MULTI-STATE PREVENTIVE MAINTENANCE AND REPLACEMENT SCHEDULING FOR MULTI-COMPONENT SYSTEMS CONSIDERING NON-FAILURE STOPS

In continuous operating units, lost production costs are high due to downtime. Economic profitability of these industries is conditional on the implementation of a proper maintenance policy to increase reliability and reduce equipment operating costs. In these industries, all systems, from the simplest to the most complex ones, require scheduled maintenance to reduce the risk of failure. Maintenance scheduling is a branch of industrial engineering that reduces maintenance costs by controlling manufacturing equipment and machinery for repairs and replacement schedules and using statistical analysis. As far as our knowledge is concerned, all article papers in this area have assumed that devices will not stop until it is damaged or inspected. However, in the real world, failure or periodic inspections are not only causes of shutdowns but also non-failure interruptions such as the type and continuity of work, the furnace cycle, prototyping and testing, final evacuation, project execution, and corrections can stop machines. This provides a good opportunity for doing some maintenance activities at this time and prevents future system shutdowns for periodic inspections which in turn increase system access and reduce maintenance costs. The duration of these stops is limited. With this new approach, a mathematical model is developed in order to optimally schedule preventive maintenance and repair activities in a multi-component system that can be maintained. The maintenance planning horizon is divided into discrete equal size periods and three possible types of activities are considered for each component. The optimal decision is searched for each component in each period to meet the desired reliability at the lowest cost. We also considered the duration of activities and the time-dependent cost of shutdowns. As a solution method, the genetic meta-heuristic algorithm is implemented in Visual Studio software and finally, to examine the efficiency of the proposed model, a numerical example is provided.