A unified framework for state and time-dependent parameter estimation of automotive engines

Engine calibration and control are important aspects of effective engine performance while optimizing engine efficiency, fuel consumption, and exhaust gases. This requires the knowledge of the parameters and states of highly nonlinear models that capture the overall dynamics of engines over the lifespan of a vehicle. Combined state and parameter estimation techniques for a nonlinear engine model have not been given much attention in the literature. Recently, a methodology based on the unscented Kalman filter and recursive least square has been reported for the combined state and parameter estimation for spark-ignition engines. However, the reported methodology is sensitive to the initial state covariance matrix. To address this issue, a new unified strategy is proposed that eliminates the careful initialization of the state covariance matrix. The performance of the proposed strategy is analysed for a nonlinear mean value spark-ignition engine model consisting of the throttle system, intake manifold system, engine speed dynamics, and fuel system. The robustness of the proposed algorithm to random internal and external noises is reported through simulation analysis of different input–output sets. © 2024 Elsevier Ltd