Experimental evaluation of reduced kinetic models for the simulation of knock in SI engines

Downsizing by turbo charging is a current approach for the reduction of fuel consumption of Spark Ignition (SI) engines. For downsized engines compression ratio has to be set as high as possible to achieve substantial gains in thermodynamic efficiency. Unfortunately, the possibility to take full advantages offered by downsizing is limited by knock phenomenon, which imposes constraints both on supercharging and compression ratios. Quasi-dimensional and multidimensional simulation can play a role of increasing importance for the design and the optimization of future engine prototypes more and more based on advanced combustion concepts, provided that well proven tools for knock simulation may be available. In this regard, a number of detailed and semi-detailed kinetic schemes have been recently proposed to simulate the auto-ignition and combustion in wide ranges of temperature, pressure and air fuel ratio typical of different experimental approaches, such as: flow reactors, constant volume bombs, rapid compression machines, shock tubes, test engines. However, at moment, the use of large kinetic models is limited, particularly in multidimensional simulation, because of the enormous calculation effort required. On the other hand, different kinds of reduced models (skeletal, global, etc.) have been proposed, but they can be effectively used only on defined ranges of temperature, pressure and air fuel ratio. Thus, to set up reduced models, proper experiments are required for each field of interest. In this scenario, an experimental procedure to evaluate the auto-ignition behavior of different fuels in conditions similar to the ones of the end gas of SI engines is proposed in this paper. A CFR engine was used because of his flexibility and of his wide diffusion. For all tests, carried out using iso-octane as fuel, inlet temperature was controlled at 423 K, the engine speed set at 900 (rpm), and the relative air/fuel ratio varied in a wide range, from rich (0.74) to lean (1.51). For each test condition, the engine was motored and compression ratio was varied until auto-ignition was induced. The internal EGR and other not measurable parameters were estimated with the aid of a 1-D commercial code, whose user combustion module was customized with a FORTRAN routine developed by the authors. Numerical simulations were carried out using two reduced mechanisms due respectively to Tanaka [13] and to Golovitchev [16], and their capability of predicting auto-ignition was evaluated by comparison with experiments.