Complicated geological processes and controlling factors always make the distribution of geochemical elements having nonlinear characteristics. Traditional linear model methods are limited due to the character of nonlinear element association. In this paper, the authors adopted aiNet to preprocess geochemical data in order to study the geochemical metallogenic element association. Fractal content-area method was used to determine the anomaly threshold of elements. Fractal content-gradient method was used to determine the zone of element concentration. Thus a series of nonlinear metallogenic element association method system to seek out geochemical anomaly was formed. The authors applied this nonlinear method for anomaly seeking of geochemical metallogenic element association to process 1∶10 000 soil geochemical survey data from Donggapu mining area of Tibet and achieved results.