Maintenance prioritization clustering for medical ventilator

Abstract

This study aims to establish prioritized maintenance level clustering that prioritizes maintenance based on the selected attributes that may lead to early termination or beyond economical repair (BER) in medical ventilators based on the asset details and corrective maintenance history dataset of 1,056 records that span from the year 2017 to 2021. These datasets are extracted from a web-based Computerized Maintenance Management System (CMMS) from maintenance Company A with more than 30 attributes or features. The method used to achieve the desired result is the unsupervised machine learning algorithm, K-Means Clustering, to establish a maintenance prioritization clustering from unlabeled data. The result obtained shows that the dataset was successfully separated into three clusters. The line chart was used to visualize the relationship between clusters and attributes. It demonstrates the distribution of the attributes within each cluster, and able to identify the patterns or trends that were then used to determine the level of maintenance prioritization as "Low", "Medium", and "High".