Design of patterns in tubular robots using DNN-metaheuristics optimization

Concentric-tube robots for minimally invasive surgery pose a potential risk of tissue rupture because of the structural instabilities caused by high value of bending-to-torsional-stiffness ratio (EI/GJ). In this study, a novel optimization method based on metaheuristic optimization accelerated by a deep neural network (DNN)-based surrogate model to obtain optimized pattern parameters is presented. The method minimizes EI/GJ while conforming to the minimum compliance constraints and geometric restrictions. The proposed optimization process utilizes a DNN trained using 855 datasets generated by finite element analysis that cover the pattern design parameter space. The pattern design parameters were derived from topology optimization. The results demonstrate that the proposed optimization method yielded pattern designs that outperformed previous designs within a reasonable time frame (less than 900 s) without requiring manual parametric study or sensitivity analysis.