Reliability based vulnerability modelling of metal-clad industrial buildings to extreme wind loading for cyclonic regions
Introduction
Extreme wind events (e.g. wind storms, cyclones, hurricanes) are one of the main natural hazards which cause damage to buildings and result in large economic losses in Australia and elsewhere (Middelmann, 2007, Holmes, 2007). The prediction of damage to buildings from extreme wind events is essential to developing policies to effectively reduce economic losses. Wind vulnerability models are used to predict the probability of damage to buildings and their contents due to wind loading. Vulnerability models play a key role in cost-benefit analysis which contributes to developing design procedures and other mitigation strategies to reduce economic losses due to severe wind events (e.g., Li and Stewart, 2011, Stewart, 2003, Stewart et al., 2014, Stewart, 2014, Vickery et al., 2006a). The models can be developed either by fitting curves to the actual damage data from historical wind damage records (i.e. empirical models and insurance data) or by using engineering knowledge to obtain the damage due to wind loading by investigating the behaviour of a building and its components (i.e. engineering models). Empirical models have drawbacks such as, lack of wind damage data (Ham et al., 2009), lack of capability to examine the changes in building design and construction methods, lack of ability to examine the effectiveness of building adaptation measures for climate change (Zhang et al., 2014). There are also a number of issues associated with utilising claim data such as; access to the insurance claim data, insurance valuation cost and the actual damage cost, and insurance claim databases that do not disaggregate losses between building exterior and interior (Pita et al., 2013). Moreover, empirical vulnerability curves are based on what has happened in the past. They cannot assess changes in vulnerability due to future changes in design standards, materials or construction practices. This highlights the need of developing vulnerability models based on engineering and structural reliability methods. It is noted however that, as with all models, engineering vulnerability models should be validated or benchmarked with empirical models based on past events where possible to give more confidence in modelling assumptions and realism.
There are several engineering vulnerability models developed for different types of structures which use reliability-based methods (Vickery et al., 2006a, Vickery et al., 2006b, Pinelli et al., 2004, Pinelli et al., 2008, Ham et al., 2009, Henderson and Ginger, 2007, Rosowsky and Ellingwood, 2002, Ellingwood et al., 2004, Unanwa et al., 2000, Lee and Rosowsky, 2005, Li and Ellingwood, 2006, Zhang et al., 2014, Ham et al., 2009, Lindt and Dao, 2009). Most of these models are developed for U.S. structure types and have mainly considered residential buildings such as single-family houses. Few publicly available engineering vulnerability models are found in the literature for Australian buildings. A suite of vulnerability curves were developed for different Australian building types by Geoscience Australia and James Cook University (Wehner et al., 2010) through the expert opinion of the Australian wind engineering community. The curves were developed based on the expert’s experience in post-event survey activity. Many of these curves are proprietary, however, some details are described by Wehner et al. (2010) and Ginger et al. (2010). Henderson and Ginger (2007) developed a reliability-based engineering vulnerability model for Australian high-set houses against wind loading. This study examined possible component and connections failures such as, roof cladding pulling over fixing, cladding fastener failure, batten joint failing at rafter and rafter joint failing at ridge. However, the features such as load redistribution based on progressive failure load paths, spatial distribution of wind load and internal pressure variation caused by the roof sheeting failure, were not considered in the Henderson and Ginger model. Consequently, it was not possible to determine the extent of roof damage at a given wind speed. Given that industrial buildings are vulnerable to extreme wind loading, particularly in the presence of a dominant openings in Australia, it is necessary to identify the extent of wind vulnerability of such buildings, and take actions to protect them against damage where appropriate. In Australia, gable roof metal clad industrial buildings are the most commonly used for manufacturing, storage and processing industries.
An engineering vulnerability model is developed in this paper for metal clad industrial buildings subject to wind loading, based on structural reliability, spatial variability, and probabilistic analysis. Roof sheeting failure is considered in this model which includes two main failure mechanisms (i) roof cladding failure at fastener (i.e. roof cladding pulling over fixing, fastener failure by tension or fastener pulling out of purlin) and (ii) purlin failure (i.e. purlin to rafter connection failure or purlin buckling failure). The external pressure coefficients are obtained from wind tunnel model testing. The vulnerability curves developed are for representative industrial buildings (i.e. hot rolled structural steel, metal-clad, gable-end industrial building) designed to current Australian building standards in cyclonic regions in Australia (North Queensland). Results are presented herein considering the effect of roof cyclone assemblage (washers) and large or dominant openings in the building envelope. Experience in recent cyclones in Australia suggests that some roller doors fail at their connections to the building, thus causing a dominant opening, leading to increased building damage (Henderson and Ginger, 2008). In this model, load redistribution after connection/component failure is incorporated based on the progressive failure load paths. This allows the model to track the timing and extent of fastener and purlin failure, which lead to loss of roof sheeting. Damage is defined as proportional loss of roof sheeting. Internal pressure is treated as a function of openings created by failed roof sheets due to wind load. The interdependency of the component failure is also considered in this model.
Section snippets
Stochastic model development
As discussed in the Introduction, a vulnerability model is developed herein for industrial buildings in cyclonic regions in Australia subject to extreme wind loading. Industrial buildings with spans of 20–40 m, lengths of 50 m or more, heights of 5–10 m, and gable-end low pitch (less than 10°) roofs are used in industrial applications in Australia. The structural systems of these buildings generally consist of portal or pin-jointed structural steel frames, spaced at 4 m to 8 m along the length of
Results
Results are presented for two wind directions namely 0° and 90°. The variation of the vulnerability due to presence and absence of large openings on the building envelope is also analysed. It is assumed that the dominant opening (e.g., a large access door) is always on the windward wall for this study. Finally the effect of using cyclone washers on roof vulnerability is also discussed. As a starting point we will assume that the industrial building in the cyclone regions in Australia has a
Discussion and future works
Due to lack of information available for input parameters, several assumptions and best estimates were unavoidable to derive the output results. One major assumption made herein is the number of fastener failures required to cause roof sheet failure. The components/connection strengths are also obtained based on component testing and expert judgments, thus require further investigations.
According to Fig. 5, the internal pressure reaches a constant value when four roof sheets fail. This roof
Conclusions
A wind vulnerability model was developed to predict the probability and extent of damage to metal-clad industrial buildings due to extreme wind loading. The model considered the spatial probabilistic characteristics of wind load and component/connection strength, and load sharing of failed components/connection for the roof envelope. Features such as load sharing and internal pressure variation with progressive failure are included in this vulnerability model. The vulnerability curves were
Acknowledgements
The authors gratefully acknowledge the financial support of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) Flagship Cluster Fund through the project Climate Adaption Engineering for Extreme Events, in collaboration with the Sustainable Cities and Coasts Theme of the CSIRO Climate Adaption Flagship.
References (42)
- et al.
Characteristics of large pressures in regions of flow separation
J. Wind Eng. Ind. Aerodyn.
(1993) - et al.
Fragility assessment for roof sheathing failure in high wind regions
Eng. Struct.
(2005) - et al.
Hurricane damage to residential construction in the US: Importance of uncertainty modeling in risk assessment
Eng. Struct.
(2006) - et al.
Safety indices for wind loading in Australia
J. Wind Eng. Ind. Aerodyn.
(1983) - et al.
Validation of a probabilistic model for hurricane insurance loss projections in Florida
Reliability Eng. Syst. Safety
(2008) - et al.
Hurricane vulnerability modeling: Development and future trends
J. Wind Eng. Ind. Aerodyn.
(2013) Cyclone damage and temporal changes to building vulnerability and economic risks for residential construction
J. Wind Eng. Ind. Aerodyn.
(2003)- et al.
Extreme value distributions for peak pressure and load coefficients
J. Wind Eng. Ind. Aerodyn.
(2008) - et al.
The development of wind damage bands for buildings
J. Wind Eng. Ind. Aerodyn.
(2000) - et al.
Reliability-based modelling of typhoon induced wind vulnerability for residential buildings in Japan
J. Wind Eng. Ind. Aerodyn.
(2014)