Predicting the Optimum Process Parameters for Seamless Tube Rolling with FEM and BPNN

The objective of the designer in tube rolling is to choose the process parameters that provide for acceptable tube diameter, wall thickness at the end of the rolling process. Nowadays, the empirical know-how of the designer is still decisive for the process parameters. Itpsilas often requires very costly trial-and-error, and it is complicated and time-consuming. The paper presents a unique reverse engineering approach to the design of process parameters. It uses finite element simulation to obtain the data as the training samples of neural network, and artificial neural network for prediction the optimum process parameters. Instead of using the process parameters as input values and rolling outcomes as output values, the BPNN model configures the tube parameters as inputs and the process parameters as outputs. A group data of tube parameters are inputted to the trained neural network, and the optimum parameters are obtained in the output. In addition, an example of continuous tube rolling is included to demonstrate the efficacy of this approach to tube rolling.