基于EMD和ICA的AR模型在两相

为了对气液两相流流型进行准确识别，提出了一种基于经验模态分解（EMD）和独立成分分析（ICA）的AR模型的流型识别方法。该方法首先提取气液两相流动态差压信号，通过EMD技术对其进行分解，对分解出来的高频模态进行小波包消噪，然后通过ICA技术实现原始信号的降维处理。对通过ICA得到的独立分量建立AR模型，将模型参数和残差方差作为特征向量，建立不同流型的模板向量。计算未知流型信号的特征向量与模板特征向量的综合Mahalanobis距离，通过比较各判别距离的大小得到流型识别的结果。对40 mm水平管气水两相流进行实验，利用文丘里管采集动态差压信号，采用上述处理过程可以对泡状流、弹状流、塞状流进行有效识别，识别率达94.3％。该方法受环境条件影响小，可以有效滤除信号中的噪声成分，识别率高，易于工程实现。

Application of AR model based on EMD and ICA in flow regime identification for gas-liquid two-phase flow

An autoregressive model (AR model) based on empirical mode decomposition (EMD) and independent component analysis (ICA) was applied to identify gas-liquid two-phase flow regime. The dynamic differential pressure signals coming from Venturi tube were first processed with normalization and EMD. The intrinsic mode functions (IMFs) with high frequency component were denoised by wavelet packet. Then the dimension of IMFs was reduced through ICA. The independent components obtained from ICA were selected to establish the AR model. With model analysis, the coefficients and the variances of residual errors were defined as the eigenvector, and templates of different flow regimes were established. The flow regime was identified by calculating the synthetic Mahalanobis distance of eigenvectors between template and pressure fluctuation signal. The experimental results of gas-liquid two-phase flow in horizontal pipes with 40 mm inner diameter manifest that this method can identify bubble flow, slug flow and plug flow with an accuracy of 94.3%. This method can be realized easily for identification of flow regime in gas-liquid two-phase flow with less influence of environment.