量纲分析与工程数据挖掘结合的掘进总载荷建模

Total loads modeling of tunnel boring machines based on dimensional analysis and in-situ data mining

近年来,越来越多的工程装备实现了施工过程中多维数据的检测与收集。同时,通过数据挖掘建模来实现工作参数的优化调控已成为一种发展趋势。如何将数据挖掘过程与所研究问题的内在机理结合,建立具有一定物理可解释性的参数预测模型并提高其泛化性,是该研究领域的难点问题之一。以全断面隧道掘进机的掘进总推力和总扭矩为研究对象,提出一种量纲分析与工程数据挖掘相结合的参数建模方法。该方法从各影响参量的物理力学本质出发,分析其中需要满足的量纲制约关系,构建具有一定可解释性和泛化性的显式模型框架,在其约束下进行工程数据挖掘建立定量预估模型。对模型在不同工况、不同工程中的预测效果进行评估,结果表明,提出的方法可以实现2类典型工况下的总推力和总扭矩建模,其计算结果在不同工程中均具有较好的适用性与预测准确度,模型可显示反映参量间的非线性影响关系。定量模型可为掘进机施工中的参数优化调控提供参考。同时,这种量纲分析与工程数据挖掘相结合的建模方法为工程装备的多参量数据挖掘建模提供了一种新思路。

In recent years, increasing number of engineering equipmenthas realized the detection and collection of multi-dimensional data in the construction process. It has become a development trend to realize the optimal control of working parameters through data mining. How to combine the process of data mining with the intrinsicmechanism of the studied problemsto establish parametric predictionmodels with certain physical interpretabilityand improve its generalization is one of the difficult problems in this field. The total thrust and torque of afull-sectiontunnel boring machine(TBM) were chosen as the research object, and a parametric modeling method that combinesdimensional analysis and engineering data mining wasproposed. Starting from the physical-mechanical nature of each influencing parameter, an explicitmodel frameworkwith certain interpretability and generalization was constructed based on the dimensional analysis, and quantitative prediction models wereestablished by engineering data mining under the constraints of this framework. The prediction effects of the model were evaluated under different working conditions for different projects. The results show that the proposed method can realize the modeling oftotal thrustand torque undertwo typical working conditions.The calculated results have good applicability and prediction accuracy among different projects, and the models can explicitly reflect the nonlinear relationsamongparameters. The findingscould provide a reference for optimalcontrol of parametersin tunneling. The modeling method combining dimensional analysis and engineering data mining provides a new perspective for multi-parameter data mining modeling of engineering equipment.