Abstract: Multidimensional funetion approximation of fatigue experimental curves is achieved by applying neural network techniques.The constructed Back-Propagation Neural Network(BPNN) learns these curves by using leaming algorithm,and this sort of leaming is a long time process.After BPNN has learned the curves suceessfully,the prediction of new curves will pro- ceed with short time. The prediction process can be linked very easily with fatigue life pxediction program. The results of fitting show that the method is verv effective,stable and convenient.

Key words: fatigue (materials), neural nets, function analysis