It is very important to evaluate damage levels of structures for prediction of fatigue fracture. Purpose of this study is to develop a methodology for evaluating fatigue damage and conditions by a nondestructive inspection technique. Low cycle fatigue tests were conducted for carbon steel specimens (0.15 and 0.45 wt.% carbon) under various fatigue conditions. During the tests, magnetic properties and a magnetomechanical acoustic emission were measured at predetermined number of cycles. The test results exhibited 480 pattern sets which describe the relationship between the measured parameters and the fatigue information such as the damage level (Cycle ratio N/N_f), the stress amplitude (σ_a) ann the stress ratio (R). A nondestructive evaluation system consisting of measuring and a neural network was proposed. The effectiveness of this system was demonstrated by comparing results of the neural network with experimental data obtained from the fatigue tests.