LAUSR

Abstract Recognition of unstable and harmful condensation regimes in liquid pipe flow system can promote a higher level of flow assurance in liquid propellant rocket engine. However, challenges are encountered in extracting distinguishable characteristics from pressure oscillation signals which commonly contains plentiful information strongly associated with various condensation regimes. This article attempts to set up a simple and practical approach of recognizing the steam jet condensation regime in water pipe flow system based on statistical features of pressure oscillation. The recognition procedure was performed in three major steps. Initially, twelve statistical features of pressure oscillation in time-domain (probability density function) and frequency-domain (power spectrum density) were chose. Subsequently, principal component analysis was implemented to obtain the clear interrelations between condensation regimes and statistical features of pressure oscillation signal, and then to extract useful features for establishing condensation regimes clusters for classification in the selected features space. Finally, least squares support vector machine was adopted to the clusters for construction of classifiers to forecast the condensation regimes automatically. The experimental results showed that the proposed approach is feasible and effective for recognizing the steam jet condensation regime in water pipe flow system by statistical features of pressure oscillation.