Structural optimization design of machine tools based on parallel artificial neural networks and genetic algorithms

This study addresses a dynamic modeling and design methodology for machine tools based on parallel artificial neural networks and genetic algorithms. Firstly, subjected to geometrical and static stiffness constraints, a machine tool optimization problem is proposed by minimizing the weighted functions of lower-order natural frequencies and frequency responses. Then, the dynamic analysis of the holistic machine tool is systematically investigated based on the proposed improved reduced dynamic model, leading to the formulation of the mathematical expression for multi-objective optimization. Utilizing genetic algorithms, the proposed optimization problem is solved after the functions between performance and design variables are approximated by employing feedforward backpropagation neural networks. Finally, an optimization example and experiments are implemented on a box-in-box type precision horizontal machine tool prototype. The designed machine tool offers expected dynamic behaviors over the task workspace. Experimental results demonstrate that the derived model is accurate and effective for the prediction of lower-order dynamics, as well as the effectiveness of the design methodology used in its development.