为适应舰船设备多源数据异构性与工况动态性，提出多源数据融合与机器学习的舰船设备故障预警方法。部署振动、温度、压力及转速传感器，构建分布式采集网络，考虑节点能耗实现三维空间传感器部署，采用嵌入式学习技术，通过线性判别变量调整多源数据权重，以融合多源数据，结合长短期记忆网络特征编码器与Softmax分类器，提取时序多源数据动态特征，实现舰船设备故障分类与预警。实验证明，该方法在线性判别变量取值为1时实现最优数据融合，在正常、恶劣海况及长时间运行工况下，主机、轴系、泵组设备融合数据质量分别达93.53%、95.26%、91.08%，预警时间裕度最高达52.18 s，特征分离度达91.44%，能够实现高精度、高可靠性故障预警，为船舶安全运行提供关键技术支撑。
To adapt to the heterogeneity of multi-source data and the dynamics of working conditions of ship equipment, a fault early warning method for ship equipment based on multi-source data fusion and machine learning is proposed. Deploy vibration, temperature, pressure and rotational speed sensors to build a distributed acquisition network. Consider the energy consumption of nodes to achieve the deployment of three-dimensional spatial sensors. Adopt embedded learning technology, adjust the weights of multi-source data through linear discriminant variables to integrate multi-source data, and combine long short-term memory network feature encoders and Softmax classifiers to extract the dynamic features of time series multi-source data. Realize the classification and early warning of faults in ship equipment. Experiments prove that this method achieves the optimal data fusion when the value of the linear discriminant variable is 1. Under normal, harsh sea conditions and long-term operating conditions, the fusion data quality of the main machine, shafting and pump set equipment reaches 93.53%, 95.26% and 91.08% respectively. The maximum warning time margin reaches 52.18 seconds, and the feature separation degree reaches 91.44%. It can achieve high-precision and high-reliability fault early warning, providing key technical support for the safe operation of ships.
2025,47(22): 180-184 收稿日期：2025-5-9
DOI：10.3404/j.issn.1672-7649.2025.22.027
分类号：U672；TP277
基金项目：陕西省教育厅科学研究计划重点项目资助（20JZ086）
作者简介：卫昆（1976 – ），男，博士，讲师，研究方向为信息系统、大数据挖掘与机器学习及智能决策
参考文献：
[1] 李维波, 高峰, 肖朋, 等. 基于WOA-RF算法的船舶柴发配电系统故障诊断[J]. 中国舰船研究, 2025, 20(2): 77-88
LI W B, GAO F, XIAO P, et al. Fault diagnosis of ship diesel power distribution system based on WOA-RF algorithm[J]. Chinese Journal of Ship Research, 2025, 20(2): 77-88
[2] 宋秦中. 基于多智能体技术的舰船电缆绝缘故障检测[J]. 船电技术, 2024, 44(7): 12-16
SONG Q Z. Insulation fault detection of marine cable based on multi-agent technology[J]. Marine Electric & Electronic Technology, 2024, 44(7): 12-16
[3] 李子睿, 崔晓龙, 王超, 等. 基于卷积自注意力网络的机械设备故障诊断方法[J]. 舰船电子工程, 2024, 44(1): 145-151
LI Z R, CUI X L, WANG C, et al. Convolutional self-attention network-based fault diagnosis method of mechanical equipment[J]. Ship Electronic Engineering, 2024, 44(1): 145-151
[4] 刘建男, 车驰东. 基于VQ-VAE的船用设备轴承故障诊断模型[J]. 船舶工程, 2025, 47(6): 53-62
LIU J N, CHE C D. Fault diagnosis model of marine equipment bearings based on VQ-VAE[J]. Ship Engineering, 2025, 47(6): 53-62
[5] 汪益兵, 韩志豪, 站翌婷. 面向智能船舶的设备故障预测与管理系统[J]. 船舶工程, 2025, 47(3): 93-98
WANG Y B, HAN Z H, ZHAN Y T. Equipment fault prediction and management system for intelligent ships[J]. Ship Engineering, 2025, 47(3): 93-98
[6] 姜景强, 施伟锋, 刘宇航, 等. 船舶发电机定子绕组匝间短路故障预警[J]. 电力系统及其自动化学报, 2024, 36(6): 145-151
JIANG J P, SHI W F, LIU Y H, et al. Warning of interturn short-circuit fault in stator winding of ship generator[J]. Proceedings of the CSU-EPSA, 2024, 36(6): 145-151
