A probabilistic method for optimization of fire safety in nuclear power plants

As part of a comprehensive fire safety study for German Nuclear Power Plants a probabilistic method for the analysis and optimization of fire safety has been developed. It follows the general line of the American fire hazard analysis, with more or less important modifications in detail. At first, fire event trees in selected critical plant areas are established taking into account active and passive fire protection measures and safety systems endangered by the fire. Failure models for fire protection measures and safety systems are formulated depending on common parameters like time after ignition and fire effects. These dependences are properly taken into account in the analysis of the fire event trees with the help of first-order system reliability theory. In addition to frequencies of fire-induced safety system failures relative weights of event paths, fire protection measures within these paths and parameters of the failure models are calculated as functions of time. Based on these information optimization of fire safety is achieved by modifying primarily event paths, fire protection measures and parameters with the greatest relative weights. This procedure is illustrated using as an example a German 1300 MW PWR reference plant. It is shown that the recommended modifications also reduce the risk to plant personnel and fire damage