摘要
为实现耙吸疏浚船疏浚产量的实时准确预测与作业控制策略优化,提出了一种新的数据驱动自适应深度学习框架。该框架通过实时评估误差指标,从候选模型库中动态选择性能最优的深度学习模型架构,以实现疏浚产量的高精度估计。构建了以航速、耙头角度和泥泵转速等多维度船舶状态及控制参数为输入,泥浆密度与流量为输出的深度学习输入-输出网络模型,应用该模型可准确预报实时疏浚装舱产量信息;进一步以疏浚产量最大化为目标,采用遗传算法对关键控制参数进行全局优化,并生成最优控制策略。研究结果表明,自适应深度学习框架选取LSTM神经网络模型预测泥浆密度和流速,预测精度较高,优化后的控制参数组合使疏浚产量提升4.42%。所提出的数据驱动耙吸疏浚船装舱产量实时预测及控制策略优化方法,可为操作人员提供准确的产量预测信息与可靠的参数优化决策支持,从而显著提升疏浚作业效率与经济效益。
To achieve real-time and accurate prediction of dredging production for trailing suction hopper dredgers(TSHD)and optimize operation control strategies,a novel data-driven adaptive deep learning framework is proposed.This framework dynamically selects the optimal-performing deep learning model architecture from a candidate model library by evaluating real-time error metrics,enabling high-precision estimation of dredging production.Specifically,an deep learning input-output network model is constructed with multidimensional key control parameters as inputs,such as vessel speed,drag head angle,and dredge pump rotation speed,and slurry density and flow rate as outputs.Based on real-time vessel status and control parameters,this model can accurately forecast real-time dredging production information.Furthermore,with the goal of maximizing dredging production,a genetic algorithm is employed to globally optimize key control parameters and generate optimal control strategies.The research results indicate that the adaptive deep learning framework,which employs an LSTM neural network model,achieves high accuracy in predicting mud density and flow velocity.The optimized combination of control parameters increases dredging production by 4.42%.The proposed data-driven method for real-time dredging production prediction and control strategy optimization provides operators with accurate production forecasts and reliable parameter optimization decision support,thereby significantly improving dredging operation efficiency and economic benefits.
作者
包昱皓
李夏
周滢
赵永生
李铭志
何炎平
BAO Yuhao;LI Xia;ZHOU Ying;ZHAO Yongsheng;LI Mingzhi;HE Yanping(State Key Laboratory of Ocean Engineering,Shanghai Jiao Tong University,Shanghai 200240,China;CHEC Dredging Co.Ltd.,Shanghai 200136,China)
出处
《中国造船》
北大核心
2025年第6期131-146,共16页
Shipbuilding of China
基金
国家自然科学基金项目(52271284)
上海交通大学深蓝计划重点项目(SL2021ZD201)
中央高校基本科研业务费专项资金。
关键词
自适应深度学习
产量预测
控制策略优化
耙吸疏浚船
adaptive deep learning
production prediction
control optimization
trailing suction hopper dredger(TSHD)
