The administration of transportation infrastructure entails addressing a multitude of obstacles arising from the intricate, fast changing and dynamic nature of the environment. This thesis focuses on improving infrastructure maintenance planning through the application of deep reinforcement learning. The research begins with a thorough examination of current approaches in transportation network simulation, revealing the omissions in current strategies for infrastructure upkeep. This is followed by the development of a simulation environment that simulates realworld conditions and integrates pavement and bridge condition models to evaluate various maintenance strategies.
The proposed framework is tested on a simulated transportation network in the United States, which incorporates a traffic model to account for dynamic changes. The DRL algorithms are then used to make maintenance policies for this environment. These policies are evaluated through extensive simulations, with a focus on reducing maintenance expenses and improving the overall condition of the infrastructure.

Truss-to-GO is the name given to a Research-by-Design idea for a novel construction technology for Lattice Structures such as trusses and space frames. The ambition is to develop an in-situ fabrication method for these structures, using a continuous Flax- FRP (Fiber Reinforced Polymer) composite-technology-based ‘Rope’ that can be compactly transported to site in a spool, and manually ‘woven’ into truss-like shapes and finally cured and rigidified. This thesis explores the first steps of such a technology, starting some of the mechanical properties that can be achieved by braided flax polymer composites as a single strut or chord in the lattice, with consideration for the pre-cure consolidation strategy for the composite. Further, the mechanical performance of a joint, or node, in the lattice is tested, and a proposal for the on-site infrastructures and fabrication workflow is developed. Finally, a simple comparison is made between the Truss-to-GO technology and some conventional lattice construction technologies, with insights about the potentials, limitations and future research topics for this ‘fantasy’ technology.

The Dutch government was mandated by the national court in May 2019 to address high nitrogen-oxide levels, threatening construction permits to be canceled if target values are not met. As the agrarian sector contributed significantly to nitrogen-oxide emissions, reductions in this sector were necessary. Ongoing protests by farmers indicate the unresolved crisis. A government report suggests that around 500-600 high-emission agricultural businesses must cease operations for a substantial impact. Consequently, large cattle farms, known as "mega barns," will become vacant. Traditionally demolished or recycled, their materials may have the potential for repurposing, especially given the projected availability of similar buildings in the next 10-20 years.

Both the European Union and the Dutch Government have set themselves targets to convert their economies from linear to circular before the year 2050. A circular economy means a system in which products and materials are kept within the loop as long as possible, reducing the need for new raw materials and production – reducing energy needs and CO2-emissions in the process. Recycling is currently coming up as a mainline strategy, however, it is considered to be less circular than other strategies, especially compared to reuse. The steel sector is seen as an exemplary industry that recycles a lot but would pose major environmental benefits when shifting the chain towards reuse. Yet, making this shift happen is withheld by certain barriers, especially on the designer’s side. The fact that information on availability, quality and quantity of reusable components is scarce in the critical early phases of the design process, is one of these main barriers.

This thesis attempts to introduce a tool that employs reverse-engineering techniques to analyze and predict the availability of structural steel components in industrial buildings, providing designers with knowledge about potentially available materials as early in the process as possible. By utilizing publicly available data, the tool enables an accurate estimation of the length, type, quantity and quality of the elements. This is done by the use of parametric design software such as Rhino3D, Grasshopper and Karamba3D. The research explores ways of making use of existing structures’ geometry and design requirements in order to predict the structural properties of the load-bearing components. The tool has been tested and evaluated on a series of cases, all of which are industrial farm halls situated in the Netherlands. This case testing has been used to improve and finetune the output results of the tool. In the end, the developed tool is able to predict steel profiles within a +/- 1 profile class range. Additional analyses are incorporated to assess cost savings, environmental benefits, and element quality.

The reuse of insulating glass plays a crucial role in promoting sustainability and circularity within the construction sector. CO2 emissions must be reduced according to the Climate Agreement. Reusing materials can help achieve this goal. Reusing glass without reheating it can provide environmental and cost benefits, making it a suitable option. In this study, 36-year-old insulating glass panels (double-glazing) from an apartment building in Amstelveen have been tested for strength through destructive testing. This is done to see what the strength of the glass is after the glass has been in the building for 36 years and has been exposed to opportunities for damage. Various patterns of damage on both the interior and exterior side of the glass panels from an insulated glass unit affects the strength that can eventually be assigned to the glass. The aim of this thesis is to be able to show whether and how insulated glass units can be reused and what the strength is of naturally aged glass after a certain lifetime. From six insulated glass units obtained, specimens measuring 150mm x 150mm are extracted and tested for strength using a Coaxial Double Ring test. Prior to the destructive tests, a number of specimens were viewed under the Keyence VHX 7000 digital microscope to see how the damages are and whether they vary between the four sides of an insulated glass unit. This revealed that the outer side of the outer panel (side #1) contains the most and homogeneous damage. Weathering, especially rain and wind, which can bring sand particles, for example, create homogeneous damage on the glass surface. The entire surface is exposed to this type of weathering. The outer side of the inner panel (side #4) also contained quite a lot of damage from probably cleaning and (human) touching. Despite the fact that the inner sides of both panels (the sides in the cavity, side #2 and #3) theoretically had little potential for weathering, damage could also be seen on these, although it was more localised, meaning it probably occurred during the production process or preparation of the specimens for this experiment. During the Coaxial Double Ring tests, 406 specimens were tested for strength. The strengths of the specimens ranged from 16.8 MPa to 243.7 MPa, with an average strength of 67.5 MPa. The average strengths of side #2 and #3 (sides in the cavity) are slightly higher than those of side #1 and #4 (outer sides), between 9% and 17%. A tin-tester was used to determine which side of each specimen was the tin-side prior to the strength tests, to include this in the analysis of the results. The specimens tested on the tin-side were found to be on average much weaker than the specimens tested on air-sides. The average measured strength of the air-side is 31% higher than that of the tin-side. In a number of tests carried out with new glass, this large difference is already present, while literature states that the difference in strength between the air- and tin-sides can be considered marginal. Fracture statistics were used to analyse the data and find a strength for each series tested. This was done using Weibull theory. The two-parameter Weibull distribution has proven to be very conservative at low failure probabilities, resulting in a very low design strength (failure probabilities of 0.8% and 0.12%) for a series of specimens when there is a large spread in data. The spread in data was often of greater influence than the values of the measured strengths. As a result, the design strengths for series tested on the air-side were usually lower than specimens tested on the tin-side, while the average strength measurements were actually higher. The greater variation in strengths of specimens coming vi from the inner side #2 and #3 results in lower design strengths than the outer sides (#1 and #4), which contain more (homogeneous) damage. The values of design strengths of the naturally aged glass tested were almost always lower than the design strength of new glass according to current standards. The design strengths of new glass are 22.5 MPa according to NEN2608 and 25.0 MPa according to EN16612. When looking at how many specimens failed at stresses lower than these values, there are only three and six respectively. So, there are very few spots in 36-year-old glass that are weaker than the design strength that can be assumed according to NEN2608 or EN16612. These lowest values are not consistently found in specimens tested on either the air-side or the tin-side, so based on this study, the influence of this can be ruled out. By this means, a more realistic and less conservative option seems to be to take the strength of the weakest specimen in a test series and then assign it to the entire panel from which that test series comes. Reuse options are discussed in the final section of this thesis. Using a number of common glass structures, it explains the ways in which weathered glass can or cannot be reused and in what function. How the results obtained can be handled and used is a point of discussion, as the values are not always representative of the strength of the entire glass panel. Further research will have to show whether the current design standards and the parameters used in them are well chosen or whether they should be different for used glass. For example, use a lower material factor of 1.35 instead of 1.6 – 1.8 when calculating the design strength or use a higher failure probability than 0.0012 to determine design strength in a Weibull distribution. The efficiency and effectiveness of the method used can also be optimised in further research.

The preservation of historical heritage buildings frequently requires the assessment of brick masonry structures. These structures exhibit complex behaviour due to the arrangement of units and the interaction between bricks and mortar. This study aims to compare two modelling techniques for simulating masonry's compressive response and understanding which technique performs better.

This study employs two block-based modelling techniques: a simplified brick-to-brick model and a detailed brick-to-brick model. The block-based modelling technique was selected to capture the intricate structural details and crack patterns of masonry specimens. Wallet specimens were evaluated to understand the compressive behaviour of masonry.

Two experimental benchmarks were chosen to validate the selected modelling approach. First, experimental results by Thamboo (2020), also modelled with a simplified brick-to-brick model by Zahra et al. (2021), were considered. However, to obtain numerical results comparable to the experimental ones, unrealistic input parameters had to be assumed (i.e., very low value of elastic modulus for the bricks); consequently, this benchmark was ultimately deemed unreliable. Hence, a second benchmark from Jafari (2021) was chosen which did not present such issues.

Considering the effect of confinement was crucial to obtain results in agreement with experiments. The detailed brick-to-brick model, explicitly modelling each brick-and-mortar joint with solid elements and accounting for mortar confinement, proved to be the only modelling approach able to simulate the second experimental benchmark. The simplified modelling technique, considering mortar joints as zero-thickness interfaces between solid elements representing the bricks, consistently underestimated the peak loads as mortar confinement was not included. This may explain why unrealistic input parameters were used by Zahra et al. (2021) to simulate the first benchmark. In general, the post-peak behaviour was difficult to simulate.

A sensitivity study was performed by varying the boundary conditions, Poisson’s ratio of mortar and integration scheme. The study on boundary conditions was performed on both the simplified and detailed models whereas the study on Poisson’s ratio and integration scheme was performed on the detailed model. The boundary conditions did not have a significant influence on the global behaviour of the specimen in the detailed or simplified models. Small variation in compressive strength and crack pattern were observed, but the effect on post-peak behaviour could not be evaluated due to numerical instability. The Poisson’s ratio of mortar was found to have a significant influence on the peak compressive strength of the masonry specimen. A higher Poisson's ratio resulted in a greater value in the peak compressive strength of masonry. A higher value of Poisson’s ratio increases the amount of confinement in mortar layers resulting in higher values of compressive strength of masonry. A high integration scheme (3×3×3) and a regular integration scheme (2×2×2) effectively captured crack propagation. On the contrary, a reduced integration scheme (1×1×1) could not do so effectively. The reason can be attributed to the higher number of integration points in the former cases which allows for better propagation of cracks.

These investigations provide precious insight into the choice of modelling techniques for simulating the compressive behaviour of brick masonry. The detailed modelling technique can be implemented to study the compressive behaviour of masonry specimens if mortar confinement is accounted for. However, further effort is necessary to reduce numerical instability to simulate the post-peak response. This can be adopted further to study the influence of, for example, bond pattern, specimen thickness (e.g., multi- vs single-wythe masonry) and specimen shape (e.g. wallets, prism, core) on the compressive behaviour of masonry.

This research investigates how to design a glass pedestrian bridge that is structurally optimised for multiple load scenario’s and feasible in terms of fabrication and construction. Linear glass elements such as panes have limited freedom of shape. On the other hand, cast glass can be produced into any shape possible. When glass is used in large quantities with large cross-sections, the annealing process can take a long time and require a significant amount of energy. The annealing time must be kept to a minimum to make cast glass a viable option. This is where topology optimisation can play its part. During topology optimisation, the layout of an element is optimised on the given constraints for the specified design domain. A compliance-based optimisation is found to be the most suitable objective function. The software package Ansys and various plugins for McNeel Rhinoceros/Grasshopper are compared on their topology optimisation tool. During this comparison, Ansys is found to be the most suitable software. This is mainly because of the possibility of implementing a cross-sectional constraint during the optimisation, which is needed for implementing a maximum annealing time. The most suitable production method for topology optimised cast glass elements is found to be kiln casting with a 3D printed sand mould. Glass is a brittle material with a low tensile strength compared to its compressive strength. The only type of bridge that can be made entirely in compression, without pre-tensioning, is an arch bridge. This bridge type is used for the final design. The loads considered during structural analysis are the vertical traffic load (distributed and point) and the vertical wind load. The bridge is made from borosilicate glass. This type of glass is used to reduce the effect of temperature on the cast structure of the bridge. Additionally, the supports and connections are designed so that they do not restrain deformation from temperature. Furthermore, the shape and place of the connections and supports ensure the possibility of unequal settlement without the formation of unacceptable stresses. Multiple variant studies are carried out. In the first variant study, optimisation is done on several 2D models to determine the best shape for the bridge's final design. The second variant study focuses on the 3D shape of the bridge and how the bridge is split into manageable segments. The final bridge comprises four cast outer elements, one cast keystone, and a float glass bridge deck. The bridge is simply supported on both sides of the canal. The connections and supports are based on compression and friction and use a polyurethane interlayer. The weight of the final optimised bridge is reduced from 32t to 11.2t with a maximum cross-sectional thickness of 210 mm to get an annealing time of approximately one month. Due to symmetry in the design, only one-fourth of the bridge is optimised. During the optimisation, seven different distributed load scenarios are implemented. For the point load, a region is excluded from optimisation. After optimisation, finite element analysis and manual verifications are performed.