Kinematic parameter identification and experimental study of principle-level variable stator vane mechanism

To enhance the adjustment accuracy of the variable stator vane (VSV) mechanism, a set of kinematic parameter identification methods is proposed. First, the kinematic model and the error models of the principle-level VSV mechanism are established. Considering the solving difficulty caused by the coefficient matrix being not full rank, the method of singular value decomposition (SVD) is employed to determine and eliminate redundant parameters. Then, three parameter identification methods, Gauss–Newton (GN), Levenberg –Marquardt (LM), and particle swarm optimization (PSO), are proposed, combining the typical sample points extracted using Latin hypercube sampling (LHS), to solve the kinematic parameters. The results indicate that the GN method achieves high identification accuracy only when the measurement noise is minimal. In practice, however, it is prone to computational singularity in the presence of noise, which limits its applicability. In contrast, the LM and PSO algorithms demonstrate better error tolerance and both yield satisfactory identification performance. Under realistic conditions, the LM method exhibits greater robustness and practical applicability. Finally, through identification experiments, the accuracy of the principle-level VSV mechanism is improved by 34.16%, greatly reducing its uncertainty. This validates the feasibility and effectiveness of mechanism parameter identification, providing an efficient approach to enhance motion precision of other mechanisms.