Parameter Optimization Using PSO for ESN-Based Robotic Belt Grinding Modeling

When the robotic belt grinding system needs to control the removal rate accurately, to optimize the grinding parameters is an important task after the model is obtained. In this paper, what is different from the previous methods is that the output of the model is not the removal rate but the workpiece feedrate v_w or the normal grinding force F_n, so the reverse resolution of the model doesn´t need to be done when to optimize the parameters. And a new object function is presented which promises for more suitable for minimizing the step values between two adjacent points. The common PSO (particle swarm optimizer) is applied to optimize the grinding parameters based on ESN (echo state network) model with v_w or F_n as the output. The results of the experiments using practical data prove that the ESN-based model with v_w or F_n as the output is feasible, the parameters´ optimization using PSO is effective, and the new object function performs better for minimizing the step values than the previous.