Abstract
Houses in general, are exposed to adverse weather conditions such as winds, snow, rainfalls, heat and
cold that can damage their roofs. Houses are built to provide shelter to humankind against these adverse
weather conditions. However, during extreme wind events such as storms, hurricanes and tornadoes, the
roof is the most fragile component of the house, in that it can be easily damaged. Strong or extreme winds
are very disruptive and have been seen over the past centuries as major agents of damage to structures such
as housing roofs, bridges, buildings and other structures. Strong winds have been categorized by the World
Meteorological Organisation (WMO) as those with a minimum speed of 17 m/s. The world meteorological
encyclopaedia of weather and climates records that the highest strong wind speed ever attained across the
globe during extreme weather events was 483 km/h or 134.2 m/s. The destructive effects of these winds
has paved the way to wind engineering through the construction of boundary-layer wind tunnels, the
development of wind codes and the drafting of extreme wind maps, in order to build structures strong
enough to resist extreme weather events. Most wind codes however, are limited to straight winds, whereas
there also exist turbulent and vortical winds, which are even more disruptive. Wind damage to housing
roofs can be mitigated if their most vulnerable components are assessed through the study of their structural
performance under extreme winds.
This research was aimed at defining and comparing the structural performance of housing roofs of
metal sheeting and tiled covers under extreme winds. The general principle in assessing structural behaviour
of housing roofs with regard to wind damage has always consisted of quantifying wind pressures acting on
roofs, assessing the damage of roofing components through laboratory experiments, and predicting strong
winds damage to roofs. Extreme winds on roofs can be quantified using various means such as the wind
tunnel laboratory, pressure chambers and fans Wall-of-Winds (WOW). Structural performance of housing
roofs is often assessed by subjecting roof samples to surface forces and evaluating their structural behaviour
as they fail under laboratory conditions. Two main methods have generally been used for predicting roofs
damage under strong winds consisting of the Monte Carlo Simulation (MCS) and the engineering-based
method. The MCS combines the probabilistic occurrence of extreme winds with the probabilistic failure of
roofing components into a log-normal distribution function. Whereas, the engineering-based method makes
use of load resistance factored design (LRFD) to predict damage. Wind codes have served as guidelines for
quantifying winds on housing roofs as they mostly rely on probabilistic distribution of winds over a region
and on pressure distributions on low-rise buildings. Most wind pressure distributions quoted in wind codes
are based on wind tunnel experiments conducted on reduced-scale building models. Although the existing
wind codes are helpful tools for predicting wind distributions on structures, most of them are limited to
straight winds.
v
Engineering wind tunnels have been built based on fluid dynamics with the concept of the
atmospheric boundary-layer. According to fluid dynamics, winds can be analyzed in terms of flows as
straight, turbulent and vortical. Most extreme wind events such as storms, tornadoes and hurricane are
turbulent and vortical. However, some wind tunnels have provided pressure coefficients of straight and
turbulent winds, except for vortical winds. Some wind codes such as Eurocode have included turbulent
winds; whereas vortical winds have only been used by meteorologists to define winds such as tornado and
hurricanes. Wind codes in general are limited to straight winds, whereas some of them include turbulent
winds. Vortical winds have not yet been included in wind codes due to challenges related to quantifying
extreme winds. As such, arrangements had to be made in order to use all types of wind flows in this study.
Some studies on housing roofs have confirmed that the cover, purlins or battens, rafters and coverto-
truss connections were the most fragile roofing components during extreme wind events. As such, most
damage to housing roofs has been assessed using these components as samples. Some studies in the
literature have revealed that the structural performance of metal sheeting roofs depended on the resistance
of cover-to-truss connections; whereas that of tiled roofs depended on tiles individual resistance and their
ability to form a whole system. Roof samples in this study were therefore constructed of cover, purlins or
batten and truss top chords or rafters. Samples were assembled in compliance with the South African
contemporary houses and building regulations (SANS 10400:2010).
The present research relied on the engineering-based method, and was conducted in three (3) phases:
quantification of wind loads on roofs, assessment of structural performance through laboratory experiment,
and prediction of damage based on structural performance. Twelve (12) Representative samples of metal
sheeting and tiled roofing systems were assessed under laboratory conditions to ascertain fragility of roofs
to strong winds. These samples consisting of normal and hurricane connections, represented six (6) South
African single-storey contemporary residential homes of flat, gable and hipped roofs respectively. These
houses comprised three (3) models of metal sheeting cover as well as three (3) of tiled roof cover. These
houses could be of normal connections or of hurricane connections; hence the use of twelve (12) samples.
The first phase which consisted of quantifying winds on metal sheeting and tiled roofs was based on
the use of the South African code (SANS10160-3: 2019), the Eurocode (EN 1991-1-4: 2005), fluid
dynamics and computational fluid dynamics (CFD). The South African code (SANS 10160-3:2019) has
been employed to define geographical features of houses for exposure to South Africa’s extreme winds
climate. Straight winds were quantified based on the South African code, turbulent flows based on the
Eurocode (EN 1991-1-4:2005), and vortical winds based on fluid dynamics. Arrangements were made for
vortical speed at the core to be equal to 134.2 m/s. Wind loads were therefore determined for speeds up to
vi
134.2 m/s. These loads have been combined with dead loads in uplift at ultimate limit state (ULS) to
determine experimental loads to be applied as surface forces on samples under laboratory conditions.
Combined uplift forces varied as per cover from one type of geometry to another, and from one flow to
another. CFD served for simulating different types of flows on house models. CFD results were compared
with wind codes and fluid dynamics results in order to validate calculations.
The second phase consisted of applying experimental loads on different samples in order to assess
their failure under laboratory conditions. Metal sheeting and tiled roof samples of the same dimensions
were subjected to surface forces that caused failure of components and connections, while loads were
recorded with corresponding deflections. Metal sheeting samples failed at average maximum load and
deflection of 11.6 kN and 55.5 mm, respectively; whereas tiled roofs failed at average load and deflection
of respectively 3.9 kN and 39 mm. Damage of metal sheeting samples was caused by removal of roof nails
and detachment of purlins from rafters by lack of clips. Whereas damage to tiled samples was caused by
breaking of individual tiles in a row parallel to battens. Experimental results were obtained in terms of loaddeflection
curves, which presented different features depending on whether metal sheeting or tiled roofs.
The third phase consisted of analysing experimental data and conducting mathematical modelling in
order to predict damage. Experimental data showed that metal sheeting roofs’ load-deflection curve was
parabolic and depended on the cover’s mechanical properties and pull-out strengths of nails. Experimental
results of tiled roofs showed a tile-to-tile succession of parabolic curves that depended on each single tile’s
breaking force and mechanical properties, as well as tile-to-tile interlocking force. It was observed that
metal sheeting roofs deflected like simply supported beams, as they were held by long roof nails. The
maximum deflection recorded by the loading device was equal to the ultimate pull-out length of nails. Tiled
roofs, however, deflected like simply supported rectangular plates that broke following a row parallel to
battens. Structural performance of metal sheeting roofs has been predicted as a parabolic load-deflection
curve which depended on mechanical properties of the cover and the nails, such that load was a function of
deflection. However, structural performance of tiled roofs was predicted as a tile-to-tile succession of
parabolic load-deflection curves which depended on mechanical properties and breaking force of each
single tile such that, the load was a function of deflection. The relationship between load and deflection has
allowed establish a relationship between speed and deflection such that, speed was a function of deflection.
Damage prediction made on metal sheeting and tiled samples were approximately similar to experimental
results.
It has been concluded that the structural performance of metal sheeting roofs under extreme winds
depended on the cover’s mechanical properties and the removing strength of roof and purlin-to-truss nails.
vii
The structural performance of tiled roofs under extreme winds depended on the resistance and mechanical
properties of each individual tile, their resistance at the edges of the roof, and their interlocking ability in
forming the cover. Damage of these systems to strong winds can be predicted in terms of load-deflection
or speed-deflection curves with results that are reliable. As such, structural performances of roofs of
different house models were determined by prediction.