基于极限梯度提升树集成学习的绞吸挖泥船泥浆浓度预测

Mud concentration prediction of cutter suction dredger based on integrated learning of limit gradient boosting tree

针对绞吸挖泥船泥浆浓度γ测量仪具有辐射性,发生故障时现场不能检修,容易造成施工中断。文章提出了一种基于极限梯度提升树(XGBoost)集成学习的绞吸挖泥船泥浆浓度预测方法,研究表明XGBoost浓度预测模型的拟合优度为0.9532,均方根误差仅为1.423,预测效果较好。在挖泥船施工时即可利用XGBoost学习挖泥船施工数据,进而建立挖泥船泥浆浓度预测模型,实现泥浆浓度的实时预测,在γ浓度仪发生故障时可有效代替其工作,提高了挖泥船施工的连续性。

Theγmeasuring instrument for the mud concentration of the cutter suction dredger is radioactive,and cannot be repaired on site when it fails,which may easily cause construction interruption.This paper proposes a slurry concentration prediction method for cutter suction dredgers based on Exterme Gradient Boosting(XGBoost)ensemble learning.The study has been showed that the XGBoost concentration prediction model has a goodness of fit of 0.9532 and a root mean square error of only 1.423.The prediction effect is good.XGBoost can be used to learn the historical construction data of the dredger during the construction of the dredger,and then the mud concentration prediction model of the dredger to be established to realize the real-time prediction of the mud concentration.When the gamma concentration meter fails,its work can effectively replaced and continuity of dredger construction greatly improved.