Intelligent advisor system for prescriptive maintenance of engineered assets using failure modes, effects and criticality analysis, knowledge graph and machine learning

Abstract

Maintenance strategies for engineered assets are shaped by the diversity of potential failure modes and their impact on functional performance and safety. As these assets comprise numerous interconnected components, they are prone to concurrent degradation and multi‐source faults. Consequently, maintenance personnel face challenges in efficiently and accurately identifying specific failure types and pinpointing root causes to plan effective interventions. This paper introduces an intelligent advisor—a question answering (QA) system—for fault diagnosis, failure mode identification and automated root‐cause analysis by integrating failure modes, effects and criticality analysis (FMECA), machine learning (ML) and a knowledge graph (KG). A linear actuator serves as a case study to validate the approach via a four‐stage implementation: (i) FMECA construction from design knowledge and known degradation mechanisms; (ii) signal processing and feature engineering from sensor data collected under multiple fault scenarios; (iii) digitisation of FMECA into a KG to represent concepts and relations; and (iv) a natural language processing (NLP) layer that enables user‐friendly interaction. Results show that the integrated framework enhances the interpretability and traceability of automated diagnostics, providing a transparent pathway from sensor‐level anomalies through KG‐based reasoning to prescriptive maintenance actions.