基于数据驱动的水下滑翔机生物污损监测方法  

Data-Driven Method for Underwater Glider Biofouling Monitoring

作　　者：王延辉[1,2] 张新海 杨明 闫彩清 杨绍琼 Wang Yanhui;Zhang Xinhai;Yang Ming;Yan Caiqing;Yang Shaoqiong(School of Mechanical Engineering,Tianjin University,Tianjin 300350,China;The Joint Laboratory of Ocean Observing and Detection,Pilot National Laboratory for Marine Science and Technology(Qingdao),Qingdao 266237,China;Qingdao Institute for Ocean Engineering of Tianjin University,Qingdao 266237,China)

机构地区：[1]天津大学机械工程学院,天津300350 [2]青岛海洋科学与技术试点国家实验室海洋观测与探测联合实验室,青岛266237 [3]天津大学青岛海洋技术研究院,青岛266237

出　　处：《天津大学学报（自然科学与工程技术版）》2023年第1期93-102,共10页Journal of Tianjin University：Science and Technology

基　　金：国家自然科学基金资助项目(11902219);天津市自然科学基金资助项目(18JCJQJC46400).

摘　　要：以我国自主研发的“海燕-L”水下滑翔机为研究对象,针对其长时间运行过程中出现的生物污损问题,提出了一种基于数据驱动的水下滑翔机生物污损监测方法,以指导试验过程中控制参数制定与任务规划.首先,通过动力学模型仿真分析,研究了生物污损对水下滑翔机单剖面运行时间的影响.然后,根据动力学分析结果,结合岸基操控条件以及水下滑翔机的通信方式,提出了一种生物污损监测方法:基于机器学习方法,利用水下滑翔机未发生生物污损时的关键数据,建立单剖面运行时间预测模型;对试验过程中水下滑翔机的单剖面运行时间进行预测,并监测其预测偏差;利用所建立的阻力系数与预测偏差的关系式,预测生物污损引起的阻力变化.最后,通过海上试验验证了所提出方法的正确性与有效性.试验结果显示,单剖面运行时间预测偏差与阻力系数均可以反映水下滑翔机的生物污损程度.当水下滑翔机正常运行时,所提出的方法对其阻力系数的最大预测误差仅为2%,可以实现对生物污损程度的准确监测.该方法只需要使用水下滑翔机运行过程中的部分关键数据,即可快速获取其性能变化,极大提高了监测效率,可以为“海燕-L”的实际海上应用提供支持.同时,该方法也可以为其他类型水下滑翔机的性能监测提供参考.This study proposes a data-driven biofouling monitoring method for monitoring the biofouling of the“Petrel-L”underwater glider independently developed by China,in order to address the biofouling problem during its long-term operation.The method will help set the control parameters and plan the mission during the experiment.First,the effect of biofouling on the operation time of the underwater glider in a single profile was investigated via dynamic model simulation.Then,based on the results of the dynamic simulation,a biofouling monitoring method was proposed by combining the shore-based maneuvering conditions and the communication mode of the underwater glider.By applying the original data obtained in the pretest period when the glider was not yet biofouled,a prediction model based on the machine learning method was established to predict the operation time of Petrel-L in a single gliding profile,and then the prediction deviation was monitored.The relationship equation between the drag coefficient and prediction deviation was proposed to predict the change in drag caused by biofouling.Finally,the correctness and validity of the proposed method were verified via sea trials.Experimental results show that the prediction deviation of single gliding profile operation time and drag coefficient can reflect the biofouling degree of the underwater glider.When the underwater glider is in normal operation,the maximum prediction error of the proposed method is only 2%,which allows for accurate monitoring of the biofouling degree of the underwater glider.The proposed method can quickly obtain the performance change in the glider merely using some key data during the operation,which greatly improves the monitoring efficiency and can provide support for the practical application of“Petrel-L”at sea.Further,the method can also provide a reference for the performance monitoring of other types of underwater gliders.

关 键 词：水下滑翔机 生物污损 状态监测 数据驱动方法 

分 类 号：TP242.3[自动化与计算机技术—检测技术与自动化装置]