返回
类型 应用研究 预答辩日期 2018-03-10
开始(开题)日期 2015-06-02 论文结束日期 2018-01-16
地点 中心楼525 论文选题来源 国家自然科学基金项目     论文字数 7.6 (万字)
题目 AUV组合导航系统容错关键技术研究
主题词 自主水下航行器(AUV),故障检测,容错滤波,组合导航,高斯过程回归
摘要 自主水下航行器(Autonomous Underwater Vehicle, AUV)作为人类探索和开发海洋的重要工具,在我国海洋发展战略中起重要作用。AUV导航系统为其提供位置、速度、姿态等信息,准确可靠的导航信息是AUV成功执行任务的保证之一。本文以AUV组合导航系统容错技术为主题,对AUV容错组合导航系统建模、多传感器异步信息融合、故障检测技术、导航传感器短时失效处理方法、AUV组合导航系统容错滤波算法等关键技术进行了研究。具体研究内容和结论如下: 1. 针对AUV导航系统较高的可靠性需求,设计了基于无反馈联邦滤波的AUV容错组合导航系统结构。为每个局部滤波器设置故障检测模块和故障隔离决策模块,有利于及时检测故障并做出应对处理,以避免故障交叉感染。分析AUV常用导航传感器/系统的工作原理,建立以捷联惯性导航系统(Strapdown Inertial Navigation System, SINS)为参考系统的组合导航系统误差模型。 2. 针对AUV多传感器组合导航系统中由于各传感器采样频率不同而造成的异步信息融合问题,提出基于多尺度的异步信息融合算法。在不考虑量测滞后的情况下,利用数学归纳法递推得到基于多尺度的系统模型,并针对有\无辅助量测信息的时刻分别推导融合估计算法;在考虑量测滞后的情况下,通过状态和观测的分块和扩维得到基于多尺度的系统模型,利用不同尺度上的观测信息对目标状态进行估计,推导以数据块为单位的融合估计算法。在异步量测的SINS、DVL和TAN构成的AUV组合导航系统中,分别针对不考虑和考虑量测滞后的两种情况进行仿真,结果表明基于多尺度的信息融合算法与基于单一尺度的信息融合算法相比提高了异步信息融合精度,从而有利于提高AUV组合导航系统多传感器异步量测情况下的定位精度。 3. 针对AUV工作环境的不可预测性,许多不确定因素会影响导航传感器量测的稳定性和准确性,甚至可能导致传感器发生故障的情况,提出基于高斯过程回归的故障检测方法。利用粒子群优化算法辅助求取高斯过程回归模型的最优超参数,将训练好的高斯过程回归模型用于预测滤波新息,并基于此构建新颖的故障检测函数,有效增大了系统无故障和有故障时故障检测函数值的差异,从而增强了该故障检测方法对渐变故障的敏感度。通过基于跑车数据的半物理仿真实验进行验证,结果表明基于高斯过程回归的故障检测方法能及时检测突变故障和渐变故障,尤其在检测渐变故障时,与残差卡方故障检测法相比具有较强的敏感性。基于高斯过程回归故障检测方法的提出有利于AUV组合导航系统在导航传感器发生故障时及时做出隔离处理,从而有助于保证复杂工作条件下AUV导航结果的准确性。 4. 针对AUV导航传感器DVL易发生短时失效的情况,提出基于偏最小二乘回归和支持向量回归的双模型复合处理方法,利用复合量测估计器的预测结果代替DVL量测信息用于组合导航。为克服失效时刻偶然因素对估计结果带来的不良影响,采用失效时刻及其前期时刻的SINS解算信息作为训练输入,针对该自变量具有多重相关性的情况,采用偏最小二乘回归建立估计器模型,以保证所建模型的稳健性。为进一步提高估计精度,引入支持向量回归这一非线性回归模型建立估计器以预测偏最小二乘回归预测后的残余部分。通过数学仿真和半物理仿真验证了所提复合量测估计器可有效提高传感器短时失效情况下的系统导航结果精度,从而有利于保障恶劣工作环境下AUV组合导航系统的可靠性。 5. 针对AUV组合导航系统高精度高可靠性的需求,提出基于模糊逻辑的智能容错滤波算法。考虑量测噪声先验知识不充分的问题,利用模糊逻辑实现量测噪声方差阵的自适应调整。在故障检测的基础上,引入量测可信度这一变量,通过模糊推理确定量测可信度的取值,并以此调节对量测信息的利用率。综合自适应局部滤波、故障检测算法和故障隔离算法,设计基于模糊逻辑的AUV组合导航系统容错滤波算法流程。通过基于SINS/MCP/DVL/TAN的组合导航系统进行仿真验证,结果表明模糊容错滤波算法可有效避免突变故障和渐变故障引起的导航精度下降问题,从而有利于提高AUV组合导航系统的容错性能。 基于上述研究,AUV组合导航系统的准确性和容错性得到较大提高,从而有利于增强AUV的可靠性和安全性。
英文题目 Research on Key Techniques of Fault Tolerance in AUV Integrated Navigation System
英文主题词 autonomous underwater vehicle (AUV), fault detection, fault-tolerant filtering, integrated navigation, Gaussian process regression (GPR)
英文摘要 As important instruments for exploring and exploiting the oceans, autonomous underwater vehicles (AUVs) are of great significance in our country’s ocean strategy. The navigation system of an AUV provides position, velocity and attitude information. Accurate and reliable navigation information of the AUV is prerequisite to the successful execution of the tasks. This dissertation focuses on the fault-tolerant technology for AUV integrated navigation systems. System design, information fusion of asynchronous multi-sensors, fault detection, the approach to deal with short-time sensor malfunctions, fault-tolerant filtering for AUV integrated navigation systems are studied. The main research work and results are as follow: 1. In order to meet the sound reliabiltiy of the AUV navigation system, the structure of the fault-tolerant system is designed by using no-feedback federated filter. Fault detetion module and fault isolation decision module which are used to avoid the cross contamination of faults are set up for each local filter. The working principles of the common navigation sensors or systems for AUVs are analyzed. With the strapdown inertial navigation system (SINS) as the reference system, the error model of the AUV navigation system is build. 2. An information fusion algorithm based on multi-scale estimation theory is proposed to deal with the situation that multi-sensors measure with different rates. In the case of without the measurement delay, the multi-scale system model is recursively got by mathematical induction. By judging whether the measurement information exists, respective fusion algorithms are derived. In the case of with the measurement delay, the multi-scale system model is obtained by partitioning and extending the state and measurement information. The target state parameters are estimated by employing the measurements on different scales. The fusion algorithm is deduced in units of data block. The proposed algorithms are used in an AUV integrated navigation system composed of asynchronous SINS, DVL and TAN. The simulation results show that the multi-scale information fusion algorithms have the higher degree of accuracy compared with the single-scale information fusion algorithms. Then the proposed algorithms can improve the positioning accuracy of AUV navigation systems with asynchronous multi-sensors. 3. Fault detection methods are studied as many uncertainties in AUV workplaces may affect the stability and accuracy of the sensors. A novel fault detection method using Gaussian process regression (GPR) is proposed to solve the problem that the gradual fault is difficult to detect timely. To avoid the local optimization, particle swarm optimization is introduced to find the optimal hyper-parameters of GPR model. The GPR model is used to predict the innovation of Kalman filter. Then a novel fault detection function (FDF) is build by employing the predicted innovations. The structure of the FDF help enlarges the difference between the FDF value of fault-free system and that of failing system. The semi-physical simulation shows that the proposed method can detect the gradual fault more quickly compared with the residual chi-squared test. Thus the navigation systems with the proposed method can quickly make a judgment and then handle the failures, which enhance the correctness of the navigation results. 4. As a common device for AUV navigation systems, the DVL has the high risk of short-time malfunctions. To solve the problem, a novel hybrid approach is presented. The approach employs partial least squares regression (PLSR) coupled with support vector regression (SVR) to build a hybrid predictor. During the DVL malfunctions, the hybrid predictor offers the estimation of the DVL measurements for information fusion. As the current and past calculating velocities of SINS are taken as the predictor’s inputs, PLSR is applied to cope with the situation where there exists intense relativity among independent variables. Since PLSR is a linear regression, SVR is used to predict the residual components of the PLSR prediction to improve the accuracy. The mathematical simulation and the semi-physical simulation show that the PLSR-SVR hybrid predictor can correctly provide the estimated DVL measurements and effectively extend the tolerance time on DVL malfunctions, thereby improving the accuracy and reliability of the navigation results. 5. To ensure the security and reliability of AUVs, an intelligent fault-tolerant filtering algorithm based on fuzzy logic is proposed. As insufficiently known priori statistics will reduce the precision of the state estimates, fuzzy logic is employed to adaptively adjust the measurement covariance matrixes of local filters online. The confidence of measurement is introduced as the regulatory factor to realize the fault isolation. The algorithm flow of the fault-tolerant filtering for AUV integrated navigation systems is designed. The simulation based on SINS/MCP/DVL/TAN integrated navigation system shows that, the proposed filtering algorithm detects and insulates both the abrupt fault and the gradual fault effectively, which enhances the fault-tolerance of the system. Based on the above research and innovative design, the accuracy and fault-tolerance of AUV integrated navigation systems have been greatly improved which helps to enhance the reliability and security of AUVs.
学术讨论
主办单位时间地点报告人报告主题
程向红课题组 2013年4月25日 中心楼525 朱倚娴 自适应Fuzzy-PID的复合控制策略
程向红课题组 2014年4月24日 中心楼525 朱倚娴 基于高斯过程回归的联邦滤波故障检测方法
程向红课题组 2017年4月10日 中心楼525 朱倚娴 A Hybrid Approach to Deal with DVL Malfunctions for Underwater Integrated Navigation Systems
程向红课题组 2017年10月27日 中心楼525 朱倚娴 组合导航系统中异步多传感器信息融合算法
程向红课题组 2014年9月26日 中心楼525 王晓飞 临近空间飞行器的INS/GPS组合导航技术研究
程向红课题组 2015年9月25日 中心楼525 胡杰 单轴旋转捷联惯导系统轴向陀螺漂移标校方法
程向红课题组 2016年10月21日 中心楼525 刘全 单轴旋转的误差分析及旋转方式
程向红课题组 2017年6月5日 中心楼525 毕校伟 基于多种群遗传算法的USV PID控制
     
学术会议
会议名称时间地点本人报告本人报告题目
惯性技术发展动态发展方向研讨会 2013年10月31日 上海海悦酒店 基于自适应模糊-PID复合控制的陀螺稳定平台
IEEE/ASME International Conference on Advanced Intelligent Mechatronics 2015年7月8日 韩国釜山BEXCO An intelligent fault-tolerant strategy for AUV integrated navigation systems
     
代表作
论文名称
一种陀螺稳定平台自适应模糊-PID复合控制方法
An intelligent fault-tolerant strategy for AUV integrated navigation systems
A novel fault detection method for an integrated navigation system using Gaussian Process Regression
A novel hybrid approach to deal with DVL malfunctions for underwater integrated navigation systems
 
答辩委员会组成信息
姓名职称导师类别工作单位是否主席备注
何秀凤 正高 教授 博导 河海大学
朱欣华 正高 教授 硕导 南京理工大学
陈熙源 正高 教授 博导 东南大学
刘锡祥 正高 教授 博导 东南大学
潘树国 正高 教授 博导 东南大学
      
答辩秘书信息
姓名职称工作单位备注
阳媛 其他 讲师 东南大学