基于TREE-LSTM算法的船舶汽轮机组变负荷故障诊断  

Research on fault diagnosis of marine steam turbine units during variable load process based on TREE-LSTM algorithm

作　　者：王灏桐 李彦军[1] 杨龙滨[1] 史建新[1] WANG Haotong;LI Yanjun;YANG Longbin;SHI Jianxin(College of Power and Energy Engineering,Harbin Engineering University,Harbin 150001,China)

机构地区：[1]哈尔滨工程大学动力与能源工程学院,黑龙江哈尔滨150001

出　　处：《舰船科学技术》2024年第17期110-115,共6页Ship Science and Technology

摘　　要：针对船舶汽轮机组变负荷过程故障诊断中的耦合参数时序特征难以捕捉以及正常参数变动的干扰等问题,引入TREE-LSTM神经网络模型以实现复杂非线性系统动态数据分类。首先建立某船舶汽轮机组仿真模型,分析并进行故障仿真;随后进行数据预处理与特征工程;最后训练TREE-LSTM模型进行故障诊断,并与SVM、LSTM等模型进行比较。TREE-LSTM模型对于船舶汽轮机组变负荷过程的故障诊断正确率为98.7%,正确率最高。由于引入时间序列与复杂神经网络拓扑结构,TREE-LSTM在处理非线性系统动态数据分类问题时效果更好。In response to the difficulties in capturing the coupling parameter time series characteristics and the interference of normal parameter changes in fault diagnosis during the variable load process of marine steam turbine units,the TREE-LSTM neural network model is introduced to achieve dynamic data classification of complex nonlinear systems.Firstly,establish a simulation model for a certain marine steam turbine unit,analyze the fault mechanism,and conduct fault simulation;subsequently,perform data preprocessing and feature engineering;finally,a TREE-LSTM model was built for training and fault diagnosis,and compared with models such as SVM and LSTM.The TREE-LSTM model has a fault diagnosis accuracy of 98.7%for the variable load process of marine steam turbine units,with the highest accuracy.It is ultimately believed that due to the introduction of time series and complex neural network topology,TREE-LSTM performs better in dealing with dynamic data classification problems in nonlinear systems.

关 键 词：汽轮机组 动态仿真 故障诊断 树形长短时记忆网络 

分 类 号：TK269[动力工程及工程热物理—动力机械及工程]