A Knowledge-Based Expert System for Predictive Maintenance and Fault Diagnosis in Heavy Machinery

Abstract

In the mining and construction industries, maintaining productivity, minimizing downtime and cutting operating costs all depend on effective heavy machinery maintenance. Whether corrective, preventive, or predictive, traditional maintenance methods frequently rely significantly on human expertise, which can result in inconsistent diagnostics and expensive delays. In order to mimic the diagnostic reasoning of seasoned technicians, this study presents a knowledge-based expert system for predictive maintenance and fault diagnosis in heavy machinery. Domain knowledge was extracted from engineers, maintenance manuals and fault logs using a Design and Development Research (DDR) framework. This information was then codified into if-then production rules in ES-Builder 3.0. Using a responsive web interface built with PHP, HTML5 and MySQL, the system combines an explanation module, knowledge base and inference engine. Controlled evaluation on 26 separate fault cases achieved 96.15% diagnostic accuracy (95% Confidence Interval (CI): 81-99%), with an average response time of 1.74 seconds and closely matched expert diagnoses for power, hydraulic and oil-level faults. The system's clear logic improved interpretability and user confidence, proving the ongoing usefulness of knowledge-driven AI in settings with limited data. The study shows that rule-based expert systems can provide dependable, comprehensible and scalable heavy-equipment maintenance solutions, offering a useful basis for upcoming hybrid AI architectures that combine adaptive learning and real-time sensor data.