Prediction of Fuel Maps in Spark Ignited Gasoline Engines using Kriging Metamodels 2020-01-0744
Creating a fuel map for engine simulation can be computationally demanding. Design of Experiments (DoE) and metamodeling is one way to address this issue. In this paper, we introduce a sequential process to generate a Kriging metamodel of an engine fuel map which accounts for different engine characteristics such as load and fuel consumption at different operating conditions. The generated metamodel predicts engine output parameters such as fuel rate and load. We first create a metamodel to accurately predict the engine output parameters. The predicted fuel map is then filtered based on constraints to obtain the optimum fuel map. The latter is then compared with the actual optimum map created using a full factorial DoE. The results show that the created fuel map metamodel is of high accuracy compared to the actual map. The fuel map metamodel is obtained with about one tenth of the number of engine simulations for a full factorial design, leading to much faster predictions.