Creating value out of text logs
GE Transportation is a division of General Electric that supplies and maintains locomotives for transportation companies. In an effort to reduce costs, streamline operations, and increase performance, managers approached Analytics Lab (A-Lab) to see if MIT students could derive useful data from the reams of messy text logs kept by the company’s mechanics. The goal of the project was to use the data to help anticipate the need for locomotive maintenance, thus improving service to customers and saving GE and its customers money.
Applying analytics and modeling
GE Transportation gave the student team three years’ worth of maintenance service sheets in electronic form. The students used multiple text-analytics methods to wrangle information from the entries, which were riddled with non-standard spellings and abbreviations. The students developed analytics to cluster the text based on a variety of characteristics and created a model that linked certain features to higher likelihoods of mechanical failure in a key component of the locomotive. The additional data enabled students to better predict equipment failures within the first 78,000 miles of locomotive travel than using GE’s quantitative data alone.
Predicting maintenance needs
“We’re very happy with the results,” said Justin Rivera, manager of data science and analytics at GE Transportation. “The students became subject matter experts for us on this niche project, and we were able to capitalize on their findings.” Now, Rivera said, GE Transportation is able to optimize maintenance schedules in the many locomotives that GE supports.