Abstract
HVAC systems account for 40-50% of building energy use, yet over 90% of installations underperform. A key, often overlooked contributor is the gap between designed and as-built ductwork: installation constraints, undocumented modifications and ageing introduce concealed structural and parametric changes that alter pressure-flow behavior and compromise indoor air quality and energy performance. Because commissioning, control, and fault detection methods typically assume a correct topology, their effectiveness is undermined by such discrepancies. This PhD develops a scalable, measurement-efficient methodology to reconstruct the as-built aeraulic network from a design prior using sparse, commissioning-compatible measurements. A graph-based simulation environment, multi-regime excitation, and a two-layer discrepancy parametrization enable systematic identification of structural and parametric deviations. A physics-informed reconstruction algorithm with ambiguity-aware hypothesis ranking is complemented by an information-theoretic sensing strategy determining minimal sensor placement and excitation requirements. Validation progresses from controlled benchmarks to lab experiments and a building-scale demonstration. Deliverables include an as-built digital-twin reconstruction framework under minimal sensing, advancing scientific insight in network identifiability, sparse sensing, and multi-regime inference, to enable more reliable and energy-efficient HVAC operation.
Researcher(s)
Research team(s)
Funding
Project type(s)