كنفرانس بين المللی مهندسی صنايع و سيستم ها

صفحه اصلی / دوازدهمین كنفرانس بين المللی مهندسی صنايع و سيستم ها

Predictive Maintenance of Hospital Medical Equipment Using Machine Learning and Simulation in Industry 5.0

نویسندگان :

Mohammad Rahmani¹ Mohammad Sheikhalishahi²

1- دانشگاه تهران 2- دانشگاه تهران

کلمات کلیدی :

medical equipment،predictive maintenance،machine learning،discrete-event simulation،Industry 5.0،maintenance optimization

چکیده :

Hospitals need maintenance policies that protect clinical service without overloading technicians or scheduling unnecessary work. This paper reports a simulation-based proof of concept that connects machine-learning (ML) failure-score estimation to discrete-event simulation (DES) for a synthetic fleet of 60 hospital devices in three criticality classes. A 630-day condition generator produced 36,000 post-warm-up device-day records from equipment age, operating hours since maintenance, previous failures, service interval, and criticality. Three classifiers were evaluated with a chronological 60/20/20 split. Logistic regression (LR) had the highest F1-score (0.203) at the fixed test cutoff and was selected for its recall-sensitive operating behavior, although random forest had the higher AUC (0.772). The uncalibrated LR output was used only as a scenario-specific decision score in a one-year DES comparing reactive, periodic, and predictive maintenance over 40 replications. Within the specified synthetic base case, predictive maintenance reduced emergency failures by 24.6%, critical-equipment downtime by 11.2%, technician overtime by 24.4%, and scenario cost by 13.1% relative to reactive maintenance. These results demonstrate internal policy behavior, not hospital effectiveness, and require calibration and validation with real equipment data. Industry 5.0 is represented through operational human, resource-economic, and resilience proxies; environmental sustainability is not measured.