Development Of Evolution Model For Hot Deformation Of 2.25CrlMo0.25V Steel

Large forgings are generally used as the core component of large mechanical equipments. Due to the poor working environment, it is necessary for the equipment to have good mechanical properties and microstructure. It is highly significant for the prediction of microstructure and the selection of technological parameters that learning the rule of microstructure evolution in the forging. 2.25Cr-1Mo-0.25 V steel is the priority material of hydrogenation reactors, which is the large core equipment in the petrochemical industry.Firstly, the 2.25Cr-1Mo-0.25 V steel forging state material with three different initial grain size has been used as the sample of the gleeble-1500 D thermal compression experiments. The stress-strain relationship of material in the process has been studied. The grains of hot compressed sample by grinding, polishing and corrosion have been observed under the optical microscope, the impact of different deformation conditions on stress-strain curve and microstructure have been analyzed.Secondly by analyzing the algorithm, network structure of the BP network and the related parameters of the simulated annealing algorithm, the dynamic recrystallization model has been established on the basis of thermal compression experiment data, which include the average grain size model, the critical strain model, the dynamic recrystallized volume fraction model and grain size model. The result of the network has been analyzed, which shows the accuracy and extensiveness of the network model.The dynamic recrystallization model established based on this BP neural network is integrated into DEFORM software. The FEM simulation of the microstructure evolution in the forging process of 2.25Cr-1Mo-0.25 V steel has been realized by secondary development of DEFORM software.Finally the cylinder upsetting experiment is carried out. The good fitness between the experiment data and the FEM simulation result of DEFORM software prove the accuracy and extensiveness of this dynamic recrystallization model established by BP network model.