Incorporating CFD to Reduce the Development Time of Low NOx Engine Technologies, 03-R8079Printer Friendly Version
Inclusive Dates: 07/01/08 07/01/09
Background - It is well known that in-cylinder NOx formation is a strong function of combustion temperature, where higher temperatures produce higher amounts of NOx. As such, the primary approach to reducing in-cylinder NOx is to use high levels of exhaust gas recirculation (EGR) to reduce combustion temperatures. At higher temperatures, the dominant NOx formation mechanism is thermal NOx , which is characterized well by the Zeldovich mechanism in virtually all CFD models. However, when high levels of EGR are used to drive down the combustion temperature, the NOx vs. temperature relationship breaks down and the Zeldovich mechanism no longer accurately predicts NOx formation. At these lower temperatures, different mechanisms for NOx formation are at work. With respect to the reduction of the NOx emissions post-cylinder, two leading NOx aftertreatment systems are lean NOx traps (LNT) and selective catalytic reduction (SCR) systems. SCR systems are of primary interest to the transportation industry because when properly implemented they offer high levels of NOx reduction with no adverse effect on fuel economy. CFD can be used to simulate the SCR system and predict the effect of design variables on the NOx conversion and ammonia slip. The objectives were to: improve the accuracy of in-cylinder NOx prediction under high EGR conditions by improving the NOx model; reduce the simulation time of SCR systems by implementing 0-D model for NOx and ammonia predictions in place of computationally intensive detailed chemistry.
Approach - A CFD model of a combustion chamber from a Caterpillar C15 engine was used. The approach for the in-cylinder NOx prediction involves CFD simulation using Converge 3-D CFD software. Different NOx formation (called "prompt NOx") models have been implemented to improve the in-cylinder NOx prediction at low temperature. The models are inserted to the Converge software using a user defined function (UDF). In SCR system modeling, the 3-D domain of SCR is substituted by a 0-D SCR model. A coupling between 0-D and 3-D was developed.
Accomplishments - Of several prompt NOx models, it was found that the prompt NOx model of De Soeto shows the best prompt NOx prediction. Based on measured and simulated data, a correction factor is implemented to the model to increase its accuracy, namely 1.086x10-2 for low-end-of-injection pressure and 1.211x10-2 for high end-of-injection pressure. The application of the corrected De Soeto version 1 model in other operating conditions results in maximum error of 4.3 percent. For the SCR model, detailed chemistry process inside the SCR system is substituted by a 0-D empirical model. The research found that the application of 0-D empirical model saves 40 percent of the calculation time.