Flame Spectrum Analysis with the Use of Artificial Neural Networks

Evaluation of the air excess ratio (lambda) in the combustion chamber can be done on the base of simplified flame spectrum analysis. Although the obtained results are valuable for the better understanding of the process, they do not let to determine the air excess ratio in different operation conditions. The research limitations have appeared due to incomprehension of chemical and physical aspects of combustion process. Light intensity during the combustion is influenced by several parameters such as engine load, rotational speed, ignition timing, temperature etc. Thus, it is extremely difficult to develop an accurate measurement method based on evaluation of radicals' emission ratio.

Artificial neural networks (ANN) can be used in order to derive air excess ratio from time-domain courses of light intensity signals measured in chosen wavelengths. This paper presents results of ANN application for different ways of evaluation of air excess ratio on the base of simplified spectrum analysis. This approach provides a possibility to obtain results that are independent of engine operation conditions.