A Model-Based Prediction Approach for Frequency Response Function of a Magnetic Adsorption Machining Robot

The adsorption machining robot is an important branch of robotic machining equipment. Influenced by the low-stiffness structure, the robot's machining efficiency is significantly limited by vibration deformations. The frequency response function (FRF) is an essential theoretical foundation for analyzing cutting stability to control vibration deformations. However, characterized by adsorption interfaces and parallel structures, modeling adsorption machining robots to predict the FRF is challenging. This paper proposes a model-based prediction approach combined with parameter identification to predict the FRF of adsorption machining robots. An initial dynamic model is first established based on the virtual joint method, considering the adsorption interface and frame flexibility. Then, the Sobol method is introduced to analyze the parameter sensitivity of the nonlinear target function to select the parameters to be identified. The number of parameters involved in identification is reduced by 50% after analysis. Finally, the genetic algorithm is used to identify the values of parameters in the initial model. The experimental results indicate that the prediction error is reduced by 53.61 % compared to the initial model. The frequency errors of the five main peaks are less than 6.86%. The proposed method can also be applied to other parallel robots.