As the energy transition accelerates, improving wind energy efficiency and forecasting becomes increasingly critical. One key challenge lies in reconstructing high-fidelity atmospheric boundary layer (ABL) flow fields from sparse measurements, especially in regions influenced by wind turbines. This thesis explores the use of Latent Diffusion Models (LDMs) to reconstruct physically plausible ABL flow fields from limited spatial data. Where previous work focused on homogeneous, neutral ABL states, this research extends the methodology to more diverse and realistic conditions by including both stable and neutral boundary layers with embedded wind farms. A conditional diffusion model is trained in the latent space of an autoencoder (AE), using both local measurements and global atmospheric labels. The model is evaluated using statistical, physical, and spectral metrics, and its ability to generalize across multiple ABL regimes is analyzed. Results show that the model can generate realistic reconstructions from extremely sparse inputs, including turbine wake structures, and can potentially serve as a tool for initializing Large Eddy Simulations (LES). These findings mark an important step toward integrating generative models into operational wind forecasting and control systems.

Steel Orthotropic Bridge Decks (OBDs) are deck plates stiffened by open or closed stiffeners. Due to the high strength/weight ratio, this type of deck plate is a popular choice for long-span and movable bridges where dead weight is important. OBDs are subjected to many traffic load cycles during the service life. The stress ranges are relatively high, with the highest stress concentrations often at welded connections. Therefore, the design of welded connections is often governed by fatigue resistance. The discovery of fatigue cracks worldwide indicates that our knowledge of fatigue resistance design in OBDs is limited. For example, unexpected cracks were observed only after several years of use in the Severn Bridge in the United Kingdom (opened in 1966), the Van Brienenoord Bridge in the Netherlands (opened in 1990), and the Humen Pearl River Bridge in China (opened in 1997). With increasing traffic, higher fatigue resistances are required. The main challenges for infrastructure worldwide are the design of safe and economical new OBDs and the accurate assessment of existing ones. The current standards for fatigue design of bridges, EN 1993-1-9: 2005, AASHTO LRFD Bridge Design Specifications: 2020, and JSSC recommendations: 2012, primarily rely on the nominal stress method or the hot spot stress method. The problems with these methods are that the nominal stress in OBDs is not well defined and that the resistances (detail categories) for the hot spot stress method are based on standard plate connections other than the details in OBDs.

This dissertation aims to provide a comprehensive explanation of the behaviour of critical welded connections in OBDs stiffened by continuous trapezoidal stiffeners based on experimental investigations and Finite Element Analysis (FEA). This investigation focuses on the most critical welded connections, specifically the welded connections between the stiffener and the deck plate, as well as the welded connections between the crossbeam and the stiffener, utilising the author’s experiments and results from the literature. The studied details in this dissertation are summarised below:
Detail Description
C2a Stiffener-to-deck plate weld, weld toe crack in stiffener
C2b Stiffener-to-deck plate weld, weld root crack in weld
C1a Stiffener-to-deck plate weld, weld toe crack in deck plate
C1c Stiffener-to-deck plate weld, weld root crack in deck plate at crossbeam
C4d Crossbeam-to-stiffener weld at stiffener bottom, weld root crack
C3a Crossbeam-to-stiffener weld, weld toe crack in stiffener at lower end of the weld

Illustrations of the cracks are shown on the cover page and in Figure 1.5.

The author carries out experimental investigations on a 5.1 × 9.4 m2 full-scale OBD specimen with a 20 mm thick deck plate and 16 mm thick webs of crossbeams for details C1c (Chapter 5), C4d (Chapter 6) and C3a (Chapter 7), and nineteen 350×200×200 mm3 small-scale stiffener-to-deck plate connections for details C2a and C2b (Chapter 3). The 20 mm thick deck plate is used because a thicker deck plate (≥ 14 mm) is becoming common in newly designed OBDs as a response to cracks found in thinner deck plates (10 mm or 12 mm). Either automatic or manual welding is used for the deck plate welds in small- and full-scale specimens (Chapters 3 and 5).

Six corresponding detail categories are established, covering the deck plate thicknesses from 10 mm to 20 mm. Among them are two connection details: C1c (Chapter 5) and C3a (Chapter 7), which were not included in either the first or the second generation of EN 1993-1-9: Eurocode 3: Design of steel structures - Part 1-9: Fatigue. The surface extrapolation approach is used to calculate the hot spot stress for the details: C2a (Chapter 3), C1a (Chapter 4), C1c (Chapter 5), and C3a (Chapter 7). The force equilibrium (pair) approach calculates the structural stress for detail C2b (Chapter 3) and the nominal stress for detail C4d (Chapter 6). Fatigue resistance is evaluated using multiple failure criteria for full-scale experiments instead of using the first visible crack as the main failure criterion. The fatigue resistances of the studied welded details are relatively high compared with the fatigue resistances in the standards. The main reasons for this are: a steep stress gradient towards the hot spot (Chapters 3, 4, 5, and 7), a possibility to redistribute loads from a weakened component to adjacent parts (Chapters 5, 6, and 7), connections loaded in cyclic compression or out-of-plane bending or in a combination of the two instead of cyclic axial tension (Chapters 3 to 7), a relatively thin plate thickness of stiffener (Chapters 3 and 7), and strict requirements for the geometry of the weld between stiffener and deck plate (Chapters 3, 4, and 5). Additionally, automatic welding shows a higher and more consistent fatigue performance for details C2a and C2b (Chapter 3).

Finite element models are built using the commercial software Abaqus. The effective notch stress, averaged strain energy density factors, notch strain and fracture mechanics are used to account for the effects of penetration depth, load ratio, initial flaw, residual stress and weld profile on the fatigue behaviour of the welded connections. The Linear Elastic Fracture Mechanics (LEFM) gives good predictions for the fatigue resistance of both the weld toe (detail C2a) and the weld root (detail C2b). A geometric corrected LEFM hand calculation model is proposed for C1a with three different weld profiles (Chapter 4). The surface crack propagation of C1a calculated by the proposed model, together with the Paris’ equation, is validated against the experiment in the literature. Probabilistic fracture mechanics analysis for detail C1a is carried out using the same method. The calculation predicts the fatigue resistances well compared with the values obtained from the experiments in the literature. The Extended Finite Element (XFE) method is used to study the crack propagation of C1c within the LEFM framework (Chapter 5). The crack arrest and the crack path of C1c for 10 mm, 12 mm, 16 mm, and 20 mm thick deck plates are correctly predicted (validated against experimental investigations).

The residual stress due to welding is numerically studied using Abaqus for detail C3a (Chapter 7), which is further used in the fatigue analysis using the Notch StrAin (NSA) and the Elastic-Plastic Fracture Mechanics (EPFM) for the fatigue initiation and the short crack propagation, respectively. An engineering framework is proposed for the three-dimensional fatigue crack propagation using the LEFM. The crack propagation is predicted using the XFE algorithm in Abaqus, which successfully predicts the long crack propagation of C3a. The fatigue initiation stage and the crack propagation stage in the available 2 stage model are analysed by the NSA and the LEFM, respectively. The author bridges the gap between the fatigue initiation and the long crack propagation using the EPFM. The 2+ stage model is therefore proposed by the author, which provides a full range fatigue analysis of metallic structures with a specific focus on welded connections (Chapter 7). The proposed model realistically considers the geometric characteristics and different material properties (e.g. heat affect zone) at the initiation location and quantitatively implements the welding-induced residual stress obtained by FEA.

Recommendations are given for the fatigue design of the details based on this study, which were used as input for the technical specification “TS 1993-1-901 — Fatigue design of orthotropic bridge decks with the hot spot stress method” as part of the second generation of Eurocodes.

An Orthotropic Steel Deck (OSD) is suspectable to fatigue and this forms an important design criterium. In this research the rib-to-crossbeam connection with a cope hole is studied using Finite Element Methods (FEM). The fatigue assessment is performed with the hot spot stress method making use of surface stress extrapolation. Two fatigue cracks are investigated: the crack initiating from lower weld end in the rib, and the crack which appears at the weld toe of the rib-to-crossbeam connection and propagating in the crossbeam at some distance from the soft toe of the cut-out.

Engineering firms widely use shell elements to model the structure, instead of solid elements. The modelling and computational time is limited when compared to the solid element. When a comparison is made between the results from the solid and shell FE models, a scatter in results is often seen. This thesis studies the difference of shell and solid element modelling in the hot spot stress calculation of OSDs. Improvement of the shell element modelling is given based on the FE analysis using various shell element modelling techniques and compared with solid element models which represent as a reference.
Three different weld modelling techniques are applied to shell FE model: IIW, Eriksson and a combination of the first two approaches. The weld modelling techniques for the shell FE model are first applied on a simple single sided fillet welded transverse stiffener connection. In this study both the deformations and hot spot stress are compared.

A small parametric analysis is performed to investigate the difference in hot spot stress for pure in-plane and out-of-plane rotation of the crossbeam. In this study the different weld modelling techniques for the shell FE models are included. In the extensive parametric analysis two different studies are performed: one based on the influence of the loading positions, the second study is based on different thicknesses of the parts of the OSD. Based on the loading positions study the critical loading positions are further investigated.

From both the fillet welded transverse stiffener specimen and the full OSD specimen it is concluded that the shell FE model in which use is made of the combined weld modelling technique, gives the largest improvement. The ratio between shell and solid FE models of the obtained hot spot stress is reduced from a factor 1.25 to a factor 1.05 for the lower weld toe crack in the rib.

For the crack in the trough, the mean value of the hot spot stress ratios is equal to 1.04 for both the old and new OSD variant and the coefficient of variation (CV) is equal to 0.8% and 2.2% respectively. For the crack in the crossbeam, the mean values are equal to 0.80 and 0.76 for the old and new OSD variant respectively. The CV is equal to 1.9% and 2.0% for the old and new OSD variant respectively. Furthermore, the stress gradient and the stress at 40mm from the weld toe show similar results when compared with the solid FE models. The coefficient of variation for both investigated cracks is low, i.e. below 5%, thus the ratio between obtained hot spot stress of the shell with combined weld and solid FE models is consistent.