ArticlePDF Available

Prediction of Long-Term Seawall Settlement Based on Least Squares Support Vector Machine


Abstract and Figures

In this study, the basic principle of the least squares support vector machine (LS- SVM) method was introduced systematically. The method was applied in a seawall engineering project to establish a prediction model of long-term seawall settlement based on the LS-SVM method, and the model parameters were optimized based on the genetic algorithm. Based on the monitoring data of seawall settlement during the construction of a soft foundation seawall project in Wenzhou, China, a sample was selected to train the model, and the predicted value of long-term settlement calculated by the model was compared with the measured sample value. The results show that, with highly accurate and reliable prediction results, the prediction model of long-term seawall settlement based on the LS-SVM method possesses great application value.
Content may be subject to copyright.
A preview of the PDF is not available
Recently, a robust least squares support vector machine (R-LSSVM) was proposed, but its computational complexity is very high compared with the traditional least squares support vector machine (LSSVM). To reduce R-LSSVM's computational complexity, an improved version, i.e., extended LSSVM (E-LSSVM), is developed in this paper. E-LSSVM and R-LSSVM are equivalent in terms of the generalization performance, but the former needs lower computational complexity than the latter. It is proved that the traditional LSSVM is a special case of E-LSSVM, and based on this fact, we know that the bias in the traditional LSSVM owns manifest physical meaning, i.e., the mean of the modeling error. To solve the mathematical model of E-LSSVM, two algorithms, DE-LSSVM (dual E-LSSVM) and PE-LSSVM (primal E-LSSVM), are proposed from dual and primal spaces, respectively. Even competing against the traditional LSSVM, DE-LSSVM takes the edge in term of the training time. In addition, the sparse problem and cross validation of DE-LSSVM are discussed as well. To verify the effectiveness and soundness of the proposed DE-LSSVM and PE-LSSVM, experiments on regression and classification problems are investigated. To be more important, DE-LSSM and PE-LSSVM are successfully applied to the fault diagnosis of aircraft engine, showing that they are eligible for potential techniques of the fault diagnosis of aircraft engine.
Full-text available
This paper presents a new model for predicting the displacement of a landslide based on the least-squares support vector machine (LSSVM) with multiple factors and a genetic algorithm (GA) is used to optimize the parameters of the LSSVM model. First, based on original monitoring displacement data, single factor GA-LSSVM models are established with and without wavelet decomposition. Second, from the analysis of the basic characteristics of a landslide, the main influencing factors of landslide displacement are identified according to their correlation coefficients. A multifactor GA-LSSVM model is then established for the prediction of landslide displacement. A case study of a landslide reveals that wavelet decomposition can efficiently improve the prediction accuracy of the GA-LSSVM model. In addition, the multifactor GA-LSSVM model performs consistently better than the single factor models for the same measurements.
After the basic theory of Asaoka's method is systematically introduced, the relevant calculation program is written based on parabola interpolation method and linear least squares simulation method. Then, detailed analyses of the effect of time interval ∆t on predicted results and related indexes are done, and the suggested scope of ∆t is given. Besides, the final settlements of each level loading for an embankment subjected to multi-loadings are forecasted. It is observed that the parabola interpolation method in the calculation code is feasible and Asaoka's method has higher application value due to its own characteristics.
Based on the fact that surface settlement was a very complex system engineering, the geo-material was complex, fuzzy and nonlinear in geotechnical engineering, a new model based on grey system was presented, grey relational was used to find the key components. The kernel function and penalty factor was determined by simulated annealing. The nonlinear relation between surface settlement and influencing factors was obtained from the key components by SVM model and the model was applied to the practical engineering. GR-SVM enhances the efficiency and the capability of forecasting. The results of experiment show that the model is valid and can be used to analysis tunnel settlement.
In this chapter we consider bounds on the rate of uniform convergence. We consider upper bounds (there exist lower bounds as well (Vapnik and Chervonenkis, 1974); however, they are not as important for controlling the learning processes as the upper bounds).
Setting of the learning problem consistency of learning processes bounds on the rate of convergence of learning processes controlling the generalization ability of learning processes constructing learning algorithms what is important in learning theory?.
Support vector machines have been very successful in pattern recognition and function estimation problems. In this paper we introduce the use of least squares support vector machines (LS-SVM's) for the optimal control of nonlinear systems. Linear and neural full static state feedback controllers are considered. The problem is formulated in such a way that it incorporates the N-stage optimal control problem as well as a least squares support vector machine approach for mapping the state space into the action space. The solution is characterized by a set of nonlinear equations. An alternative formulation as a constrained nonlinear optimization problem in less unknowns is given, together with a method for imposing local stability in the LS-SVM control scheme. The results are discussed for support vector machines with radial basis function kernel. Advantages of LS-SVM control are that no number of hidden units has to be determined for the controller and that no centers have to be specified for the Gaussian kernels when applying Mercer's condition. The curse of dimensionality is avoided in comparison with defining a regular grid for the centers in classical radial basis function networks. This is at the expense of taking the trajectory of state variables as additional unknowns in the optimization problem, while classical neural network approaches typically lead to parametric optimization problems. In the SVM methodology the number of unknowns equals the number of training data, while in the primal space the number of unknowns can be infinite dimensional. The method is illustrated both on stabilization and tracking problems including examples on swinging up an inverted pendulum with local stabilization at the endpoint and a tracking problem for a ball and beam system.
Prediction of seawall settlement based on BP neural network-real time tracing algorithm model Yangtze River
  • Yuejun Shen
  • Peng Qin
Investigation and analysis on post-construction settlement of coastal levees in Zhejiang Province Water Resources and Hydropower Engineering
  • Fang Wu
  • Weidong Zhu
Comparison of theoretical settlement and actual observation settlement of seawall for reclamation project in West Great Pond Small Hydro Power
  • Hui Xie
  • Hua Zhang
Research on data normalization for SVM training
  • Tang Rongzhi
  • Huichuan
  • Sun Haitao
TANG Rongzhi, DUAN Huichuan, SUN Haitao. Research on data normalization for SVM training. Journal of Shandong Normal University (Natural Science), 31(2016):60-65.