Coded by CheolJeongPark: Advcaned Manufacturing Laboratory(AML).
Check My Portfolio and Research Summary

Machine Learning Prediction : Rail Temperature.

0. READ ME!

These codes are for new AML lab member studying machine learning and anyone who interested in machine learning prediction for rail temperature.

We use python(mostly jupyternotebook) on our code. These codes are developed by Sung Uk Hong, CheolJeong Park and Jongwon Yoon.

Raw data(rail temperature and climate data) used on these codes are not provided! (Becacuse of paper issue)

If you need more detailed information or collaborative research, please contact us!

professor Seong J. Cho : scho@cnu.ac.kr
Sung Uk Hong : hsu12375@gmail.com
Cheol Jeong Park : pffiro@gmail.com
1ongwon Yoon : jongwon3498@naver.com

1. Abstract

Rail temperature is critical feature at train industry. Thesedays, where air temperature has increased becasue of abnormal climate, train system is suffered by rail temperature. At high rail temperature, the risk of buckling on rail increases and the train operates at slow speed to keep the trian from derailment resulted from buckling. Train industries have monitored rail temperature by directly measuring rail temperature to maintain the rail and train safety. However, it is difficult to measure temperature across the full section of the rail, because of cost or maintenance of measurement system. The train industries now focus on predicting rail temperature.
On this study, we developed rail temperature prediction model using machine learning with the data provided by the meteorlogical agency. We developed our model by 4 steps as shown below.

1. Data acquisition at measurement system(desigend by AML)
2. Data wrangling and analyzing features
3. Select machine learning algorithms and evaluate models
4. Select the model showing the best performance
5. Analyzing features for predictin rail temperature

We evaluated our model's performance with r-square and mean absolute error. Our model showed high performance at the whole range of the rail temperature, especially at high rail temperature(over 40℃), compared to reported rail temperature prediction models. We also explains what features are important for rail temperature at different temperature ranges.

2. Outcome

Two papers related to this study are currently being reviewed and one is being prepared for publication. We presented relevant research at 3 conferences(2018 Asian Conference on Railway Engineering and Transportation, 2020 Spring Korean railway Conference, 2020 Asian Conference on Railway Engineering(postponed to Nov 2021)). We also won the Excellent Paper Presentation Award from 2020 spring korean railway conference!.

For more detailed infromation Please contact us by e-mail shown 0.READ ME! at above.