Predictive Maintenance in action -

targeting cost reduction and asset up-time

Jun 2018, Sydney

Eamonn Barrett

Delivering reductions in breakdown and cost across a machine fleet through predictive mantenance and reinforcement learning simulation

Results

Simulation and preliminary results indicate that the move towards AI-governed predictive maintenence across metropolitan areas will reduce loss of sales by 72%. It will also improve maintenance efficiencies and reduce seasonal spikes in machine failure. 

A major wholesale distributor was looking for ways to improve its maintenance across a range of more than 50,000 machines. These machines were sales points, meaning downtime was a resulting in a loss of sales as well as reduction in customer satisfaction and urgent maintenance calls.

Remi AI used AutoML, Simulation and Reinforcement Learning to provide strong predictive capabilities around component failure in over >50,000 machines. Individual optimal prediction models were found through AutoML approaches for each component across the fleet of machines. These forecasts were then used in simulation and using reinforcement learning methods, an optimal predictive maintenance strategy was developed will still maintaining a strong reactive maintenance policy. This was then validated against historical breakdown data.

 

Simulation and preliminary results indicate that the move toward AI governed predictive maintenance across metropolitan areas will reduce loss of sales by 72%. It will also improve maintenance efficiencies and reduce seasonal spikes in machine failure.