Abstract
Industrial robots have become a crucial tool for milling large and complex surfaces. The dynamic characteristics of a robotic structure significantly influence milling accuracy and efficiency. Accurate identification of these dynamic parameters during operation is essential for vibration suppression and enhancing machining performance. Notably, the dynamic characteristics of a robot in operation differ from those in a static state. Operational modal analysis (OMA) enables the identification of structural dynamic parameters under operating conditions, but it necessitates input signals that approximate white noise, which is not met by the excitation forces during typical robot operations. Moreover, OMA is primarily applied to fixed structures, such as buildings and bridges, whereas the dynamic characteristics of robots vary as their poses change during operation. To address these challenges, this paper proposes an autonomous modal analysis method that considers the dynamic spatial sensitivity of robots and the randomness of the excitation frequency band and direction. Firstly, to mitigate the impact of pose changes on modal parameter identification, a method for predicting the sensitivity of natural frequencies based on structural modal shapes is proposed. This approach limits the range of the robot's self-excitation motion based on dynamic spatial sensitivity analysis. Secondly, the necessary condition for the randomness of the excitation force direction is established, requiring that the torque projection matrix be of full rank. Building on this, a broadband random signal meeting the requirements of white noise is generated through multi-joint random acceleration and deceleration movements. Finally, the efficacy of the proposed autonomous modal analysis method is validated through multi-joint combined self-excitation modal analysis experiments.