噪声环境下基于小波熵的声发射识别

Recognition of acoustic emission based on wavelet entropy in noisy condition

针对声发射技术在旋转机械故障检测中的强噪声干扰问题,提出了一种基于小波熵的声发射检测算法.该算法首先给定一个合理的阈值,对声发射信号进行小波分解.然后进行分帧处理,使信号在较短的时间间隔内保持特性基本不变,从而求出每一帧信号的小波熵.通过比较每一帧信号的小波熵值与阈值的大小,判断该信号为声发射帧还是噪声帧.为了检验算法的检测效果,在转子实验台上获得碰摩声发射信号,并在测试数据上叠加不同信噪比的高斯白噪声和非平稳噪声,进行声发射识别.实验结果表明:该算法具有较高的识别正确率;在低信噪比环境下,通过调整阈值的可调参数可以有效提高识别的正确率.

In acoustic emission（AE） technique,to avoid the serious noise disturbance in the fault diagnosis of rotary machine,a recognition algorithm based on wavelet entropy（WE） is proposed.First,through setting up an appropriate threshold,an AE signal is decomposed by wavelet transform.Secondly,the AE signal is divided into some equal frames to keep the characteristic approximately constant in a short time interval,and the WE of each frame can be calculated.Thirdly,through comparing the value of each WE with the threshold,it can be determined whether the frame belongs to AE frames or noise frames.To test the recognition efficiency of this algorithm,a rub impact AE signal obtained from a rotating test stand is added with white noise and non-stationary noise at various signal-to-noise ratios（SNRs）,followed by AE recognition.The experimental results indicate that this algorithm has a high recognition efficiency.In a low SNR environment,the recognition efficiency can be improved by adjusting the parameter of the threshold.