By: Ade Chandra Nugraha, Suhono Harso Supangkat, I Gusti Bagus Baskara Nugraha, Yunendar Aryo Handoko
Digital transformation triggers the application of information technology in assisting daily activities. The use of information technology triggers data processing that generates information to assist policy making. The digitalization megatrend is leading the ongoing reshaping of transportation and mobility systems. In this new industry trend, where data is the new currency, system conditions are assessed through data-driven approaches that envision future behavior and/or failures that may trigger interventions, in what is referred to as a predictive framework. One of the applications of information technology in the maintenance process is the application of Predictive Maintenance (PdM).
PdM is a maintenance strategy carried out by utilizing data and technology to predict maintenance needs on a system or equipment before significant damage occurs. In the context of rail geometrics, PdM can help identify rail geometric problems before serious damage occurs that can cause accidents or disruptions in rail usage. By performing proper maintenance on a railroad track, it can help reduce maintenance costs, extend rail life, and improve the safety and efficiency of the railway system. Condition based method (CDM) is one of the methods that help implement PdM. CDM for railway transportation supported by a reliable system and efficient monitoring system and using integrated data, can improve the reliability, availability, maintainability, and safety of railway track components and structures.
Currently, most railway infrastructure managers in Indonesia conduct rail monitoring by measuring geometric irregularities with special train EM120. Although this method provides accurate measurements of rail irregularities, the use of gauge trains is very costly and time-consuming. For this reason, in recent years, much research has focused on developing new methods for track condition monitoring using dynamic measurements of in-service railway vehicles. One of the dynamic impact measurements of rolling stock is by measuring sway using accelerometers.
The motivation to find a model to detect rail irregularity using in-service railway vehicles is one of our challenges. By referring to the following figure and assisted by a data driven approach, which implements certain machine learning models, decision making for rail maintenance can be developed and certainly supports the implementation of predictive maintenance models.
Through a data driven approach, a machine learning model will be used that contains specialized algorithms in generating information from the data. Machine learning model is conceptually built by data, model and learning. Based on the purpose of machine learning in data driven application, data becomes the core of machine learning. From the collection of data owned, data patterns will be obtained which will be used as a model of machine learning algorithms for defined needs. To provide a better precision value, the machine learning model needs to be trained and automatically perform learning by optimizing the parameters of the model.