Optimal Design Of Plate Spring Based On Improved BP Neural Network

The fuel assembly plate spring is an important component of the fuel assembly of pressurized water reactor.It mainly provides appropriate compression force to ensure the safe and reliable operation of the fuel assembly,and makes up for the difference of axial growth of each component in the reactor cavity to ensure the structural integrity of the fuel assembly.In this paper,the global optimization ability of genetic(GA)algorithm is used to improve the error back propagation(BP)neural network,and an optimization design method of plate spring based on GA-BP neural network is established.Taking the width,width ratio,thickness,thickness ratio and height of the plate spring as the optimization design variables,the better stiffness performance as the objective function,and the performance parameters such as strength and compaction force as well as the process dimension as the constraint conditions,the mathematical model of the optimization of the plate spring was established,and the structural parameter combination to improve the performance of the plate spring was obtained.Among them,the objective function and constraints such as strength and compression force are nonlinear mapping relations between structural parameters and performance parameters,which are obtained by the GA-BP neural network performance prediction model of plate spring.The main work of this paper is as follows:(1)The plate spring model is simplified,and the pressing process is simulated by using finite element software.The results of simulation and test are in good agreement,which verifies the reliability of the simulation results.The orthogonal test method was used,and the parameterized modeling script was written,and the neural network data set was obtained by finite element calculation.(2)Based on the BP neural network,the strength,stiffness and compression force prediction models of the plate spring were built respectively.The neural network was trained with the preprocessed data set,and the nonlinear relationship between the structural parameters of the plate spring and the above performance parameters was obtained.According to the calculation results,it is found that BP neural network has the disadvantages of weak generalization ability and easy to fall into local minimum.(3)In view of the shortcomings of BP neural network,genetic(GA)algorithm was used to improve it,and the prediction models of strength,stiffness and compression force of plate spring based on GA-BP neural network were built respectively.The nonlinear relationship between structural parameters and performance parameters of plate spring was obtained by using the same data set to train them.Comparing the two neural network models,it is found that GA-BP neural network is superior to BP neural network in both fitting ability and generalization ability.(4)Based on GA-BP neural network performance prediction model of plate spring,an optimization mathematical model of plate spring was established.Through finite element calculation,the optimized rear plate spring is compared with the existing one,and it is found that the stiffness performance of the optimized rear plate spring is better,and the overall performance is improved compared with the existing plate spring.