Creating a Virtual Tyre Temperature Sensor

University essay from Lunds universitet/Matematisk statistik

Abstract: To accurately determine the efficiency and range of an electric vehicle, one must be able to estimate the rolling resistance of the car. This is currently done using standardized methods developed under laboratory settings, where transient aspects and effects of varying temperatures are excluded. Given the strong correlation between tyre temperatures and the rolling resistance, determining this temperature is of great interest. This Master's thesis investigates the development of a virtual sensor for predicting the tyre temperature during dynamic driving using recurrent neural networks (RNN). The primary goal is to examine if a virtual sensor can predict the tyre temperature within ±2 °C of the actual temperature using on-board vehicle signals. The study also evaluates the significance of these signals in the model. Two different RNN architectures, Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), were trained and evaluated. The LSTM performed slightly better and results indicated that the model can predict the tyre temperature within the ±2 °C interval around 90\% of the time. The most crucial features contributing to model performance were identified as vehicle speed, ambient temperature, brake pedal position, accelerator pedal position, and road inclination. To improve on this results, a few interesting future research areas were suggested.

  AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)