed forward neural network(NN) was applied for the inverse mapping to quantify the inner flaw in addition with a principal component analysis process. Solving of obverse problem utilizes finite element method(FEM). To cope with the huge computer burden necessary to generate training data for NN of three dimensions of giant nodes in DC field, a novel fast forward solver based on databases was proposed for the rapid computation of DCPD signals. The numerical results indicated that the quantitative NDT of metallic foam can be well performed by using the NN approach and the proposed fast solver can significantly decrease the computer resources but with a satisfactory accuracy.