噪声与振动控制2025，Vol.45Issue(3)：132-137,150,7.DOI:10.3969/j.issn.1006-1355.2025.03.021
基于MA-TCN的导管架式海洋平台故障诊断
Fault Diagnosis of Jacket Offshore Platform Based on Multi-head Attention TCN
摘要
Abstract
Aiming at the imperfection that the conventional neural networks used in noise detection in marine environ-ment insufficiently utilize time series data,a fault diagnosis method for jacket offshore platforms based on Multi-head Atten-tion Mechanism Time Convolutional Networks(MA-TCN)was proposed.Firstly,the original vibration signal was input to the model,and the time series features were extracted by using the time convolution network.The problem of gradient disap-pearance during network training was alleviated by residual structure and dilated convolution.Then,the multi-head attention mechanism was utilized to re-weight network features and emphasize the useful feature information for fault diagnosis.Fi-nally,the model features were fused and output to realize the fault diagnosis of the jacket offshore platform.Combined with the 11 working states simulated by the sea test,the feasibility of the proposed model was verified.The results of the model were also compared with those of the models mentioned in other literature.The results indicate that the fault diagnosis rate of the proposed model is above 95%,which is better than the other models.关键词
故障诊断/导管架平台/深度学习/时间卷积网络/多头注意力Key words
fault diagnosis/jacket platform/deep learning/time convolutional networks/multi-head attention分类
信息技术与安全科学引用本文复制引用
吴文凯,高军伟,车鲁阳,段琳,官晟..基于MA-TCN的导管架式海洋平台故障诊断[J].噪声与振动控制,2025,45(3):132-137,150,7.基金项目
山东省自然科学基金资助项目(ZR2019MF063) （ZR2019MF063）
