Intelligent systems in maintenance planning and management

Abstract

Maintenance can be divided into reactive and preventive. In reactive maintenance damage is corrected after it has occurred. On the other hand, predictive maintenance anticipates the damage in the future. Predictive maintenance can be periodic with fixed time intervals or predictive with forecasted failure times. Maintenance planning is intelligent when maintenance needs are predicted and optimised. Intelligent maintenance systems require data collection, data transfer, data storage, data processing and Decision Support Systems to be in place. Machine learning algorithms in forecasting and optimisation can take increasing quantity of collected data into consideration in intelligent maintenance planning. Two types of forecasting models—time-series and causal methods can be used for intelligent maintenance planning. Optimisation algorithms can be divided into local and population search methods. Hybrid methods combining more than one algorithm have been used efficiently for maintenance planning. Maintenance planning is important in the transport sector and various models and methods have been applied both in road and vehicle maintenance. In practice, it is important to have accurate deterioration and cost models in place. In research, both data-driven and mathematical models have been applied, but data-driven methods are becoming more practical as computation is increasingly more feasible.