隧道洞室地基稳定性进化支持向量机识别

Stability recognition of tunnel cavern foundation based on support vector machines with genetic algorithms

针对隧道洞室地基稳定性的影响因素较多,计算复杂的情况,为了快速评价洞室地基的稳定性,选择合理的洞室开挖方案,加快施工进度,提出一种基于支持向量回归(SVR)算法的洞室地基稳定性识别方法。支持向量机(SVM)具有结构简单、学习泛化能力强等优点,采用遗传算法结合交叉验证法选择模型参数,提高了预测精度。以重庆小什字车站洞室地基为例,验证了建立的进化支持向量回归模型能够快速、准确的获取不同方案下的洞室地基安全系数,进而评价其稳定性,且预测结果比模糊神经网络预测结果要好,结果表明该方法是可行的。

Considering the complexity of stability calculation of tunnel cavern foundations, in order to acquire cavern foundation safety factors of different schemes quickly and accurately then to choose the reasonable scheme and improve the efficiency of construction, a cavern foundation stability predicting model based on SVR algorithm is put forward. Support vector machines have the advantage of simple structure and excellent performance in machine learning. In order to improve the accuracy of predicting, genetic algorithm and cross validation methods are used to select the model parameters. Take Xiaoshizi tunnel cavern foundation in Chongqing as a example, the efficiency of the model is verified. Result proves the method is feasible.