Accurate measurement of thermal conductivity is not quite as straightforward as the simple steady-state theory would suggest.

The most commonly used and reliable apparatus for measuring thermal conductivity of insulation and other poor thermal conductors are guarded hot-plates and heat-flow meters. These instruments create a steady-state temperature gradient across a specimen of the material, by sandwiching it between an isothermal heated-plate and cold-plate. Additional thermal guards are positioned so as to minimise any loss or gain of heat from the heated-plate or from the edges of the specimen. The amount of power required for the heated-plate to create a particular temperature gradient will be proportional to the thermal conductivity of the material.

When considering the measurement of insulation products, the low heat-fluxes involved mean that the specimen and plates need to be thermally isolated, to a high degree, from the ambient environment. There are also associated problems such as thickness effects due to heat transfer by radiation, and the effects of moisture content and material homogeneity. With higher thermal conductivity materials such as plastics and masonry products, there are other factors including the effect of contact resistance between the specimen and the temperature sensor, which need to be taken into account.

Each of the four measured parameters (heat flux, temperature drop, specimen thickness and area) are directly related to the calculation of thermal conductivity, which means that any uncertainty in these parameters is transferred directly to the overall measurement uncertainty for thermal conductivity. National laboratories, such as NPL, would expect to measure an insulation material at ambient temperatures and obtain agreement to ± 1%, accredited laboratories within ± 3% and other test laboratories to within ± 5%. For higher conductivity materials and higher temperatures the agreement would not be as good.

Details of Thermal Measurement Services can be found at Thermal Performance.