As one of the main material of machinery and equipment parts, metal material is widelyused in automotive, marine, aerospace, nuclear industry and other fields. Due to the effect oftensile, compression, the alternate loading in the long-term use process, these devices internalmicrostructure changes gradually. In the early stage, they appear dislocations, slip bands andfatigue crack, however, with the cracks becoming more numerous and large, the structure ofmetal material is eventually destructed. Therefore, this thesis uses ultrasonic nonlineartechnique to research micro metal fatigue damage, and has the certain reality meaning.This thesis describes the principle of the ultrasonic nonlinear testing of metal fatiguedamage, and researches the way of ultrasonic nonlinear technique for the aluminum specimentensile fatigue damage, at the same time, analyzes the ultrasonic nonlinear parameters effectscausing by signal stimulus, couplant and window function. Through the analysis ofrelationship of ultrasonic nonlinear parameters and stress, this thesis fits the relationship usingsecondary polynomial and exponential. The fitting error data show that the single ultrasonicnonlinear parameters are not well characterization of fatigue damage. Aiming at this problem,this thesis selects three typical position of specimen fundamental, second harmonicsmagnitudes and ultrasonic nonlinear parameters to be characteristics through differentcombinations, and uses BP neural network to recognize the fatigue damage. The results showthat BP neural network model which adopts different position fundamental and secondharmonics magnitudes as characteristics has better recognition results, and the maximum errorrate of recognition reaches0.8%. |