The compression tests of TiC-Ni alloy were carried out by Gleeble-2000 simulator, and the flow stress data were obtained at different temperature, strain rates and various true strain. Based on the experimental datas and network knowledge, an artificial neural network with back propagation algorithm was established and knowledge based on constitutive relations model was developed after training. Error analysis shows that the artificial neural network model for constitutive relationship has higher predicted precision, and it can be used for guiding the reactive hot press process and applying in finite element simulation of TiC-Ni.