针对风电机组叶片排水孔堵塞或被雷击穿孔等问题,提出一种非接触式的声学检测方法.该方法首先对采集到的信号转化为时频图,利用中值滤波和自适应阈值的方法将时频图二值化,根据二值化时频图中哨音轮廓特点,提取轮廓信号时域和频域等 9 个参数作为特征向量,提出了动态半径的支持向量数据描述异常检测模型(dynamic radius support vector data description,DR-SVDD).将DR-SVDD和SVDD的异常检测模型用于风机叶片哨声诊断,验证了该方法的有效性.
Based on the Detection Method of Wind Turbine Blade Whistle Abnormal Sound
A non-contact acoustic detection method was proposed to solve the problem of wind turbine blade drainage hole being blocked or perforated by lightning.In this method,the collected signals are first transformed into time-frequency graphs,and the time-frequency graphs are binarized by means of median filtering and adaptive threshold.According to the whistle contour characteristics in the binarized time-frequency graphs,9 parameters of the contour signal in time domain and frequency domain are extracted as feature vectors.A dynamic radius support vector data description(DR-SVDD)anomaly detection model is proposed.The DR-SVDD and SVDD anomaly detection models are applied to the whistle diagnosis of fan blades,and the effectiveness of this method is verified.
wind turbineacoustic detectiondrainage hole pluggingsupport vector data description
NETL
NSTL
万方数据
