Application of the phase-averaged pre-processing method in identification of oscillatory two-phase flow patterns based on textural features

Abstract

Oscillatory two-phase flow as a common two-phase flow occurs extensively within the engine's thermal management system. However, it is found that the oscillatory two-phase flow patterns at the same location have different flow states caused by the strong turbulence under fast moving speed conditions, making it difficult to propose systematic identification and judgement criteria. Common image processing methods applied to tube flow are unsuitable for the analysis of oscillatory two-phase flow with particularly complex flow patterns. To solve this problem, this paper presents a study aiming to determine common rules from the instantaneous flow patterns by eliminating the random variability and making the oscillatory two-phase flow pattern images more representative. Applying the phase-averaged pre-processing and textural feature method to the instantaneous oscillatory two-phase flow patterns, which are captured by the designed visualized test bench, the lowest identification results based on the Support Vector Machine (SVM) could achieve an accuracy of 92.5%. The results suggest that the study presented in this paper offers an effective method for the analysis and identification of oscillatory two-phase flow patterns. Visualized experimental results in the related fields of two-phase flow will be more comprehensive with objective descriptions and representative flow pattern images.