Abstract:In the course of multistage forging for producing gear blank,how to do preforging design of gear blank directly affects metal flowing patterns of the finish forging parts,the filling situation of the forging die cavity,the quality of the forging products and die life.Firstly,using characteristics of high learning efficiency and generalization ability and high prediction accuracy of the extreme learning machine (ELM) network,the ELM network model between the size parameters of preforging,finisher forming force and maximum die stress of finish forging was established.And then,by using genetic algorithm (GA) with the function of overall situation optimization in order to improve the prediction accuracy and stability,the reasonable size of finsher force of die was arrived for the obtained preforging parts.The optimal shape and size parameters of preforging parts are determined at specified conditions.