基于深度SSD改进模型的传动设备状态在线监测研究  

Research on On-line Monitoring of Transmission Equipment Status Based on Improved Deep SSD Model

作　　者：王宜 周大可[2] WANG Yixian;ZHOU Dake(Wuhu Machinery Factory,Wuhu 241000,China;School of Automation Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211100,China)

机构地区：[1]国营芜湖机械厂,安徽芜湖241000 [2]南京航空航天大学自动化学院,南京211100

出　　处：《计算机测量与控制》2024年第3期99-105,共7页Computer Measurement &Control

摘　　要：针对现有传动设备在线监测算法存在的检测精度低、效率差等问题,提出一种基于改进SSD网络模型的在线检测算法;先对故障集进行预处理,通过滤波调制、共振解调等环节滤除原始故障集的噪声干扰;以VGG-16为基础设计了SSD网络结构,同时增加了辅助卷积层和预测层;对SSD网络模型进行改进,引入了注意力机制模块和特征增强模块,改善模型各层的数据共享性能同时提高了模型的数据训练效率;基于通道拼合方式对故障数据进行多尺度特征融合,并优化SSD模型的各层金字塔结构,以更好的匹配先验框及选择最佳的损失函数;实验结果显示,提出算法的传动设备故障检测率达到98.8%,参与对比的3种传统算法故障检测率分别为94.2%、93.6%和93.7%,同时提出算法的检测效率也优于传统算法。Aiming at the problems of low detection accuracy and low efficiency of existing online monitoring algorithms for transmission equipment,an online detection algorithm based on improved single shot detector(SSD)network model is proposed.Firstly,the fault set is preprocessed,and the noise interference of the original fault set is filtered by filtering modulation and resonance demodulation.The SSD network structure is designed based on VGG-16,and the auxiliary convolution layer and prediction layer are added.To improve the SSD network model,the attention mechanism module and feature enhancement module are introduced to improve the each layer data sharing performance of the model and improve the data training efficiency of the model.The multiscale feature fusion of fault data is carried out based on the channel fusion method,and the pyramid structure of each layer of the SSD model is optimized to better match the prior frame and select the optimal loss function.The experimental results show that the transmission equipment fault detection rate of the proposed algorithm is 98.8%,and the fault detection rates of three traditional algorithms are 94.2%,93.6%and 93.7%,respectively.Meanwhile,the detection efficiency of the proposed algorithm is better than that of the traditional algorithm.

关 键 词：深度SSD 传动设备 在线监测 辅助卷积 数据训练 先验框 损失函数 

分 类 号：TM11[电气工程—电工理论与新技术]