WCIPT8 - 8 th WORLD CONGRESS ON INDUSTRIAL PROCESS TOMOGRAPHY Iguassu Falls, PR, Brazil, September 26 to 29, 2016 ISIPT - The International Society for Industrial Process Tomography 1 Annular flow pattern recognition using statistical data analyses of Electrical Impedance Tomography J. Polansky , M. Wang University of Leeds, School of Chemical and Process Engineering, LS2 9JT, Leeds, UK jiri.polansky@icloud.com ABSTRACT Collecting very large amount of data from experimental multiphase measurement is a common practice in almost every scientific domain. There is a great need to have specific techniques capable of extracting synthetic information, which is essential to understand and model the specific flow phenomena. The intention of developing a method for recognition of flow regime using decomposition mathematical technique comes from the fact that each regime is characterised by typical dynamic behaviour. To recognise the flow dynamic structures, means indeed the recognition of the prevalent regime moreover indicates the actual flow conditions of the monitored area. The direct approach of Proper Orthogonal Decomposition (POD) as introduced by Lumley and the Linear Stochastic Estimation (LSE) as introduced by Adrian are used to identify typical multiphase flow instability. The present approach of statistical data-analysis extends the current evaluation procedure of Electrical Impedance Tomography (EIT) applied on air-water flow measurement. Wavelet Transformation and Kalman Filtering was used as complementary techniques for motion of fluid and flow structures detection and decomposed EIT signal similarity estimation. The paper demonstrates the capability of EIT measurement techniques combined with POD/LSE post-processing for studying annular flow patterns in vertical and horizontal pipeline. Keywords Proper orthogonal decomposition, gas-liquid flow, Annular flow, Electrical Impedance Tomography 1 INTRODUCTION 1.1 Flow classification and applications Considering a gas-liquid two phase flow (Thomas et al., 2013), the liquid and gas are regarded as the continuous and dispersed phases respectively. Gas-liquid flows are commonly observed in many industrial processes such as oil and gas (Freese 2006), chemical and pharmaceutical (Holland 1995), transportation and nuclear industries (Lavicka 2013). The relative distribution of the gas and liquid phases can take many different configurations depending on the process conditions, such as the flow rates of the gas and liquid. The configuration of the gas and liquid phases is known as the flow regime. The flow regime describes the pattern of the inner structure of the flow and important hydrodynamic features such as volume fraction, phase and velocity distributions. Two-phase flow regimes are often determined subjectively using direct methods such as the eyeballing method, high-speed photography method and the radioactive attenuation method. Empirical flow regime maps such as the Baker chart (Baker 1954) are commonly used for approximate and rapid identification of the flow regime under specific operating conditions. However, due to their approximate and subjective nature these techniques are not able to identify the prevalent multiphase flow regime with the required degree of accuracy. Statistical analysis of the signal has also been used for identification of flow regimes (Faraj 2015) 1.2 Flow pattern estimation The prediction of flow patterns for fully developed gas-liquid flows typically employs mechanistic models that use different pressure drop and void fraction estimation procedures for each flow pattern (Hewitt 1969). Accurate prediction of heat transfer, void fraction and pressure drop in gas-liquid flow is