MA – AI based tool for the automated application of characteristic curves and maps in vehicle thermal management

The aim of this master’s thesis is to develop an AI-based tool for the efficient calibration of characteristic curves, maps and parameters for thermal management in modern vehicles.
An existing Simulink model of a vehicle’s refrigeration cycle emphasizes the evaporator and compressor to analyze air conditioning efficiency. This Simulink model will be expanded with an airflow model that includes the radiator blinds opening angle and fan speed, as well as a characteristic curve-based control strategy of the actors.
After expanding the Simulink model this 1D simulation is used to generate the required training data for the AI system. On this training data, advanced machine learning techniques such as deep learning with regression algorithm, and reinforcement learning for control optimization, will be employed. Feature engineering will ensure the training data captures critical parameters such as refrigerant pressures, temperatures, mass flow rates, and component efficiencies. The optimization process will involve training and testing multiple algorithms, including neural networks, gradient boosting methods, with performance evaluated against a predefined cost function based on metrics like energy efficiency. The process continues with optimizing the model, selecting the best-performing algorithm based on the cost function, and generate system-specific maps for various ambient temperatures and drive cycles. The first application of the developed tool is to optimize the airflow control at the front end of a vehicle by changing the actuation of the fan and radiator blinds.

Bearbeiter: Giriraj Kokare

Betreuer: Friedrich Schlieter (IAV)

Verantwortlicher: Prof. Dr.-Ing. Martin März