Systems Reliability/Safety Analysis Using System Behavior Model

研究的主要内容是开发一种基于系统行为模型获取系统故障信息的方法，以防止系统分析人员的主观判断和经验导致分析错误。在项目的第一年，检查了键合图的适用性，以便从系统设计假设中获得系统行为模型。键合图可以从能量流的角度统一地表示一个由机械、电气和液压元件组成的系统。系统的动力学行为可以用系统状态方程表示，系统状态变量表示能量存储，输入变量表示能量供应，特征函数表示元素。由于部件的特征函数在设计初期只能定性地表示，本文提出了一种利用稳态状态条件定性地评价部件失效条件下过程变量偏差的方法。在第二年，开发了一种使定性评价更有效的方法。首先，用树形图表示由系统键合图得到的系统状态方程。树状图由节点和分支组成，前者表示键合图元素、输入变量和系统状态变量，后者表示节点之间的输入输出关系，并具有其源节点的输出值。通过沿树状图传播由组件失效假设引起的偏差，可以评估节点上的所有偏差。可处理多种失效情况。因为一个节点对应一个系统状态，所以所有的系统状态都可以从树状图的节点值中求值。只要对系统行为的假设是正确的，就保证了树图评价结果的有效性。此外，还考虑了分析结果在系统故障诊断中的应用。然而，由于无法找到一种有效的方法来减少能够解释观测到的系统状态的候选故障的数量，因此利用树图进行故障识别仍然是一个有待解决的问题。少

The main of the research is to develop a method of obtaining information on the system failure based on the system behavior model to prevent analysis errors caused by the subjective judgment and experience of system analysts.In the first year of the project, the applicability of bond graphs is examined for obtaining a system behavior model from the system design assumptions. Bond graphs can represent a system composed of mechanical, electrical and hydraulic components in a unified way from the viewpoint of energy flow. The system dynamical behavior can be represented as system state equations in terms of system state variables denoting energy storage, input variables denoting energy supply and characteristic functions denoting elements. Since component characteristic functions can be represented only qualitatively at the early stage of design, a method is developed for the qualitative evaluation of deviation in process variables caused under a component failure condition using the stea … More dy state condition.In the second year, a method is developed for making the qualitative evaluation more efficient. Firstly, system state equations obtained from a system bond graph are represented using the tree graph. A tree graph is composed of nodes and branches : the former represent bond graph elements, input variables and system state variables, while the latter represent input-output relations and between nodes and has the output value of their source node. By propagating a deviation caused by a component failure assumption along the tree graph, all the deviations at nodes can be evaluated. Multiple failure conditions can be dealt with. Since a node corresponds to a system state, all system states can be evaluated from the node values of a tree graph. As long as the assumptions on the system behavior are correct, the validation of the evaluation result from the tree graph is guaranteed. Further, application of the analysis result to the system failure daignosis is considered. However, the failure identification using the tree graph remains a subject to be solved in the near future because an efficient method cannot be obtained to reduce the number of possible candidates which can explain the observed system condition. Less