Failure criterion and seismic fragility evaluation of isolated nuclear power plant piping system using BP neural network

Abstract

The pipeline connecting the isolated nuclear island and the non-isolated turbine building is prone to significant relative displacements during intensive earthquakes. Generally, elbows closest to the isolation layer are considered to be at particularly high risk of seismic damage. The failure criterion is the critical part in determining the damage state of elbows throughout the loading cycle. Therefore, in this study, a three-tier, four-category failure criterion was firstly proposed in response to the damage patterns of elbows. A BP neural network model was subsequently developed using extensive simulation data derived from a verified numerical model, accompanied by influence analysis of the three failure indices. Finally, a fragility evaluation was conducted based on the seismic response of the piping system, and a comparison is made between the new criterion and the existing ones. The results indicate that the new failure criterion demonstrates comprehensiveness and safety, enabling a more detailed division of damage before elbow failure and providing more conservative estimates upon failure occurrence. Additionally, the study finds that under the current design of nuclear power plants, elbows across isolation layers fail to meet the requirements for safe shutdown earthquakes in certain directions. By adjusting the elbow parameters, the resistance of the elbows to damage under seismic conditions can be improved to some extent.