چكيده به لاتين
Many studies have been carried out to measure multiphase flow of water, oil and gas by scientists with a variety of fields of study, including mechanical engineering, nuclear engineering, chemistry and other sciences. Multiphase flow metering is one of the most important issues in the oil and gas industry since the early 80's. In a general definition, the multi-phase flow of water, oil and gas is formed of several immiscible phases from several simultaneous paths in a multi-phase flow system. The three-phase flow of oil, gas and water is commonly seen in crude oil and natural gas industries. The complex and variable flow patterns make it difficult to measure these flows and turn them into a hot topic for studying. Due to the importance of measuring the multiphase flow in the upstream and downstream industries, as well as the economics of using multiphase flow meters, instead of commonly used field separators, it is important to use multiphase flow meters to be of great importance. In this research, the multiphase flow measurement methods, which are: component fraction, component velocity, mass flow rate and volume of components, as well as various models of computational fluid dynamics simulation were investigated. The two-phase flow loop was designed and manufactured using existing equipment. In the two-phase flow loop, the water flow was pumped from the water tank using a centrifugal water pump and after measuring the temperature and pressure by a temperature and pressure transmitter, the water flow rate was measured by the electromagnetic flow meter. The air flow was supplied by the screw air compressor. After measuring the temperature and air pressure using a temperature and pressure transmitter, the flow rate of the air was measured by turbine flow meter. Then, water and flow mixed at mixing zone and two-phase flow was formed. Two-phase flow rate was measured using Coriolis and Orifice flow meters. After calibration of Coriolis flow meter using electromagnetic flow meter, the effect of air volume fraction and Reynolds number of two-phase flow on the performance of coriolis and Orifice flow meters were studied. In order to investigate the performance of coriolis and Orifice flow meters, the distance between the meters and mixing zone was assumed to be 100 times greater than diameter of the pipeline to ensure the two-phase flow development. The Coriolis Flow meter diagram versus air volume fraction of two-phase flow was obtained. It was observed that by increasing the mass flow rate of the single phase flow of water which passing through the Coriolis flow meter, the Coriolis flow meter factor was closer to one and while air volume fraction of the two-phase flow increases, the Coriolis flow meter factor will be greater than one. Air volume fraction of two-phase flow was determined using the obtained average density by Coriolis flow meter. Air volume fraction of two-phase flow also determined by the artificial computational intelligence such as one layer neural network, two and three layer neural network which optimized by genetic algorithm and support vector machine with radial basis kernel functions and their performance were compared together. Between one layer neural network training algorithms, the Bayesian regularization training algorithm had the best performance and its mean squared error for test data was equal to 3.9%. Performance of the support vector machine with radial basis kernel functions was better than one layer neural network with Bayesian regularization training algorithm and its mean squared error was equal to 3.25% for test data. In order to achieve better performance in prediction of air volume fraction of two-phase flow, number of neurons in each layer of neural network optimized and number of layers on neural network increased and optimized with genetic algorithm. The three layer neural network optimized by genetic algorithm had the best performance for air volume fraction prediction and its mean squared error was equal to 0.9% for test data. The cumulative error percentage graph for the predicted results by the three-layer neural network optimized with genetic algorithm, showed that 95 percent of the points had the cumulative deviation percentage less than 5%. Similarly, three-layer neural networks, optimized with genetic algorithm, can play an important role in many industrial applications with extreme precision to act as a soft sensor under various operating conditions. The Orifice plate flow meter performance was investigated for two-phase flow which passes through it. It was observed that while the Reynolds number of two-phase flow increases, discharge coefficient of the Orifice plate will increases. The incremental slope of discharge coefficient of the Orifice plate flow meter was higher for low Reynolds number of two-phase flow which two-phase flow pattern was plug. For bigger Reynolds number of two-phase flow, the flow pattern of two-phase flow was stratified and the slope of variation of the orifice discharge coefficient versus the Reynolds number of two-phase flow was lower. Discharge coefficient and pressure drop of the Orifice plate flow meter and two-phase flow velocity were validated using computational flow dynamics. The control volume consists of a pipeline with a length of 1.2 meters, the Orifice plate being located at the distance of 1 meter from the inlet boundary, with quad meshes with the boundary conditions governing the input boundary of the pipe inlet was mass flow inlet and the boundary of the outflow at the pipe outlet, and The surface of the tube was considered as wall. Solution was considered as steady and gravity acceleration was equal to 9.81 m/s2. In order to ensure the independence of the simulation results from the mesh cells, the mash independency analysis was done and it was observed that the average difference between the simulation results, including the two-phase flow velocity in the central axis of the pipe, with the number of simulation cells equal to 213300 and 251400 is equal to 0.5%, and the number of simulation cells of 213300 was considered as the optimal number of meshing geometry cells. The simulation of the control volume with the Eulerian-Eulerian multiphase model was performed in steady condition and turbulence models such as the standard K-Epsilon, SST K-Omega, STN K-Omega and RNG K-Epsilon were compared in terms of performance, and it was found that the discharge coefficients obtained for the Orifice plate with the standard K-Epsilon turbulence model had the best performance. Difference between experimental discharge coefficients obtained in this study and standard K-Epsilon model was 2%. One of the most important achievements of this research was the quantitative and qualitative measurement of two-phase flow using Coriolis and Orifice plate flow meter and Artificial Intelligence techniques, and also providing a suitable field to manufacture multiphase flow meters.
