基于小波包与Elman神经网络的气力提升装置流型识别技术研究

Flow pattern identification for airlift devices based on wavelet packet and Elman neural network

气力提升装置流型对气液流动特性及提升系统性能均有很大的影响,但由于气液两相交界面形态以及截面含气率动态变化、气液两相速度复杂难测等原因,致使提升管流型亦交替变化且不易识别。针对这一难题,提出了基于小波包分析与Elman神经网络的流型辨识策略:利用小波包分析方法提取提升装置压差信号各频带能量特征值,借助Elman神经网络辨识技术,以各频带能量为Elman网络输入变量,以流型为输出变量,通过对Elman神经网络进行大量数据训练。从而对提升系统流型进行辨识。实验结果表明,该方法对流型辨识精度达到了92.6%,比BP网络高6.5%,能有效对提升管流型进行辨识。

Airlift devices＆#39; flow pattern has a great influence on performances of gas-liquid flow and a lifting system. However,because of unpredictable gas-liquid speed and dynamic interface morphology,and dynamic varying of section gas ratio,the flow pattern of lift pipes alternately changes and it becomes too difficult to identify.Aiming at this problem, a flow pattern identification strategy was put forward based on wavelet packet and Elman neural network.The lift device pressure difference signal＆#39;s frequency band energy eigenvalues were extracted with the wavelet packet analysis.Then,with the help of Elman neural network identification technology,the frequency band energy was taken as Elman network＆#39;s input variable,the flow pattern was taken as the output variable.A lot of experimental data were used to train Elman neural network.At last,flow patterns were effectively identified with this neural network.Experimental results showed that this method＆#39;s identification accuracy reaches 92.6%,it is 6.5% higher than that of the BP network;this method can effectively identify flow patterns of air lift deives.