Levels of Autonomy for Predictive Maintenance: A Structured Approach with Digital Twin Integration

Autonomous predictive maintenance (PMx) is increasingly recognized as essential for optimizing reliability, minimizing downtime, and reducing operational costs. However, the concept of “autonomy” remains loosely defined, leading to inconsistent implementation across industries. At the same time, digital twins (DTs) hold promise for enabling advanced autonomy by offering real-time data integration, dynamic simulation, lifecycle monitoring, and advanced analytics. Yet, the absence of a structured roadmap for development prevents these capabilities from driving autonomous PMx solutions. This paper addresses both of these challenges with a dual contribution. First, it adapts a well-established Levels of Autonomy (LoA) framework-originally developed for autonomous vehicle standards-and maps it to defined informational requirements (IRs) and functional requirements (FRs) in PMx, offering a standardized method for systematically advancing autonomy. Second, we illustrate how to apply these guidelines to support DT-based PMx solutions across an asset or system's lifecycle, progressively fulfilling LoA-specific IRs and FRs at each stage. An illustrative example focused on a pedestrian steel truss bridge contextualizes how this progressive approach fosters more structured and uniform DT implementation and higher PMx autonomy.