In order to solve the problems of cracks and fracture easily appearing on the ears of hinge beams during the operation of cubic-anvil high pressure apparatus, and excessive tonnage of press required during the forging process of hinge beams, with the goal of reducing forging damage, forming load and stress peak during the forging process, the forging process parameters of hinge beam were optimized. The influences of each process parameter on forging damage, forming load and equivalent stress peak of forgings were obtained by using the range and variance analysis of orthogonal experiments, the three response objectives were converted into corresponding grey correlation degrees by combining grey correlation analysis with entropy weight method, and the agent model of process parameters and grey correlation degrees was established by extreme learning machine to develop a process parameter prediction system which could rapidly evaluate the advantages and disadvantages of the process parameters.Finally, based on the improved particle swarm algorithm within the feasible domain, the optimal process parameters were obtained. The results of relevant physical and chemical analysis show that the pore and inclusions in the finished cross-section of hinge beam and the microstructure at the maximum stress location meet the requirements.