- We worked on-site with future users of the tool to ensure all of their specific needs and unique constraints were considered.
- The tool was developed in multiple stages, allowing requirements to be reconsidered after each stage so that development was always producing what was most relevant at the time.
- The BOSS Platform allowed us to rapidly develop Aurizon’s new tool and provides access via a web interface, allowing users to log in from anywhere with an internet connection.
The Locomotive Maintenance Productivity division of a large American Rail Service Provider is constantly investigating new ways to provide greater value and reduce costs. The Engine Overhaul component of Locomotive Maintenance is a large part of the provider’s service, and accounts for significant costs. Engine Overhaul scheduling in the past has been based on calender cycles – so once an engine has been in service for a set length of time, it is scheduled for overhaul.
Since calender cycle-based maintenance does not take into account the health of an engine, the provider switched to undertaking overhaul based on a threshold cumulative motoring energy output (MWh) of an engine since overhaul/installation. This metric was used to better estimate engine condition. However, the provider understands that motoring energy threshold is not necessarily a direct proxy for component condition (e.g. an engine doing small yard work with variable engine use may accumulate more damage than an engine running in optimal running conditions across country).
The long-term aim for the provider is to schedule timing and workscope of maintenance and overhauls based on component condition (i.e. to implement a condition-based strategy rather than a cycle-based strategy).
Four stages were taken to achieve the required outcomes.
- Integrate the provider’s datasets and confirm assumptions made around joining the disparate datasets.
- Generate descriptive statistics around scrap rates (the response variable) and potential predictors generated from the datasets. Predictors statistics were generated to see how engine variables changed as engines aged. Predictors included:
- Locomotive statistics (motoring energy accumulated, time spent in different temperature and operating conditions, operating locales, etc),
- Oil history
- Automated engine fault readings
- Engine operating conditions (sensor readings such as temperatures and pressures)
- Determine statistically significant predictors of component scrap rates.
- Recommend next steps that will allow the provider to undertake condition-based maintenance.