Answers are grounded in this paper and cite their sources.
Study implements a semantic validation framework for HVAC diagnostics, indicating effective anomaly detection and relevance.
Despite increasing application of machine learning in building operations, systematic validation of detected anomalies against real maintenance activities remains limited. This study develops a semantically validated diagnostic framework that integrates unsupervised clustering of operational data with generative AI-based interpretation of maintenance records. Application to a dedicated outdoor air system identified anomalous regimes exhibiting substantially greater control variability than normal operation (Cohen's d=2.47, p<0.001). Generative AI extraction of maintenance semantics demonstrated strong agreement with human coding <ns0:sc>(k=0.81) </ns0:sc>and high reproducibility (94% exact match across independent runs). Temporal analysis revealed significant enrichment of anomalies near documented HVAC work orders (OR=2.25, p=0.003), supporting diagnostic relevance. Incorporating embedding-based semantic similarity further improved validation precision by 69% (p=0.008), nearly doubling correct identification while reducing false positives. By leveraging only standard building infrastructure and requiring no labeled training data, the framework establishes a scalable approach to weakly supervised validation of unsupervised HVAC diagnostics and advances reproducibility standards for generative AI-assisted building analytics.
No takes yet. Share an insight, caveat, or question.
Imani et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: