Closed-Loop Simulationsmodell der Fahrzeugklimatisierung

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2023; Aachen : RWTH University Aachen Online-Ressource : Illustrationen, Diagramme (2023). = Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2023
This thesis describes the development of a closed-loop model for the virtual application of the air con-ditioning system for vehicles. Virtual application describes the design and development of the control behavior of a system by a virtual model. To be able to observe the behavior of the controlled system, the model must be in a closed-loop state, since this is the only way to represent the complete control loop. Within the literature research conducted at the beginning of the thesis, mainly models where found in which either the control system, the air conditioning system in connection with the air vents or the cabin model were simplified to an extent that a closed-loop state is not possible. First, different parts of the air conditioning system were investigated by measurements to enable their modeling. To measure the air and temperature distribution of the HVAC Unit, a test setup of the distribution tract was built to measure the distribution of the flow variables for different representative flap positions. Since the accuracy of the existing measurement methods was not sufficient, special attention was paid to the development of a new measurement method. Furthermore, a measurement of the leakage enables the integration of the leakage air rate into the simulated models. In addition to the distribution tract, a test setup was created to measure the delivery characteristic of the blower. Training data is needed as a basis for building machine learning models. In this work, the models used to generate this data are referred to as baseline models. Derived equivalent pressure drops for diffusers and air ducts significantly reduced the computational time of the baseline CFD model used for the re-quired number of data points, allowing an HVAC fluid model to be derived in STARCCM+. In addition, a baseline blower model was also created in STARCCM+ and validated using the blower characteristics. Due to the complex flow processes in the area of the blower tongue, the delivery characteristic of the blower can only be calculated with the required accuracy by scale-resolving simulations. Based on the validation of the baseline models, reduced metamodels from the field of machine learning were built using the Python Toolbox P7. Here, the air handling unit consists of several submodels based on Gauss-ian processes. The initial model of the distribution tract does not achieve the necessary accuracy for the virtual application. To increase the accuracy, a resampling algorithm and a linear error condition were designed and implemented in an extended model. The different submodels were interconnected in a HVAC-metamodel. The modeling used results that have no information about the multi-dimensional flowfield at the cabin vents. The flow field of the outlets is relevant for the thermal comfort in the cabin. To compensate for this effect and to avoid having to calculate the outlets in the overall vehicle model due to the different geometric length dimensions, a reduced model was derived for the outlets based on a principal component analysis. The flow fields were decomposed into their characteristic eigenmodes and eigenvalues and the eigenvalues were represented by a regression model. A model of a Porsche 911 4S (992) with a high level of detail was built as the controlled system in this work. The model consists of a weathering model, vehicle model and cabin model. The software of the climate con-troller was derived as an FMU model and coupled with the complete vehicle model, the reduced mod-els and a residual bus simulation. For representative load cases, the closed-loop composite model was validated using a climatic wind tunnel measurement. The load cases correspond to the VDI specifica-tions for summer and winter load cases for the representation of heating and cooling processes. The validation demonstrated the suitability of the model for virtual application.
Published by RWTH University Aachen, Aachen