Regression-based prediction of flow-induced dominant vibrational frequencies in two-phase flow regimes

An extensive spectrum of vibrational frequencies exists in two-phase Flow Induced Vibrations (FIVs) due to the presence of diverse flow patterns. The distinct behavior of flow regimes poses a significant challenge in predicting Dominant Vibrational Frequencies (DVFs) of two-phase FIVs. This research aims to develop correlations for predicting DVFs in Stratified, Slug, Wavy, Annular, Elongated bubble, and Dispersed bubble flow regimes. Numerical and experimental investigation is carried out to observe the DVFs of different flow patterns on a horizontal 90-degree bend. DVFs are extracted from the Fast Fourier Transform (FFT) response of the bend. A third-degree polynomial regression technique is adopted to identify the optimal fitting polynomials of DVFs relative to gas and liquid superficial velocities in each flow regime. This study has successfully developed thirdorder regression polynomials to predict DVFs across diverse two-phase flow regimes for corresponding superficial velocities of gas and liquid. The proposed polynomials for all six flow regimes demonstrated Root Means Squared Error (RMSE) below 2%. Higher R-squared values affirmed the efficacy in elucidating the variance in DVFs. The findings significantly elevate the prediction of DVFs across diverse flow patterns offering a critical advancement in two-phase flow-induced vibrations.