A Neural Network Based Approach for the Intake Air Mass Flow Prediction in SI Engines

The correct measurement of the intake air mass flow is fundamental for the Spark Ignition (SI) Engines to guarantee an efficient balancing with the fuel one. In the aeronautical field, this measurement is essential to ensure the correct functioning of engines based on the aircraft's altitude. The technological growth of SI engines coupled with advanced control and estimation techniques improved engines in terms of fuel consumption and pollution emission reductions, increasing their performances. Typically, model-based estimation techniques have been employed for the manifold air pressure (MAP) and flow (MAF) virtual measurements taking into account the two principal approaches for MAF determination in engine control units called speed-density and alpha-speed. Furthermore, Neural Networks can be employed to predict these variables, avoiding the presence of unmodeled dynamics in model-based approaches. In this work, a methodology based on Feedforward Neural Networks (FNNs) to predict MAP and MAF is presented. The developed Networks are able to predict these two fundamental operative variables of SI engines in transient conditions after training based on steady-state data obtained through a well-known intake manifold dynamical model (IMDM). The proposed approach allows the costs reduction related to expensive experimental tests and, virtual sensors based on FNNs can be employed as redundancies in measurements.