通过将遗传算法和支持向量机相结合,建立了用于PC斜拉桥主梁标高误差预测的自适应GA-SVM模型。采用遗传算法对传统的支持向量机进行改进,利用其强大的全局搜索能力求解出支持向量机的最佳参数组合。将该模型用于梅溪河斜拉桥施工过程主梁标高误差预测。结果表明:该模型具有较高的学习和泛化推广能力,给出的预测结果令人满意。 By combining genetic algorithm with support vector machine, an adaptive GA-SVM model is established for predicting elevation error of main beam of PC cable-stayed bridge. The genetic algorithm is used to improve the traditional support vector machine, and its best global search capability is used to solve the best combination of support vector machines. The model is used to predict the elevation error of the main girder during the construction of Meixi River cable - stayed bridge. The results show that this model has high learning and generalization ability, and the forecast result is satisfactory.