Powertrain modeling and engine torque estimation using nonlinear observers.

摘要

The need for improvements in the on-line estimation of vehicle performance variables is greater nowadays as a result of more stringent emission control legislation and performance requirements. Two model-based Sliding Mode engine load torque observers, with and without integrators, are developed to improve on-board diagnostics to detect different types of malfunctions without additional expensive sensor installations. Along with the observers, an engine model and a transmission system model are also derived.; An engine model with time-varying inertia is developed using the Lagrangian approach. This engine model is capable of compensating for the velocity-related inertia torque and calculating more accurate engine performance estimations, especially for engines operating at high speeds. For constant inertia engine models, a methodology using the least squares method to compute an optimal constant inertia value is proposed. The advantages and disadvantages of both methods: the least squares method and the traditional average method are discussed. A transmission system model for the Ford Automatic Overdrive (AOD-E) system is developed using the Newtonian approach. The transmission model incorporates all rotating inertial components, including a Ravigneaux planetary gear set, a hydraulic control system, and the torque converter assembly with a direct-drive lock-up clutch. The transmission system model is capable of simulating both transient and steady-state operations.; By combining the engine and the transmission system models, two Sliding Mode engine load torque observers, with and without integrators, are developed and tested through computer simulations. The Sliding Mode observer is known for its ability to compensate noises as well as modeling errors. Several possible modeling error sources of the observers are discussed. An algorithm for selecting the linear gain and the switching gain matrices for the Sliding Mode observers are presented. The stability and robustness criteria of the observers are also presented. Moreover, the effects of switching gain changes on the observer are examined.; The results of computer simulations show the robustness and adequacy of the observers under various modeling errors and operation situations. Without additional sensor installations, the observers present an inexpensive alternative to enhance vehicle on-line diagnostics ability.