Analytics in the Rail Transportation Industry

Value Proposition

Properly designed and positioned on-car analytics can be done without a man-in-the-loop to deliver accurate and continuous information to operations managers:

  • Up-to-date condition of rail cars / critical components / cargo environment
  • Accurate predictions of time-to-failure for key parts of sub-components
  • Real-time track information and change detection – with geolocation data
  • Improved asset utilization
  • Reduced maintenance down-time
  • Optimized maintenance labor and supply and spare part inventory
  • Timely awareness of pending maintenance and safety issues
  • Location and condition of cargo / predicted conditions / time to unacceptable conditions

Background

Lone Star Analysis is an analytics company with 15-years of experience delivering accurate predictive insight to operations managers and executives of complex organizations in uncertain environments. Our modeling and simulation capability has evolved from man-in-the-loop executive decision-making support to predictive models at the edge of the network – without a man-in-the-loop. This cutting-edge capability enables operations managers to execute critical monitoring and analytics at the point of need and enables real-time information to be delivered for timely operational decisions without requiring enormous downloads of data or bandwidth.

Predicting Component Failure

Lone Star’s approach to delivering immediate, measurable, and sustainable value through edge analytics is to build and deploy virtual models of system or component failure at the point of need. These models are deployed on a gateway in an operational environment and fed sensor data of system conditions. The models can run on selected intervals – from months to milliseconds – to assess the probability of system or component failure as a function of sensor and external environmental inputs. Entire rail cars can be evaluated, if desired, from wheel health through air conditioning systems. Some example sensor data follow:

  • Temperature
  • Humidity
  • Vibration
  • Corrosion
  • Noise (dB) / Insufficient Lube
  • Airflow
  • Current
  • Voltage
  • Tachometer
  • Misalignment
  • Pressure (ft)
  • Flow (GPM)
  • Harmonic Distortion
  • Sigma Currents
  • Output Reflections
  • Moisture Detection
  • Metal Particle Detection
  • Oil Condition Sensor
  • Shock
  • Acceleration (X Dimensions)
  • Others Available

Lone Star has demonstrated the ability to predict system or component failures with one to two-week notifications or more, while systems are still operating within design specifications. These notifications are backed up with true cause-and-effect transparency, enabling prescriptive action recommendations to prevent failures. Information from real-time analytics enables “just-in-time” condition-based maintenance on systems while eliminating unscheduled maintenance downtime. This insight also enables the optimization of spare part inventories and extends the useful life of systems by minimizing component degradation.

Integration Options

Integration options for the output of Lone Star predictive maintenance models can span the gamut from simple text message notifications of critical failure alerts to full integration with existing asset management systems. Customer needs, existing infrastructure, and operating environments drive the requirements.

Technical Details

  • Predictive maintenance models are built with Lone Star’s AnalyticsOS (AOS) Architect software and run on AnalyticsOS (AOS) Edge software.
  • AOS models can be tailored for various environments and applications quickly, enabling fast time to value realized.
  • AOS Edge is pre-integrated on Intel, HP, and Dell-based gateways running Linux to collect and process sensors and other information.
  • Lone Star can identify the ideal sensor set-up to monitor and predict the performance of a given system, or build a model around existing sensor suites.
  • Models are transparent, auditable, and easily explainable.