In order to better understand the variation of flow structure with delay time, we propose the element area (EA) of attractor morphology parameter in this paper. First, the conductance fluctuating signals and adaptive optimal kernel time-frequency representations of different gas-liquid flows are shown, we can find that flow pattern evolution is always accompanied by the numerical and frequency changes of large amplitude fluctuation (LAF). Then three kinds of signals, i. e., rossler signal, white noise and sinusoidal signal with multi-components, are used for analyzing the simulations, and the results indicate that the greater the frequency of LAF, the smaller the delay time of first crest of EA( peak ) is, and that the more the LAF, the bigger the peak value of first crest of EA(hpeak) is. Additionally, we use the above rule to analyze the conductance fluctuating signals measured from upward gas-liquid two-phase flow experiments and the signal length is selected to be 10 s for analysis. When the water superficial velocity is fixed to be 0.1138 m/s and the gas superficial velocity is gradually increased, we find that the peak is constant and hpeak changes up and down at bubble flow. When the flow pattern evolves into bubble-slug transition flow, the peak begins to turn bigger, and when the flow pattern evolves into slug flow, the peak becomes constant again while the hpeak increases monotonically with the gas flow rate increasing. The peak begins to become smaller as the flow pattern evolves from slug flow into churn flow, and we can find that the peak and hpeak of transition flow are alike. The peak and hpeak of bubble flow and churn flow are also alike because their dynamical mechanisms are similar but the downward trend of bubble flow is more gently than that of churn flow. When the water superficial velocity is fixed to be 0.2719 m/s, we can find similar variations of peak and hpeak to the above. Finally we determine the fall ratio (Rf) which is the ratio of the difference between the first crest and the first trough of EA and the hpeak, and then quantitatively distinguish three typical flow patterns, i.e., bubble flow, slug flow and churn flow by the Rf - peak distribution.