摘要
水电机组在运行过程中会出现不同品质的声音,其部件也都具有自身特定的声纹,声纹中含有反映设备健康状态的丰富信息。通过在灯泡贯流式水电机组灯泡头、灯泡体、转轮室等位置布置声频传感器,采用基于时域特征峭度的卷积窗-短时傅里叶变换技术对采集的音频信号进行特征提取,能够准确识别出不同工况下机组的声纹,从而对设备进行异常报警和故障诊断。
Different quality sounds will appear during the operation of hydropower units,and their components also have own specific voiceprint,which contain rich information reflecting the health status of equipment.Acoustic sensors are arranged at the bulb head,bulb body and runner chamber of bulb tubular hydropower units.The convolution window short-time Fourier transform technology based on time-domain feature kurtosis is used to extract the features of the collected sounds,which can accurately identify the voiceprint of the unit under different working conditions,and can carry out abnormal alarm and fault diagnosis for the equipment.
作者
陈茗
胡边
李靖
CHEN Ming;HU Bian;LI Jing(Hunan Polytechnic of Water Resources and Electric Power,Changsha 410131,China;Wuling Power Co.,Ltd.,Changsha 410004,China;Hydropower Industry Innovation Center of State Power Investment Co.,Ltd.,Changsha 410004,China;Engineering Technology Research Center of Hydropower Intelligent of Hunan Province,Changsha 410004,China)
出处
《电工技术》
2022年第18期28-30,共3页
Electric Engineering
基金
国家自然科学基金(编号51775185)
湖南省高新技术产业科技创新引领计划(编号2020GK2094)
国家电力投资集团统筹科研项目(编号TC2020SD01)。
关键词
水电机组
声纹识别
时域峭度
特征提取
故障诊断
hydropower units
voiceprint recognition
time domain kurtosis
feature extraction
fault diagnosis
