Iterative approach in model reduction and its validation for laminar and turbulent combustion.
Detailed chemistry based simulations are computationally expensive especially in the industrial scale combustion systems. Model reduction techniques provide an advantage to include the detailed chemistry effects with a reduced computational cost. Generally, in model reduction techniques, a prior knowledge about the combustion system is necessary. On contrary, REDIM generation needs a minimum knowledge about the combustion system or the information about the combustion system can be provided through the gradient estimates.
In this work, the focus is to develop an iterative methodology to provide gradient estimates in a generalized way so that the REDIM can be iteratively converged very close to the detailed manifold according to the gradients of the reduced coordinates provided by the physical combustion system in the each iteration step. The iterative methodology provides the automatic evolution of REDIM according to the diffusion-related (gradients) information received from a real physical system with minimum a-priori knowledge requirements about the system configuration. Additionally, in this iterative procedure, REDIM reduced calculations do not provide gradient information within the whole physical boundary of system states. Therefore, a gradient estimate method is developed to extrapolate the gradient data on those REDIM mesh points where scattered gradient data is not available.
Use the advantage of REDIM where gradient information can be provided in a generic way so that the manifold can evolve accordingly. A generalised iterative methodology is devised where gradient information is provided by the REDIM reduced CFD calculations itself.
To answer the question how the scattered gradient information from reduced CFD calculations can be mapped on the REDIM mesh points. For this, a new gradient estimate method to map the scatter gradient data on REDIM mesh is developed.
Examine the iterative methodology in laminar and turbulent combustion especially in partially-premixed combustion where the simulation of mixed-mode based combustion is often challenging using model reduction techniques.