摘要
风机叶片表面在出现损伤时,叶片在空气中转动过程中,产生的气动音频信号有异于叶片无损时的气动音频信号,其信号特征参数与叶片的损伤类型存在一定的关系,但风机叶片在空气中旋转产生的气动音频信号往往是非线性非稳态的,对于这种信号的特征提取存在一定困难,实验证明,正常叶片气动音频信号与故障叶片的气动音频信号的频带存在显著差别,基于此,提出一种基于ITD固有时间尺度分解的时频分析方法,先将信号自适应的按频率从高到低分解为若干PRC旋转分量,计算每一频带内的时域信号的能量,构造初始特征向量,再使用PCA对特征向量进行降维,简化计算,提高分类器识别效率,最后将降维简化后的样本特征向量输入到支持向量机进行训练识别,并从特征提取时间和识别率上与传统EMD分解作比较,结果表明,该算法降低了特征提取的计算量,减少了特征提取时间,且有更高的识别率。
In the event of damage on the surface of the fan blade,the pneumatic audio signal generated during the rotation of the blade in the air is different from the pneumatic audio signal when the blade is not damaged.The signal characteristic parameters have a certain relationship with the damage type of the blade,but the pneumatic audio signal generated by the rotation of the blade in the air is often nonlinear and non-steady.There is a certain difficulty in the feature extraction of this signal.Experiments show that there is a significant difference between the frequency band of the pneumatic audio signal of the normal blade and the pneumatic audio signal of the faulty blade.Based on this,a time-frequency analysis method based on the inherent time-scale decomposition is proposed.Firstly,the signal is adaptively decomposed from high to low into several PRC rotation components,and the energy of the time domain signal in each frequency band is calculated.Construct the initial feature vector,then use PCA to reduce the dimension of the feature vector,simplify the calculation,improve the recognition efficiency of the classifier,and finally input the reduced-dimensional sample feature vector into the support vector machine for training and recognition.Compared with the traditional EMD decomposition from feature extraction time and recognition rate,the results show that the algorithm reduces the computational complexity of feature extraction,reduces feature extraction time,and has a higher recognition rate.
作者
刘登
崔宏维
姚恩涛
Liu Deng;Cui Hongwei;Yao Entao(College of Automation Engineering,Nanjing University of Aeronautics&Astronautics,Nanjing 210016,China)
出处
《电子测量技术》
2019年第23期68-73,共6页
Electronic Measurement Technology
