A preliminary study on text mining operator logbooks to develop a fault-frequency model

Textual data in operator logbooks represent an untapped opportunity to retrieve information about the maintenance routines of HVAC equipment and control infrastructure. This paper presents a case study in which seven years' worth of work order logs from 44 buildings on a university campus were analyzed. After extracting HVAC-related terms such as fan, AHU, VAV, stuck, and leak from custom operator descriptions, the apriori algorithm was used to derive association rules that define the coexistence tendencies of the terms in a work order (e.g., coexistence of the terms radiator and leak). Based on this analysis, a preliminary HVAC work order frequency model was put forward. The results indicate that the annual work order intensity per 1000 m2 (10,764 ft2) was about 4. More than 70% of the HVAC-related work orders were issued to address zone-level problems. Among the AHU-level work orders issued for a physical subcompo-nent, more than 90% were related to fans. Future work is planned to analyze the HVAC-related work order patterns with numeric data from automation and controls networks.