Training Based Testing Of Temperature Sensors In Diesel Engine Aftertreatment System
Ryöti, Markus (2020)
2020
Konetekniikan DI-ohjelma - Master's Programme in Mechanical Engineering
Tekniikan ja luonnontieteiden tiedekunta - Faculty of Engineering and Natural Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2020-11-03
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202010307732
https://urn.fi/URN:NBN:fi:tuni-202010307732
Tiivistelmä
The goal of this thesis is to research the possibility for creating an automated test for verifying the correct installation of temperature sensors in a diesel exhaust aftertreatment system. Both model and training based solutions are discussed. The implementation is done by training multiple machine learning models and comparing their results.
Data was specifically collected for this research. Data collection consisted of taking measurements from a test engine using different sensor combinations. In total there were four different sensor combinations and for each combination 25 recordings were taken. The recordings had variating starting temperatures. The data was then labeled accordingly for supervised learning.
Results show that classification of sensor installations can be achieved with high accuracy. All the used models provide promising results while logistic regression model seems perform the best. More important and limiting issue is the data gathering process for training and testing the models.
Data was specifically collected for this research. Data collection consisted of taking measurements from a test engine using different sensor combinations. In total there were four different sensor combinations and for each combination 25 recordings were taken. The recordings had variating starting temperatures. The data was then labeled accordingly for supervised learning.
Results show that classification of sensor installations can be achieved with high accuracy. All the used models provide promising results while logistic regression model seems perform the best. More important and limiting issue is the data gathering process for training and testing the models.