Preventive maintenance scheduling with Predictive AI Analytics
INDUSTRY
Transportation/Railroad
use case
Predictive Maintenance
Solution Type
Real-time Analytics, Machine Learning
.svg)
The Preventive Maintenance Scheduling project was undertaken for a US railroad firm, with the goal of enhancing railway safety and operational efficiency. By designing and implementing a real-time. Warm Bearing Prediction Model, the project leveraged AI to process continuous streams of track sensor data, enabling predictive maintenance scheduling and reducing track incidents.
The railroad firm faced a critical issue with its existing maintenance scheduling system. The previous model relied on a Big Data instance within Teradata, which used historical and training data to generate alerts. However, this system could only run once a day, using information that was already a day old by the time it was processed. As a result, the alerts and maintenance schedules were often outdated and ineffective, leading to increased risk of track incidents and compromised railway safety.
The challenge was to develop a solution that could process sensor data in real-time, ensuring that maintenance schedules were up-to-date and reflective of the current conditions on the tracks.
To address this challenge, the project team designed a new Warm Bearing Prediction Model using machine learning algorithms implemented with Apache Spark. The focus was on creating a system that could process streaming sensor data continuously, enabling real-time updates to maintenance schedules.
Key steps in the approach included:
Implementing Apache Spark to handle continuous streams of sensor data, allowing for real-time analysis and decision-making.
Developing ML algorithms capable of predicting potential track issues based on the sensor data, enabling proactive maintenance scheduling.
Ensuring seamless integration with existing railway systems to provide real-time alerts and updates to maintenance teams at various stations.
The resulting solution was a robust Warm Bearing Prediction Model that significantly improved the railroad firm’s ability to predict and prevent track incidents. By continually processing streaming sensor data, the new model was able to update maintenance schedules in real-time, providing timely alerts to stations and maintenance teams. This proactive approach ensured that potential
issues could be addressed before they led to track incidents.
The implementation of the new predictive model led to a remarkable improvement in the firm’s maintenance scheduling process. Model run-times were reduced by over 60%, allowing for faster and more efficient processing of sensor data. Moreover, the efficacy of the maintenance schedules improved by over 1200%, dramatically enhancing the firm’s ability to prevent track incidents and ensuring a safer railway operation.
This project not only resolved the issues associated with the outdated maintenance model but also set a new standard for predictive maintenance in the railroad industry.
