提出基于神经网络的船舶通信网络异常数据识别方法，为去除冗余数据，将通信网络数据映射至统一数据区间，并切片为多模态数据模式。由联合关联规则，提取通信网络多模态数据特征，作为基于粒子群优化模糊神经网络的网络异常数据识别模型的输入样本，经粒子群算法调节模糊神经网络连接权重后，将模型网络模型训练为稳定状态，在此状态下通过Softmax分类器，结合数据特征隶属度与模糊规则，分析船舶通信网络数据特征是否异常，输出多类型异常数据识别结果。实验结果表明：该方法可有效去除船舶通信网络异常数据内存在的冗余数据，具备较好的预处理效果；提取船舶通信网络数据特征时的偏差数值极小，特征提取能力好；可有效识别不同类型的船舶通信网络数据内的异常数据，且识别精度较高，应用性较强。
In order to remove redundant data, the communication network data is mapped to a unified data interval and sliced into multi-modal data patterns. By the combined association rules to extract modal data communication network more features, as a fuzzy neural network based on particle swarm optimization of network anomaly data identification model of input samples, the particle swarm algorithm to adjust the fuzzy neural network connection weights, after the model training network model for steady state, in the state by Softmax classifier, combined with the membership degree of data features and fuzzy rules, this paper analyzes whether the characteristics of ship communication network data are abnormal, and outputs the recognition results of multiple types of abnormal data. The experimental results show that the method can effectively remove the redundant data in the abnormal data of ship communication network, and has a good preprocessing effect. The deviation value of extracting ship communication network data feature is very small, and the feature extraction ability is good. It can effectively identify abnormal data in different types of ship communication network data, and the recognition accuracy is high, and the application is strong.
2022,44(17): 148-151 收稿日期：2022-04-28
DOI：10.3404/j.issn.1672-7649.2022.17.030
分类号：TN913
作者简介：田银磊(1972-)，男，硕士，副教授，主要从事物联网技术研究
参考文献：
[1] 何凯彦, 李维波, 许智豪, 等. 基于双CPU的双以太网与双RS-422交互通信技术[J]. 中国舰船研究, 2020, 15(3): 177–184
[2] 孙万峰, 李维波, 李齐, 等. 基于DMA与中断方式的组态监控系统容错通信技术[J]. 中国舰船研究, 2021, 16(4): 199–207
[3] 周毅. 智能船舶网络风暴测试及抑制技术[J]. 船海工程, 2021, 50(3): 41–44+48
[4] 王鹏宇, 程郁凡, 徐昊, 等. 基于卷积神经网络联合多域特征提取的干扰识别算法[J]. 信号处理, 2022, 38(5): 915–925
[5] 赵英, 王丽宝, 陈骏君, 等. 基于联邦学习的网络异常检测[J]. 北京化工大学学报(自然科学版), 2021, 48(2): 92–99
[6] 鞠东豪, 李宇, 王宇杰, 等. 基于堆栈稀疏去噪自编码器神经网络的舰船辐射噪声目标识别算法研究[J]. 振动与冲击, 2021, 40(24): 50–56+74
[7] 徐千驰, 王彪. 基于小波包分析和深度学习的舰船辐射噪声识别[J]. 船舶工程, 2021, 43(5): 29–34+43
