Probabilistic Models of Design Wind Loads in South Africa

Botha, Jacques (2016-10)

Thesis (DPhil)--Stellenbosch University, 2016.

Thesis

ENGLISH ABSTRACT: The formulation of design wind loads on structures is subject to multiple sources of uncertainties,both inherent in the physical process itself and due to the engineering models used in the formulation. Accurate reliability modelling of this process is required to ensure that the desired reliability performance of structures is achieved when using wind loading standards. This dissertation presents an investigation of wind load uncertainties and the development of a probabilistic wind load model based on a rational and transparent reliability basis. A background investigation of the wind load formulation and relevant wind engineering models was performed in order to identify sources of uncertainties. Through this process the primary wind load components relevant to the South African wind load formulation were identified. These components are the time variant free-field wind pressure, and the time invariant pressure coefficients and terrain roughness factors. The traditional approach to probabilistic modelling of wind loads was investigated. Relevant existing probabilistic models were investigated and compared. The lack of substantiating background information relating to the existing models calls into question the trustworthiness of results obtained using those models. This serves as the primary motivation for the research presented in this dissertation. The reliability basis of the investigation was therefore developed in a rational and transparent manner. To this end the traditional reliability approach was extended through the use of hierarchical Bayesian models to quantify wind load uncertainties on the component level. The three primary wind load components were investigated independently. This investigation follows the investigation by Kruger (2011) of the South African strong wind climatology and the development of a new characteristic wind speed map. The results from that investigation were used to quantify the free-field wind pressure uncertainties due to the inherent variability of the strong wind climate and the conservative bias introduced through the use of a wind speed map. Lacking sufficient sources of information to quantify the time invariant component uncertainties, a rational reliability method for the use of comparison of wind loads standards as an indicator of wind load uncertainties was developed. This methodology was used in conjunction with direct sampling of limited available observed data to quantify the uncertainties inherent in pressure coefficients and terrain roughness factors. The new component models were compared to the corresponding distributions in the existing models. In all cases the new models resulted in higher reliability requirements than the existing models. Full probabilistic wind load models which include secondary factors were then developed using engineering judgement and Bayesian updating of the existing models. Multiple reliability assessments were performed using the new models to investigate the reliability performance of the South African wind load formulation. The results conclusively show that the current wind load partial factor of 1.3 does not provide adequate reliability performance. Finally, the effect on the average design wind load across the country due to the combined implementation of the new characteristic wind speed map and a potentially updated wind load partial factor was quantified.

AFRIKAANSE OPSOMMING: Die formulering van ontwerp windlaste op strukture is onderhewig aan verskeie bronne van onsekerheid, beide inherent in die fisiese proses self en as gevolg van die ingenieursmodelle wat gebruik word in die formulering. Akkurate betroubaarheids modellering van hierdie proses is nodig om te verseker dat die gewenste vlak van betroubaarheid van strukture bereik word wanneer windlas standaarde gebruik word. 'n Ondersoek van windlas onsekerhede en die ontwikkeling van 'n probabilistiese windlas model wat gebaseer is op 'n rasionele en deursigtige betroubaarheids basis, word in hierdie verhandeling aangebied. 'n Agtergrond ondersoek van die windlas formulering en relevante windingenieurswese modelle is uitgevoer om bronne van onsekerheid te identifiseer. Deur hierdie proses is die primêre windlas komponente relevant aan die Suid-Afrikaanse windlas formulering geïdentifiseer. Hierdie kom-ponente is die tyd-afhanklike vryveld winddruk, en die tyd-onafhanklike druk koëffisiënte en terrein ruheid faktore. Die tradisionele benadering tot probabilistiese modellering van windlaste is ondersoek. Relevante bestaande probabilistiese modelle is ondersoek en vergelyk. Die gebrek aan stawende agtergrond inligting wat verband hou met die bestaande modelle bring die geloofwaardigheid van die resultate verkry deur die modelle in gedrang. Dit dien as die primêre motivering vir die navorsing wat in hierdie verhandeling aangebied word. Die betroubaarheids basis van die ondersoek is dus op 'n rasionele en deursigtige wyse ontwikkel. Met die oog hierop is die tradisionele betroubaarheids benadering uitgebrei met gebruik van hiërargiese Bayesiese modelle om windlas onsekerhede op die komponent vlak te kwantifiseer. Die drie primêre windlas komponente is onafhanklik ondersoek. Hierdie ondersoek volg die ondersoek deur Kruger (2011) van die Suid-Afrikaanse sterk-wind klimatologie en die ontwikkeling van 'n nuwe karakteristieke windspoed kaart. Die resultate van daardie ondersoek is gebruik om die vryveld winddruk onsekerhede, as gevolg van die inherente variasie van die sterk-wind klimaat asook die konserwatisme weens die gebruik van 'n windspoed kaart, te kwantifiseer. Weens 'n gebrek aan voldoende bronne van inligting om die onsekerhede van die tyd-onafhanklike komponente te kwantifiseer is 'n rasionele betroubaarheids metode vir die gebruik van vergelyking van windlas standaarde as 'n aanduiding van windlas onsekerhede ontwikkel. Hierdie metode is in samewerking met direkte steekproefneming van beperkte beskikbare waargenome data gebruik om die onsekerhede inherent in druk koëffisiënte en terrein ruheid faktore te kwantiffiseer. Die nuwe komponent modelle is vergelyk met die ooreenstemmende statistiese verdelings in die bestaande modelle. In alle gevalle het die nuwe modelle gelei tot hoër betroubaarheids vereistes as die bestaande modelle. Volle probabilistiese windlas modelle wat sekond^ere faktore insluit is daarna ontwikkel met gebruik van ingenieurs oordeel en Bayesiese opdatering van die bestaande modelle. Betroubaarheid assesserings is uitgevoer met die nuwe modelle om die betroubaarheid van die Suid-Afrikaanse windlas formulering te ondersoek. Die resultate wys onweerlegbaar dat die huidige windlas parsiële faktor van 1.3 onvoldoende betroubaarheid voorsien. Ten slotte is die effek op die gemiddelde ontwerp windbelasting regoor die land as gevolg van die gekombineerde implementering van die nuwe karakteristieke windspoed kaart en 'n potensieel opgedateerde windlas parsiële faktor gekwantifiseer.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/100058
This item appears in the following collections: