Research On Nuclear Pipeline Load Identification And Optimization Based On Genetic Algorithm And Neural Network

Nuclear power is an important part of China's clean energy power generation.As of 2018,nuclear power generation accounted for only 5.2%,which is still far away from developed countries.Therefore,vigorously developing nuclear power can maintain the continuous optimization and upgrading of the industrial structure.In the nuclear power system,nuclear pipelines are an important part.In order to ensure the safety and stability of nuclear pipelines,it is necessary to carry out load identification and optimization research on nuclear pipelines.On the basis of reviewing and summarizing the current research status of nuclear pipeline load identification and optimization of pipeline-spring support system,the four algorithms are studied and analyzed using genetic algorithm(GA),multi-island genetic algorithm,BP neural network and GA-neural network Pros and cons.Test examples show that the multi-island genetic algorithm converges faster than the genetic algorithm,and the average genetic algebra is reduced by 23.6%;compared with the fitting effect of the GA-neural network and the BP neural network,the average error rate drops from 0.96% to 0.63%.By establishing the load identification model of the truss bridge,the displacement of the truss bridge is analyzed under the action of multi-point transient loads.According to the displacement of the bridge,the BP neural network and GA-neural network are trained.The results show that the fitting effect is the best and the error is very low when training 100 times.On this basis,the SiPESC.FEMS pipeline analysis system was used to perform load identification and analysis on the transient load and temperature transient load of the nuclear pipeline.The calculation results show that the training effect of the GA-neural network is still very good,and the average error rate is less than 1 %.According to the structural analysis and testing of a discharge nuclear pipeline in actual engineering problems,the natural frequency of the nuclear pipeline system,the response spectrum displacement and acceleration are obtained,which shows that the nuclear pipeline analysis system has the ability to analyze complex practical problems.Further,using the SiPESC.FEMS pipeline analysis system,a pipeline-spring support system model is established,by analyzing the displacement and modal of the pipeline under different spring stiffness(1000 ~ 10000 N / mm).The results show that as the spring stiffness increases,the natural frequency of the pipeline decreases;when the spring stiffness is 4000 ~ 7000 N / mm,the displacement of the pipeline and the axial force at the position of the maximum axial force in the system tend to stabilize.By analyzing the changes of different spring bracket positions(0 ~ 1.71m),it is obtained that with the movement of the spring bracket,the displacement of the pipeline first decreases and then increases,and at the same time,the axial force of the pipeline at the maximum axial force also decreases first and then increases.Large,but different from displacement is that the "inflection point" of displacement is in the middle of the pipe,that is,at 0.855,and the "inflection point" of axial force is at 0.428.Using the SiPESC.OPT multi-objective genetic algorithm,a multi-objective optimization model with different loads,different spring stiffnesses and different spring support positions was established to study the multi-objective optimization of nuclear pipelines.In dual-objective optimization,by optimizing the minimum displacement of the pipeline and the minimum axial force at the maximum axial force as the objective function;then the thermal expansion displacement is added as the third objective function,and the Pareto frontier of all solutions is obtained.And analyze the difference between multi-objective optimization and single-objective optimization from the optimization results.