为解决传统的船体结构损伤识别效率低下、准确性不高等问题，提出一种适用于船体结构的智能诊断识别与预测模型。首先对采集到的振动冲击信号进行预处理并在时域维度进行特征分析和提取。然后利用大量样本数据，采用改进的卷积神经网络（CNN）模型对船体结构损伤类型和严重程度进行分类识别，采用ARIMA时间序列模型对船体结构损伤时间进行预测，进而实现船体结构健康状态的及时预警和提前预知，为船体结构预防性维修提供参考。通过实际案例应用表明，提出的基于改进卷积神经网络和ARIMA模型的船体结构损伤识别和预测方法，可以充分利用船体结构健康监测大数据，有效提高故障的识别效率和准确率，为船舶装备的智能运维和健康管理提供可行的途径。
To address the issues of low efficiency and accuracy in traditional ship structural damage identification, an intelligent diagnostic recognition and prediction model suitable for ship structures is proposed. Initially, preprocess the collected vibration impact signals and conduct feature analysis and extraction in the time domain. Subsequently, utilizing a large dataset, an improved convolutional neural network(CNN)model is employed for the classification and identification of ship structural damage types and severity. An ARIMA time series model is applied to predict the time of ship structural damage. This approach enables timely warning and advance knowledge of the health status of ship structures, providing a reference for proactive maintenance. Practical case applications demonstrate that the proposed ship structural damage identification and prediction method based on the improved CNN and ARIMA models can effectively utilize big data from ship structural health monitoring, significantly enhancing the efficiency and accuracy of fault identification. It offers a viable approach for intelligent operation and health management of maritime equipment.
2024,46(19): 170-175 收稿日期：2023-12-8
DOI：10.3404/j.issn.1672-7649.2024.19.031
分类号：TH17
基金项目：国家重点研发计划资助项目（2018YFF0214705）
作者简介：宋庭新(1972-),男,博士,教授,研究方向为装备智能运维与健康管理等
参考文献：
[1] 阮旻智. 舰船装备维修决策建模与优化技术[M]. 北京: 科学出版社, 2018.
[2] YUAN M, WU Y T, LIN L. Fault diagnosis and remaining useful life estimation of aero engine using LSTM neural network[C]//IEEE International Conference on Aircraft Utility Systems, 2016: 135-140.
[3] ZHANG C, LIM P, QIN A K, et al. Multiobjective deep belief networks ensemble for remaining useful life estimation in prognostics[J]. IEEE Transactions on Neural Networks and Learning Systems, 2016, 28(10): 2306-2318.
[4] 贾成芬, 唐小洁, 张攀. 基于HSMM的船舶机械故障演化预测模型[J]. 舰船科学技术, 2022, 44(22): 174-177.
JIA Chengfen, TANG Xiaojie, ZHANG Pan. Fault evolution prediction model for marine machinery based on HSMM[J]. Ship Science and Technology, 2022, 44(22): 174-177.
[5] 孙华峰. 基于小波神经网络的船舶动力系统故障预测方法研究[J]. 舰船科学技术, 2018, 40(6): 67-69.
SUN Huafeng. Ship power system fault prediction based on wavelet neural network[J]. Ship Science and Technology, 2018, 40(6): 67-69.
[6] 陈方圆, 邹永久, 张鹏, 等. 基于ARIMA-KF模型的船舶系统设备状态参数预测[J]. 科学技术与工程, 2021, 21(35): 15255-15261.
CHEN Fangyuan, ZOU Yongjiu, ZHANG Peng, et al. Prediction of equipment state parameters of ship system based on ARIMA-KF model[J]. Science and Technology and Engineering, 2021, 21(35): 15255-15261.
[7] 闫洪胜, 左洪福, 孙见忠. 考虑预测与健康管理的民机维修成本效益仿真评估方法[J]. 南京航空航天大学学报, 2020, 52(6): 912-922.
YAN Hongsheng, ZUO Hongfu, SUN Jianzhong. Cost-Benefit evaluation for aircraft maintenance based on prognostic and health management [J] Journal of Nanjing University of Aeronautics and Astronautics, 2020, 52(6): 912-922.
[8] 沈保明, 陈保家, 赵春华, 等. 深度学习在机械设备故障预测与健康管理中的研究综述[J]. 机床与液压, 2021, 49(19): 162-171.
SHEN Baoming, CHEN Baojia, ZHAO Chunhua, et al. Research review of deep learning in mechanical equipment fault predictionand health management[J]. Machine Tool & Hydraulics, 2021, 49(19): 162-171.
[9] 徐翠锋, 王绍, 景亚鹏, 等. 基于振动信号降噪的钢板件损伤定位[J]. 科学技术与工程, 2021, 21(22): 9440-9446.
XU Cuifeng, WANG Shao, JING Yapeng, et al. Damage localization of steel plate based on vibration signal denoising [J] Science, Technology and Engineering, 2021, 21 (22): 9440-9446.
[10] 蒋佳炜, 胡以怀, 丁梦豪, 等. 基于相对混乱度特征选择的船舶风机智能状态监测与故障诊断[J]. 上海海事大学学报, 2023, 44(1): 88-94.
JIANG Jiawei, HU Yihuai, DING Menghao, et al. Intelligent condition monitoring and fault diagnosis of marine fans based on feature selection of relative chaos[J]. Journal of Shanghai Maritime University, 2023, 44(1): 88-94.
[11] 肖俊青, 金江涛, 李春, 等. 基于深度学习风力机齿轮箱的故障诊断[J]. 太阳能学报, 2023, 44(5): 302-309.
XIAO Junqing, JIN Jiangtao, LI Chun, et al. Fault diagnosis of wind turbine gearbox based on deep learning[J]. Journal of Solar Energy, 2023, 44(5): 302-309.
[12] GLOCK A C. Explaining a random forest with the difference of two ARIMA models in an industrial fault detection scenario[J]. Procedia Computer Science, 2021, 180(1): 476-481.
[13] 谢旭阳, 余刃, 王天舒, 等. 基于卷积神经网络和迁移学习的电动泵故障诊断方法研究[J]. 四川兵工学报, 2021, 42(7): 239-245.
XIE Xuyang, YU Ren, WANG Tianshu, et al. Fault diagnosis method for electric pump based on convolutional neural network and transfer learning[J]. Sichuan Ordnance Engineering Journal, 2021, 42(7): 239-245.
