AFR Of Engines Model Predict Control Based On NARMAX And RBF Neural Network

With the increasingly improvement of people’s life, cars have become the indispensabletraffic tools in today’s world, but that is inconsistent with the depletion of energy, seriouspollution and people’s growing environmental awareness. Therefore, how to achieve theprecise control of engine air-fuel ratio (AFR) has become a hot topic in the academic world.Traditional ways of air-fuel ratio control was established with a large number ofexperiments，while it cannot control the air-fuel ratio accurately when the engine work in thetransient engine operating conditions. So in this paper, we have made up an identificationmodel with nonlinear prediction method to research the engine control, on the basis of thein-depth study with the AFR of Spark ignition (Spark ignition, SI) engine. Finally, we findout an engine control method with high control precision, strong robustness, adaptive abilityand is suitable for engineering application simultaneously. Compared with the PI controlmethod which is used before, the new method which combines the RBF with NARMAXmodel adopted by this paper can improve the AFR control precision of the SI enginesignificantly.Using RBF (Radial basis function, RBF) neural network model to identify the SI engineAFR system and adopting random amplitude signal as excitation signal can make the engineshow all the dynamic characteristics. Utilizing the fading memory RLS method to realize theparameters self-adaption of the engine, can also set the model parameters according to theexisting situation constantly. Thus it makes the model have a series of advantages such as asmaller amount of calculation and a higher modeling precision.The AFR systems of SI engine is modeled and analyzed by using NARMAX model(Nonlinear autoregressive moving average with exogenous input, NARMAX). We found thatthe second-order NARMAX model can realize the identification of the engine model byoptimizing the model. At the same time, the second order NARMAX model can bedecomposed into the linear part and the nonlinear part. Utilizing the fading memory RLSmethod can also realize online self-adaption of the engine parameters. Meanwhile, the modelprecision is improved effectively by the introduction of output feedback item andinput-output cross-terms in the model structure.According to the characteristics of those models above, this paper puts forward a RBF and NARMAX Joint model. Combining the two models can consummate each other, whichowns small amount of calculation of the RBF and the actual physical meaning of theNARMAX model parameters together. So it can realize the AFR control of the SI enginewith high precision, small amount of calculation and strong robustness etc. Throughsimulation experiments, the results verify that this algorithm is really effective and it canimprove the control precision of the SI engine AFR significantly...