【摘要】 将改进的具有双群特性及带变异算子的粒子群优化算法与小波分析结合优化神经网络预测地基沉降量。针对粒子群算法易陷入局部极小值的缺陷,将粒子总群分成两个子群,分别对两个子群进行不同的搜索策略以增强算法的全局和局部搜索能力。其中一个子群采用变惯性权重进行局部细搜索;另一个子群采用大的惯性权重进行全局搜索,并与小波分析去噪结合,优化神经网络参数,对地基累计沉降数据进行预测。实验结果表明这种划分使算法有较强的全局和局部搜索能力,同时提高了预测精度。更多还原

【Abstract】 The authors used the improved Particle Swarm Optimization(PSO) algorithm that has two subgroups and a mutation operator binding wavelet analysis to optimize neural network parameter to forecast the foundation settlement.Since the basic particle swarm optimization easily falls into the local minimal value,the authors divided the particle swarm into two subgroups.In one subgroup the inertia weight of the particle swarm optimization algorithm decreased when the iterations increased.And in the other subgroup the particle swarm optimization algorithm adopted big inertia weight to do the overall situation search.And the authors used this improved algorithm binding wavelet analysis to optimize the neural network parameter to forecast foundation settlement.The experimental result indicates that this method has strong global and local search ability,and has high forecast precision.更多还原