摘要
对于大部分机械设备或者结构来说,在线结构健康监测对于其能否正常运行非常关键。由于结构本身因素和受到环境影响,在结构运转过程中所采集到的信号含有大量噪声,这可能会对特征提取造成很大影响,从而影响最终结果,如果进行传统的结构在线健康监测,需要经过一系列降噪处理才能得到正确监测结果。同时对于复杂结构,其故障信号十分复杂,较难实现自动判别故障,一般需要专家帮助分析。使用采集到的磨机中主减速机的振动信号在所建立的一维卷积神经网络模型中进行测试,发现无需进行降噪处理和依赖专家评估,由模型自动获得的测试结果与实际结果完全一致,由此初步证明在结构健康监测中引入卷积神经网络是一个有效的方法。
For most mechanical devices and structures, on-line structural health monitoring (SHM) is critical to their normal operation. In traditional on-line structural health monitoring process, the structure itself and the environment influence will lead to a large amount of noises in the collected signals, which may have a great impact on the final results of feature extraction. Therefore, a series of noise reduction processing is necessary before the correct monitoring results can be obtained. At the same time, for complex structures, the fault signal is also very complex, and it is difficult to realize automatic fault identification. Generally, experts are needed to help analyzing the fault. In this study, the vibration signal of the main reducer in a grinding machine is collected and tested in the model established by 1D convolutional neural network (CNN). The test results are completely consistent with the actual results without the requirement of denoising and expert’s accessing. It can be preliminarily proved that the CNN is an effective method for SHM.
作者
吴磊
纪国宜
WU Lei;JI Guoyi(State Key Laboratory of Mechanics and Control of Mechanical Structures,Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China)
出处
《噪声与振动控制》
CSCD
2019年第4期200-204,共5页
Noise and Vibration Control
关键词
振动与波
在线结构健康监测
一维卷积神经网络
强噪声信号
特征提取
vibration and wave
on-line structural health monitoring (SHM)
1D convolutional neural network (CNN)
strong noise signal
feature extraction
