A design framework for optimizing forming processing parameters based on matrix cellular automaton and neural network-based model predictive control methods

Abstract

• A matrix CA method is developed to quickly describe microstructure variations with processing parameters. • Neural network-based model predictive control (NNMPC) method is used to online optimize processing parameters. • Based on matrix CA and NNMPC method methods, a material design framework is developed. • Ideal microstructures of a Ni-based superalloy can be obtained by hot deformation based on this design framework. The advanced modelling/simulating method and the effective optimization/controlling strategy relying on knowledge-based systems are highly demanded in industrial manufacturing of alloy components. In this work, based on the advantages of cellular automaton (CA) simulation and neural network-based model predictive control (NNMPC) methods, a material design framework is developed to optimize processing parameters for the designed target microstructures of alloys. In this framework, a matrix CA simulation method is developed to accurately and quickly describe the variations of microstructures with processing parameters. NNMPC, which is an effective control method for nonlinear and multi-objective system, is utilized to online optimize processing parameters according to the designed target microstructures. Based on the optimized processing parameters, the hot compressive deformation tests of a Ni-based superalloy are conducted to verify the effectiveness of the developed framework. The experimental results well agree with the simulated/designed ones, which implies that the developed material design framework can effectively optimize processing parameters for the designed target microstructures. Also, the developed material design framework is used to obtain the uniform and fine microstructures of a Ni-based superalloy.