动态模糊神经网络在大坝变形预报中的应用

针对静态模糊神经网络的局限性,提出了在线动态建模的模糊神经网络方法.当新增样本进入训练集之后,根据新样本对模型的贡献大小,在已有模型的基础上进行动态修正,这样可以减少建模的计算时间.新方法实现了增加样本而矩阵阶数不增加,避免了矩阵求逆运算,理论上可以提高计算效率.实例表明动态模糊神经网络方法是可行的,可实现持久预报,具有较强的适应能力和较高的预报精度,可应用于在线实时变形预报及相关领域.

Application of Dynamic Fuzzy Neural Network to Dam Deformation Prediction

To overcome the limitation of static fuzzy neural network model for dam deformation prediction, the dynamic fuzzy neural network approach is proposed. When a new sample is added to the training set, the online dynamic revising to the original model is performed according to the contribution of the new sample, The novel approach can help reduce the modeling time and enhance the computation efficiency theoretically, for it need not compute the inverse matrix, and the increment of matrix size is avoided whenever a new sample is added to the training set. Experiments indicate that the dynamic fuzzy neural network model is feasible in revising the original model with high speed and predicting the dam deformation with high precision. It can adapt itself continually to the new samples with new conditions. Therefore, it can be used for online deformation prediction and similar areas.