Abstract
Asset intensive industries such as Energy, Oil & Gas are continuously challenged to mitigate the risk of equipment failure to avoid unplanned costs, impacts on production, safety and environmental implications. It is a constant battle to balance the cost of preventative maintenance measures against the risk of equipment failures.
Predictive analytics is the application of powerful statistical modelling techniques to derive forward-looking ‘strategic actionable insights’ from historical data, as well as being capable of providing real-time ‘decision support insights’ on specific equipment.
It is applied to best advantage by mining multiple disparate data sources such as performance logs, sensor data, incident data, financial data, text based maintenance logs, human resources data and environmental data.
This paper describes an approach to:
Accurately predict which subtle combinations of characteristics tend to lead to increased frequency of failures.
Unearth patterns in maintenance operations over time that could point to opportunities for improvements.
Identify the characteristics that tend to increase ownership cost and downtime over the life of a system.
Predict what parts are likely to fail in the near future.
Identify ‘at risk’ parts that have not yet failed so that they can be replaced just-in-time to avoid unscheduled downtime due to failure.
Mine thousands of text based logs that describe the maintenance performed on systems to determine what important observations are being logged by the maintenance team.
- ■
Scheduled maintenance can be adjusted and optimised on a dynamic basis so as to maximise equipment availability and avoid equipment failure.
- ■
Costly unnecessary maintenance can be avoided.
- ■
Maintenance resources can be deployed in an optimal fashion.
- ■
Analytical insights can be used to improve equipment design and maintainability.
