基于改进Mallat算法及包络线的气阀故障诊断

The fault diagnosis of the air valve on the basis of improved Mallat algorithm and envelope curve

气阀故障微弱信号特征不明显,常规方法难以通过计算结果来判断气阀状态,且传统Mallat算法存在频率混叠及边界震荡现象,难以提取出气阀故障特征。为此,提出一种结合改进Mallat分解算法及包络线的故障特征提取方法。改进传统的边界处理方式,对信号右边界进行数据延拓,用二阶Volterra模型对延拓信号进行预测,用递推最小二乘法求取预测系数,用非抽样算法对信号进行小波包分解。对分解得到的各频带信号进行奇异值降噪处理,通过奇异熵增量曲线选择降噪阶次,画出信号的上、下包络线,用于提取气阀故障特征。对仿真信号及工程信号的处理表明,用改进的Mallat算法对信号进行分解,消除了边界震荡及频率混叠现象,成功提取了阀片破损的微弱故障特征,取得了良好的效果。

In light of the difficulty of identifying the weak signal characteristic of the air valve failure,a method to extract the fault characteristic on the basis of improved Mallat decomposition algorithm and envelope curve was formulated.The traditional method to deal with boundary was improved.The extension of data on the right boundary of signals was conducted.The second-order Volterra model was adopted to predict the extended signal.The recursive least square method was applied to derive the prediction coefficient.The non-sampling algorithm was used to carry out the wavelet packet decomposition of the signals.A singular-value noise reduction of signals of different frequency bands obtained by decomposition was conducted.The singular-entropy incremental rate curve was adopted to select the denoise order and draw the upper and lower envelope curves which were used to extract the valve fault characteristic.The treatment of the simulation signals and engineering signals showed that decomposition of signals through the improved Mallat algorithm eliminated the boundary oscillation and frequency aliasing,the weak fault characteristic of the broken valve plate was successfully extracted and a desirable effect was reached.