地下洞室施工损伤区物理参数神经网络预测模型

Neural Network Prediction Model for Physical Parameters of Damaged Zone in Underground Cavern Construction

摘要: 地下洞室爆破施工过程中，爆破损伤区的物理参数对结构的稳定性和安全性具有重要影响。传统的物理参数确定方法依赖于大量的现场试验和经验公式，存在一定的局限性。为提高损伤区物理参数的预测精度，提出了一种基于拉丁超立方体采样的损伤区物理参数的神经网络预测模型。利用拉丁超立方体采样方法计算大量有限元模型，建立了包括顶点位移、掌子面位移等多种测量点在内的输入特征集，并采用神经网络模型进行训练和预测。实验结果表明，该方法能够有效地预测地下洞室爆破施工过程中损伤区岩体的弹性模量变化，并且具有较高的预测精度。与传统方法相比，神经网络方法不仅减少了人为干预，还能够快速适应不同的施工环境，具有较强的应用潜力。最后，讨论了该方法的工程应用前景，并提出了进一步研究的方向。

Abstract: During the underground cavern blasting construction, the physical parameters of the blasting damaged zone have an important impact on the stability and safety of the structure. The traditional method of determining physical parameters relies on a large number of field tests and empirical formulas, which has certain limitations. In order to improve the prediction accuracy of the physical parameters of the damaged zone, this paper proposes a neural network prediction model for the physical parameters of the damaged zone based on Latin hypercube sampling. A large number of finite element models are calculated using the Latin hypercube sampling method, and an input feature set including multiple measurement points such as vertex displacement and face displacement is established. The neural network model is used for training and prediction. The experimental results show that this method can effectively predict the change of the elastic modulus of the rock mass in the damaged zone during the underground cavern blasting construction, and has a high prediction accuracy. Compared with the traditional method, the neural network method not only reduces human intervention, but also can quickly adapt to different construction environments, and has strong application potential. Finally, this paper discusses the engineering application prospects of this method and proposes the direction of further research.