Individual Cylinder Air-fuel Ratio Balancing Control Based On Genetic Algorithm Identification

Individual cylinder air-fuel ratio balancing control can effectively reduce the fuelconsumption, exhaust emission and improve maximum output torque in multicylinder gasoline engine. But existing balancing control methods are complex andcostly which need to install a liner oxygen sensor(UEGO) in every exhaust portduring parameter identify process. This will also affect exhaust flow then cause theparament differ from normal engine. In order to avoid those drawback, this paperproposed a method which need only one UEGO to finish the identify experiment anddid following works:First, after analysised the existing time domain modeling method, this paperbuilded the production and transmission model of individual cylinder AFR variationwhich made it possible to identify the parameter using only one UEGO.Then, the AFR variation model was rewrited and the unknow parameters werereduced to14in order to simplify the identify process. And a steady multi-step inputmethod was created to handle the disturbance which may existing in the experiment.The identify data was gotten from the simulation combine engine model in GT-Powerand control model in Simulink.After that, in order to solving this identify problem which Least Square methodcan not be used, this problem was transformed into a optimization problem whichseeking the minimum value of a constrained multivariate function, so that the geneticalgorithm can be used to search the answer. And a lot of modifications was made toimprove the adaptability of basic genetic algorithm, for example proposed multi-targetevaluation function, modified crossing operator, small habitat competition, adaptivevariation, chaotic local search and so on. And the identify result was pretty good.Finally, the individual cylinder air-fuel ratio balancing control model was builtbased on Kalman filter theory using the parameters gotten form the identify process.Simulation was performed by the model constructed in Simulink and combined withthe engine model in GT-Power. Sampleing noise was plused in to simulate realexperiment. Controlling result showed that the AFR variation among differentcylinders can be reduced from±10%to±1%in2seconds by the control methodproposed in this paper. And the result also proved the effectiveness of the geneticalgorithm based identify method.