As a widely recognized optimization method, BP neural network can provide scientific guidance for the formulation of reasonable process parameters. However, due to the randomness of its own weights and thresholds, the prediction accuracy remains to be further improved. The forming and manufacturing of heterogeneous welded sheet is a new extrusion connection method. There are many factors affecting the bonding quality, which brings trouble to the evaluation of bonding strength and quality. In this paper, orthogonal experiment, finite element simulation and process experiment were used to design and verify the key process parameters that affected the bonding strength of heterogeneous sheets. BP neural network and genetic algorithm neural network were used to predict the bonding strength. The results showed that the genetic algorithm neural network model has higher reliability, and the prediction accuracy was 99.5 %. Compared with the traditional BP neural network, the prediction accuracy was improved by 5.78 %, and the error was reduced to 0.5 %. It has good generalization ability, and provides a new way for intelligent reliability evaluation of high performance heterogeneous sheets extrusion manufacturing.