Roll force prediction by combined FEM and ANN in the hot rolling process under nano-lubrication condition

The article focuses on predicting rolling force (RF) in the hot rolling process by incorporating nano-lubrication conditions through strategies available in finite element method (FEM) software. Additional work includes RF prediction by artificial neural network (ANN) modelling. The experimental RF were collected from an experimental rolling mill (ERM) for hot rolling IS 2062 E250 steel. The simulations were performed using SIMUFACT and the results were validated by experimental data. The prediction was made accurately by adjusting the lubrication panel available in the software. Then, FEM and ANN were performed to predict RF at various combinations of process parameters. FEA simulated hot rolling of IS 2062 E250 with varying process parameters. It incorporates changing 5 process parameters for FEA as roll speed, reduction%, nano-lubrication condition, billet size (thickness), and rolling temperature. Then, the accumulation of data for further application in ANN modelling. The validation of FEM software was successful with an error of less than 5%. According to the validation, the lubrication corresponding to the nano-lubrication setup was fixed in the software. The mesh was optimised and detected the least RF deviation at 1.8 mm mesh. The ANN was performed from the simulation data. The ANN performed better for a single hidden layer model for 8 neurons. The training, validating, and testing of the model resulted in an R-square value of 0.991, 0.987, and 0.994, respectively. The root mean square error (RMSE) value obtained was 0.985. Therefore, the ANN model developed was performing better. The article uses the combined strategy of FEM and ANN, with available lubrication panels in FEM software for accurate RF predictions.