为有效诊断舰船用柴油机机械微磨损故障，设计新型船用柴油机机械微磨损故障诊断系统。此系统传感模块通过电感式传感器采集柴油机机械表面油液磨粒的静电信号后，由油液分析芯片启动基于MACNN的微磨损故障类型识别模型，使用多层卷积神经网络，全面提取静电信号的多维波动特征图后，以重组的方式将多维特征图转换为一维向量，通过序列注意力机制，学习重组后油液磨粒静电信号一维向量的序列特征，识别特征所属微磨损故障类型，完成柴油机机械微磨损故障诊断。若机械磨损严重或表面存在污秽，便会驱动图像分析模块，进行图像采集配合诊断。经测试，此系统对多种微磨损故障类型的诊断结果无错分情况，诊断结果有效。
In order to effectively diagnose the micro-wear fault of marine diesel engine, a new micro-wear fault diagnosis system for marine diesel engine was designed. The sensing module of the system uses inductive sensors to collect the static electricity signals of the oil abrasive particles on the diesel engine mechanical surface, and then the oil analysis chip starts the MACNN based micro-wear fault type recognition model. After comprehensively extracting the multi-dimensional fluctuation feature map of the static electricity signal, the multi-dimensional feature map is converted into a one-dimensional vector by restructuring, and the sequence attention mechanism is used, Learn the sequence features of the one-dimensional vector of the recombined oil abrasive static signal, identify the micro-wear fault type of the feature, and complete the diesel engine mechanical micro-wear fault diagnosis. If the mechanical wear is serious or the surface is dirty, the image analysis module will be driven for image acquisition and diagnosis. The test shows that the system has no wrong classification for the diagnosis results of various micro-wear fault types, and the diagnosis results are effective.
2023,45(8): 142-145 收稿日期：2022-10-23
DOI：10.3404/j.issn.1672-7649.2023.08.027
分类号：U664.121
基金项目：校本科技类重点课题（2022-KJZD-003）
作者简介：李媛媛(1988-),女,硕士,讲师,研究方向为机电一体化
