Abstract:In two-phase flow study, accurate predication of two-phase flow regime is one of the keys for calculation of heat and mass transfer and flow resistance between the two phases. The present study proposes a new flow regime feature extraction&nbsp;method based on non-stationary conductivity probe signals. Two types of discriminative network models, the probabilistic neural network (PNN) and nonlinear support vector machine (SVM), are established for intelligent flow regime identification.&nbsp; A series of 132 experimental datasets&nbsp; of air-water two-phase flow taken in&nbsp; vertical-downward two-phase flow are used for model training.&nbsp;The wavelet packet energy and time-domain characteristic parameter are selected for the&nbsp;eigenvectors of bubbly, slug, churn-turbulent and annular flow regimes.&nbsp; They are obtained by&nbsp;wavelet packet decomposition and time-domain feature statistical analysis methods.&nbsp;The wavelet packet decomposition results&nbsp; show that the&nbsp;fluctuation of the conductance signal in two-phase flow are mostly&nbsp;in low -frequency.&nbsp;The time-domain characteristic parameters, including amplitude, standard deviation, arithmetic mean, rectification mean, peak factor, impulse factor and margin factor, are derived from the reconstructed waveform of low-frequency signals. Compared to&nbsp;the characteristics of the wavelet packet energy, the intervals of the time-domain characteristic parameters are wider, which makes it beneficial for flow regime classification.&nbsp;Since the spread of radial basis functions has a great influence on the performance of probabilistic neural network.&nbsp;Sensitivity analysis on&nbsp;the influence of different "spread" values on the probabilistic neural network model hwas done. It is found that&nbsp;the spread value of 0.3 gives the smallest training error and generalization error.&nbsp; The performance of PNN and nonlinear SVM applied to&nbsp;flow regime identification is evaluated&nbsp;by comparing the percentage of right identification results with the flow visualization results. It is found that the differences between the model identification and the flow visualization happen mostly in the boundaries of flow regime transition. The overall accuracy of flow regime identification using the nonlinear SVM method is 90.47%, which is higher than 83.33% using the PNN method. This verifies the superiority of nonlinear SVM in solving small samples and nonlinear flow regime identification and classification problems.Keywords:Flow regime identification, PNN, nonlinear SVM, WPD &nbsp;

Vertical-Downward Two-Phase Flow Regime Identification by Probabilistic Neural Network (PNN) and Nonlinear Support Vector Machine (SVM)

Paper Proceedings: https://asmedigitalcollection.asme.org/ICONE/ICONE28/volume/85246

2021 28th International Conference on Nuclear Engineering (ICONE28)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.