 Source (www.bygg.ntnu.no/batek/dr_ing/okland/summaryokland.htm)
Summary
Moisture is certainly the most frequent reason for damage to building structures in Norway. Moisture transport in building materials involves many transport mechanisms and has a very complicated theory. The moisture content in the material is closely connected to the heat- and air balance of the structure. One of the transport mechanisms that needs more research is natural and forced convection.
The objectives of this work have been to increase knowledge of the effects natural and forced convection on the hygrothermal performance of highly-insulated building structures. The focus has been on wood timber-frame structures insulated with mineral wool. This is the most common building structure used in Norwegian single-family houses. The work was divided in three main parts; a thorough study of relevant literature, computer modelling/simulations and measurements.
A network model was used to study air leakages in buildings. Stochastic behaviour was included to study the importance of workmanship, material properties and climate variation. The measurements were conducted to find material properties to use in the computer programs. Measurements on full-height wall systems have also been done in both climate chamber and a test-house.
The climate chamber measurements included six timber frame walls with 300 mm of glas fibre insulation. One wall was perfectly insulated, while the other included different air layers in the insulation cavity to study bad workmanship. The walls were initially supplied with water in order to include effects of built-in moisture. Measurements of the temperature profile through the walls clearly indicated stronger natural convection in walls with air gaps along lining material than for perfectly insulated walls or walls with air breaks. The six walls were also examined during exposure to forced convection. The walls were exposed to a maximum of 10 Pa pressure difference. The air was flowing into the wall through a slit in the lower part on the warm side and out of the wall at upper part on the cold side. The air was also controlled to flow in the opposite direction. An estimation of the accuracy of moisture-content measurements in wood showed that it was difficult to identify relatively small convection effects without using HAM computer programs.
The perfectly insulated test-house walls had 150 mm insulation. Results from the test-house measurements clearly showed an effect of forced convection. The effect was surprisingly high for walls that did not have exposure to air leakage similar to that in forced convection measurements in climate chambers. The reason may be variation in the climate exposed to these walls. The measurements indicated that the walls were exposed to small pressure differences. The walls had not been heated significantly by convection, which gives a high amount of condensation.
A two-dimensional model called SIMPLE-FULUV was developed and used to model composite systems with both an air gap and a porous material. The simulations, that included Navier Stokes equations in the air gap, indicated that there was both an upwards and downwards air flow in the air gap. This was not found with other models using Darcy's equation both in the air gap and the porous material. The deviations were especially high for a wall with a large air break. Navier Stokes equations used in the new model did required a very short time-step and thus are difficult to use in annual moisture content modelling with existing computer tools.
Simulations with the LATENITE model were compared to measurements of natural convection effects in climate chambers and the agreement was found to be satisfactory. The effect of natural convection on moisture content of timber-frame members was not larger than the accuracy of the measurements. The maximum simulated local Sherwood number in spruce for the six wall systems was 1.15 for the wall with air gaps on both sides the insulation. This means that natural convection could increase the moisture content in spruce from 0.2 kg/kg to 0.23 kg/kg. The effects of natural convection on moisture content were much larger in the insulation than in the wood. The additional simplified SIMPLE-FULUV simulations agreed reasonably well with LATENITE and the measurements. The effects of natural convection were however significantly higher in the upper parts of insulation and spruce for these simulations. The effects were especially high for the case with an air break along the top plate.
Additional simulations with LATENITE were performed on a wall with air gap at the cold side of the insulation. The effects of natural convection were significantly reduced when built-in moisture was removed, reduced air permeability was used or when a vertical air barrier was installed. Long-term simulation with real climate from V?rnes exposed to a wall with an air gap on both sides of the insulation gave a maximum local Sherwood number equal to 1.08. The width of the air gaps was half of that used in the laboratory wall and the built in moisture was in wood. This indicates that natural convection does not influence traditionally-designed wall systems very much as far as the hazard of mould growth is concerned.
The investigations of air leakages showed that network models were useful to find the importance of leakage in different building structures. These calculations indicated that roof structures was most sensitive to air leakage. A stochastic model based on the Monte Carlo method was developed to find the importance of different climate parameters and workmanship. The simulations showed that the climate seems to be more important for the amount of condensation than the different variations of workmanship defects. Several of the variables that determine the convection effects have a typical stochastic behaviour, but it remains to see if this work will encourage more use of stochastical models in the future.
The simulation models are useful to find the effect of convection since this effect can be turned off in the simulations. The simulations are also effective to investigate many cases of long term performance of structures that are too time-consuming for a measurement investigation. The network model was useful to investigate climatic influence of a whole building. Combined with a stochastic model this was a very powerful tool to find the hazard of moisture damage. Measurements are necessary to verify simulations. This thesis shows that highly insulated building structures should be built with mineral wool with low air permeability. One must also try to avoid large air spaces in the insulation cavity. | 