Numerically modelling the multi-physics phenomenon of combustion is challenging as it involves fluid flow, chemical reaction, phase change, energy release, etc. Combining numerical models for all these phenomena into a single solver ensuring scalability and performance is a daunting task. Based on the hierarchical meshing technique building cube method (BCM) we present a numerical framework "CUBE" for modelling internal combustion engines. The framework efficiently combines a fully compressible flow solver, chemical reaction and combustion model, a particle-in-cell based liquid spray model, and an immersed boundary method for geometry treatment. The flow, temperature fields and the transport of reacting species an all speed Roe scheme is adopted discretization of the advective flux. The solver is coupled with the equilibrium chemical reaction library CANTERA to model combustion. The parcel model-based particle source-in-cell (PSI-cell) method is adopted for modelling liquid fuel spray and its evaporation. Validation of the numerical framework is carried out by using experimental data of a model internal combustion engine known as the Rapid Compression Machine (RCM). Evaluation of the framework with strong scaling analysis shows good scalability.

Efforts to realize the LES of a full-scale combustor for liquid rocket engines are reported. LS-FLOW-HO, which is being developed as a large-scale computation platform in JAXA, has dramatically reduced the computation cost to less than 1/20 compared to conventional in-house solvers, enabling LES of a single-injector combustor or a sub-scale combustor with multiple injectors to be performed in about days to weeks using the JAXA's supercomputer system JSS3. Preliminary results were obtained for the LES of a full-scale combustor with more than 500 injectors, although some issues remain, such as suppression of unphysical pressure oscillations in the early stages of the simulation. Thermoacoustic coupling associated with the acoustic modes specific to the actual combustor geometry is a phenomenon that can only be reproduced by such a large scale and high-fidelity simulation, and it is expected to be a means of predicting combustion stability before conducting tests.

The world highest performance computer "FUGAKU" has been operated to enable an EXA scale simulation. For applying it to combustion problems, novel approaches of numerical models and methods are expected. Here the latest progress and prospect of their investigation for FUGAKU scale combustion simulation are reported in some topics, such as flow simulation methods by immersed boundary approximation, combustion models by detail reaction and flamelet approaches and a coupling model for spray and flow.

It is necessary to create new technologies more than ever in order to comply with the increasingly strict fuel efficiency and exhaust gas regulations of automobiles. In order to create new technologies, it is necessary to understand the combustion phenomena in the engine in detail and to verify the ideas, and for that purpose, it is effective to utilize numerical simulation. Even in recent years, CPU and network performance have been continuously improved and the prices have been declining, making it easier to use high-speed computers for simulation. Currently, quasi-steady CAE is being actively used in product research and development by each company, but in the future it will be necessary to handle unsteady and transient phenomena that take place in reality, and for that purpose, further computer performance is needed. In such a situation, Fugaku, the world-wide highest-end computer started its operation in 2020, and it has the potential to perform calculations that were previously impossible. We analyze engine combustion on Fugaku using HINOCA, which was constructed in "SIP innovative combustion technology" and has been evolved at AICE (the research association of Automotive Internal Combustion Engines), and success in obtaining results that could not be obtained with conventional computers. By evaluating them, we plan to explore new ways to utilize next-generation HPC (High Performance Computing) that will contribute to the realization of carbon neutrality. In this paper, we will explain the application contents for Fugaku usage, the outline of HINOCA, and the plan of concrete calculation contents such as cycle-to-cycle variation calculation and PM/PN calculation.

Numerical simulations of combustion fields are indispensable recently not only for the understanding of combustion physics but also for the efficient development and optimal design of combustion devices. Because of the lack of computing power and the immature state of turbulent combustion models, however, the accuracy of the present numerical simulations of combustion fields is insufficient. It is strongly expected that the Supercomputer Fugaku, which has maintained its place at the top of the high-performance computing world for the fourth consecutive term, will be great use to researches in the combustion community. One of the most important issues in such combustion researches is the prediction and suppression of combustion instability, which often induces flashback, generates combustion noise and damages combustors. In this article, our recent work on the combustion instabilities by large-eddy simulation (LES) using the Fugaku is introduced, together with some other combustion researches related to the use of High Performance Computing (HPC) systems.

The multi-physics and multi-scale simulation of thermal fluid - structure interaction on an oxy-fuel coal gasifier based on a large-eddy simulation of coal gasification and a heat conduction simulation on coal gasifier pressure vessel is described in this report. The LES employing the Eulerian-Lagrangian manner to describe the motion of coal particles was validated with comparing with the experimental data related to the semi-closed carbon capture system and it was revealed that the LES could quantitatively capture the feature of the coal gasification characteristics. Moreover, the LES coupled with the VOF method to simulate the motion of the high viscosity molten slag layer was also performed. This gas-solid-liquid three-phase reacting flow simulation is expected to be a reliable technique to optimize the gasification conditions for stabilizing the operation. The thermal fluid - structure interaction simulation strategy was developed precisely and its performance on the large-scale parallel computing was evaluated. The multi-phase and multi-scale simulation has a great potential as a digital twin to assess a practical-scale clean energy system.

An ammonia coaxial jet diffusion flame stabilized with hydrogen has been investigated. It has been demonstrated that the method of this type of burner maintains the NH₃ flame stably even when the coaxial high-speed air flows about 10 m/s. In this study, the flame structure and NOx formation/reduction mechanism were investigated by experiments and calculations. The calculations were performed using three schemes, and the best scheme was selected by comparing the calculation and experimental results. As a result, (1) the results obtained using the CRECK-Mech scheme well agreed with the experimental results of the flame structure and NOx emission level. (2) The formation zone of NOx was determined by separating thermal NOx and fuel NOx. Thermal NOx was only formed near the hydrogen flame close to the burner rim at high temperature region. The amount of thermal NOx was considerably smaller than that of fuel NOx. (3) When OH and H radicals were supplied by turbulent, a large amount of NOx was induced. Concurrently, the mixing of NOx and NH₃ and the reduction reaction were promoted, consequently reducing NO to N₂. The reduction reaction of NOx is found to intensely generate on the slightly inner side of the high temperature region (i.e., NH₃-fuel rich region), thus inhibiting NOx increase. Therefore, arranging the reduction zone where NO can sufficiently react with N, NH, and NH₂ is important in burner design. In addition, NOx formation/reduction mechanism in the NH₃/H₂ combustion was clarified.

To provide a better understanding of the ignition process using nanosecond repetitively pulsed discharges (NRPDs), ignition trials of quiescent lean propane-air mixtures (equivalence ratio of 0.7) in a constant volume combustion chamber were conducted. Both the post-discharge and the following flame kernel development were visualized and analyzed by high frame rate imaging of the chemiluminescence. NRPD produces a gas motion resulting in a jetting phenomenon and that influence on the time of flame kernel development. For fixed inter-electrode gap and constant total energy, the ignition time was minimized at a certain pulsation frequency, with higher pulse repetition ratio being detrimental and resulting in longer ignition delay times. The effect of energy deposition, pulse repetition frequency (PRF), and inter-electrode gap was attributed to the competition between characteristic recirculation time from the discharge-induced flow field and the inter-pulse time.