The study is motivated by the fact that a substantial share of reliability-relevant information about complex technical systems is stored not only as structured measurements but also as semi-structured textual records, including incident narratives, maintenance requests, defect reports, expert statements, and technical event descriptions. The aim of the paper is to develop a method that integrates such information into a formal framework for risk forecasting and intervention prioritization. A semantic-evidential graph method is proposed. Textual data are transformed into evidential estimates of degradation and failure, embedded into a failure-intensity model, and complemented with a graph signal reflecting inter-component dependencies. The experiment uses open aviation data provided by the USA National Transportation Safety Board and the MaintNet maintenance corpus. After filtering and aggregation, 42,956 family-month observations were obtained. The proposed method was compared with three baseline models: a structured model, a frequency-based model, and a topic-based model. On the 2024-2025 test set, the topic-based model achieved the best ranking metrics, whereas the proposed method delivered the best F1 score and the highest decision-making utility under a severely constrained monthly intervention budget. The results confirm that integrating semi-structured textual information into a mathematical risk model improves the quality of reliability-management decisions for complex technical systems.
No takes yet. Share an insight, caveat, or question.
Konnikov et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: