New model targets train derailments
Though many train derailments are minor, others can have serious consequences that include extensive property damage, injuries or deaths, and environmental disasters. In an age when many hazardous materials are transported by rail, the danger of train derailments is especially acute.
“Track geometry defects such as uneven tracks are one of the two primary causes of freight train derailments on main tracks, so they have a critical role in railroad safety,” said Hadi Meidani, an assistant professor in the Department of Civil and Environmental Engineering at the University of Illinois at Urbana-Champaign, during a recent Roadway Safety Institute seminar.
Because track geometry defects have been identified as a major cause of train derailments, the Federal Railroad Administration has set track safety standards requiring defects to be corrected or protected within a prescribed time limit. However, predicting when track defects are at risk of exceeding safety limits is a major challenge; although a large volume of track measurement data is currently being collected, track deterioration is a complex process.
In his presentation, Meidani outlined research focused on developing new models that account for the complex nature of track geometry deterioration. The new predictive model for track geometry defects will make reliable predictions on when a defect will exceed safety parameters. In addition to improving safety, this model could enable more efficient planning of track maintenance.
“Our idea was to see if we could use current information to predict when track defects would move beyond the cautionary ‘yellow’ level into the more severe ‘red’ level defect using a survival analysis probability model,” Meidani said. “Essentially, this type of model calculates the probability of failure across time, telling us how much time we have before a defect enters ‘red’ levels and fails, and how long it could survive before failing.”
To develop their model, researchers took data provided by rail companies, cleaned it, and determined which factors impact track geometry defects. In this case, they found only railroad class and defect amplitude had statistical significance, so those were the variables they included in their simplified model. Next, they tested their model to determine its accuracy and found that it could correctly predict failure more than 70 percent of the time.
“Our model had a higher percentage of correct predictions than the segment-based method currently being used by rail companies, as well as a lower percentage of wrong survival predictions, which is critical for safety,” Meidani said.
Researchers are now working to optimize the model’s resolution, determine whether there is an ideal track inspection policy that minimizes costs while improving safety, and quantify the relationship between better prediction models and improved safety.