Limits of mixture dilution in gas engines

Abstract

Natural gas find application in transport as well as for stationary power generation. These have a lower efficiency compared to the most widely used power plant, the diesel engine, however running on natural gas also have some distinct advantages. Gas that are optimized to the same level as their diesel counterparts can achieve lower specific CO2 emissions due to the higher H/C ratio of the fuel. This is of great interest since CO2 is a greenhouse gas. A further motivation for the use of natural gas as a fuel is the need for diversification of fuel supply. Heavy duty gas in particular are predominantly used with a fuel mixture. This mixture dilution can be done with either excess air, these are called lean burn engines or with recycled exhaust gases, called lambda 1 + EGR engines. When using mixture dilution at low engine load, engine efficiency is increased by reducing throttling losses. Due to the mixture dilution, peak combustion temperatures decrease and the formation of NO is reduced. Mixture dilution however can only be applied to a certain extent. When the mixture is diluted excessively, the emission of unburned hydrocarbons increases sharply due to engine misfires, partial burns and other combustion instabilities. When the mixture is diluted, its burning velocity decreases. Lean burn therefore typically apply a combustion system (combustion chamber shape) which increases turbulence to boost the overall burning rate. The generally accepted 'model' of the combustion process is one where flames locally propagate with the laminar burning velocity and where the overall burning rate can be found by multiplying the flame area with the burning velocity. Developing dedicated gas with help of engine simulation codes (e.g. GT-power, AVL Boost, Ricardo Wave and others) requires knowledge on the flame propagation process to predict the burning rate or heat release. Generally a basic assumption of a spherical flame front which propagates outward from the spark plug is used. The propagation speed of the flame front (relative to the unburned mixture) or turbulent burning velocity, ST, in that case is taken proportional to the turbulent velocity fluctuation which, in turn, is assumed to scale with engine speed. The underlying assumption here is again that combustion occurs in flamelets that locally propagate with the laminar burning velocity and that flame surface is created by turbulent velocity fluctuations. For mixtures however this assumption is less validated. The predictive capabilities of engine simulation codes for burn are therefore less. In the study presented in this thesis, the limits of combustion in a heavy-duty gas engine were explored with the aim of increasing understanding of the combustion process of these mixtures. This knowledge can then be used to improve the predictive capabilities of engine modeling codes. The questions to be answered here were 1. Does the turbulent burning velocity still scale with engine speed for very mixtures ? 2. Is the combustion regime that is applicable for very mixtures is still 'classical' flame propagation ? 3. Is the assumption of spherical propagation justified for mixtures in gas ? The investigation was split in a number of separate parts which targeted different aspects of the combustion process and its interaction with the flow phenomena. A single cylinder test engine featuring optical access to the combustion chamber was designed and constructed for this purpose. Engine tests in the same engine, only with the transparent parts replaced by full metal ones, were conducted to explore the limits of its operating range. The heat release and mass burning rate were determined from in-cylinder pressure measurements. The turbulent burning velocity ST was determined using the classical assumption of a spherically propagating flame front. The laminar burning velocity SL for the same mixtures was measured using the technique of constant volume combustion in the EHPC (Eindhoven High Pressure Cell). This resulted in a correlation for the laminar burning velocity SL as function of pressure, temperature and mixture equivalence ratio. A similar, but less extensive correlation was determined for mixtures diluted with inert components. The diluent in this case was a mixture of nitrogen and carbon dioxide with similar thermal properties as recycled exhaust gas for a stoichiometric gas engine. These correlations were used to determine the ratio of turbulent to laminar burning velocity for various engine operating conditions. It was shown that the burning velocity ratio for mixtures scales approximately with engine speed, confirming the general model. To evaluate the applicable combustion regime for mixtures, both the process of flame propagation and also the properties of the turbulent flow inside the engine cylinder were investigated. The applicable combustion regime can be derived from typical velocities and spatial scales of the combustion process. The flow inside the engine cylinder was measured using Particle Image Velocimetry. To enable this, the engine was equipped with a window in the piston crown and also the upper part of the engine liner was replaced by a transparent piece. The turbulent velocity fluctuations were determined from ensembles of images recorded at the same crank position but in different engine cycles. The cycle-to-cycle fluctuation was separated from the in-cycle turbulent fluctuations using a spatial filtering technique. Spatial autocorrelation functions were used to determine the integral length scale of the flow. The applicable combustion regime was demonstrated in a Borghi diagram that was constructed from the laminar burning velocity SL, the turbulent velocity fluctuation u', the integral length scale of the flow and an estimate for the flame thickness. The combustion in this engine was typically found in the 'corrugated flamelets regime' for stoichiometric mixtures. For the combustion process of very mixtures the 'thin reaction zones regime' was more applicable. This finding confirmed that combustion in a gas engine for these mixtures still occurs in the form of 'classical' flame propagation. The experiments with optical access were then extended with visualizations of the combustion process using Mie scattering on the same oil droplets that were used for the velocity measurements. Since oil droplets are not visible in the burned zone, the contour of a two-dimensional slice of the inflamed volume could be made visible. These experiments showed that the assumption of a spherically propagating flame is largely valid for stoichiometric mixtures. For mixtures however it was found that the shape of the inflamed volume is far from spherical. For many images islands of unburned mixture were found in the burned zone and islands and peninsulas of burned products were seen in the unburned mixture ahead of the flame. This indicates a very irregular, highly corrugated three dimensional structure. The contour of the inflamed zone was investigated further with help of fractal analysis. An attempt was made to quantify the flame surface from the images and to compare this to the surface needed to justify the mass burning rate found from the pressure signal analysis and laminar burning velocity. Here it was found that the surface determined from the fractal analysis was a factor 2-5 too small. This was attributed to the limited information on the three dimensional structure of the inflamed volume and also to the fact that deformations larger than the integral length scale of the flow were not accounted for by the fractal analysis.