Abstract: According to production data of a 1580 rolling mill, an integrated method combining finite element analysis and neural networks was presented for hot rolling. In the method, plastic deformation during the rolling process was firstly modeled by a finite element method, and then a neural network provided parameter adjustment for the finite element model, so the integrated model had the advantages of neural network and finite element methods. At the same time, intelligent chaos particle swarm optimization (CPSO) was used to optimize weights and thresholds of the network. A comparison between simulation results and actual production data proved the validity of the integrated model.