摘要
在盾构机实际施工过程中,对于刀盘扭矩的预估是盾构装备设计与施工调整的重要依据。地铁盾构施工区间往往穿越多个地层并历经复杂的地质环境,盾构机每掘进1环将会产生海量的地质参数与机械参数数据,对于庞大的数据体系,传统的统计分析方法已经无法满足数据量的需求。文章提出用机器学习算法结合数据预处理方法来对复合地层下的刀盘扭矩进行建模预测。鉴于盾构数据具有时间属性的特点,文章采用PSO-SVR与LSTM时序算法进行对比分析。首先采用数据预处理的方法进行数据清洗和特征筛选,之后将样本数据用于模型建立,最后对模型的适用性进行对比分析。分析结果表明,LSTM时序分析算法在盾构掘进参数预测方面具有更好的表现效果,平均误差率仅为5%左右,可以满足工程层面的需求。
In the actual construction process of shield machines,the estimation of cutter head torque is an important basis for the design and construction adjustment of shield equipment.Shield construction section often passes through multiple strata and goes through complex geological environment.Each link of shield tunneling will produce a large number of geological and mechanical parameters.For a big data system,traditional statistical analysis cannot meet the demand of the data volumes involved.In this paper,machine learning algorithms combined with data preprocessing is proposed to model and predict cutter head torque in composite formation.In view of the time attribute of shield data,this study used PSO-SVR and LSTM algorithm for comparative analysis.Firstly,the data preprocessing method was used for data cleaning and feature selection,then the sample data was used for model establishment,and finally the applicability of the model was compared and analyzed.The analysis shows that LSTM time series analysis algorithm has better performance in the prediction of shield tunneling parameters,and the average error percentage is only about 5%,which can meet the needs of engineering level.
出处
《现代城市轨道交通》
2022年第2期46-51,共6页
Modern Urban Transit
基金
国家自然科学基金项目(51978322)。
关键词
地铁
盾构隧道
刀盘扭矩
大数据分析
粒子群优化
LSTM
metro
shield tunnel
cutter head torque
big data analytics
particle swarm optimization
LSTM
