Development Of eva‎luation Model For Turnouts Using Machine Learning

9/20/2023 12:00:00 AM

In the rail industry, component deterioration rates are high due to widespread use, limited maintenance due to lack of funding, and the growing demand for rail access. Currently, infrastructure managers lack a decision support tool or model that can help them effectively and efficiently in making decisions about turnout failure. In this thesis, the development of the eva‎luation model of the turnout device in railway lines using classification techniques based on decision tree and machine learning to determine the need for maintenance in railway turnouts has been discussed. Recently, many studies have proposed the use of machine learning techniques to determine the performance condition of a line component, detect the need for maintenance and possible failure modes. It is expensive and impractical to install monitoring devices on all line components throughout the network. The aim of this thesis is to develop an intelligent model that uses the data available in the railway and gives interpretable results. Therefore, using the technical-geometrical data of 126 turnouts in line 1 of Tehran subway, a model based on decision tree and machine learning was developed, and finally by optimizing it, we achieved an acceptable accuracy of 86% in determining the condition of turnout. The approach used in developing intelligent models and explaining their results has a wider application and can be used for other types of line components and maintenance planning scenarios.

يادگيري ماشين , راه‌آهن

Machine Learning , Railway

Danial Ghaffari

Dr. Zakeri