Fuel lean premixed combustion has potentials to achieve high efficiency and low emissions, however ignitability of lean mixture and, flame stability and controlling in this combustion are significant issues to be addressed. The main issues for lean burn intermittent combustion engines  are (i) the mixture preparation for lean combustion requires expensive or premium technology and (ii) achieving this combustion over a wide range of load and speed is difficult for a smooth running of the engine. Various combustion technologies such as PCCI or HCCI, DICI, DISI, etc., are being explored by automotive OEMs. The common objective of these various concepts is to operate the engine in diesel cycle for high efficiency with premixed charges prepared inside the cylinder for low emission levels. These bring in a number of technological challenges, specifically to achieve PCCI or HCCI combustion because the combustion onset is controlled by chemical kinetics induced autoignition phenomena, which is strongly influenced by in-cylinder turbulence, charge temperature and pressure. Thus, keeping the engine at its desired performance level over a range of loads is of significant challenge and methods are yet to be devised for this.
Computer aided engineering constitutes a vital part in advanced manufacturing techniques. It allows for quantum leaps in precision and cost saving. Engines development have long benefited from computational modelling, in fact the computational fluid dynamics (CFD) applied to turbulent combustion plays an important role to address the challenges of lean combustion for automotive engines. Nonetheless, the efficacy of this approach depends strongly on the accuracy, fidelity and robustness of the combustion models used in the simulations. A combustion model with the above attributes must also be simple to implement and use in a multi-dimensional CFD code to answer what if? type questions during the engine development programme. Also, the combustion model must have the ability to represent the change in the physics that could arise due to changes in the operational parameters, fuel type or fuel-blend. For example, a change in the engine load can lead to a change in the combustion characteristics, which may result in different effects of competing thermochemical and turbulence processes. These processes and their interaction are nonlinear and thus it is not easy to achieve the above desired attributes of combustion sub-models unless the model parameters are tied closely to the physics of the problem rather than treating them as empirical or semi-empirical parameters in the common usual way. This physics-based modelling approach is expected to be robust and accurate, and one such modelling method has been developed at CUED, which has been applied to a number of continuous combustion systems. Initial application of this method to spark-ignited gasoline engine is encouraging . It is hoped that this modelling approach may evolve in due course to be a ``truly predictive'' method, which is very much required for the development of future engines, because an often-stated aim of CFD simulations is to minimise or even attempt to eliminate the cut-and-try design process .
The common methods used to model in-cylinder combustion include G-equation, FSD (flame surface density), BML (Bray-Moss-Libby) and EBU (eddy break up) approaches. These approaches are reviewed in  and invariably presume that the local Damkohler number is very large and thus the flame can be treated as a thin surface. The in-cylinder pressure variation with crank angle computed using these combustion models in RANS (Reynolds-averaged-Navier-Stokes) [5-16] and LES (large eddy simulation) [17,18] approaches agreed quite well with the experimental measurements in HCSI [5-9, 11], HCCI [14, 15,19], GDISI [10,13,16] engines. The GDISI engines can have wall guided and spray guided fuel injection and involve partially premixed or stratified charge combustion depending on the spark advance angle. Some of the above studies have also attempted to address this mixed mode combustion, but the coupling between mixing and chemical reaction, which plays an important role in partially premixed charge combustion, has been ignored largely. The HCCI condition can be envisaged for both PFI (port fuel injection) and DI (direct injection) of gasoline or a blended fuel. Despite a good agreement between measured in-cylinder pressure vs crank angle variation and its computed values using these combustion models, some amount of model tuning was involved to achieve the desired level of agreement. It is unclear if these models with the same model parameters would work for another set of operating conditions . However, these models performances are acceptable for PFI-SI engine combustion since the in-cylinder Damkohler number is large resulting in flamelet combustion. The basic premises of these models become questionable for fuel-lean combustion conditions relevant for PCCI engines.
The bulk of these modelling studies used Unsteady RANS methodology. There are some studies using LES methodology as well. The choice between the URANS and LES for in-cylinder reacting flow calculation strongly depends on the objectives of a particular study. If one is interested in the cycle-to-cycle variation then LES is the only choice but if the interest is only on the statistics then URANS is adequate. Also, to construct these statistics (averaged over many cycles), LES must be run for many (at least few tens of) cycles to have meaningful comparison with experiments since the relationship between the resolved field and the underlying turbulent field is statistical. Obviously, running LES for many cycles will incur heavy computational cost. Furthermore, in combusting in-cylinder flows with premixed and partially premixed reactants the essential rate controlling processes such as chemical reactions and molecular diffusion occur at small scales, which are to be modelled using sub-grid scales. Thus, favouring LES over URANS must be evaluated carefully. If the large-scale flow dynamics becomes the rate controlling process (for example for cycle-to-cycle variation) then LES will offer greater advantages.