基于小波包和GBDT的瓦斯传感器故障诊断

针对瓦斯传感器的常见故障类型,提出了一种基于小波包和GBDT的瓦斯传感器故障诊断方法.该方法首先使用3层小波包分解对瓦斯原始故障信号进行分解;然后利用LDB算法削减得到重构信号能量,经归一化处理后作为输入分类器的特征向量;接着利用由梯度提升技巧和决策树构成的GBDT分类器作为故障模式的训练和识别器;最后通过瓦斯传感器诊断实例验证了该方法的有效性.实验结果表明,采用该方法进行瓦斯传感器故障诊断相比其他方法具有更高的诊断精度和更好的样本泛化能力.

Fault Diagnosis for Gas Sensor Based on Wavelet Package and GBDT

Aiming at the common fault types of gas sensor,a fault diagnosis method of gas sensor based on wavelet package and GBDT(gradient boost decision tree) is proposed.Firstly,the original fault output signals are decomposed by three-level wavelet package decomposion.Then local discriminant base (LDB) is used to cut the decomposed signals to obtain the reconstructing signal energy.Secondly,the normalizing reconstructing signal energy is inputted to train the GBTD classifier which consists of gradient boosting machine and decision tree.Finally,the GBDT classifier is used to diagnose different fault modes of gas sensor.Also,the effectiveness of the proposed method is demonstrated by a simulated fault diagnosis example of gas sensor.The experimental results show that compared with other methods,the proposed method has a higher diagnosis accuracy and a better smaple generalization ability in gas sensor fault diagnosis.