基于LSTM的绞吸挖泥船泥浆浓度动态软测量

Dynamic Soft Sensing of Slurry Concentration of Cutter Suction Dredger Based on LSTM

受恶劣的施工环境的影响,绞吸挖泥船在疏浚过程中应用的泥浆浓度计易发生故障。为保证绞吸挖泥船在泥浆浓度计损坏时能继续施工作业,提出一种基于长短时记忆(LSTM)神经网络的泥浆浓度动态软测量方法。利用实测施工数据建立泥浆浓度软测量模型,并将所建LSTM模型与BP神经网络(BPNN)模型、KNN模型和循环神经网络(RNN)模型进行对比,验证基于LSTM神经网络的绞吸挖泥船泥浆浓度软测量模型的预测效果。试验结果表明,该方法能实现对绞吸挖泥船泥浆浓度的实时准确预测,预测精度高且稳定性好,具有良好的工程实用价值。

Affected by the harsh construction environment,the slurry concentration meter of the cutter suction dredger is prone to malfunction during the dredging process.In order to ensure that the cutter suction dredger can continue the construction operation when the slurry concentration meter is damaged,a dynamic soft sensing method of slurry concentration based on long short-term memory neural network(LSTM)is proposed.The soft sensing model of slurry concentration is established by using the measured construction data,and the LSTM model is compared with BPNN model,KNN model and RNN model.Experimental results show that this method can realize real-time and accurate prediction of the slurry concentration of cutter suction dredgers,with high prediction accuracy and good stability,and has good engineering practical value.