Model based control and efficient calibration for crank-to-run transition in SI engines

The latest emission regulations demand drastic reduction of tailpipe hydrocarbon emission of Spark Ignited, Port Fuel Injected (SI PFI) automotive engines during crank-to-run transition. This results in sharply increased calibration effort and associated cost based on the current control architecture of production engines, during this extremely fast transient operating regime under the constraints of substantial emission reduction and good engine startability.; A new approach using model based control is proposed for this problem. The engine start behavior and overall fuel dynamics characterization during crank-to-run transition are thoroughly investigated. A scheduled in-cylinder fresh air charge predictor is constructed. A nonlinear input correction function, invoked prior to engaging fuel dynamics control, is developed. In order to cover a wide range of engine coolant temperatures by means of scheduling, the linear spline modeling technique is applied to air and fuel dynamics modeling, identification and control design. A new criteria, from the class of subspace methods, is introduced to evaluate system order and model quality.; Finally, a predictive fuel dynamics control scheme is realized to overcome individual cylinder fuel dynamics effect by systematically combining the scheduled in-cylinder fresh air charge predictor, the direct inversion of a fuel dynamics model and an inverse correction function. By means of an intelligent mode scheduling of the in-cylinder fresh air charge predictor with misfire and poor-start detection, a fault tolerant predictive fuel dynamics control results. It has been demonstrated that the calibration effort of start fuel control during engine start and crank-to-run transition is reduced significantly for production inline-4 cylinder engines. In addition to solving this practical problem using model based control, this dissertation research also raises several theoretical questions worthy of further research.; In summary, this dissertation research makes following contributions: (1) individual in-cylinder fresh air charge prediction and individual cylinder fuel dynamics compensation; (2) accommodated misfire and poor-start; (3) significantly reduced calibration effort; (4) introduction of the Linear Parameter Varying Linear Splines (LPV-LSP) technique for gain scheduled modeling, identification and control design; and (5) introduction of a novel method to identify ARMA models using subspace methods.