Nhảy đến nội dung

Topology-based geometry optimization for a new compliant mechanism using improved adaptive neuro-fuzzy inference system and neural network algorithm


Van Bang Dinh, Ngoc Le Chau, Nam T. P. Le, Thanh-Phong Dao

Source title: 
Engineering with Computers, 2021 (ISI)
Academic year of acceptance: 

In precision engineering, compliant mechanisms are growingly promising mechanisms in designing micro/nano positioners and manipulators due to emerging advantages of free friction, no joint, and decreased assembly. Nevertheless, compliant mechanisms have flexible configurations with nonlinear behaviors, the design, analysis, and optimization are becoming challenges, and a systematic design method is still limited. Therefore, this paper proposes a new multi-phases optimization design method for compliant mechanisms. In the suggested method, the topology optimization is integrated with finite element method, intelligent modeling, and neural network algorithm. First, the solid isotropic material with penalization-based topology is used to design a new compliant mechanism. The numerical simulations are conducted. Next, the parameters of adaptive neuro-fuzzy inference system are optimized by the Taguchi to achieve an improved ANFIS (IANFIS) model. The IANFIS approaches are used to predict behaviors of the developed mechanism. The results confirmed that the developed IANFIS has a highly accurate prediction in comparison with other regression models. Particularly, the metric values of IANFIS models are relatively good. Particularly, the R2 value is approximately 1 while the MSE, RMSE, and SD values are approximately 0. Last, the neural network algorithm is extended to search the optimal geometry sizes for the compliant mechanism. In the size optimization, two scenarios for are taken into consideration. For the scenario 1, the displacement, rotation angle, parasitic, and stress of the mechanism are found about 1.9977 mm, 0.8232 degrees, 0.1666 mm, and 13.94 MPa, respectively. For the scenario 2, the displacement, rotation angle, the parasitic, and stress are approximately 1.8501 mm, 0.8237 degrees, 0.1429 mm, and 11.8193 MPa, respectively. The results of size optimization showed that the displacement of the mechanism is enhanced by 12.94% and the rotation angle is improved to 4.5E+11% in comparison to the initial topology. The statistic results of Friedman and Kruskal–Wallis found that the accuracy and efficiency of proposed method are superior to those of other methods with p-values less than 0.001. The proposed method is applicable to other industrial systems.