Mechanical Characteristic Test of the Space Docking Mechanism Based on Machine Learning

During the development of space docking mechanism, various specialized and costly equipment are required to conduct ground tests. In this paper, a method is proposed for constructing a mechanical characteristic test platform for spatial docking mechanisms using an industrial manipulator and a three-dimensional force and torque sensor. Our work encompasses the development of mathematical models for testing, analysis of testing method stability, implementation of gravity compensation during the acquisition of three-dimensional forces and torques, and the proposal of an automatic calibration method for gravity compensation parameters based on multiple linear regression. Additionally, we constructed a backpropagation (BP) neural network to extract mechanical characteristic test results of the target docking mechanism from comprehensive mechanical properties. The testing platform described in this paper has been successfully established and validated. The proposed mechanical characteristic testing method for docking mechanisms demonstrates advantages including rapid hardware platform deployment and cost-effectiveness, while maintaining adaptability to meet diverse testing requirements for various docking mechanisms. This platform proves capable of accommodating comprehensive mechanical assessments across different docking configurations.