Improving identification of combustion engine to use in model-based calibration

The calibration of modern internal combustion engines is complex and requires a lot of time and cost on the test bench. A large number of calibration parameters and a tradeoff between fuel consumption and emissions make calibration a complex multi-objective optimization problem. The purpose of this article is the development of calibration and model-based optimization techniques for internal combustion engines to reduce calibration time and cost and improve optimization accuracy. This paper focuses on empirical modeling of engines and optimization concerning emissions standards and fuel consumption. To identify the combustion models, a modeling toolbox is developed with an intelligent identification method. To optimize the data collection, the design of experiment methods is reviewed and appropriate methods are selected to collect information with the minimum data from all over the design space. Finally, a study is performed on the EF7 engine in the test room of IPCO and it is shown that the number of experimental data is reduced from 5500 data to 1500 data with the help of the design of experiment by Sobol method and modeling by a deep neural network. so, the virtual engine can replace the real engine in the calibration process.