PRESCRIPTIVE MAINTENANCE OF KOMATSU DUMP TRUCK HD785-7 BASED ON MACHINE FAILURE DATABASE USING NAÏVE BAYES CLASSIFER IN FULL MAINTENANCE CONTRACT PT ABC SITE SANGKULIRANG

Abstract

This research develops a Prescriptive Maintenance (RxM) model to improve the Physical Availability (PA) of Komatsu Dump Truck HD785-7 under a Full Maintenance Contract (FMC) at PT ABC Site Sangkulirang, addressing the issue of low PA that affects operational efficiency and mining productivity. By applying the Naïve Bayes Classifier and the DMAIC (Define, Measure, Analyze, Improve, Control) framework, historical maintenance data from the Daily Activity Report (DAR) in Lakoni systems were transformed into predictive insights to guide maintenance prioritization. A total of 924 data records were processed, with 682 used for training and 242 for testing the classification model. The model achieved 97.93% accuracy, 100% precision, and 94.12% recall, demonstrating its effectiveness in identifying units requiring preventive maintenance. Integrating the RxM model into maintenance operations reduced unplanned breakdowns, optimized technician allocation, and improved scheduling decisions, with supporting tools like dashboards and updated Standard Operating Procedures (SOPs) ensuring long-term sustainability. This study highlights the potential of machine learning in enhancing asset management within the mining industry.