摘要
泵机组故障诊断的难点在于信号特征向量的提取,而故障特征往往淹没在复杂的噪音中。本文利用自适应小波函数对采集到的振动信号进行降噪,滤掉了无关的噪声信息,根据振动能量的分布,对降噪过的信号进行四层小波包分解,提取出的特征向量分布明显。最后将分类特征向量输入神经网络进行训练,测试的结果证明,该方法识别精度高、速度快,具有良好的应用前景。
The adaptive wavelet function was used to denoise the collected vibrate signal and filter those unrelated noises,then based on the signal energy distribution,the denoised signal was decomposed to four layers with wavelet package and made the characteristic vector distribution more obvious.The classified characteristic vector was brought to BP neural network for training.The testing result proves this method＇s high accuracy in recognition and speed,and good application prospect.
出处
《化工自动化及仪表》
CAS
北大核心
2010年第4期36-38,共3页
Control and Instruments in Chemical Industry
基金
解放军后勤工程学院博士生创新基金
重庆市自然科学基金资助项目(CSTC
2008BB7142)
