A manual adaptation of the engine control becomes more and more difficult due to the rising number of engine control settings. Alternatively, model-based approaches can be used. However, these are rather demanding especially if the emissions themselves are to be considered. Here, dynamic neural networks are proposed which allow the application of very fast measuring techniques at engine test stands. The calculation of the static control maps is then done offline by means of the dynamic models. This approach allows an adaptation to any given legal test cycle.
Print ISSN: 0178-2312
Volume: 51, 05/2003
Pages: 213