Development and Validation of a New Generation Rail Wear Model Using Emerging Big-Data Analytic Techniques

Start: January 2017, End: August 2018

Principal Investigator/Researcher

Dr. Allan M Zarembski, Joseph Palese (Graduate student) University of Delaware

Project Description

The proposed research is to develop a more comprehensive rail wear degradation model utilizing emerging big-data techniques that are relevant to the types of data readily available to the railway industry. This data represents a large set of data captured through automated inspection. A major US Class 1 railroad with over 20,000 miles of track has already agreed to provide data from their rail profile measurement systems (ORIAN) onboard three of their track geometry cars. This would represent over 60,000 miles of rail wear data taken per year, corresponding to three or more rail wear measurements per year for a multi-year period.

The goal is to replace the generally “simplistic” wear models in use today with a higher end- wear forecasting model that accounts for the key influencing factors for rail wear: (1) traffic information, (2) rail type, (3) level of lubrication, (3) curvature, (4) grade, and (5) other influencing factors. Big data analysis techniques such as Data Mining, Data Fusion, and Sensor Fusion will be used.

It is expected that the resulting model will be used in maintenance planning and management of the rail infrastructure to allow for modeling large sets of rail wear (and profile) data, along with other railway physical asset and inspection data to predict effective replacement points based on rail wear standards. As such, it is a strong fit to the theme of the UTC activity.