改进灰色人工神经网络模型的超高层建筑变形预测

针对灰色人工神经网络模型初始化权值和阈值的随机性导致易产生误差积累和过拟合的缺陷,该文利用遗传算法的全局优化能力训练灰色人工神经网络模型的权值和阈值,构建了基于遗传算法的灰色人工神经网络超高层建筑物变形预测模型。结合长沙北辰新河A1超高层建筑变形监测实例,用该文所提模型与灰色人工神经网络模型分别进行变形数据的处理分析和预测。实验结果表明,该文提出的模型具有更好的预测精度,预测趋势也更加逼近实际测量结果。

The GM-BPNN prediction research in the deformation forecasting of the super high-rise building

In view of the randomness of initializing the grey theory and back propagation neural network(GM-BPNN)model's weights and threshold and the defects of error accumulation and over fitting,the global optimization ability of genetic algorithm was used to train the GM-BPNN model's weights and threshold;then the GM-BPNN prediction model based on genetic algorithm(GA-GM-BPNN)was proposed for the deformation forecasting of super high-rise building.Taking the A1 super high-rise building in Changsha Beichen Xinhe as an example,the deformation data processing and prediction of GA-GM-BPNN and GM-BPNN model were analyzed.The results showed that GA-GM-BPNN model has the better prediction precision,and its forecast trend is more close to the actual measurement results.