Skip to Main content Skip to Navigation

Model-based System Engineering for Powertrain Systems Optimization

Abstract : Powertrain systems optimization in modern automobiles relies on model-based systems engineering to cope with the complex automotive systems and challenging control design requirements. Two prerequisites for model-based powertrain optimization are the powertrain simulator and the control design, which ensures a desirable powertrain operation during driving cycles. This thesis revolves around these prerequisites and belongs to the model-in-the-loop phase of the control development lifecycle. It first aims at identifying control-oriented powertrain systems models, particularly linear black-box models because of the merits they present in terms of accessibility to linear control design and facility of integrating changes in the powertrain system technical definition. It also aims at identifying and controlling powertrain systems featuring transport time delay because integrating the delay in the model and control design is crucial on the former’s system representability and on the latter’s optimality. Based on these premises, we address the powertrain from the engine air-path perspective. We first identify a linear black-box state-space (SS) model of a gasoline engine air-path, using an identification algorithm based on subspace methods. Different model orders and algorithm parameters are tested and those yielding the best identification and validation results are made clear, which leads to an 85% time gain in future similar identifications. While this part considers the air-path as a whole, the rest of the work focuses on specific air-path components, notably the electric throttle (ET), the heat-exchanger, and the exhaust gas recirculation (EGR). Regarding the ET, we inspire from the physical laws governing the throttle functioning to construct a linear-parameter-varying (LPV) mathematical SS model, which serves to set the regression vector structure of the LPV black-box ARX model, which is representative of an ET test bench and reflects its nonlinearities and discontinuities as it varies from one functioning zone to another. To address the questions of heat and mass transport time delays in the engine air-path, we refer to the heat exchanger and the EGR respectively. Recasting the infinite-dimensional hyperbolic partial differential equations (PDEs) describing these transport phenomena as a time-delay system facilitates the adjoint system identification and control design. To that end, a space-averaging technique and the method of characteristics are used to decouple the hyperbolic PDEs describing the advective flows in a heat exchanger, and to reformulate them as a time-delay system. Reducing the error between the output temperature of the model and that of a heat exchanger test-bench is what seeks the gradient-descent method used to identify the parameters of the time-delay system, which surpasses the PDEs in terms of identification accuracy and computational efficiency. On the other hand, the EGR is addressed from a control-oriented perspective, and the PDEs describing the mass transport phenomenon in its tubular structure are recast as a SS system subject to output delay. To regulate the burned gas ratio in the intake gas, the amount of recirculated gas is controlled using two indirect optimal control approaches, taking into account the model’s infinite-dimensional nature and accompanied with the Augmented Lagrangian Uzawa method to guarantee the respect of the input and state constraints, thus resulting in a controller of superior performance than the initially existing PID. In general, this thesis is located half-way between the academic and the industrial sectors. By evaluating the eligibility of integrating existing system identification and control theories in real automotive applications, it highlights the merits and demerits of these theories and opens up new prospects in the domain of model-based powertrain systems optimization.
Document type :
Complete list of metadatas

Cited literature [105 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Monday, March 30, 2020 - 12:14:13 PM
Last modification on : Tuesday, October 6, 2020 - 12:38:02 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02524432, version 1




Sandra Hamze. Model-based System Engineering for Powertrain Systems Optimization. Automatic Control Engineering. Université Grenoble Alpes, 2019. English. ⟨NNT : 2019GREAT055⟩. ⟨tel-02524432⟩



Record views


Files downloads