| Two-phase flows are very often seen in industrial applications such as energy conversion systems, filtration, spray processes, natural gas networks and nuclear reactor cooling. Changes in pressure loss, heat and mass transfer rates, momentum loss, rates of back mixing, and vibration of pipes all vary greatly with two-phase flow patterns. Therefore, it is quite important to identify the flow patterns. The early experimental methods for detecting flow regimes were mostly based on direct observation, and were greatly subjective. In order to increase the objectiveness, indirect methods of observations were elaborated, based on the statistical analysis of the fluctuating characteristics of the flow. Further, criteria for flow regime transition were defined on a theoretical basis. However, due to the complex nature of two-phase flow, theoretical analyses have not been able to describe the system perfectly.; In this project, a new method is developed for identifying two-phase flow regimes in a vertical channel, from the void fraction signals gathered by an optical fiber probe system. Flow regimes are detected from signal patterns, using hidden Markov model (HMM). HMM has been increasingly used from the 1960's in speech recognition and numerous other fields. HMM is a very strong tool for comparing two or more signals, which is a key to solve several problems in practice. In the present work, for two-phase flow pattern identifications, HMM was applied as follows: the data corresponding to two-phase flow conditions with clearly known regime (i.e. conditions far away from transition regions) are used as references to detect and classify the conditions with unclear regime (i.e. conditions close to transition regions). Reference conditions were selected using the map provided by Taitel et al. (1980), together with photos taken during the experiments. For each regime, three reference conditions were selected. Flow pattern likelihoods were calculated for each test condition. Two-phase flow maps were extracted from the numerical results, using the maximum likelihood and total maximum likelihood. In the maximum likelihood approach, the regime for a given test condition was determined by selecting the highest likelihood value among all reference conditions. In total maximum likelihood approach, the sum of likelihood values of reference conditions of the same regime was used for comparison. Then all the conditions were categorized and reflected on Taitel's map and the transition boundaries were depicted.; Void fraction signals were collected from sixty conditions of air-water two-phase upward flow. Seven different homogenous velocities (0.5 m/s to 5 m/s) and nine void fractions (10% to 90%) were simulated in a test section. The polycarbonate tube was 2 meters long and 19 mm in diameter. Air and water flowrates were monitored and compensated based on the pressure measured at the tube entrance.; Void fraction signals were collected using a single step index multimode optical fiber probe located at the center of the tube, and a National Instruments data acquisition board. For each condition simulated in the test section, one minute worth of data was sampled at a frequency of 99 kHz. In addition, for each test, flow patterns were captured using a Nikon digital camera. (Abstract shortened by UMI.)... |
