【作者】 杨婷；

【导师】 张汝波；

【作者基本信息】 哈尔滨工程大学 ， 计算机应用技术， 2009， 硕士

【摘要】 智能水下机器人(AUV)由于工作环境的复杂性、不可预测性,一旦出现故障,轻则造成机器人失效无法完成水下作业任务,重则导致机器人沉没丢失,造成严重损失,因此如何提高其软件可靠性及故障诊断能力已成为AUV研究领域的重要课题。本文以某智能水下机器人为背景,利用可靠性分配与分析方法,对AUV智能规划与决策控制系统软件各模块进行了可靠性分配及故障模式分析研究,然后针对AUV主要的软件故障模式,建立了基于FMEA的贝叶斯诊断网络。本文首先分析了水下机器人的软件体系结构,在此基础上,给出了基于重要度、执行情况和复杂度的可靠性分配方法对水下机器人软件模块进行可靠性分配。其中在计算调用系数和复杂度系数时,根据软件实际运行情况,给出了结合Thayer和MEMBOW两种复杂性系数计算方法来权衡复杂度系数,并进行了详细的实现过程。在得出水下机器人各软件模块的可靠度后,结合FMEA分析了水下机器人主要软件模块的失效模式及影响。然后针对传统FMEA中RPN计算存在的不足,将模糊集理论与灰色关联理论结合对FMEA进行了改进。最后,针对常见软件故障模式,本文建立了基于FMEA的贝叶斯诊断网络。该网络模型由故障原因,故障模式以及故障影响三层结构组成,并且对单一故障和多故障进行了实验。实验表明,采用上述方法能有效地提高水下机器人系统的可靠性以及故障诊断能力,确保水下机器人在复杂海洋环境下安全可靠地航行与作业,为进一步的容错控制提供依据,在水下机器人技术中有着重要的现实意义。更多还原

【Abstract】 In view of the complexity and unpredictability of working conditions, Autonomous Underwater Vehicle (AUV) once takes place of breakdown or runs into trouble, it will result in great losses, such as underwater mission failure or submerge damage which causes serious consequences. Therefore it is very important for AUV to ensure its software reliability and failure diagnosis ability in the AUV research area.As for software modules of AUV intelligent planning and decision-making control system, a method using software reliability of distribution and analysis technology is researched. Then in the base of underwater vehicle’s main failure mode, a Bayesian Network based on FMEA is proposed which is applied into fault diagnosis of AUV.Firstly, the software architecture of AUV is discussed, and then one reliability allocation method based on importance, operational practice and complexity of AUV software modules is described. And during computing transfer coefficient and complexity coefficient, a new method based on Thayer and MEMBOW is applied to solve the problem of complexity coefficient computation for AUV software modules.Secondly, after obtaining the reliability of AUV software modules, main software failure modes and effects are analyzed by using FMEA method. Allowing for the insufficiency of traditional FMEA on computing RPN, a FMEA method is improved according to the fuzzy set theory and the gray relational theory.Finally, taking the common software failure mode into account, a three-layered topological Bayesian diagnostic Network based on FMEA is proposed. This network model consists of failure mode, cause and effect. Meanwhile considering single failure and complex failure, we carry on experiments separately according to Bayesian net. As a result, experiments show that the above methods can effectively improve the reliability of the AUV system, as well as fault diagnostic capabilities. Furthermore they are ensured that AUV can be capable of navigating and completing underwater task safely and reliably in a complex marine environment. The results also provide the basis for fault-tolerant control which has great practical significance in AUV technology.更多还原