Found on Artificial Neural Network and Genetic Algorithm Design the ASSEL Roll Profile

This article introduces a method of optimizing design of ASSEL roller profile. It obtains the feed angle and toe angle for all rolling gauges with genetic artificial neural network (GANN), and to optimize weight coefficient and network structure by training of weight value of genetic neural network. The model was tested on the imported ASSEL mill. The results indicate that the inaccuracy of feed angle and toe angle obtained based on the model were both less than plusmn2%. With obtained feed angle and toe angle, ASSEL roller characteristic parameters of each gauge can be calculated via roller profile design formulas provided by MEER of Germany. The ASSEL roll profile characteristic parameter values are determined by searching all common characteristics and approximate attribute in the possible gauges with parallel optimization method, one of the generic algorithms (GA).