Improvement Of Reliability Analysis Method And Its Application On Breakwater Stability Of Nuclear Power Plant

Breakwater in nuclear power plant are provided primarily for nuclear power plant and water drainage facility and pumping station such as circulating cooling water and nuclear safety water to avoid attack from wave. Its foundation is usually seabed subsoil which has poor engineering properties and would easily suffer from geological disasters under external forces such as earthquakes. The stability of breakwater has a great influence on the normal operation of the plant and the water drainage facility and pumping station. It is not enough to evaluate merely the stability with the safety factor because the strength parameter variability of the breakwater foundation and embankment material is large. Currently there is little reliability study of breakwater in nuclear power plant. A new sampling method was employed in traditional response surface method and improved Kriging model to construct performance function. Reliability index was then calculated by using Hasofer-Lind reliability method.Traditional quadratic polynomial response surface method is firstly employed to study the reliability of breakwater.In reliability analysis, iteration is needed in searching for design point when using response surface method. During each iteration, a number of sampling points are generated around the current design point and used to form the approximate explicit performance function. For complex problem, generating sampling point is time consuming. An improved response surface method is proposed and named as cumulative sampling method, where both new sampling points and the existing sampling points are put together and employed to obtain new responses surface. Cumulative sampling method can further utilize the existing sampling points and avoid reliability iteration into a local saddle point or convergence failure. If only the sample points close to the limit state surface are selected in the cumulative sampling, the efficiency can be further improved. The effectiveness and efficiency of the proposed method are examined by two examples. Then this method is employed to analyze the reliability of a breakwater in a nuclear power plant.The quadratic polynomial is often used to form performance function in traditional response surface method, but it is hard to fit a limit state surface with high degree nonlinearity and implicit form and is sensitive to sampling center and sampling method. The particle swarm optimization is used to improve the Kriging model for its global optimization capability, which solves the problem of related parameter optimization, and has better global optimization capability to simulate the limit state surface with high degree nonlinearity and implicit form. The improved sampling method named cumulative sampling method s is employed. Case study shows that improved Kriging model based on sampling cumulative method is good at fitting limit state surface and the efficiency and precision has been improved greatly compared with traditional response surface method. If only the sample points close to the limit state surface are selected in the cumulative sampling, the efficiency can be further improved. Based on two examples of slope stability analysis, it’s suitable to cumulate samples when their factor of safety is within0.9and1.1. The improved Kriging model based on cumulative sampling is not sensitive to sampling center and sampling method and has better adaptability. At last the method is employed to analyze the reliability of a breakwater in a nuclear power plant.