In situ adaptive tabulation of vapor-liquid equilibrium phase change models for the numerical simulation of high-pressure transcritical flows with multiple components

Abstract

Vapor-liquid equilibrium (VLE) models constitute a set of first-principles thermodynamic frameworks tailored to address the complexities of transcritical multiphase flows. They excel at accurately capturing phase transitions occurring under high-pressure conditions, a challenge that remains elusive for alternative modeling approaches. However, VLE-based computational fluid dynamics (CFD) simulation is computationally very expensive for multi-component systems, which severely limits its applications to real-world systems. This limitation severely constrains the practical applicability of VLE in real-world scenarios. In response, this thesis presents the development of a novel ISAT-VLE method based on in situ adaptive tabulation (ISAT). Our primary objective is to significantly enhance the computational efficiency of VLE-based CFD simulations while concurrently reducing memory consumption. To achieve this, we introduce ISAT-VLE solvers, tailored to accommodate both fully conservative (FC) and double flux (DF) schemes. Innovative techniques are proposed to eliminate redundant records within the ISAT-VLE table, further optimizing the method's performance. This thesis encompasses a series of simulations, including high-pressure transcritical temporal mixing layers and shock-droplet interactions, utilizing the ISAT-VLE CFD solvers. Our results demonstrate that the new method achieves a remarkable speed-up factor, ranging from approximately 10 to 60, while ensuring that ISAT errors remain well-controlled within a tight margin of 1%. In our exploration of ISAT-VLE in parallel computing, we encountered a notable performance degradation. To address this challenge, we have developed an innovative ISAT method designed explicitly for parallel computing environments. This novel ISAT approach adopts a hybrid MPI-MPI model to enable shared memory support and employs it to construct a concurrent binary tree, facilitating efficient concurrent read and write operations. This innovation allows us to establish a shared ISAT table within each computing node. Moreover, we have introduced a load balancing algorithm based on shared memory, enabling dynamic workload allocation to minimize imbalance across processors. Additionally, we have implemented a merged shared and local table ISAT strategy to further enhance overall performance. The outcome of these enhancements has been a remarkable transformation in both computational performance and scalability. Specifically, when simulating 3D shock-droplet interactions with 128 processors, our approach has achieved an impressive nearly sixfold increase in performance compared to the original ISAT methodology. Furthermore, the utilization of shared ISAT tables optimizes the usage of table records while concurrently reducing memory consumption. These advancements collectively signify a significant stride forward in the efficiency of parallel computing using ISAT methodologies. Then we investigate supercritical CO2 systems which play a pivotal role in semi-closed sCO2 cycles, holding great promise as the next-generation power cycle. We use the VLE model to investigate the multicomponent effects on the sCO2 systems. VLE-based thermodynamic analyses show that a small amount of combustion-relevant impurities (e.g., H2O, CH4, and O2) can significantly elevate the mixture critical point of the sCO2 systems. As a result, the so-called “supercritical” CO2 systems might be in the subcritical two-phase zone where phase separation occurs. At the relevant conditions in this study (100-300 bar), phase separation only has a small influence on the CO2/H2O/CH4/O2 mixture density, but has a considerable influence on the heat capacity of the mixture. VLE-based CFD simulation of a laminar premixed sCO2 shock tube shows that expansion waves can trigger significant condensation in the systems and the latent heat of the condensation can change the temperature and density fields in the systems. To understand the phase separation during mixing, VLE-based large-eddy simulations (LES) of turbulent jet-in-crossflows in the sCO2 systems are conducted, and the results show that when two subcritical gas or supercritical gas-like streams mix, the mixture can partially condense to subcritical liquid phase. Higher pressure, lower temperature, and higher H2O concentration can enhance the phase separation phenomenon in the systems.