CIVIL ENGINEERING 365 ALL ABOUT CIVIL ENGINEERING



AbstractPrioritization of maintenance and rehabilitation interventions according to their urgency, expected improvements, costs, and downtimes is a necessary task in facility management. This is particularly of interest in critical facilities requiring nondisrupted operations or minimal stoppages such as hospital buildings. Maintenance scheduling is currently performed in hospitals in a purely subjective manner where experts categorize the importance and priority levels of different interventions and plan them according to their associated cost. Despite its popularity and workability, this process can be criticized for its high dependence on the expert’s knowledge and experience as well as its sole dependence on cost as the main driver for applying interventions. Therefore, in this paper, an objective methodology is developed to select the most suitable interventions and accordingly group the different actions together. This is expected to minimize the overall downtime and disruption caused by the maintenance and rehabilitation works in healthcare facilities. Introducing a combined artificial intelligence-based methodology, a triobjective optimization model is primarily utilized to select the interventions expected to maximize the overall performance of the facility as well as inhibit the minimum levels of costs and downtime. The output is further fed into an unsupervised machine learning model where interventions are grouped together according to their relevancy on a hybrid clustering approach, integrating between hierarchical and K-means clustering algorithms. This yields an action plan for use by maintenance personnel to schedule the disruptions of critical spaces (i.e., intensive care units and operation rooms) within the hospital building. This helps ensure a smooth and continuous operation in the facility as well as prevent any harm to facility occupants that could result from maintenance and rehabilitation works.



Source link

Leave a Reply

Your email address will not be published.