Predicting underwater unmanned vehicle dynamic recovery process in nonlinear watershed based on BP neural network

基于BP神经网络的非线性流域UUV动态回收过程预测

School of Marine Science and Technology, Northwestern Polytechnical University, Xi’an 710072, China

Received: 24 March 2023

Abstract

Because of a nonlinear watershed′s interference during the recovery of an unmanned underwater vehicle (UUV), a closed-loop control method for optimizing the recovery path of the UUV based on the BP neural network is proposed. The paper uses the computational fluid dynamics (CFD) method to simulate the hydrodynamic coefficients for recovering the UUV relative to a submarine in different paths. The numerical simulation results are used as the initial data for training the BP neural network. Using the Latin super-law, the location of the nonlinear watershed is randomly sampled. Hydrodynamic coefficients of the UUV in the nonlinear watershed at sampling points are predicted based on the BP neural network. The results show that the error predicted by the neural network through root mean squares is within 10%. Through combining the prediction results of the neural network with the UUV longitudinal maneuverability equation, the error of the recovery speed and steering interval is compared with the theoretical recovery path. The closed-loop control method of UUV dynamic recovery in the nonlinear watershed is optimized.

摘要

针对水下无人自主航行器(UUV)回收过程中流域存在非线性干扰问题, 提出了一种基于BP神经网络优化UUV回收路径的闭环控制方法。采用计算流体力学(CFD)方法模拟UUV相对于潜艇以不同路径进行回收的水动力系数, 将数值模拟结果作为训练BP神经网络的初始数据, 利用拉丁超立方法对非线性流域的位置随机采样, 采用神经网络输出UUV在采样处的水动力系数, 实现非线性流域内UUV动态回收过程的水动力系数预测。结果表明: 通过均方根检验神经网络预测水动力系数误差均在10%范围内。将神经网络预测结果与UUV纵向操纵性方程结合, 对比回收速度和操舵间隔与理论回收轨迹的误差, 优化UUV动态回收路径的闭环控制方案。