锚杆抗拔强度模型准确性分析

Analysis of the Accuracy of Anchor Rod Pull-out Strength Model

锚杆与周围土体的极限黏结强度估算是锚杆设计的必要参数。现有规范基于锚杆与岩土体类型建议了极限黏结强度特征值取值范围,但该方法所隐含的模型不确定性未知,给锚杆设计带来潜在风险。鉴于此,利用文献收集大量实测锚杆抗拔数据,建立数据库;采用机器学习方法对数据集进行分类,以评估规范公式在不同影响因素下的模型准确性。研究发现:现行规范方法平均低估极限黏结强度约15%~50%,且预测精度离散性高,超过60%。基于数据分析引入模型校正因子,对规范公式进行校正。校正模型平均精度无偏,精度离散性中等。

The estimation of ultimate bond strength between anchor rods and the surrounding soil is the necessary parameters in anchor rod design. Based on the types of anchor rod and rock mass, the existing specifications suggest the value range of the ultimate bond strength eigenvalue, but the model uncertainty implied by this method is unknown, which brings potential risks to the anchor rod design. In view of this, a large number of measured anchor rod pull-out data collected from the literature are used to establish the anchor rod pull-out database. The machine learning method was used to classify the dataset to evaluate the model accuracy of the normative formula under different influencing factors. It is found that the current standard method underestimates the ultimate bond strength by 15 % ~ 50 % on average, and the prediction accuracy has a high dispersion of more than 60 %. Based on the data analysis, the model correction factor is introduced to correct the standard formula.The average accuracy of the calibration model is unbiased, and the accuracy dispersion is medium.