Neural network approximation of internal-combustion engine characteristics

The most preferable approach to develop mathematical models of complex technical objects is based on experimental data approximation. It is reasonable to use data mining systems, in particular, fuzzy inference systems and artificial neural networks (ANN) as an approximation tool.
The article presents the results of applying a feedforward ANN to developing a mathematical model of an internal combustion engine. The mathematical model is developed through approximating the following basic data: internal combustion engine speed characteristics, efficiency and exhaust toxicity indicators. During computing experiments, the approximation error of the engine characteristics versus the model structure and parameters has been investigated. The developed model allows to solve further problems connected with analysis and optimization of the engine's working processes for specified traction and high-speed modes of vehicles.