Skeletal Model Reduction with Forced Optimally Time Dependent Modes

Abstract: Sensitivity analysis with forced optimally time dependent (f-OTD) modes is introduced and its application for skeletal model reduction is demonstrated. f-OTD expands the sensitivity coefficient matrix into a lowdimensional, time dependent, orthonormal basis which captures directions of the phase space associated with most dominant sensitivities. These directions highlight the instantaneous active species and reaction paths. Evolution equations for the orthonormal basis and the projections of sensitivity matrix onto the basis are derived, and the application of f-OTD for skeletal reduction is described. In this framework, the sensitivity matrix is modeled, stored in a factorized manner, and never reconstructed at any time during the calculations. For demonstration purposes, sensitivity analysis is conducted of constant pressure ethylene-air burning in a zero-dimensional reactor and new skeletal models are generated. The flame speed, the ignition delay, and the extinction curve of resulted models are compared against some of the existing skeletal models. The results demonstrate the ability of f-OTD approach to eliminate unimportant reactions and species in a systematic, efficient and accurate manner.