This paper introduces an integrated approach for the reduction of large chemical kinetic mechanisms. This approach consists of three main steps. In the first step, the DRG (directed relation graph with species- specific error threshold) method is used to efficiently eliminate redundant species. After that, the PCA (principal component analysis of sensitivity matrix) approach is applied to further remove unnecessary reactions. During the reduction process, different improved error functions are defined to monitor the performance of the generated mechanism by checking its capability to satisfy the requested error tolerance. In the third step, a single-objective, GA (genetic algorithm)-based optimization procedure is established for the calibration of the Arrhenius parameters of the selected reactions to have the mechanism fit the desired experimental data. Application of the integrated reduction approach to a detailed mechanism that includes 1083 species and 4635 reactions for a tetra-component gasoline surrogate mixture, consisting of n-heptane, iso-octane, toluene and 1-hexene, is presented, and a reduced TRF (toluene reference fuel) mechanism with 164 species and 839 reactions is eventually obtained. Finally, the derived reduced TRF mechanism is validated for some typical combustion applications: the spatially homogeneous constant-volume auto- ignition, perfectly stirred reactor (PSR), variable pressure flow reactor (VPFR), 1-D planar freely propagating laminar premixed flame and HCCI (homogeneous charge compression ignition) and DICI (direct-injection compression ignition) engine combustion. Generally, in spite of the reduced size, our TRF mechanism exhibits satisfying capability in predicting not only global parameters, such as ignition delay time and extinction residence time, but also detailed profiles of species concentration, including both major fuel components and some key intermediate radicals. Moreover, numerical simulation of HCCI and DICI engine combustion are compared with experimental results, and generally good agreement is observed. In a word, the proposed mechanism reduction approach successfully reduce the original detailed mechanism of gasoline surrogate to a reduced TRF mechanism, while still maintaining comparatively high prediction accuracy compared to the original detailed mechanism
