Transfer Learning-Based Stiffness Modelling of Spherical Scissor Linkage Remote-Centre-of-Motion Mechanism for Structural Optimisation.

BACKGROUND

The spherical scissor linkage remote-centre-of-motion (SSL-RCM) mechanism has multiple joints and a low stiffness, severely affecting its load-bearing capacity and motion accuracy. Accurate stiffness modelling of the SSL-RCM mechanism is key to guide its design.

METHODS

This paper uses the virtual joint method to establish a stiffness model considering passive joints. Applying transfer learning, small sample finite element data is extended by fitting a large dataset of simulation models, yielding a high-accuracy stiffness model for the SSL-RCM mechanism. A multi-objective genetic algorithm considering stiffness factors is used for structural optimisation of the SSL-RCM mechanism.

RESULTS

Comparative results show that the stiffness model achieves a 12.2% higher fitting accuracy than the baseline model, and an SSL-RCM mechanism with improved overall performance is obtained.

CONCLUSIONS

The stiffness model establishes an analytical relationship between the geometric dimensions and corrected stiffness of the SSL-RCM, which can be used for structural optimisation and design guidance.