A novel approach to assess the cyclone impacts on the brick veneer house of Australia using hybrid natural language processing

Tropical cyclones represent one of the most severe natural hazards affecting residential buildings in Australia, causing extensive structural and non-structural damage. Brick veneer houses, as the predominant residential construction type in Australia's existing housing stock, are subject to a range of defects, from localised elemental damage to complete structural failure. Structural retrofitting is a timely requirement for these cyclone-affected houses to enhance structural resilience while complying with updated modern design standards. Hence, a structured analysis of cyclone-related defects in the existing housing stock remains a prerequisite for implementing optimal retrofitting solutions. Numerous post-cyclone damage assessments have been conducted. However, no established structured defect database is present that facilitates detailed defect information. Thus, this study proposes a structured cyclone defect database using a hybrid Natural Language Processing (NLP) approach to extract and structure the available defect data sources. The database comprises both qualitative and quantitative information on element-level defect variations, failure mechanisms, the effect of construction cohort on identified defects, and state-based vulnerability of cyclone defects. Analysis conducted on the structured-defect database reveals that roof systems are consistently identified as the most vulnerable building element, which accounted for more than 68% of the recorded defects, followed by defects associated with wall systems and openings. Wind-induced failure is the predominant failure mechanism affecting all the elements, while contributing to more than 60% of identified defects. Debris impact and wind-driven water ingress are the other two identified failure mechanisms, collectively accounting for 30% of the defects. Temporal trends in damage patterns were also highlighted through the age-cohort analysis. State-based comparative analysis further revealed similar vulnerability patterns with regional variations, reflecting differences in cyclone exposure conditions, construction practices, and regulatory evolution. Overall, the developed structured database presents a valuable resource for cyclone vulnerability modelling, prioritisation of retrofitting solutions, and future development of predictive tools for residential resilience.