A calibration method for the end pose of shield segment erector based on hybrid identification algorithm and hierarchical identification strategy

The kinematic control accuracy of shield segment erector is a key factor affecting the precise positioning of segment grabbing and the quality of lining assembly. It is necessary to calibrate the kinematic parameters of its end pose. However, the kinematic parameters are coupled with each other, and the identification convergence speed is slow, and the accuracy is low, resulting in inaccurate kinematic calibration results. To address the aforementioned issues, this paper proposes a kinematic calibration method for the segment assembly machine that incorporates a hybrid identification algorithm and a hierarchical identification strategy. Firstly, based on the MD–H model, an end-effector position kinematic error model was established, and the concept of prior constraints was introduced. A method for adaptive scale compensation using geometric constraints was proposed to ensure the accuracy of visual position detection. Secondly, to reduce the parameter coupling effect, a hierarchical identification strategy based on parameter sensitivity analysis was proposed. A hybrid identification algorithm combining the IPSO algorithm and the LM algorithm was constructed, and its feedback adaptive learning mechanism was combined to effectively improve the identification accuracy and speed. Finally, kinematic parameter calibration experiments were conducted on the full-scale segment assembly intelligent assembly test platform independently built by our team. The results show that the method proposed in this paper reduces the absolute error of the end-effector position by 54.66% and the variance by 42.31%. Compared with the classic LM algorithm, the proposed method reduces the mean of the calibrated positioning error by approximately 10.1%, while simultaneously reducing the error variance by approximately 27.6%, demonstrating its advantage in identification stability and achieving synergistic optimization of identification accuracy and convergence speed.