基于高斯过程的大坝变形预测模型

A Gaussian process based model for predicting deformations of dams

鉴于高斯过程在处理高维数、小样本和非线性等复杂的问题的突出表现,提出基于高斯过程的大坝变形预测新模型.详细分析基于GP的大坝变形预测模型构建过程,并建立预测结果的方差及置信区间的估计方法.将GP模型与多种大坝预测模型进行精度对比,并进行结果可靠性分析.通过实例分析可知,GP模型预测结果的精度指标优于其余方法,并且实测值均在估计的置信区间内.因此,证实GP模型是一种精度及可靠性高的大坝变形预测新方法.

In view of the excellent performance of the Gaussian process model in dealing with complicated problems of high dimensions,small samples,and nonlinearity,a Gaussian process based model is proposed for predicting the deformations of dams.The Gaussian process based model for predicting the deformations of dams is analyzed in detail and an estimation method for the variance and the confidence interval is proposed.The Gaussian process based method is compared with a few prediction models for dams and a reliability analysis of numerical results is conducted.Verification examples show that the Gaussian process based method has a higher computational accuracy than other methods and the measured values are all within the confidence interval.Therefore,it is validated that the Gaussian process based model is a new prediction method for the deformations of dams with high accuracy and reliability.