关键词: 深水群桩基础, 支持向量机, 蚁群算法, 轴力预测, ACO-SVM模型

Abstract: As for large-scale pile group foundation with deep water, relationship between axial force of pile shaft and environmental factor is complex and nonlinear due to complex environment. In light of advantages of support vector machine(SVM) method in solving small sample size, nonlinearity, and high dimension, we use the method to analyze measured data of axial force in pile group foundation of Suzhou-Nantong bridge, and to predict axial force for a period. Then, we look for optimal parameters by using ant colony optimization(ACO) and establish ACO-SVM model, which can avoid optionally choosing parameters. Meanwhile, we establish prediction models based on traditional SVM and RBF neural network and compare prediction results of the 3 models. The results show that, CO-SVM model is of high reliability, high accuracy and strong generalization ability, superior to SVM and RBF. Finally, CO-SVM model can be applied to predict axial force in large-scale pile group foundation with deep water.

Key words: deep-water pile group foundation, support vector machine, ant colony algorithm, axial force prediction, ACO-SVM model