State-of-the-art techniques of modeling of glass melting furnaces give a time averaged solution of the whole (coupled) model. This method seems to be satisfactory to obtain accurate heat fluxes between the glass and combusiton space, and thus to have a good boundary condition for the glass model to evaluate al the values of interest, such as glass flow patterns, temperatures, batch shape, quality indices, seed counts, and so on. However, in modeling combusiton chambers of regenerative furnaces, we can see processes that, because of their nature, are time-dependent and cannot be represented by a time-averaged steady state. This paper presents results of simulations of two designs of a regenerator. Some drawbacks of traditional regenerator designs are identified and improvements suggested.
