论文部分内容阅读
针对非等时距以及BP神经网络会陷入局部极小值的问题,本文提出了基于非等时距加权灰色组合GA-BP模型的新方法。首先对原始变形监测数据采用非等时距的加权灰色组合线性模型处理,然后利用处理后的数据求出残差进而建立GA-BP神经网络模型进行残差修正。通过对比非等时距GM(1,1)以及非等时距加权灰色线性组合模型的结果,发现基于非等加权灰色线性组合GA-BP神经网络模型可以有效地提高模型精度,具有一定的应用价值。
In view of the problem that the non-isochronous distance and the BP neural network will fall into local minima, this paper proposes a new method based on the gray-combining GA-BP model with non-equidistance. First of all, the original deformation monitoring data is processed by weighted gray combined linear model with non-equal interval time. Then the residuals are obtained by using the processed data to establish the GA-BP neural network model for residual error correction. By comparing the results of non-equidistant GM (1,1) and non-equidistant weighted gray linear combination model, it is found that GA-BP neural network model based on non-equal weighted gray linear combination can effectively improve model accuracy and has some applications value.