Numerical Modeling of Multiphase Flows with Applications to the Automotive Industry
MetadataShow full item record
Multiphase flows have become an issue of deep concern in the automotive industry. Two specific problems of concern to the ABC Group and Ford Motor Company have shown that the current multiphase models available in commercial software are not able to satisfy the requirements of the automotive industry for numerical simulations: namely, the need to solve complex industrial problems and the need to provide accurate solutions for these problems. The first problem of concern involves degassing in coolant surge tanks. Bubbles mixed in the coolant constitute a risk factor that influences the cooling performance, and the degree of mixing needs to be strictly controlled. Therefore, simulating the degassing rate accurately is of critical importance for the design of a coolant surge tank. The second case of concern focuses on a multiphase-flow-induced vibration and noise problem in a dynamic valve. This problem relates directly to the noise, vibration and harshness (NVH) performance of a hybrid vehicle. The dynamic motion of a dynamic valve, which is controlled by a balance between the spring force and the pressure force of the diesel-air multiphase flow acting on the valve, results in a complicated generation and propagation of mechanical and fluid-dynamic noise. This is a complex problem that combines multiphase flows and fluid-structure interaction (FSI). Motivated by these two problems posed by our industrial collaborators, this research has been conducted specifically in order to improve both the predictive accuracy of and degree of complexity of problems that can be addressed by the multiphase flow models presently available in commercial software. Moreover, the improvements obtained here are used to solve the two industrial problems described above. Towards this purpose, the ANSYS-FLUENT software is utilized for the current research because of its overall computational capabilities, its ease of use, and its widespread application in the automotive industry. However, to address the two aforementioned problems, it was necessary to improve the modeling capabilities of the ANSYS-FLUENT software. For this purpose, a number of model improvements were implemented in this study. Firstly, a multiphase flow model in ANSYS-FLUENT was improved through the utilization of user-defined functions (UDFs). Secondly, UDFs were also implemented to generalize the solution procedures in the FSI model. Thirdly, a Scheme command file was used to the FSI model to provide a significantly improved solution algorithm based on an implicit solver that is capable of addressing problems requiring the strong coupling between the flow solver and the FSI solver. Validations of the key methodologies (viz., the multiphase flow, the dynamic meshing capability and the FSI) of the integrated modeling system were carried out before being applied to the two specific problems of concern described above. Firstly, a bubbly turbulent flow in a vertical pipe was simulated and then compared with the experimental measurements in order to validate the improved multiphase flow model. The new multiphase flow model is based on the original Eulerian model (sometimes referred to as the two-fluid model in the literature) in ANSYS-FLUENT, but this model incorporates a number of new interfacial force models which were implemented using a UDF. The "wall peak'' observed in the radial distribution of the gas volume fraction and the shift of this peaks towards the core (center) of the flow (to give a "core peak'') as the volume fraction of the gas increases are captured correctly in our simulations, in contrast to the simulation results reported by other researchers. The immiscible model was also tested and found to be able to capture the major features of the bubbly flow in the vertical pipe. A turbulent flow over a square cylinder with a prescribed motion was used as the second benchmark test case. The purpose of this test was to validate the capability and accuracy of the dynamic (or moving) mesh model in ANSYS-FLUENT. Two methods (namely, smoothing and layering) were used to generate the dynamic mesh. Both of these dynamic meshing methods yielded good performance in terms of the mesh quality. The predicted shedding of the Karman vortex street behind the cylinder and the “lock-in” phenomena over a small range of reduced velocities agreed well with the experimental measurements. The layering method for dynamic mesh generation was found to give a higher computational efficiency than the smoothing method. Validation of the FSI model was more complicated than that of multiphase flow model, as models for FSI depend critically on the specific physical characteristics of the problem and these characteristics dictate whether a weak coupling or strong coupling is required for the solution. To validate the model for a weak-coupling FSI problem, a turbulent flow over and past an autonomous-oscillating square cylinder was used as the benchmark test case. The vortex-induced vibration occurring behind the square cylinder for this case constitutes the classic weak-coupling FSI problem. A UDF was implemented to simulate the FSI using a dynamic mesh. A layering method was used to generate the dynamic mesh. The predicted shedding of the Karman vortex street behind the cylinder and the ``lock-in'' phenomenon over a small range of reduced velocities was found to agree well with the experimental measurements. For a strong-coupling FSI validation, we used a laminar flow in a heart valve as a benchmark test case. This valve has a free rotating leaflet controlled by periodic variations of the unsteady inlet velocity. To address this case, a new implicit solution methodology was proposed and implemented using a command file (written in Scheme) in order to solve this large-displacement problem. Simulation results indicate that the periodic motion of the leaflet was predicted fairly accurately even when friction was ignored. The resulting implicit FSI model demonstrates the capability of the new modeling tool in solving strongly coupled FSI problems similar to that associated with the operation of a dynamic valve. The improved and validated multiphase flow model was applied firstly to a simulation of the degassing problem in a coolant surge tank. Grid sensitivity tests and time step sensitivity tests were conducted in order to determine optimal values for various parameters to be used in the simulation. Parametric tests, including inlet velocity tests, bubble diameter tests and liquid viscosity tests were undertaken. A degassing rate was proposed and defined as a metric for the degassing process. Geometries for a single-chamber and three connected chambers were simulated. It was found that bubbles are degassed stage by stage owing to the buoyancy effect. The degassing rate is a dynamic variable related to many physical properties, such as the inlet velocity, the liquid viscosity and the bubble size. A larger inlet velocity, a greater liquid viscosity, and a smaller bubble size result in a lower degassing rate. The effect the drag force models is not obvious although the drag force is the most important interfacial force influencing the degas rate. The effect of the lift force models is even weaker than that of drag force for determination of the phase interaction. It is concluded that our proposed integrated modeling system provides reasonably good results for the degassing process in a surge tank. Finally, the multiphase-flow-induced vibration and noise problem in a dynamic valve was simulated using our proposed integrated modeling system. For a two-phase flow through the valve, the simulations showed that the deformation and breakup of gas bubbles in the gap between the poppet and the valve seat generate vibrations that arises primarily from the force imbalance between the spring and the two-phase fluid-flow-induced forces on the poppet. A spectral analysis of the transient pressure force on the poppet revealed the presence of a strong cyclical behavior consisting of two major components. There was a low-frequency peak located at about 87 Hz and associated with the frequency of the poppet vibration (and which we interpret to be the source of the mechanical noise) and a high-frequency peak located at about 450--970 Hz which is associated with compressibility effects and the unsteady vortex motions in the spring chamber. The poppet vibration and noise are influenced by various factors such as the flow condition, the spring system properties, and the geometry of the valve. Larger bubbles and a lower inlet velocity result in larger displacements in the poppet in a non-equilibrium condition and induce a greater loading on the spring due to the higher pressures. These pressures in turn amplify the poppet vibration and noise. The detailed simulations and subsequent analysis of the complex interactions that occur in the turbulent multiphase fluid motion through a moving poppet valve allowed deeper physical insights and an improved understanding of both the source and properties of the vibration and noise generated in this complicated dynamic system.
Cite this version of the work
Chenguang Li (2015). Numerical Modeling of Multiphase Flows with Applications to the Automotive Industry. UWSpace. http://hdl.handle.net/10012/9450