Parameter optimization of the bio-inspired robot propulsion through the deep learning based reduced order fluid-structure interaction model. (1st July 2022)

• Record Type: Journal Article
• Authors: Ying, Zixiang; Wang, Linxiang; Melnik, Roderick
• Is Part Of: Ocean engineering. Volume 255(2022)
• Publication Date: 2022-07-01
• Subjects: Bio-inspired locomotion -- Fluid-structure interaction -- Reduced order modeling -- Proper orthogonal decomposition -- Neural networks; Ocean engineering -- Periodicals; Ocean engineering; Periodicals; 620.4162
• DOI: 10.1016/j.oceaneng.2022.111436
• Languages: English
• ISSNs: 0029-8018

Abstract

Abstract: In this paper, an effective model (POD-NIROM) is proposed, which makes full use of Long short-term memory Neural Network (LSTM NN) and proper orthogonal decomposition (POD) to predict the fluid dynamics around the moving boundary, as well as the soft robot locomotion. This is the first time that the proposed model has been used to optimize the propulsion parameters of the robotic fish, in which the body stiffness of the robot has been taken into consideration. To discuss the reliability of the presented model, the prediction and simulation of robot swimming performance are compared with experimental measurements, the results show a consistent trend. Finally, the trained model is used to optimize the robot's propulsion parameters. The results show that the robot has the best propulsion speed and propulsion force when the dimensionless wavenumber k ˜ is about 0.7. Compared with the high-fidelity model, the average relative standard deviation of the present model is 2.60 %, but the calculation cost is reduced by 99.4 % . Reasonable prediction and efficient calculation make the proposed POD-NIROM has great potential in the process of bio-inspired robot swimming prediction or bio-inspired robot propulsion parameter optimization. Highlights : ● The POD-NIROM is an effective method to optimize the driving parameters of the robot ● The proposed POD-NIROM can quickly predict the swimming performance of the robot, taking the propulsion parameters as input. ● For simplicity, body-average stiffness is introduced into the model to describe swimmers with soft bodies or elastic joints. ● The reliability of the proposed model is demonstrated by comparison with the experimental data. ● When the dimensionless wave number is 0.7, the robot studied in this paper is in the optimal propulsion state