基于改进粒子群优化算法的BP神经网络在大坝变形分析中的应用

BP神经网络以其对非线性系统的强大映射能力而被广泛应用于模糊性、随机性强的大坝变形预测分析中。传统的BP神经网络由于初始权值和阈值的随机性,容易导致网络在训练过程中极易陷入局部最小值,同时存在网络收敛速度慢等缺点。针对传统算法的不足,采用改进的粒子群算法(IPSO)对BP网络的初始权值和阈值给予优化,建立大坝变形预测的IPSO-BP模型,并与PSO-BP网络模型进行对比。结果表明,改进的IPSO-BP模型具有收敛速度更快、预测精度更高的优点。该方法可供大坝安全监测和预警分析参考。

Application of improved particle swarm optimization algorithm based BP neural network to dam deformation analysis

With the strong nonlinear mapping capability，BP neural network is widely applied to the analysis of dam deformation prediction with strong fuzziness and randomness. Due to the randomness of the initial weight and threshold，the conventional BP neural network is prone to make the network fall into the local minimum in the training process，while it has the defect of slow network convergence speed. Aiming at the defect of the conventional algorithm，the BP network initial weights and thresholds are optimized with the improved particle swarm algorithm （IPSO），and then a IPSO-BP model for dam deformation prediction is established，which is compared with that of the PSO-BP network model. The result shows that the improved IPSO-BP model has the merits of faster convergence speed and higher prediction accuracy. The method can provide a reference for dam safety monitoring and early warning analysis.