Impact of using air temperatures versus surface temperatures for the determination of U-values

Floor U-value models are based on the assumption that room temperatures are

homogeneous throughout - see Chapter 2.3. This section illustrates that in reality room temperatures are not homogenous as also reported by Gauthier (2014)and MING XU (2001), and discusses the impact of this on U-value estimation. Inhomogeneity of room air

temperatures is likely caused by a complex interplay of convective currents in the room from heating elements and from ventilation and air infiltration and from different surface

temperatures.

At present it is poorly characterised where temperatures should be measured in non-homogenous spaces for the purpose of in-situ floor U-value estimation and this is further complicated by the practicalities of placing sensors, particularly in occupied houses. For walls, Siviour (1982) suggests internal air temperatures should be measured within 500mm from the sensor while Doran (2001) placed air temperature sensors 10mm away from the heat-flux sensor on the wall and a few centimetres away by BRE (2014a). No

recommendations were found for floors though Thomas (1999) described temperature sensor locations approximately 12mm above the solid ground floor surface fixed on nylon wire on a timber structure. The uncertainty of temperature determination in

non-homogenous spaces (as also discussed in Section 3.3), could make comparison between modelled and in-situ measured whole floor U-values more challenging.

The effect of temperature sensor height on U-value estimation was investigated in a position at the middle of the Salford EH floor (location 7, see Figure 42.). U-values were estimated in accordance with Equation 48. for air-to air U-values and Equation 47. where surface

temperatures were used, with RSi assumed to be 0.17 m²KW^-1 (BSI, 2007) and RSe set to 0.

Measurement uncertainties were obtained as described previously in Equation 50., which reflects the natural variation of the U-value and ±5% uncertainty associated with

temperature measuring locations. The observed heat-flow q and the external air

temperature Tea were the same in each Up-value estimate.Figure 42. shows the variation of the estimated Up-values (black data points) with temperatures measured at different heights (grey data points), ranging from 0.64 ±0.04 to 0.89 ±0.05 Wm^-2 K^-1 and indicates a general trend where estimated Up-values derived from air temperatures decreased when the height of the measured air temperature in the room increased, corresponding to the observed temperature gradient.

The un-adjusted surface-to-air Up-value (0.88 ±0.05 Wm^-2K^-1) was similar to the estimated Up-value with air temperatures measured at 100mm height, i.e. 0.89 ±0.05 Wm^-2 K^-1. However, Figure 42. also highlights the impact of the surface resistance on the estimated Up-value: including the standard surface resistance of 0.17 Wm^-2K^-1 lead to a relative change of −14% in estimated U-value from 0.88 ±0.05 Wm^-2K^-1 to 0.77 ±0.04 Wm^-2K^-1, which was similar to the 600mm air-to air U-value estimate of 0.79 ±0.04 Wm^-2K^-1. If assuming that the 600mm air temperature is a proxy for ambient temperature, a surface boundary layer thermal resistance (RSi) of 0.15 m²KW^-1 was estimated in location 7,¹² slightly below the assumed RSi of 0.17 m²KW^-1. Further research is required whether (a.) the 600mm air

temperature is an appropriate proxy for room ambient temperature and (b.) whether this is also the case for other locations on the floor, especially closer to the perimeter where there might be less laminar flow and possibly reduced boundary layer thickness. Additionally, it is unknown what the thermal resistance of the actual boundary layers are in occupied

dwellings, where the air is constantly disturbed.

While it isn't possible to identify the true heat-flow paths from these measurements, these results do highlight the potential impact of temperature sensor location on U-value estimation, though some values are within the margins of measurement uncertainty.

Figure 42. Estimated in-situ U-values in location 7 (middle of the floor) at the Salford EH, with differently estimated U-values (black data points) when derived with different internal temperatures (grey data points) at different heights in the room (x-axis). Error margins are in accordance with Equation 50.

12 Simplified calculation: RSi= (Tair-Tsurface)/q

Buoyancy effects tend to increase internal room temperatures as the height from the floor increases: temperatures at 100mm height were 2.4ºC to 4.3ºC lower than those observed at 1100mm and 1700mm height respectively¹³ in the Salford EH study. In general, using higher measured temperatures increased the apparent internal-external temperature difference (∆T), resulting in lower U-value estimates. However, U-values estimated from temperatures measured at the surface - before adjustment- and at just 100mm height do not appear significantly different; though diverge once the former is adjusted with the surface resistance. This is contrary to observations by Stinson (2012) who did not find significant differences in U-value when low-level skirting air temperature was used (2.5± 0.3Wm^-2K^-1, analysed in accordance with Equation 48.), compared to the U-value derived from internal surface temperatures (2.4± 0.2 Wm^-2K^-1, analysed in accordance with Equation 47. with RSi

adjustment). It is unknown which space heating method was used in this study; it is expected that heating method might influence U-value estimation. For instance in co-heating tests, space heating is usually provided by electrical fan heaters and circulation fans are used to mix the internal air (Wingfield, 2010b) to create less of a vertical temperature gradient.

However such heating method is not representative of occupied spaces and is likely to change the heat-flow paths being measured.

The surface resistance adjustment used in surface-to-air U-value estimations assumes constant radiation and laminar (non turbulent) airflow across a sealed building element, but for example in reality external surface thermal resistances are likely to change with changing wind-speeds (see Chapter 3). Studies reported stack airflow through gaps and cracks of floors into the internal spaces (see discussion Chapter 2.2); such infiltration through gaps between floorboards is likely to disrupt the laminar flow, changing the effective surface resistance of the floor, in addition to contributing to mass-transfer heat-flow. However the impact of stack airflow on heat-transfer and floor surface resistances is unknown and difficult to characterise.

It might be argued that a variable surface resistance (in different floors and for different locations on the floor) might be more appropriate than a constant value, depending on air infiltration through the floor, but this may be difficult to define and apply. Any

overestimation of this surface thermal resistance would lead to underestimated U-values when U-values are adjusted for with RSi. As illustrated, the significant differences in suspended timber ground floor U-values estimated via methods with and without the inclusion of a large surface thermal resistance, and possible variations in the actual surface resistances, could lead to the inaccurate estimation of U-values.

This also makes comparison of estimated U-values between models and in-situ measured sources as well as between different sources which use different methods with different floor characteristics (e.g. different floor infiltration rates) more challenging. As variation in surface resistances can impact on estimated U-values, a better understanding of this factor, and how it is affected by different floor system variables, would enhance estimation of in-situ floor U-values. Additionally, there is also an uncertainty associated with characterising in-situ heat-flow via internal air temperature measurements in non-homogenous spaces, however further study of both these issues is beyond the scope of this PhD study.

On the basis of this work, estimating U-values from surface temperature to external air temperature with RSi adjustment was considered preferable for the following reasons:

• Regardless of where temperatures are measured, all of the obtained U-values are 'valid' U-values though not necessarily all are representative of the heat-flow path through floors, which remains undefined at present. Measuring closer to the floor's surface might be more representative of the floor's heat-flow path compared to high up in the room and further away from the floor surface. However measuring on the floor surface requires adjustment with RSi, with associated uncertainties.

• Having undertaken different temperature measurements, it is more practical and convenient to monitor surface temperatures than internal air temperatures; air temperature sensors need to be suspended from the ceiling or tripods which is not practical in the middle of the room and disruptive in occupied dwellings. Suspended temperature sensors are prone to sagging/falling of or more exposed to

occupant/researcher influence while in the room.

• The use of floor surface temperatures might be more replicable as the floor surface is clearly defined.

Given the uncertainty around temperature measurements, for the purpose of this PhD and its U-value estimates, the ISO-9869 estimated ±5% air temperature error was therefore included in all U-value estimates and comparisons, including where surface temperatures were used (when this should not apply according to the ISO-9869 standard). This is to reflect some uncertainty arising from the addition of the surface thermal resistance and its impact on the estimation of U-values. It is however acknowledged that the actual effect could be much greater or smaller than the ±5% error allowed for; further research would be required to investigate this.