| 引用本文: | 谢鸥,宋爱国,苗静,孙兆光,沈晔湖.仿生机器鱼近壁面流场识别的人工侧线方法[J].哈尔滨工业大学学报,2021,53(9):164.DOI:10.11918/202010083 |
| XIE Ou,SONG Aiguo,MIAO Jing,SUN Zhaoguang,SHEN Yehu.Near wall flow recognition method for bionic robot fish based on artificial lateral line[J].Journal of Harbin Institute of Technology,2021,53(9):164.DOI:10.11918/202010083 |
|
| 摘要: |
| 针对仿生机器鱼目标近距离作业时的环境识别难题,提出一种基于人工侧线(ALL)的近壁面流场识别方法。理论分析了ALL感知近壁面流场环境的可行性,建立了ALL虚拟压力传感器阵列并采用计算流体动力学(CFD)方法计算并提取了不同参数条件下(来流速度v,靠壁距离d和游动频率f)仿生机器鱼的体表压力数据,建立了基于多层前馈神经网络的来流速度和靠壁距离预测回归模型,并对模型结构和数据特征进行了优化。结果表明:壁面效应将引起鱼体周围流场结构的非对称分布,鱼体头部和尾部的侧线传感器对流场参数的辨识度高,消除弱相关的特征对来流速度和靠壁距离预测指标的影响小且有利于降低预测模型的复杂度。研究结果能够为水下机器人环境识别的信息提取及处理提供理论方法。 |
| 关键词: 仿生机器鱼 人工侧线 神经网络 壁面 流场识别 |
| DOI：10.11918/202010083 |
| 分类号:TP242 |
| 文献标识码:A |
| 基金项目:国家自然科学基金(51975394); 苏州市科技计划项目(SNG2017054); 江苏省高等学校自然科学研究项目(18KJB510043) |
|
| Near wall flow recognition method for bionic robot fish based on artificial lateral line |
|
XIE Ou1,2,SONG Aiguo1,MIAO Jing2,SUN Zhaoguang2,SHEN Yehu2
|
|
(1. School of Instrument Science and Engineering, Southeast University, Nanjing 210096, China; 2. School of Mechanical Engineering, Suzhou University of Science and Technology, Suzhou 215009, Jiangsu, China)
|
| Abstract: |
| In view of the environment recognition problems for bionic robot fish working near the target, a near wall flow recognition method based on artificial lateral line (ALL) is proposed in this paper. Firstly, the feasibility of near wall flow field environment recognition by means of ALL was analyzed theoretically. Then, the ALL virtual pressure sensor array was established. The surface pressure data of bionic robot fish were calculated and extracted under different parameters (inflow velocity v, near wall distance d, and swimming frequency f) by using computational fluid dynamics (CFD) method. Finally, the regression model of inflow velocity and near wall distance based on multilayer feed forward neural network was established, and the model structure and data characteristics were optimized. Results show that the wall effect caused asymmetric distribution flow around the fish. The lateral sensors at the head and tail of the fish had high identification for flow field parameters. Eliminating the weak correlation features had little effect on the prediction indexes of inflow velocity and near wall distance, and was helpful to reduce the complexity of the prediction model. The results can provide theoretical and methodological basis for information extraction and processing of underwater robot environment recognition. |
| Key words: bionic robot fish artificial lateral line neural network wall flow recognition |
