Although harmonic drives have been adopted by all sorts of industrial environments, the mathematical expression of its dynamics has not yet been fully solved. An intelligent algorithm based on data acquisition is used to analyze the kinematic parameter of the flexible manipulator offline to improve the accuracy of predicting output torque. In this study, a method is proposed to capture the dynamics of harmonic drive systems, such as temperature, speed, and load. A friction model is proposed, and the model parameters are estimated by MCMC (Markov chain Monte Carlo) approximation. The accuracy of the estimated parameters in different conditions is calculated, and the sensitivity to parameter uncertainty of the proposed model is considered.

This study focuses on the parameter identification of a heavy-duty manipulator used in the remote maintenance of the China Fusion Engineering Test Reactor (CFETR). Accurate modeling of the manipulator's dynamics needs investigation of the effects of hysteresis, velocity, and other variables on output torque independently. For this reason, we estimate the undetermined model parameters using the Stochastic Gradient Hamiltonian Monte Carlo (SGHMC) method. In addition, a global sensitivity analysis is performed to assess the precision of the model's output and the relative significance of its input variables. Experiments are conducted to determine the output torque, hysteresis displacement and velocity of the CFETR's heavyduty manipulator. Our findings indicate that the SGHMC method significantly improves the efficacy of parameter identification while maintaining a high level of accuracy, resulting in a significant reduction of approximately 8% in the root mean square error (RMSE) of the output torque. In addition, the analysis of first-order and total-effect sensitivity indices reveals the influential parameters on the output torque. The sensitivity analysis offers valuable insights into the significance of parameters and system optimization. Considering hysteresis deformation, this study presents a method for modeling and parameter estimation of the output torque in a heavy-duty robotic arm. The developed method contributes to the solution of practical problems and provides the groundwork for future research on SGHMC and parameter estimation algorithms.INDEX TERMS Hysteresis, HMC method, parameter estimation.