Critical load effects on bridges using heavy vehicle weigh-in-motion data

Stap, Tyron (2022-04)

Thesis (MEng)--Stellenbosch University, 2022.

Thesis

ENGLISH ABSTRACT: This study forms part of a set of research working towards updating South Africa’s bridge design codes. According to literature, Weigh-In-Motion (WIM) data is commonly used to develop traffic load models. Using traffic load models in conjunction with statistical distribution families, the future critical load effects (LE) for a bridge are predicted for the design life, i.e., characteristic values. Previous authors have typically grouped vehicles within the WIM data set as "mixed-axle vehicle groups" while characteristic values are determined using the Extreme Value (EV) distribution family. However, it is argued that using a mixed-axle vehicle group violates the identically and independently distributed (iid) requirements of EV theory. The aim of this thesis is to determine whether the vehicle subsets from the data are in better agreement with the requirements of EV theory as opposed to the mixed-axle vehicle group. In this study, three splitting levels were investigated to determine if the iid nature of the data sets would improve by sub-dividing the vehicles in the WIM data files. At the first level, no splitting took place, hence, the data sets on this level represented the mixed-axle vehicle groups commonly used in literature. At this point in the study, it was decided to limit the study to focus on normal traffic loading caused by single-vehicle events. On the second splitting level, the vehicles recorded in the WIM data files were separated into different groups based on the number of axles each group had. The groups on this level were referred to as the "sub-axle groups". The final splitting level involved grouping the vehicles into groups based on the geometry of the vehicles found within each sub-axle group. These groups were referred to as the "sub-categories of sub-axle groups". The cause of the underlying distributions found in the previous level could be identified within each splitting level. By splitting the WIM data set more iid and non-iid data sets were introduced into the study. This led to an overall increase in the percentage of data sets with an underlying Fréchet distribution. Splitting had led to obtaining multiple vehicle subsets some of which were in better agreement with EV theory requirements than the mixed-axle vehicle group and while others were not. However, splitting had allowed for the identification of erroneous vehicle records which improved the iid nature of the mixed-axle vehicles data sets. In the end, splitting had resulted in the mixed-axle groups’ data sets better adhering to the requirements of EV theory. Overall, splitting the vehicle data sets allowed for a better understanding of the causes of the underlying distributions and improved the iid nature of the LE data sets for the mixed-axle group. However, splitting is not recommended for the sole purpose of predicting characteristic values for bridges. Instead, it is better suited for identifying potential erroneous vehicle records to allow additions to the filtration used during the cleaning of the WIM data file.

AFRIKAANSE OPSOMMING: Hierdie studie vorm deel van 'n stel navorsing om Suid-Afrika se brugontwerpkodes te opdateer. Volgens literatuur word Weeg-In-Beweging (WIB) data algemeen gebruik om verkeerslasmodelle te ontwikkel. Deur gebruik te maak van verkeerslasmodelle in samewerking met statistiese verspreidingsfamilies, word die toekomstige kritieke laseffekte (LE) vir 'n brug vir die ontwerplewe voorspel, dit wil sê kenmerkende waardes. Vorige skrywers het gewoonlik voertuie binne die WIB-datastel as "gemengde-asvoertuiggroepe" gegroepeer, terwyl kenmerkende waardes bepaal word deur gebruik te maak van die Ekstreem Waarde (EW) verspreidingsfamilie. Daar word egter aangevoer dat die gebruik van 'n gemengde-as voertuiggroep die identies en onafhanklik verspreide (iov) vereistes van EW teorie oortree. Die doel van hierdie tesis is om te bepaal of die voertuigsubversamelings uit die data beter ooreenstem met die vereistes van EW-teorie in teenstelling met die gemengde-as voertuiggroep. In hierdie studie is drie splitsingsvlakke ondersoek om te bepaal of die iov aard van die datastelle sal verbeter deur die voertuie in die WIB datalêers te onderverdeel. Op die eerste vlak het geen splitsing plaasgevind nie, dus het die datastelle op hierdie vlak die gemengde-as voertuiggroepe verteenwoordig wat algemeen in literatuur gebruik word. Op hierdie stadium van die studie is 'n besluit geneem om die studie te beperk om te fokus op normale verkeerslading wat veroorsaak word deur enkelvoertuiggebeure. Op die tweede splitsingsvlak is die voertuie wat in die WIB-datalêers aangeteken is, in verskillende groepe geskei gebaseer op die aantal asse wat elke groep gehad het. Die groepe op die vlak word verwys na as "sub-asgroepe". Die finale splitsingvlak het die voertuie gegropeer gebaseer op die geometrie van die voertuie wat binne elke sub asgroep gevind word. Hierdie groepe bestaan bekend as die "subkategorieë van sub-asgroepe". Die oorsaak van die onderliggende verdelings wat in die vorige vlak gevind is, kon binne elke splitsingsvlak geïdentifiseer word. Deur die WIM-datastel te verdeel, is meer iid- en nie-iid-datastelle in die studie ingebring. Dit het gelei tot 'n algehele toename in die persentasie datastelle met 'n onderliggende Fréchet-verspreiding. Splitsing het gelei tot die verkryging van veelvuldige voertuigsubstelle waarvan sommige beter ooreenstem met EV-teorievereistes as die gemengde-asvoertuiggroep en terwyl ander nie. Splitsing het egter voorsiening gemaak vir die identifikasie van foutiewe voertuigrekords wat die iid aard van die gemengde-asvoertuie se datastelle verbeter het. Op die ou end het splitsing daartoe gelei dat gemengde-as groepe se datastelle beter voldoen het aan die vereistes van EV teorie Oor die algemeen het die verdeling van die voertuigdatastelle 'n beter begrip van die oorsake van die onderliggende verspreidings moontlik gemaak en die iid-aard van die LE-datastelle vir gemengde-asgroepe verbeter. Splitting word egter nie aanbeveel vir die uitsluitlike doel om kenmerkende waardes vir brûe te voorspel nie. In plaas daarvan is dit beter geskik om potensiële foutiewe voertuigrekords te identifiseer om toevoegings tot die filtrasie wat tydens die skoonmaak van die WIM-datalêer gebruik word, toe te laat.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/124645
This item appears in the following collections: