project summary: 

Australian detached dwellings represent a significant portion of the national housing stock, yet many perform poorly under increasing climate extremes, driving homeowners to seek ways to reduce high heating and cooling loads. Envelope retrofit can cut these loads by between 33% to 83%, making it one of the most effective improvement pathways. Despite this potential, retrofit decision-making remains complex, expert-dependent, and inaccessible to everyday homeowners. 

This research develops an envelope-retrofit decision support system that simplifies and streamlines retrofit decisions for detached dwellings within their site contexts. The project begins by characterising the decision context through analysis of Australian standards, guidelines, and benchmark values, supplemented by interviews with homeowners and practitioners to identify key decision criteria, including cost. It then establishes an accessible data collection and fusion process combining archival records, plans, remote-sensed imagery, and handheld mobile-device scans, with each attribute tagged for provenance and confidence.

Established segmentation methods classify envelope elements, openings, materials, and performance characteristics while preserving confidence information. A knowledge graph integrates standards, dwelling attributes, uncertainties, and retrofit options to support uncertainty-aware reasoning and strategy ranking. A large language model, constrained to the reasoning trace, translates ranked recommendations into plain language. The system is validated on a purposive stratified sample of dwellings across three South Australian climate zones and four construction periods, with practitioners assessing technical validity and homeowners evaluating usability. 

The study expects to help households reduce energy costs, improve thermal performance, and strengthen resilience to climate extremes by combining digital capture, data fusion, uncertainty-aware reasoning, and constrained language-model explanations.