Abstract

Public direct current fast chargers generate heterogeneous operational snapshots, service records, and platform-specific diagnostics that cannot be treated directly as field-supported failures. Existing monitoring identifies abnormal states, while conventional Failure Mode and Effects Analysis (FMEA) ranks an already defined failure taxonomy; neither explicitly operationalizes a qualification step between raw diagnostics and maintenance mechanisms. This study develops a pre-FMEA engineering-evidence framework integrating data normalization, subsystem classification, evidence registration, service-report correlation, manufacturer-document verification, and structured score review. The bounded case comprised 30 ABB Terra T54HV diagnostic datasets collected across multiple public sites during a January–December 2025 collection window (19,866 records and 877 unique keys) and 30 associated field service reports. The evidence set retained a common cross-source schema and linked diagnostic context to documented interventions; it is not claimed to represent all charger populations or a complete failure census. Of the diagnostic keys, 546 formed a common cross-source core and 217 varied across sites. Three primary service categories—communication and authorization, global interlock, and voltage-related cases—accounted for 25 of 30 documented interventions (83.3%). Their internally reviewed screening RPNs were 105, 90, and 72; no score was independently validated. The framework supports inspectable maintenance prioritization by preserving a source-to-mechanism evidence path, excluding snapshot repetition from occurrence, disclosing overlapping mechanism links, and retaining unresolved scores as provisional.