建筑物沉降预测的改进Verhulst模型研究

The Improved Grey Verhulst Model for Forecasting Settlement of Building

为进一步提高灰色Verhulst模型的预测精度,将LS-SVM算法与灰色Verhulst模型相结合,对灰色Verhulst模型的参数估计方法和预测方法进行了改进。该方法采用LS-SVM算法,构造以背景值序列和原始序列为训练样本的LS-SVM,将Verhulst模型参数的估计问题转化为灰色LS-SVM的参数估计问题,依据LS-SVM算法求得灰色LS-SVM的参数,进而得到Verhulst模型的参数估计,方法上遵循了结构风险最小化原则,适合Verhulst小样本建模的特点。将改进的模型应用于软土地基建筑物的沉降预测,结果表明本文的方法是可行的且有效的,比传统方法预测精度高。

In order to enhance the forecasting accuracy of grey Verhulst model, an improved grey Verhulst model based on LS-SVM algorithm was presented. This model combines LS-SVM algorithm with the grey Verhulst model to improve the parameters estimation method and the forecasting method. This method constructed the LS-SVM with background value and raw data series as the training sample according to the character of grey difference equation, and converted the Verhulst model parameter estimation problem into a LS-SVM parameter estimation problem, then the parameters in the LS-SVM were solved based on the LS-SVM algorithm and the Verhulst model parameters estima- tion were also obtained. The method follows structural risk minimization principles, suitable for Verhulst model of small samples. The improved model was used to forecast settlement of building with soft soil foundation. The experiment shows that the method in this paper is feasible and effeetive, and has high forecasting accuracy over traditional method.