| dc.description.abstract | During the work presented in this thesis, the Shut-down flow model of the FlowManager? Cooldown and Restart Simulator has been developed into a transient multiphase flow model that is able to handle shut-down flow and restart flow for horizontal pipelines with uniform discretization.It has been shown that the momentum equation solver originally used in the Shut-down flow model fails to transfer momentum across velocity sign changes. As a result of this, the original momentum equation solver makes the Shut-down flow model incapable of handling the negative velocities that one may experience during shut-down and restart flow. The reason why the original momentum equation scheme fails for velocity sign changes has been identified.A new momentum equation scheme based on the upwind Finite Volume Method has been derived and implemented in the FlowManager? source code by the author. It has been demonstrated that the new solver allows the Shut-down flow model to handle negative flow velocities. Verification simulations have been performed by comparing the old and the new solvers for a two-phase gas/oil flow case. The simulations suggest that the new solver yields results that are consistent with those of the old solver. Furthermore, it is concluded that the new solver is considerably better than the old solver in terms of computational efficiency.In order to see how the Shut-down flow model compares to other multiphase flow models, the results from simulations using FlowManager? 4.3.1, OLGA® 5.0 and the Shut-down flow model has been presented and discussed. The simulation results were compared at steady state. All models were run for three simulation cases: Single-phase gas flow, single-phase oil flow and two-phase gas/oil flow.For single-phase flow, it was found that the Shut-down flow model produced results that were consistent with those of the other models. In terms of pressure drop predictions, the differences between FlowManager? and the Shut-down flow model were negligible for both of the single-phase flow cases. The same was true for the phase velocity predictions. Thus, the Shut-down flow model performed in an excellent manner. It has been concluded that the correspondence between FlowManager? and the Shut-down flow model is as expected, as the two models compute the wall friction in the same way for single-phase flow.For the two-phase gas/oil flow simulations noteworthy discrepancies were seen between the respective simulation models. The pressure drop of the Shut-down flow model was considerably lower than what was predicted by FlowManager?. The OLGA® pressure drop prediction was in between the two, however slightly more in accordance with the FlowManager? prediction. The spread between all models was considered significant.Studying holdup and phase velocity plots for the two-phase flow case, it was found that the correspondence between OLGA® and FlowManager? was far better than what was seen for the Shut-down flow model simulations. It has been concluded that it is the rather simplistic approach taken by the Shut-down flow model in determining the friction terms that is the reason for the two-phase discrepancies. While FlowManager? and OLGA® use empirical multiphase flow correlations, the Shut-down flow model flow model does not. It has been recommended that appropriate empirical correlations are introduced in the Shut-down flow model in order to model the friction terms more accurately.The way that the Shut-down flow model determines time step convergence has been reviewed. Previously, a threshold value for the pressure correction p' was used. One of the problems with this approach is that it does not determine to what extent the governing equations are satisfied. To correct this, the author has implemented methods that calculate mass, momentum and energy residuals in the Shut-down flow model source code. Following suggestions in literature, the normalized sum of absolute momentum residuals for all control volumes is used to determine convergence. The momentum residual at any iteration is compared to the initial momentum residual, that is.From the simulation results, the author has argued that the convergence criteria that are used are of significant importance when it comes to the CPU time that is used by the simulation models. Comparing the required number of SIMPLE iterations per time step used by the Shut-down flow model for the single-phase gas flow case and the two-phase gas/oil flow case, it was seen that the single-phase simulation required considerably more SIMPLE iterations than the two-phase simulation. However, inspecting the size of the residuals at the first iteration of the first time step, it was found that the initial residual was considerably larger for the two-phase case. While it has been concluded that using the momentum residual to determine convergence certainly is a model improvement, this approach still fails to recognize the size of the initial residual.Shut-down flow and restart flow simulations have been performed. From the results of the shut-down flow simulation case it was pleasantly concluded that the Shut-down flow model is able to simulate shut-down flow successfully. From the phase velocity plots it was readily seen that both the gas phase and the liquid phase come to rest as the flow inlet is closed off. Furthermore, as the phase velocities decreased, it was seen that the pressure decreased until it was the same as the outlet pressure throughout the pipe segment.From the results of the restart flow simulation case it was concluded that the Shut-down flow model is also able to successfully simulate restart flow. It was seen that the discrepancy between the final restart flow simulation solution and the final solution obtained when the Shut-down flow model is started with a steady state solution as an initial condition is negligible. This is interpreted as a strength of the Shut-down flow model. | nb_NO |
