基于GA-BP神经网络对控制棒水力缓冲器的性能预测和结构参数优化

Performance Prediction and Structural Parameter Optimization of Control Rod Hydraulic Buffer Based on GA-BP Neural Network

为通过遗传算法(GA)改进的反向传播(BP)神经网络模型预测控制棒组件水力缓冲器的缓冲性能,并实现结构参数优化,本研究对一种特定的控制棒组件水力缓冲器静水中的落棒进行模拟试验,变换试验可调参数,设置不同的试验工况,获取了大量的试验数据,通过GA-BP神经网络对控制棒组件落棒过程的最大冲击力进行预测,并构建优化数学模型,使用非线性规划函数(fmincon)进行求解,获得更优的结构参数组合。结果表明:GA-BP神经网络模型相较于BP神经网络模型具有更高的预测精度,通过fmincon函数可以实现对控制棒组件最大冲击力优化数学模型的快速求解。因此,本文的优化方法可为水力缓冲器的结构优化设计提供一定的参考。

In order to predict the buffer performance of hydraulic buffers of control rod assemblies by back-propagation(BP)neural network model improved by genetic algorithm(GA)and to realize the optimization of structural parameters.In this study,we simulated the falling rod in hydrostatic water for a specific control rod assembly hydrodynamic buffer.By changing the adjustable parameters of the test and setting up different test conditions,a large number of test data were obtained.The maximum impact force of control rod assembly in the process of rod falling was predicted by GA-BP neural network,and an optimized mathematical model was constructed.The nonlinear programming function(fmincon)is used to solve the problem,and a more optimal combination of structural parameters is obtained.The results show that the GA-BP neural network model has higher prediction accuracy compared with the BP neural network mdoel,and the fmincon function can realize fast solution of the optimal mathematical model of the maximum impact force of the control rod assembly.Therefore,the optimization method in this paper can provide some reference for the structural optimization design of hydraulic buffers.