The development of PEMS measurement techniques facilitates research into the environmental friendliness of vehicles under real operating conditions. The interpretation of measurement results, such as pollutant emissions, represented by very large sets of diverse data, requires complex numerical analyses. Sophisticated statistical methods are effective, but the interpretation of results requires the participation of an expert with highly specialized knowledge. The use of data mining methods creates broad perspectives and significantly increases the possibilities for the analysis and interpretation of experimental results. In this article, research was conducted on the transformation of a large training set obtained from measurements into a small set of Kohonen network neurons. For the trained Kohonen network, studies were conducted on recognizing specific vehicle state patterns measured during a road experiment.PEMS measured pollution emissions pattern recognition using Self-Organizing Maps (SOM)
Abstract
The improvement of PEMS enables better estimation of ecological quality of vehicles. However, interpretation of divers sort of data, e.g. exhaust emissions, gathered during on road testing requires complex numerical analyses. Advanced statistical analyses have already proven their applicability in this domain but the interpretation of the results requires the participation of experts with extended knowledge. The usage of the data mining methods brings the perspective in the ease of interpretation of obtained results.
In this article, the Self-Organizing Maps were used to transform the multiplicity of data gathered during the on road experiment into the reduced set of representative data. Trained SOM was tested in the recognition of the vehicle states measured during the on road experiment. (…)
The article contains 2267 characters.
Source: Logistyka Magazine 4/2012