Finite Volume Analysis and Neural Network Modeling of Wear During Hot Forging of a Steel Splined Hub

Hot forging of a CK45 steel splined hub was simulated using the finite volume method. The work-piece and die are assumed to be thermo-viscoplastic and thermo-rigid body, respectively. The simulation model results were validated by comparison with die and part measurements. Using a finite volume model and the Archard wear prediction method, locations of maximum wear depth were identified. The parameters of Archard’s model were the normal contact pressure, the relative velocity at die- and work-piece interfaces and the die hardness. Using results from the finite volume simulation, an Artificial Neural Network (ANN) was used to model the wear process. Seven exemplars were used for training the neural model. A network with 2 × 15 × 1 architecture selected to include two inputs, the forging temperature and flash thickness and one output, wear depth. The results indicate both the suitability of the finite volume method for modeling the hot forging process, and that the neural network model predicts die wear depth with better than 98% accuracy.