抄録
We have developed a failure propagation prediction and source detection system as an engineering tool for high reliability design of large scale plants. The system operates with four basic methods: the fault location method, the failure propagation prediction method, the nonfailure paths search method, and the risk assessment method. These methods are used not by way of a traditional tree model, but by way of a network model, i. e., a digraph with time and probability. First, we show that the network model helps us to grasp easily the structure of objective plants and also estimate the failure propagation process.
Next, we describe the actual system to apply our methods to two kinds of post-petroleum energy plants. Presenting the result of their application we show that our methods are useful in gathering materials for preventive maintenance design. We shall know the identifiability of failure origin, the validity of sensor location, the time allowance for counter operations, the assessment of hazards, the confirmation of functioning equipment, and so on.
Then, we point out the important matters such as a malfunction of gas flow oscillator and a gas leak from reboiler that demand special attention for preventive maintenance design of the post-petroleum energy plants.
