Safety Design Support of Complex Systems using Integrated Models Based on Physical Analogy

本研究的目的是利用基于物理类比规则的集成模型，建立由电气系统、机械系统、流体系统等不同系统组成的复杂系统的系统安全设计支持方法。第一阶段，平成13，对系统事故原因的推导方法、检测部件故障的传感器配置、故障诊断方法进行了研究。有必要考虑硬件、软件和人为行为作为故障的原因。一个集成的系统模型由以键合图表示的硬件动作和以输入输出关系表示的软件和人的动作组成，这种输入输出关系改变了硬件的特性。在综合模型的基础上，推导出可能导致系统事故的因果序列。考虑保护系统的作用，导出了系统失效的发生条件。通过一个简单的化工厂实例，验证了该方法的优越性。在此基础上，提出了一种考虑部件失效时间顺序的简易系统故障发生概率评估方法，并与马尔可夫分析方法进行了对比验证。在此基础上，基于键合图给出的集成模型，得到了任意部件故障识别所需的监测点，提出了最优检测顺序，并对其有效性进行了验证。第二阶段，平成14，完成了安全计算机辅助设计框架，并在案例分析中进行了综合评价。在综合系统模型框架下，采用动态贝叶斯网络作为诊断方法。对于作为系统安全设计基础的纵深防御来说，异常事件的识别及其损失评估是非常重要的。绘制了系统中潜在扰动的传播路径，并在考虑防御系统的情况下，给出了存在风险的发生概率。安全措施框架旨在减少不可接受的风险，并通过对汽化器系统的分析验证了其有效性。将显示物理行为模型的随机模型和显示系统动态状态转移的部件失效模型统一起来的动态贝叶斯网络用于故障诊断，可以方便地进行系统状态推定和异常状态诊断

This research aims at establishment of the system safe design support method by the integrated model based on a physical analogy rule to the complex system which consists of different systems, such as an electric system, a mechanical system, and a fluid systemIn the first stage, Heisei 13, the derivation method of system accident causes, the sensor allocation which detects component failure, and its diagnostic method were examined. It is necessary to take into consideration hardware, software, and human action as a cause of failure. An integrated system model is composed of hardware actions expressed by bond graph and software & human actions represented as input-and-output relation which changes the characteristics of the hardware. Cause-effect sequences which may lead to the system accident are derived based on the integrated model. Considering the effect of protective system, the system failure occurrence conditions are derived. The merits of the proposed method were confirmed throu … More gh a simple example of chemical plant. Moreover, the easy evaluation method of the system failure occurrence probability was proposed by considering component failure time order, and the merits were verified as compared with the Markov analysis. Furthermore, monitoring points required for the identification of any component failure can be obtained based on the integrated model presented in bond graph and the optimal inspection order was proposed, and the validity was confirmedIn the second stage, Heisei 14, the framework of safe computer-aided design was completed and comprehensive evaluation in a case analysis was performed. Moreover, the dynamic Bayesian network was considered as the diagnostic method using the framework of an integrated system model. For the depth in the defense which is the foundations of a system safe design, identification and its loss evaluation of an abnormal event are important. The potential disturbance propagation path in a system is drawn, and the occurrence probability in consideration of the defense system gives a present risk. The framework on safety measures aims at reduction if a risk is not acceptable, and its validity was checked in the analysis of a vaporizer system. The dynamic Bayesian network which unified the stochastic model showing the physical behavior model and component failure models which show dynamical system state transition was used for failure diagnosis, and system state presumption and diagnosis of an unusual state could be performed easily Less