Multi-Modelling of Torque and Air/Fuel Ratio Based on Engine Operating Regions

This paper addresses multi-model identification techniques and their applications in engine torque and air/fuel ratio estimation. Mult-models are currently being considered for automotive applications because of their modularity, and ease of understanding and implementation. The novelty in the paper is the application of structure identification processes to determine empirical engine multi-models efficiently and with improved quality. Two different approaches, stepwise regression with F-statistic test and orthogonal least squares with error reduction ratio (ERR) analysis, have been applied to determine the model structure. Based on the input/output measurements from a real-time RT-Wave® virtual engine model, two MISO torque and Air/Fuel Ratio models are developed. The model is then validated by output error for its prediction ability of the virtual engine behaviour.