Tomographic Approach to Automatic and Non-Invasive Flow Regime Identification
Doctoral thesis, Peer reviewed
MetadataShow full item record
Original versionPradeep, C. Tomographic Approach to Automatic and Non-Invasive Flow Regime Identification. Doctoral dissertation, Telemark University College, 2015
This thesis mainly presents the use of process tomography to study and measure flow parameters in two phase horizontal and near horizontal flows. All static measurements were made using a horizontal separator section while dynamic flow measurements were performed using the flow facility at Telemark University College. Most of the study was based on Electrical Capacitance Tomography (ECT) and Electrical Resistance Tomography (ERT) measurements. Unlike typical tomographic applications, here, the focus was on extracting information from measurements and not constructed tomograms. Artificial Neural Network (ANN) algorithms based inferential models were first developed for interface level estimation of layered flows. The results were tested and validated with both static and dynamic measurements. Separate models for oil-air, water-air and oil-water two phase combinations were developed to compare with the measurements. After having very satisfactory estimations with 12 electrode sensor data, the study was extended with the possible reduced number of electrode sensors. Here the selections were 6, 5 and 4 selected electrode combinations. Corresponding measurements of the selected electrodes were employed in the estimations. The speed of the estimation with the reduction of electrode of the sensor was also studied separately. The experimental analysis demonstrates that interface estimates of the layered flows are possible with ANN based algorithms. It is further evident that even with the reduced 6, 5 and 4 electrode sensor arrangements, acceptable results can be observed quickly but with some increased uncertainty. The possibility of using as a redundant system is also an added advantage of having this type of estimators separately. Some Gamma measurement results presented by (Vestøl 2013) were tested with the ECT tomographic technique. Here, tomograms were used in this study. For oil-air two phase flows the comparisons were acceptable despite the low resolution with ECT. But, ECT does not produce acceptable tomograms when water was the dominant medium. Slug flow parameters, such as slug translational velocity, slug front and tail velocities, liquid film thickness, slug frequency and slug length, were estimated using ECT and ERT measurements. The cross-correlation technique was used in the slug translational velocity calculation. This calculated slug translational velocity, then used in the other parameter estimations. Highspeed camera images captured provided that the estimations are acceptable, though the capturing frequency (100Hz) is not high enough. Time series of differential pressure measurements were also captured along with the tomographic measurements for the verification purpose. They were separately studied with different techniques for the slug flow parameter estimations. Power Spectral Density (PSD) was applied in the calculation of slug frequency. Results show a good agreement between ECT and ERT based estimations and differential pressure based calculations. Possible use of capacitance measurement data matrix properties for the flow regime detection was tested with experimental data. Here independent capacitance measurements of each frame are arranged in a symmetric matrix and the eigenvalues of them are calculated. How the eigenvalues are related to the flow parameters such as liquid fractions are studied here. A good agreement with both measurement data and mathematical model based results are given by (Fang & Cumberbatch 2005). A solid relationship between leading eigenvalue and the volume fractions could be observed. Possible identifications of flow regimes with the second and third dominant eigenvalues were also discussed with the experimental results. Dynamic time warping technique which is commonly applied to align two time series signals are used in the liquid slug length estimations. The fusion of both tomometric and differential pressure measurements is done with this approach. Then, the relationship between liquid slug length and warped pressure peak was also investigated. The relationship between differential pressure and liquid slug length could be identified as linear. The results are proven with different experimental data.
Has partsPaper I: Pradeep, C., Ru, Y. & Mylvaganam, S. (2012). Neural Network-Based Interface Level Measurement in Pipes Using Peripherally Distributed Set of Electrodes Sensed Symmetrically and Asymmetrically. IEEE Transactions on Instrumentation and Measurement 2012, 61(9) p. 2362-2373. Full text not available in TEORA due to publisher restrictions. The published version is available at http://dx.doi.org/10.1109/TIM.2012.2199191
Paper II: Pradeep, C., Ru, Y. & Mylvaganam, S. (2012). Reverse Flow Alarm Activation using Electrical Capacitance Tomometric (ECTm) Correlation. 2012 IEEE Sensors Applications Symposium (SAS 2012) Proceedings, p. 228-233. Full text not available in TEORA due to publisher restrictions. The published version is available at http://dx.doi.org/10.1109/SAS.2012.6166311
Paper III: Pradeep, C., Ru, Y., Vestøl, S., Melaaen, M.C. & Mylvaganam, S. (2014). Electrical capacitance tomography (ECT) and gamma radiation meter for comparison with and validation and tuning of computational fluid dynamics (CFD) modeling of multiphase flow. Measurement Science and Technology 25(7). Full text not available in TEORA due to publisher restrictions. The published version is available at http://dx.doi.org/10.1088/0957-0233/25/7/075404
Paper IV: Pradeep, C., Ru, Y. & Mylvaganam, S. (2012). Multimodal Tomometry for Slug Detection in two Phase Flow. Proceedings of the 6th International Symposium on Process Tomography, Cape Town, March 26-28, 2012.
Paper V: Pradeep, C., Ru, Y. & Mylvaganam, S. (2012). Co-operative sensor fusion using time warping in multimodal tomometry for process control. IEEE Conference on Control, Systems & Industrial Informatics (ICCSII), Bandung, Indonesia, Sept. 23-26, 2012. Postprint version. The published version is available on http://dx.doi.org/10.1109/CCSII.2012.6470504