论文部分内容阅读
水库裂缝开合情况对于水库的安全运行极为重要。将主成分分析法与广义回归神经网络结合在一起,进行水库裂缝开合度的预测。结果表明:应用主成分分析与广义回归神经网络相耦合的模型可以很好的反映环境因子(水压力因子、温度因子、时效因子)与水库裂缝开合度之间的非线性函数映射关系。同时利用Matlab软件对新疆某寒区水库裂缝的开合度进行了实例分析和预测。预测结果显示,水库裂缝开合度的最大相对误差分别8.14%,相关性系数为0.984 7,具有较高的预报精度。通过主成分分析与广义回归神经网络相耦合的方法,有效的消除了原指标间的相关性,降低了神经网络的输入,提取了对因变量解释性最强的成分,使广义回归神经网络的输入层节点数由原来的8个减少到2个,起到了网络结构的简化,增强了网络的稳定性。耦合模型弥补了最小二乘回归无法有效识别和消除因子间多重相关性影响的不足,为水库裂缝开合度、大坝位移等指标预测提供了新的思路和方法。
The opening and closing of reservoir cracks is very important for the safe operation of reservoirs. The principal component analysis method and generalized regression neural network are combined together to predict the degree of opening and closing of reservoir fractures. The results show that the model coupled with principal component analysis and generalized regression neural network can well reflect the nonlinear function mapping relationship between environmental factors (water pressure factor, temperature factor, aging factor) and reservoir opening degree. At the same time, Matlab software was used to analyze and predict the opening degree of reservoir fractures in a cold area in Xinjiang. The prediction results show that the maximum relative error of reservoir opening degree is 8.14% and the correlation coefficient is 0.984 7, which has high forecast accuracy. By the method of principal component analysis coupled with generalized regression neural network, the correlation between the original indexes is effectively eliminated, the input of neural network is reduced, the component with the most explanatory of the dependent variable is extracted, and the generalized regression neural network The number of input layer nodes is reduced from 8 to 2, which simplifies the network structure and enhances the network stability. The coupled model makes up for the deficiencies that least-squares regression can not effectively identify and eliminate the multiple correlativity between factors, and provides a new idea and method for the prediction of reservoir opening degree and dam displacement.