基于连续蚁群算法和小波支持向量机的位移反分析

Displacement back analysis based on continuous ant colony algorithm and wavelet support vector machines

提出一种基于连续蚁群算法和小波支持向量机的位移反分析模型。一方面利用具有良好时域、频域分辨能力和非线性学习功能的小波支持向量机建立反演参数和位移之间的非线性关系,避免了大量的数值计算,提高了预测精度;另一方面利用全局优化的连续蚁群算法代替传统的优化算法,避免优化过程中目标函数陷入局部最优,提高了反演的精度。应用该模型对三峡工程永久船闸高边坡4种介质弹性模量进行位移反分析,并利用反演所得参数进行监测点位移预测,计算值与监测值吻合较好,表明该方法适合解决具有非线性和不确定特性的岩土工程问题,在位移反分析中具有良好的实际应用价值。

A displacement back analysis model is proposed by combining the continuous ant colony algorithm （CACA） and wavelet support vector machines （WSVM）. On the one hand, the model establishes the nonlinear relationship of inversion parameters and displacement, minimizing numerical computation and improving forecasting accuracy with the use of WSVM, which has a good time domain, a good frequency domain, good resolving power, and nonlinear self-study capability. On the other hand, the global optimization performance of CACA rather than that of traditional optimization methods is used in the model, thereby avoiding the objective function trap in the local optimum and improving the inversion precision. The model is applied to displacement back analysis of the elastic modulus of four mediums of the Three Gorges Project （TGP） permanent lock high dope. The inversion parameters are used to predict the displacement of the monitoring points. The calculated displacement agrees with the detected displacement, indicating that the model is suitable for solving geotechnical engineering problems with nonlinearity and uncertainty. It has practical value in displacement back analysis.