基于高斯过程机器学习方法的隧道围岩分类模型

Classification of Rocks Surrounding Tunnel Based on Gaussian Process for Machine Learning

针对现有围岩分类方法的局限性,基于工程实例,利用分类性能优异的高斯过程机器学习模型建立围岩类别与其主要影响因素之间的非线性映射关系,进而提出一种基于高斯过程的隧道围岩分类模型,实现不同情况下围岩分类的合理识别。将该模型应用于川藏公路二郎山隧道围岩分类,研究结果表明,隧道围岩分类的高斯过程机器学习模型是科学可行的,与人工神经网络模型、支持向量机模型相比较,该模型具有参数自适应化的优点,能方便快捷地给出合理可靠且具有概率意义的围岩分类评价结果,可对围岩分类结果的不确定性或可信度进行定量化评价。

Aiming at the limitations of traditional methods of classifying surrounding rocks, a Gaussian process-based model for classification of surrounding rocks is proposed. The nonlinear mapping relationship between the classification of surrounding rocks and influencing factors is easily established using the Gaussian process model for machine learning, which possesses excellent classification performance based on making use of the historical knowledge of real engineering projects. The model has been applied to the classification of the surrounding rocks of the Erlangshan Tunnel on the Sichuan-Tibet Highway. The results of this case study show that the model is feasible and that reasonable, reliable and probabilistic results for classification of surrounding rocks can be obtained quickly by using the proposed model. Compared with other machine learning technologies, such as artificial neural networks and support vector machines, the proposed model has the beneficial characteristics of self-adaptive parameter determination and an uncertainty evaluation of predicted results.