Adaptive learning engine load estimation

In this paper we design new adaptive learning estimation algorithms for engine load. The flow into the engine is estimated via a speed-density calculation, wherein the intake manifold temperature is estimated on-line. First we present the ideal gas law with multiplicative uncertainty factor which can be associated with intake manifold mass or temperature to describe the pressure in the intake manifold. Using the error between measured and modeled pressure we estimate a prediction error which is used together with the tracking error to design adaptive algorithms with improved identifiability and convergence rate. The tradeoff between the speed of adaptation and the quality of the estimation signal necessitates the robustness enhancement which is achieved by sigma-modification with the sigma factor depending on the prediction error estimate. Then under the transient the estimated parameter converges to its a priori value, but under 'steady-state', sigma -modification is not active and the adjustment law is driven by both tracking and prediction errors. Further improvement is made by introducing an adaptive feedforward part instead of a single a priori value in the algorithm. Learned values of estimated parameters for high and low pressures are stored in a static map during steady-state. Then under subsequent transients learned values are used directly in the feedforward part of the algorithm that allows to reduce the convergence time of estimated parameters.