EFI controllers approximate complex, non-linear relationships between fuelling parameters, engine states and ambient conditions using linked tables or 'maps'. Determination of a single fuelling value generally requires interrogation of a number of maps and linear interpolation to determine intermediate values. Some types of Artificial Neural Network (NN) can model multi-dimensional arbitrary functions to any desired degree of accuracy, and interpolate smoothly. This paper demonstrates how Multi-Layer Perceptrons can approximate fuelling maps and trim tables, to produce smooth surfaces that model the data with good interpolation performance.
Neural networks - potential for enhanced control of automotive electronic fuel injection systems
Neuronale Netze bieten Potential für eine verbesserte Regelung elektronischer Kraftstoffeinspritzsysteme
1999
14 Seiten, 16 Bilder, 16 Quellen
Aufsatz (Konferenz)
Englisch
Neural networks - potential of enhanced control of automotive electronic fuel injection systems
Kraftfahrwesen | 1999
|Neural Network Control of Automotive Fuel-Injection Systems
British Library Online Contents | 1994
|Neural network control of automotive fuel-injection systems
Tema Archiv | 1994
|Electronic fuel injection reduces automotive pollution
Tema Archiv | 1972
|Progress in electronic fuel injection control systems for automotive diesel engines
Kraftfahrwesen | 1986
|