Abstract: Four differential pressure signals of gas-liquid two-phase flow pattern in a rod bundle channel were collected and the multiple scale marginal spectrum entropies of four kinds of flow patterns were calculated, then the identification and dynamics characteristics of four kinds of flow patterns were analyzed. The experimental results show that multiple scale marginal spectrum entropy can distinguish four kinds of flow patterns from the whole, and the dynamics characteristics of different flow patterns can be revealed quantitatively from the detailed scale of frequency domain. The four kinds of flow patterns can be accurately distinguished by using multiple scale marginal spectrum entropy rate and spectral entropy mean joint distribution, and the bubbly and churn flow that it is difficult to distinguish such transitional flow also has a good effect. Combining with the support vector machine, the method has the advantages of fast operation and high recognition rate, its accuracy rate is up to 98.11%, and is suitable for online identification of flow patterns.