Neural Networks
Prediction of Car Cabin Temperature Using Artifcial Neural Network
Author
Tien-Yun Lee
Supervisors:
Prof. Kai-Uwe Bletzinger,
Technische Universität München
Dr.-ing Thomas Schneider,
P+Z Engineering GmbH
Dr.-ing Jan Reger,
P+Z Engineering GmbH
Abstract
An automatic climate control system (ACC) was developed to be installed in the car, in which the ventilation system should react automatically on the current cabin air temperature, in order to keep the car cabin temperature within a comfortable range. However the measurement of the cabin temperature is not available and there is no enough storage to run simulation software.
The aim of this thesis is using the artificial neural network to predict the current cabin air temperature as a solution for an ACC system. An artificial neural network is a mathematical model built up by functions and equations, takes only small storage. It is a machine learning tool, by feeding and training the artificial neural network with proper samples, it can give good prediction of the cabin air temperature. Without experiment data, all the samples are calculated by THESEUS-FE, a thermodynamic simulation software using finite element based solver, developed by P+Z Engineering.
In this thesis, how to design the artificial neural network and what kind of samples should be used in order to give a good prediction is discussed. In the end of the thesis it will show that a well trained artificial neural network has very similar performance as THESEUS-FE in predicting the cabin temperature on a virtual driving situation.
|
Download full text [pdf, 2.5 MB] |