SIA OpenIR  > 海洋机器人卓越创新中心
水下机器人故障诊断与试验测试方法研究
Alternative TitleFault Diagnosis and Trial-testing of Underwater Vehicles
徐高飞1,2
Department海洋机器人卓越创新中心
Thesis Advisor王晓辉 ; 刘开周
Keyword水下机器人 自适应故障诊断 区间预测 内部异常分析 试验测试
Pages159页
Degree Discipline模式识别与智能系统
Degree Name博士
2019-05-24
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract本文针对水下机器人研发和应用中的实际需求,以辅助水下机器人的研发和运行维护、提高水下机器人的安全性和可靠性为目标,对水下机器人的故障诊断和试验测试方法进行了研究,主要研究内容如下:(1) 针对现有部分故障诊断方法需要系统的精确数学模型或完备的训练数据、对环境和系统自身的变化适应性差、难以区分故障和海流干扰的影响等问题,提出了一种具有在线学习能力的推进系统故障诊断方法。首先,通过分析相关性的变化趋势,在线估计推进系统的时延。然后,利用作业过程中采集的数据,对控制量与转速之间的关系进行在线建模。为提高建模精度,提出了一种具有记忆能力的多中心粒子群算法,对模型阶次和建模数据量进行在线优化。为适应作业过程中环境和系统自身状态的变化,设计了模型在线更新机制。基于该在线更新机制,提出了一种自适应故障检测方法。最后,通过海试数据和水池测试,验证了算法的有效性。该方法的主要环节均通过在线学习实现,与传统方法相比,该方法不需要水下机器人的数学模型和系统的详细信息,不需要进行离线训练,能够根据作业环境和系统自身的变化进行自适应调整,且对海流干扰的影响具有一定的鲁棒性。(2) 针对现有部分故障诊断方法一般采用固定阈值,对算法的自适应能力和性能造成影响的问题,提出了一种新的小样本强振荡时间序列区间预测算法,用于故障诊断过程中模型残差的自适应分析。该算法通过三次样条插值获取原始序列的上下包络线,根据序列振荡强度对包络线与模拟序列的间距进行动态拓展,基于新信息优先原理对包络间距序列预测值进行自适应调整,最终获得区间预测结果。为验证算法性能,采用来自不同各领域的应用实例进行了测试。通过与近期相关领域具有代表性的7种算法进行对比,表明本文算法在小样本强振荡时间序列预测方面具有明显优势。通过对推进器转速模型残差进行预测,表明该算法能够较好的满足对水下机器人故障诊断的工程应用需求。(3) 针对水下机器人中广泛存在位置相邻或对称布置的成组推进器这一情况,利用这种情况带来的优势,提出了一种基于同组推进器状态变量之间相似性分析的推进器故障诊断方法。首先选择合适的相似性度量指标对推进器电流序列之间的相似性进行度量,然后分析了电流序列之间相似性的概率分布特征。为检测相似性概率分布特征的异常变化,基于贝叶斯决策提出了一种相似性异常故障特征增强方法。接下来,采用本文提出的小样本区间预测方法,对相似性异常故障进行自适应检测。最后,基于对相关性变化的分析,实现故障的定位。通过海试数据以及在海试数据中注入故障,验证了该方法的有效性。(4) 针对实际应用中单独进行传感器故障诊断难以实现这一问题,通过分析水下机器人的常规作业状态,给出了近似稳态条件的具体描述,并分析了近似稳态条件下传感器的信号特征。在近似稳态条件的基础上,分别提出了基于小波细节信号分析和基于动态优化指数平滑的两种传感器故障诊断方法。通过海试数据以及在海试数据中注入故障,验证了算法的有效性。(5) 针对现有复杂系统故障诊断方法严重依赖系统详细信息和专家知识的问题,为了能够在缺少详细设计信息的情况下实现水下机器人系统内部异常现象影响因素的分析,提出了一种完全基于数据的分析方法。首先采用最大信息系数和Pearson系数,对运行数据中不同变量之间可能存在的线性或非线性关系进行分析,筛选出与异常现象有较强相关性的变量。接下来,通过时延估计对具有相关性的变量之间的因果性进行分析,找出有可能导致异常现象的变量。通过对海试中出现的异常现象进行分析,验证了该方法的有效性。(6) 针对在实验室条件下对水下机器人进行测试的具体需求,在现有数字化仿真测试和半物理仿真测试的基础上,论述了水下机器人试验测试的概念,并对相关理论进行了分析。在理论分析的基础上,设计并研制了一套完整的水下机器人试验测试系统,并在相关水下机器人的研制中取得了初步应用成果。本文的研究内容完全面向实际应用展开,本文中提出的算法将在我国正在研制的全海深载人潜水器中进行实验应用,本文中设计的试验测试系统将在近期完成最终集成并投入实际应用。
Other AbstractIn view of the actual needs in the development and application of underwater vehicles, fault diagnosis and trial-testing methods of underwater vehicles are studied in this dissertation, with the goal of assisting the development, operation and maintenance of underwater vehicles and improving the safety and reliability of underwater vehicles. The main research contents of this dissertation are as follows: (1) In order to solve problems of some traditional fault diagnosis methods such as need accurate vehicle model or comprehensive training data, difficult to adapt to changes in the environment or the system itself, difficult to distinguish influence of fault and sea current disturbances, a thruster system fault diagnosis method with online learning ability is proposed. Firstly, time delay of the thruster system is estimated online by analyzing the changing trend of correlation. After that, online modeling of the relationship between control voltage and rotating speed is implemented use the data acquired during operation. In order to improve the modeling accuracy, a multi-center particle swarm optimization algorithm with memory ability is proposed to optimize the model order and modeling data volume. To adapt to the changes of environment and system status during operation, an online update mechanism of the model is designed. Based on the online update mechanism, an adaptive fault detection method is proposed. Finally, the effectiveness of the proposed algorithm is verified by sea trial data and pool tests. The main steps of the proposed method are implemented through online learning, compared with traditional methods, the proposed method does not require mathematical model and detailed information of the underwater vehicle, need no offline training, can be adaptively adjusted according to the changes of operating environment and the system itself, robust to the disturbance of sea current. (2) In order to solve the problem that existing fault diagnosis methods generally adopt fixed thresholds, which affect the adaptive ability and performance of the fault diagnosis algorithm, a new interval prediction algorithm for small sample strong oscillation time series is proposed to realize the adaptive analysis of model residuals. The proposed algorithm obtains the upper and lower envelopes of the original sequence by cubic spline interpolation, expands the distance between envelopes and the simulated sequence according to the sequence oscillation intensity, and adaptively adjusts the predicted value of the spacing sequence based on the new information priority principle, and finally obtains the interval prediction result. To verify performance of the proposed algorithm, tests were carried out using application examples from different fields. Compared with five representative algorithms in the recently related field, the proposed algorithm has obvious advantages in the prediction of small sample strong oscillation time series. By predicting the residuals of the thruster speed model, it is shown that the proposed algorithm can meet the engineering application requirements of underwater vehicle fault diagnosis well. (3) Grouped thrusters positioned adjacent or symmetrically widely existed in underwater vehicles, take advantage of this phenomenon, this dissertation proposed a thruster fault diagnosis method based on similarity analysis between state variables of the same group of thrusters. Firstly, a proper similarity measure metric is selected to measure the similarity between thruster current sequences, and then the probability distribution characteristics of the similarity between current sequences are analyzed. In order to detect the anomalous variation of the similarity probability distribution, a similarity anomaly fault feature enhancement method based on Bayesian decision is proposed. After that, the small sample interval prediction method proposed in this dissertation is used to adaptively detect similarity anomalies. Finally, based on the analysis of the changes of correlation, the fault location is achieved. The effectiveness of the proposed method is verified by sea trial data and the injection of faults in the sea trial data. (4) In practical applications, it is difficult to achieve sensor fault diagnosis alone. By analyzing the routine operation scenarios of underwater vehicles, a detailed description of approximate steady-state conditions is given in this dissertation, and characteristics of sensor signals under approximate steady-state conditions are analyzed. Based on the approximate steady-state conditions, two sensor fault diagnosis methods based on wavelet detail signal analysis and dynamic optimization exponential smoothing are proposed respectively. The effectiveness of the proposed algorithm is verified by sea trial data and the injection of faults in the sea trial data. (5) Existing complex system fault diagnosis methods rely heavily on system details and expert knowledge. In order to analyze the influencing factors of internal anomalies in underwater vehicle in the absence of detailed design information, a data-based analysis method is proposed. Firstly, the maximum information coefficient and Pearson coefficient are utilized to analyze the possible relationships between different variables in the running data, and to select the variables that have strong correlation with the anomalies. After that, the causality between variables with correlation is analyzed by time delay estimation to find variables that may cause the abnormal phenomena. The effectiveness of the method is verified by analyzing the anomalies appearing in sea trial. (6) To meet the specific needs of testing underwater vehicles under laboratory conditions, this dissertation discussed the concept of trial-testing based on existing digital simulation testing and semi-physical simulation testing systems, and analyzed related theories. On the basis of theoretical analysis, a trial-testing system for underwater vehicles was designed and developed, and preliminary application results were obtained in the developing of several underwater vehicles. The research of this dissertation is completely for practical applications. Algorithms proposed in this dissertation will be test and applied in the full ocean depth human occupied vehicle being developed in China. The trial-testing system designed in this dissertation will complete the final integration and put it into practical use in the near future.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/25162
Collection海洋机器人卓越创新中心
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
徐高飞. 水下机器人故障诊断与试验测试方法研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2019.
Files in This Item:
File Name/Size DocType Version Access License
水下机器人故障诊断与试验测试方法研究.p(9311KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[徐高飞]'s Articles
Baidu academic
Similar articles in Baidu academic
[徐高飞]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[徐高飞]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.