Characterisation and analysis of uncertainties in building heat transfer estimates from co-heating tests

In recent years, measurement protocols for the estimation of the total aggregate building heat transfer coefficient (HTC) have provided sufficient empirical evidence to indicate that buildings often do not perform as intended. However, little research has been carried out into the associated uncertainties. Within this context, this paper reviews sources of uncertainty associated with co-heating tests; characterises these uncertainties and their impact on HTC estimates; and devises a method for the calculation of HTC uncertainty. The method proposed was applied to 14 co-heating tests, showing estimated total uncertainty ranging between 2.2-21.1 Image 1 (or 4.6-26.7% of the measured value) with a mean of 10.1 Image 1 (or 8.7%). The natural variation of HTC and often-observed inaccuracy of design calculations (the ‘prediction gap’) suggest that more accurate measurements may be of little benefit. Additionally, results suggest that weather conditions, challenging building design and poor experimental technique can all significantly contribute to HTC uncertainty. However, when suitable buildings are tested by experienced technicians and under suitable weather conditions, HTC estimates from the co-heating protocol are likely to provide a useful tool to assess and understand real-world building fabric performance.