盾构渣土作为护壁泥浆原材料配比预测研究

Research on the Prediction of Raw Material Ratio of Shield Muck as Wall Protection Mud

针对渣土利用为护壁泥浆使用最大化配比设计问题,依托苏州地铁8号线盾构工程,选取盾构穿越粉质黏土地层的渣土、膨润土和水作为护壁泥浆的原材料,通过室内试验测试配制的护壁泥浆性能。将测试结果作为数据集,对比分析了多种机器学习算法,比选了盾构渣土作为护壁泥浆原材料配比预测方法。结合现有护壁泥浆使用要求,给出了渣土最大化再利用配比。研究结果表明:①AdaBoost集成的决策树算法相比于决策树、随机森林算法,准确率更高,拟合优度均大于0.8,可作为配比设计的预测方法;②水对比重影响最大,渣土对含砂率影响最大,膨润土对黏度、失水量与泥饼厚度影响最大。在盾构渣土再利用领域,为采用机器学习作为护壁泥浆原材料配比设计的方法进行了有益的探索与尝试。

To address the issue of maximizing the utilization ratio of muck utilization for wall protection mud,this study is based on the shield project of Suzhou Metro Line 8.Muck of the shield through the silty clay layer,bentonite,and water were selected as the raw materials for wall protection mud,and the performance of the prepared wall protection mud was tested through laboratory tests.Using these test results as a dataset,a variety of machine learning algorithms were compared and analyzed,and the prediction method for the ratio of raw materials for wall protection mud was selected.Considering the requirements for existing wall protection slurry,an optimal muck reuse ratio was proposed.The findings indicate that:①The AdaBoost ensemble decision tree algorithm outperforms decision tree and random forest algorithms in terms of accuracy and has a goodness-of-fit greater than 0.8,making it a suitable predictive method for proportional design.②Water has the most significant effect on specific gravity,muck has the greatest influence on the sand content,and bentonite has the greatest influence on viscosity,water loss,and mud cake thickness.This research represents a valuable exploration into the use of machine learning for designing the raw material ratios in wall protection mud for shield muck reuse.