Development of a Neural Network Model of an SCR Catalytic Converter and Ammonia Dosing Optimization Using Multi Objective Genetic Algorithm 2011-01-1332
In this paper, a mathematical model of the SCR catalytic converter is replaced with the neural network model to accelerate the optimization process. The Euro steady state calibration test data set is used to simulate the inlet properties of the SCR catalytic converter. For each chosen condition, a separate neural network is developed. In order to generate sufficient data to form a neural network for each condition, the original mathematical model was run several times at the temperature and inlet NOx concentration of each condition with a range of different ammonia concentrations. Subsequently, using MATLAB® software, the neural network model is trained and tested for each condition. Ammonia dosing optimization is performed using multi objective genetic algorithm module of MATLAB®. The optimization objectives are NOx reduction percentage and the outlet ammonia concentration of the SCR catalytic converter. It is convenient that the NOx is reduced as much as possible while ammonia concentration does not exceed 25 ppm.
Citation: Majd Faghihi, E. and Shamekhi, A., "Development of a Neural Network Model of an SCR Catalytic Converter and Ammonia Dosing Optimization Using Multi Objective Genetic Algorithm," SAE Technical Paper 2011-01-1332, 2011, https://doi.org/10.4271/2011-01-1332. Download Citation
Ehsan Majd Faghihi, Amir H. Shamekhi
K N Toosi Univ. of Technology
SAE 2011 World Congress & Exhibition
Diesel Exhaust Emission Control, 2011-SP-2318