Gas-liquid two-phase flow is complex and has uncertainty in phase interfaces, which make the two-phase flow look very complicated. Even though the flow behavior (e.g. coalescence, crushing and separation) of single bubble or bubble groups in the liquid phase looks random, combining some established characteristics and methodologies can find regularities among the randomness. In order to excavate the nonlinear dynamic characteristics of gas-liquid two-phase flow, the authors developed an improved matrix pencil (IMP) method to analyze the pressure difference signals of the two-phase flow. This paper elucidates the influence of signal length on MP calculation results and the anti-noise-interference ability of the MP method. An IMP algorithm was applied to the fluctuation signals of gas-liquid two-phase flow to extract the mode frequency and damping ratio, which were combined with the component energy index (CEI) entropy to identify the different flow patterns. It is also found that frequency, damping ratio, CEI entropy and stability diagram together not only identify flow patterns, but also provide a new way to examine and understand the evolution mechanism of physical dynamics embedded in flow patterns. Combining these characteristics and methods, the evolution of the nonlinear dynamic physical behavior of gas bubbles is revealed. (C) 2016 The Chemical Industry and Engineering Society of China, and Chemical Industry Press. All rights reserved.