The current healthcare facility design workflow is hindered by inefficient, manual layout revisions and fragmented stakeholder communication, leading to regulatory inconsistencies that necessitate a more automated, data-driven methodology. This project explores how rule-based computational logic can generate compliant schematic layouts that respond dynamically to standardized clinical briefs. Starting from International Health Facility Guidelines (iHFG) data, a hierarchical cellular automata system was developed in a custom pipeline that outputs unique spatial configurations per planning zone-balancing strict functional adjacency with geometric performance. ... Read more

This research investigates the feasibility of developing a framework for a user-friendly daylight analysis tool that complies with NEN-EN 17037 (calculation method 1, Daylight Factor) and supports designers in the early design phase without relying on time-consuming simulations or detailed 3D modelling. This study explores whether Machine Learning (ML) can be used to automate the extraction of spatial features from early-stage architectural drawings and to predict daylight performance based on a simplified geometric representation.This research addresses the research question: How can a framework be developed for a user-friendly daylight analysis tool in accordance with NEN-EN 17037, using Machine Learning to eliminate the need for time-consuming simulations and 3D models and provide feedback to designers in the early design phase? A multimodal neural network using a U-Net architecture with a feature-reinforcement stream was trained on solid-angle encoded floorplans and demonstrated the ability to reproduce daylight factor distributions with high structural similarity. Feature extraction models (YOLO-based object detection and instance segmentation) were trained and showed promising results for detecting window and room geometries in orthogonal floorplans, although performance decreased for irregular or stylistically inconsistent drawings.A data conversion pipeline was developed to convert detection outputs into a watertight geometric representation, suitable for solid angle calculations. This pipeline integrates floorplan and facade information, reconstructs outer and inner walls, matches and scales window data, and generates a complete solid-angle matrix as input for the daylight prediction model. A proof-of-concept interface was developed to evaluate usability using the Five-E usability principles. Expert feedback indicated that the workflow aligns well with early-design needs, offering sufficient transparency and user control, though trust in ML-based daylight predictions remains limited without further validation.Overall, the results demonstrate that an ML-enabled workflow for early-design daylight assessment is technically feasible. It holds potential for reducing analysis time while maintaining interpretability. The prototype establishes a coherent end-to-end framework, from 2D drawings to encoded geometric inputs and preliminary daylight predictions. At the same time, the research identifies key limitations, such as handling irregular geometries, missing overhang and incorporating sloped-roofs. Also dataset constraints require further development before the tool can be applied in the early design phase. The study provides a foundation for future advancements toward more accessible, efficient, and practice-oriented daylight analysis tools. ... Read more

This research explores the integration of generative artificial intelligence (AI) into early-stage architectural design through a creative assistant that translates natural language briefs into 3D volumetric apartment layouts. Existing AI research shows promising results in generating spatial configurations, but they fall short in supporting real-time interaction, spatial awareness, and editable outputs. To address this gap, a new methodology is proposed that integrates OpenAI’s GPT-4o model into Rhino and Grasshopper, allowing architects to co-create volumetric layouts through a conversational interaction. The assistant was incrementally fine-tuned on realistic architectural data from the RPLAN dataset to learn spatial logic, reasoning, geometry generation, and evaluation. A custom creativity evaluation metric was developed to test the model’s accuracy and novelty performance. Accuracy was measured by how well the output aligned with the given prompt. and novelty was measured by how original it was compared to the closest reference layout and the training dataset. A residential apartment design case study was conducted to test the full workflow from reasoning to generating and evaluating. During testing, the assistant showed the ability to iteratively respond to feedback without explicit training. This demonstrates the potential of “centauric” design workflows, where human intuition and machine intelligence collaborate in real time. The final model in the pipeline achieved an average novelty score of around 69%. This system contributes a new foundation for human-AI co-design, possibly enhancing design quality, adaptability, and efficiency in future architectural practice. ... Read more

The environmental impact of building materials, expressed as embodied carbon (EC) or, in the Dutch context, through the MilieuPrestatie Gebouwen (MPG) indicator, is increasingly governed by regulatory standards and sustainability objectives. However, current assessment tools for measuring embodied carbon are typically used as late-stage compliance checks, rather than being integrated into the iterative design process of a new project. Moreover, publicly accessible data on whole-building environmental impact is scarce, limiting the ability to benchmark and contextualize new designs.
This research proposes a computational tool that facilitates early-stage environmental impact assessment of building materials, integrating three calculation methodologies: Embodied Carbon, MilieuPrestatie Gebouwen, and a circularity-informed indicator (EC+). The EC+ indicator extends conventional EC assessment by incor-porating a limited set of circularity factors: biogenic carbon storage, material replacements, and potential end-of-life benefits.
Unlike conventional tools that rely on the assessment of individual materials, the developed framework oper-ates at the level of elements. An element can be defined as a predefined assembly of one or more building materials arranged in functional layers, corresponding to a specific building part (such as: a floor, a roof or an external or interior wall). In the context of this research, an element was subdivided into three compo-nents, which were termed ’sub-elements’. These sub-elements are categorised as top, structural, or bottom. Sub-elements are manually composed from individual materials to ensure structurally realistic and technically robust assemblies. Complete elements are generated from one top, one structural, and one bottom sub-element in an automated process designed to ensure efficiency and consistency. This process is visualized in figure 1.

Assessing environmental performance at the element level offers significant advantages. It improves impact completeness by including all functional layers, enables consistent integration of circularity parameters such as modularity and connection type, and supports early-stage evaluation even when full material compositions are not yet available. Additionally, it allows physical, structural and circularity-related properties to be assigned to each element. This enables to filter out undesired configurations later in the process, based on project specific requirements.
The tool is furthermore supported by a structured element database, the purpose of which is to facilitate the storage of the aforementioned data. This enhances assessment efficiency and accuracy, and supports early-stage evaluation and cross-project data reuse by preserving validated elements.
This research also developed a methodology to generate synthetic whole building impact data. Elements, sorted by building part, are combined into complete building configurations (see figure 1). All valid combinations tions between elements are made according to the project building part ratio and user requirements, enabling the calculation of environmental impact values for a large number of hypothetical buildings. This synthetic dataset forms a robust reference against which new building designs can be evaluated and contextualised.
The research was conducted using a research-through-design methodology, integrating academic principles with practical industry requirements. The dataflow of the tool consists of three modules: input, process and output (see figure 2). It utilises minimal early design inputs, such as gross floor area and material categories, to estimate whole-building environmental impact values. The process module of the tool was developed in Grasshopper, with supplementary support from Excel and Power BI, enabling real-time calculation, data structuring and saving, visualisation, and live performance feedback.

The validation process confirmed that the tool performs with high accuracy and meets the key requirements defined at the outset of the research. These requirements were initially established and subsequently refined through a thorough review of the relevant literature and current practices in the building sector.
The developed tool offers a reliable and extensible foundation for translating sustainability ambitions and regulatory obligations into practical application by embedding environmental impact feedback into early-stage workflows. This integration supports iterative decision-making and encourages the use of low-carbon and more circular materials before critical design decisions are finalised. The tool’s real-time feedback functionality allows users to explore impact variation under project input uncertainty.
Although the tool was developed with the involvement of different stakeholders, it has not yet been applied in a real-world design process. Consequently, its impact on design outcomes remains to be empirically validated. ... Read more

The Dutch residential building sector must transition toward energy-efficient renovations to address climate change and reduce energy consumption. However, the early design phase, where critical decisions about energy renovation scenarios are made, often presents significant challenges. Project developers face issues such as unclear prioritization of decision criteria and inefficient processes for evaluating and comparing renovation alternatives. These challenges frequently lead to delays and hinder the selection of optimal solutions. This research aims to identify and address the key problems in the decision-making process during the early design phase of residential energy renovations. The central research question is: “How can the decision-making process in the early design phase of energy renovations in residential buildings be improved to enable project developers to make efficient decisions that consider environmental, economic, and social factors?” The study identifies critical bottlenecks in the early design process, such as fragmented criteria selection and the absence of a systematic approach for evaluating renovation scenarios. A structured decision-support framework is proposed, focusing on defining relevant criteria, integrating stakeholder inputs, and ranking alternatives using a balanced and transparent weighting method. The added value of this framework lies in enabling project developers to make quicker, well-informed decisions by simplifying complex processes and clarifying priorities. While a computational tool may be developed as a final product, this research prioritizes understanding and improving the decision-making process itself. Validation through a case study ensures practical applicability and relevance to the challenges faced by project developers. By providing a systematic and innovative approach to decision-making, this study enables faster and more practical decision-making processes. It contributes to the efficient planning and execution of energy renovations in residential buildings, supporting the sector’s transition to sustainability and climate resilience. ... Read more

Continuous or repetitive exposure to physical and psychological stressors can lead to a range of health problems caused by resulting malfunctional allostasis. The field of neuroarchitecture focuses on the relationship between architecture and neuroscience, and the specific neurological changes shown in biomarkers that can result from the built environment. In extreme environments, where exposure to stressors is heightened, the deliberate use of physical architecture and indoor environmental quality - for a positive influence on short and long-term stress response - is of even greater significance. Previous research investigated architecture as cause for stress, post-stress relaxation and for physical recovery. Still, architecture as acute countermeasure (reducing relative stress reaction) based on individual neuroendocrine response to prevent a shift to malfunctional allostasis is not well investigated. The thesis aims to help closing that gap by investigating the effectiveness of visual quality features as countermeasure to stressor exposure shown in biomarkers. It focuses on inter-individual reaction and the alignment of neuroarchitecture with functional requirements of architecture in extreme environments. The methodology for this is three-fold. A research framework was developed considering study design (including necessary infrastructure, stressor simulation, study conduction and data collection), and study outcomes (data analysis and data application). The framework was applied to a pilot study with human subjects focussing on colour correlated temperature (CCT) and heart rate variability (HRV). In addition to HRV, basic physical data, the participant’s stress over the last month, chronotype, perceived stress/workload during the experiment and test performance were recorded. The study data were processed through different computational models (including multi-criteria decision analyses and Bayesian linear effects models) analysing inter-individuality, the effect of confounding variables and the use of transient stressor simulation. The findings - while not generalizable due to the small data sample – indicate baseline-, stress-, and recovery-HRV were higher under warm CCT, but the HRV-change from stress to recovery was greater under blue CCT. Despite this, CCT did not show a significant counteracting effect reflected in the change from baseline-HRV to stress-HRV. The most relevant findings relate to the development and exploration of the methodology, providing directions to address inter-individuality of stress-response and cause-effect ambiguity in cross-sectional research on the effect of visual quality as countermeasure. The data were further used for the demonstration of a future application - using a microgravitational space station as case study - detailing Multi-Sensor Data Fusion and Bayesian Reinforcement Learning for a system that considers confounding influences, spatial limitations, functional requirements, accessibility, individual- and group-needs and long-term trends in biomarkers. The research framework can be extended and applied to future neuroarchitectural studies, the results of which could inform the development of adaptive systems in extreme environments. The aim of this thesis was the development of research methods and their preliminary validation that can contribute to future research on neuro-adaptive architecture as countermeasure to physiological stress response. ... Read more

For a long time, timber has taken the backseat to steel and concrete for largescale structures, but due to sustainability interests many developments have been made to improve its less desirable qualities. With the introduction of engineered wood products, large timber sections required for fire resistant design became economical and the timber structural behavior became more reliable. However, there are still many challenges when it comes to the seismic design of tall timber structures. Current modelling strategies are time-consuming to implement, provide inconsistent results, and do not account for the passive parameters which impact the seismic behavior of the structure throughout its lifetime. Since timber is a natural material, it is subject to varying levels of moisture content which impacts its stiffness. Furthermore, given the lightweightedness of timber structures (compared to concrete or steel alternatives), any changes to the mass distribution of the structure drastically changes its dynamic response.
This research project develops a computational workflow for the seismic analysis of tall timber structures which integrates the seismic analysis model seamlessly into the main design workflow, simplifies the process for setting the parameters in a semi component-level model for seismic analysis, includes a lifetime analysis option which considers the variables which impact the structural performance of the structure over time, and provides the engineer with component-level data over time. Python class objects are used to develop this computational workflow inside of the Grasshopper environment for Rhino, using the OpenSeesPy library for analysis. It follows the analysis standards provided by the (recent drafts of the) Eurocodes and supporting research papers. It has been developed to align with the most prominent tall timber construction types, as defined through the literature review. The analysis script utilizes the modelling strategy put forth by Rinaldi et al. (2021) to determine the effective stiffness of a cross-laminated timber wall, and its implementation was validated with a comparative analysis to the results of that research. The implementation of the full workflow and its impact on the design process is demonstrated through a case study, with results confirming the importance of including lifetime variables for analysis. This research increases the timeframe of analysis that the engineer can perform on tall timber structures such that the initial structural design can be informed by future predicted events, allowing for more resistant designs. ... Read more

Decision-making in early-stage design often lacks robust methods for evaluating circularity, resulting in outcomes that may not fully realize their potential for efficiency. This research presents the development of a computational tool or “Intelligent Design Assistant” that employs Graph Neural Networks (GNNs) to deliver real-time assessments of life-cycle performance and material usage for modular designs. By utilizing user-defined, simplified early-stage representations, the tool provides actionable insights into both design and environmental performance. A central point of this approach is the adoption of a graph-based framework where each building module is represented as a node, and its interactions with neighboring modules are captured through connecting edges. This framework not only reflects the intrinsic properties of each module, but it also dynamically evaluates how a module’s characteristics evolve based on its spatial and functional relationships. Although the study focuses on laminated veneer lumber (LVL)—selected for its extensive environmental data—the scalable machine learning model is designed to be applicable to a wide range of construction methods and materials. Through experimental validation, the integration of GNNs has been shown to enhance early design decision-making by providing real-time feedback. The model achieves an accuracy of approximately 85% -90% under conditions similar to the training data. This capability enables designers, clients, and other stakeholders to engage in informed discussions about design modifications and circularity measures well before detailed construction planning begins, thereby promoting more sustainable and circular design practices across the industry. ... Read more

In this master’s thesis, the pressing need for novel, bio-based and fully circular construction materials is examined. Building on existing TU Delft research, the work focuses on cellulose and lignin as the principal constituents of a new wood-like composite, using waste streams as the source. The experimental setup, inspired by a range of studies on lignin-reinforced materials, employs a hot-pressing procedure. The literature review highlights a clear gap: the combined use of cellulose and lignin, each derived from by-products, to form an innovative material matrix, with the goal to make use of lignins adhesive qualities.
The methodology includes an innovative computational method designed to provide controlled variability of structural properties, that can be directly integrated into a component design. Many material variations, like moisture content, lignin types, C/L-ratio, pre-treatment and recyclability have been explored. Principal findings from this study include identifying optimal cellulose-to-lignin (C/L) ratios for distinct mechanical performances—3:2 for flexural strength and 2:3 for compressive strength. Additionally, Soda lignin shows the best mechanical performance and the plates can be successfully reprocessed. Computational simulations using Rhino 8, Karamba3D and Wallacei proved effective in predicting and optimising mechanical properties, significantly streamlining material development. Multidimensional scaling was used to map high-dimensional material data into a two-dimensional space, clustering formulations by performance and revealing trade-offs (e.g., stiffness vs. toughness) to guide blend selection.
Microscopic analyses further supported the viability of lignin as a natural adhesive, while highlighting areas needing improvement, such as brittleness and susceptibility to warping and blistering. A link was established between mechanical testing data and component modelling, which opens up possibilities to optimise both design and material composition for sustainable and resource efficient structures.
Overall, this research successfully demonstrates the potential of fully circular waste-based cellulose-lignin composites and paves the way for scalable, high-performance, sustainable building materials. ... Read more

This research examines how dynamic façade variables influence the embodied and operational carbon of mid- to high-rise residences during the early design phase, addressing growing environmental concerns from urban densification. By analysing façade design and parametric architectural strategies, the study aims to identify sustainable solutions that minimize environmental impact while complying with regulatory standards.

A combination of literature review and computational simulations was used to evaluate different façade typologies. The literature review identified common façade systems and their carbon footprints, while a case study applied this knowledge to a realistic scenario. Using parametric modelling tools such as Grasshopper, energy simulations were conducted to assess carbon impacts. An optimization process then identified the most sustainable façade configurations, highlighting key trends and design considerations.

The findings reveal that material selection, façade design, and energy efficiency significantly impact the total carbon footprint of buildings. Among the façade types analysed, aluminium unitized façades have the highest embodied carbon emissions due to the carbon-intensive nature of aluminium production. In contrast, prefabricated timber façades have the lowest embodied emissions, benefiting from a lower carbon footprint and carbon sequestration potential. Concrete façades fall in between, with their high weight contributing to greater embodied carbon despite lower emissions per kilogram. The relationship between window-to-wall ratio (WWR) and embodied carbon varies by material; a higher WWR increases emissions for aluminium and timber façades, whereas for concrete façades, it reduces embodied carbon as glass replaces carbon-intensive concrete elements.

Operational carbon emissions are highly dependent on façade orientation. North-facing façades require the most heating due to limited solar exposure, while south-facing façades benefit from passive solar heating but require more cooling. The most effective way to reduce operational carbon is by improving glazing insulation (lowering U-values), especially in colder orientations. Increasing the Rc-value of insulation has only a minor effect when WWR is high, as window heat transfer dominates. With an assumed 2% annual improvement in energy efficiency and grid decarbonization over a 75-year lifespan, operational carbon emissions are expected to decrease by 50%, making embodied carbon an increasingly dominant factor.

Considering both embodied and operational emissions, timber façades emerge as the most sustainable option, particularly when paired with optimized glazing and insulation values. Aluminium façades have the highest total carbon footprint, with embodied emissions accounting for nearly half of the total impact even in efficient configurations. Concrete façades present a unique trend, where reducing WWR can sometimes increase total emissions due to the high embodied carbon of concrete relative to glazing. These results emphasize the need for an integrated approach to façade design, balancing material selection, insulation levels, glazing performance, and orientation to minimize total carbon impact.

This study acknowledges several limitations, including reliance on a single simulation program, uncertainties in future energy grid decarbonization, and a limited range of material and façade options. Future research should explore additional materials, occupant behaviour models, and renewable energy integration to enhance sustainability assessments. Further validation using multiple simulation methods, diverse climate models, and broader material databases would improve reliability and deepen understanding of façade performance across different environmental contexts.
... Read more

With decision-making becoming increasingly data-driven in early design stages due to global environmental challenges, and qualitative metrics remaining crucial in facade design due to their huge architectural impact, performance-driven design exploration frameworks are emerging as powerful tools to explore vast design spaces based on geometry typology and approximated performance. However, those used in facade design based solely on Self-Organising Maps (SOM), face significant challenges in computational efficiency. While more advanced frameworks used in the AEC sector combining SOM and Multi-Layer Perceptrons (MLP) address this, they still face limitations in prediction accuracy, convergence speed, interpretability, reliability, and usability, reducing their effectiveness in decision-making. This thesis aimed to overcome these limitations, by developing a novel framework integrating SOM and Kolmogorov-Arnold Networks (KAN), applied in the design process of an aluminium-based biocomposite curtain wall facade. The results demonstrate that substituting heavy performance simulations with fast approximations using KAN leads to significant enhancements in computational efficiency. In addition, comparative analysis revealed that KAN outperforms MLP in prediction accuracy on highly-complex performance metrics, with much faster convergence and smaller architectures. Furthermore, KAN proved to be faster and more intuitive to train, as well as more consistent in predictions. Moreover, KAN enhanced interpretability between geometry and performance, enabling designers to focus on relevant design variables and adjust them strategically toward optimal performance, providing an integrated solution with transparent decision-making and faster processing compared to traditional sensitivity analysis tools. Finally, a novel approach to design exploration has enabled the integration of less-geometry related design variables, enhancing optimisation capabilities, as well as proven to balance human-AI interaction more efficiently than traditional frameworks, making design exploration more interactive, thereby more effective and intuitive. Ultimately, the SOM-KAN framework has proven to advance the facade design process by facilitating more efficient decision-making in early design stages, leading to superior architectural and sustainable solutions. ... Read more

This research is an early exploration into the potential of deep neural networks predicting physically-based building performance metrics on 3D urban geometry. Specifically, this work aims to predict annual solar irradiation using networks trained on point clouds. It is expected that the proposed method will allow designers to optimize their designs based on solar performance, due to the significantly lower inference time, in comparison to traditional simulation models. Furthermore, this research paves the way for generative models, which include the evaluation of performance such as solar irradiation, wind and acoustics.

Prior research has suggested several methods to predict solar irradiation on 2D building data and low resolution 3D buildings. These researches have in common that there is a lack of real observed irradiation data on buildings. Therefore, this research proposes several methods to efficiently synthesize an irradiation dataset based on 3D building geometry, which samples are larger in scale and resolution in comparison to earlier work.

Based on the synthesized datasets, several models have been trained to predict the irradiation values. Experiments have shown that a finetuned version of the model is able to predict irradiation with an average RMSE of 21 kWh/m2 with an average inference time of 0.7 seconds, on a patch of 100x100 meters. Furthermore, this thesis provides an analysis on how patch size and point sampling technique affect the prediction error of the model.

Finally, this research provides an implementation of the network in a user-friendly client-server ecosystem, that can assist architects and engineers to fine-tune their building designs in urban environments. ... Read more

Urbanization and climate change have intensified urban heat island effects, significantly affecting outdoor thermal comfort (OTC) in cities. Traditional methods of assessing the influence of urban environments on OTC often rely heavily on computational simulations, neglecting the integration of user-centric data and comprehensive environmental monitoring. This research addresses the critical gap by employing a multi-domain, mixed-method approach to evaluate the influence of building facades on outdoor thermal comfort, specifically in the Acquabella district of Milan. To address the gap, this study suggests a structured workflow beginning with a participatory workshop and a long-term survey to gather demographic data and user perceptions, identifying areas of significant thermal discomfort. These insights inform the simulations conducted using ENVI-met software, which model the Physiological Equivalent Temperature (PET) across different facade geometries and material properties. Real-time validation is achieved through a thermal walk, which includes field measurements and on-site evaluations to capture situational thermal perceptions. The research findings underscore the substantial impact of facade materiality on thermal perception. The study also reveals that aesthetic preferences and psychological factors play a crucial role in thermal perception, highlighting the necessity of integrating these aspects into urban design strategies. These findings are synthesized into a decision matrix that guides the design of retrofit strategies. Proposed solutions focus on modifying facade materials, increasing vegetation, and optimizing the built environment based on user feedback. Simulations of these design proposals using ENVI-met demonstrate significant improvements in thermal comfort, validating the effectiveness of the proposed interventions. This thesis presents a comprehensive framework for assessing the multi-domain influence of facades on outdoor thermal comfort, advocating for the integration of user-centric data and participatory approaches in urban design. The methodology and findings offer valuable insights for policymakers, urban planners, and researchers aiming to enhance urban thermal environments and address the challenges posed by climate change and urbanization.
... Read more

The design of residential care facilities for individuals with dementia profoundly impacts their quality of life and wellbeing. Dementia-friendly architecture, thoroughly reviewed in literature, provides guidelines and assessment tools to evaluate residential spaces and enhance living conditions. Key to residents' wellbeing is their autonomy and control over their environment, which can be facilitated by optimizing wayfinding within indoor spaces. Effective spatial layouts, particularly those offering good visual access, not only promote autonomy but also improves social integration by enabling residents to see and be seen by others.

This MSc thesis investigates the feasibility of artificial intelligence (AI) to support the design of dementia-friendly architecture, focusing particularly on wayfinding—a critical element of environmental design for individuals with dementia. The study quantitatively assesses the relationship between floor plan layouts and wayfinding ease using the isovist method, linking floor plan geometry with the navigational experiences of dementia patients. The assessment was done in accordance with an established Dementia Design Principles (DDP) environmental assessment tool recognized within universal design guidelines.

A computational framework was developed to evaluate wayfinding quality using visual access analysis, which were integrated into a machine learning model. This model was trained on a dataset of 256 floor plans, employing features derived from two distinct sources: spatial metrics such as distances and centrality from the Swiss Dwellings dataset, and compactness and distance-based features extracted via Grasshopper, a visual scripting tool in Rhino 3D. The model was tested using two supervised machine learning algorithms—Random Forest (RF) and Artificial Neural Networks (ANN)—and achieved consistent accuracy rates between 70-80% using 14 features and 2 multiclass outputs describing the visual access quality. This demonstrates AI's potential as a decision-support tool in the early stages of architectural design, offering architects insights into the wayfinding quality of their designs.

The goal of this research is to develop a digital framework that can be leveraged by architects to link early-stage concept design ideation to the specialist validation of final designs to help guide the design of layouts towards DDP-compliance and reduce risk of design changes, ultimately designs that are easier to navigate by people living with dementia and enhance the quality of living. ... Read more

As the consumer electricity prices rise, European policymakers are increasingly focused on decarbonizing the power grid, which requires homeowners and local administrators to adopt renewable energy sources amidst a complex set of often conflicting objectives and constraints.

This paper introduces an innovative application of deep reinforcement learning (DRL) for long-term strategic planning of rooftop photovoltaic systems and battery energy storage within the residential sector, aiming to balance environmental and financial objectives considering the ever-evolving system condition and uncertainties inherent in the market.

The problem is modeled as a Markov Decision Process (MDP), facilitating sequential decision-making across 25 annual steps. The DRL environment incorporates a comprehensive set of variables identified through extensive literature review and market analysis. To account for their long-term dynamics, scenarios were simulated using appropriate stochastic and propabilistic processes for agent's training. A policy-based DRL agent is evaluated, exploring various residential and technological scenarios, including three single-family houses, different PV models and various optimisation scopes.

Moreover, a deployment workflow and a user interface are developed to support real-world decision-making applications. Furthermore, a separate DRL model is crafted to simulate battery management system's charging and discharging protocol.

The findings suggest that deep reinforcement learning offers a promising solution for addressing this complex problem. It offers enhanced flexibility in decision-making and helps mitigate investment risks. ... Read more

In response to the European Commission's 2020 Renovation Wave strategy aiming to double the annual rate of energy renovation for buildings by 2030, this thesis addresses the challenge of optimizing retrofitting planning using planning optimization algorithms. Despite the economic and practical hurdles identified in existing research, limited attention has been given to optimizing the timing and sequence of retrofitting actions. Recognizing the complexity introduced by uncertainty factors such as material degradation rates,this study draws parallels between retrofitting and predictive maintenance problems. This thesis is divided into two main sections: the first establishes the theoretical foundation, covering retrofitting principles, Markov Decision Processes, and planning algorithms like Q-learning and Value Iteration; the second applies these theories to practical implementation to develop a methodology for retrofitting planning optimization. Detailed methodologies are discussed, emphasizing building energy performance and degradation factors, with conclusions drawn to inform the simulation of building performance over time. This comprehensive approach aims to advance the current understanding and application of planning optimization in the building retrofitting domain. ... Read more

The growing threat of climate change highlights the necessity for long-term solutions in the construction industry. Buildings account for a large share of worldwide energy consumption and carbon dioxide emissions, so there is an urgent need for creative techniques to improve energy efficiency and reduce their ecological footprint. This thesis focuses on this by creating a computational optimization workflow for incorporating hempcrete, a low-carbon construction material, into high-performance structures during the initial stages of design. This approach uses a multi-objective optimization process to offer optimal solutions adapted to various climates and building types, optimizing energy efficiency and daylight while limiting global warming potential. Architects and engineers can get greater performance and sustainability results by experimenting with different layout options and design parameters using parametric modelling, energy analysis, and optimization algorithms. The suggested workflow provides a systematic technique to facilitate decision-making during the key design steps, promoting hempcrete implementation and accelerating the shift to performance-driven architectural design in response to climate change problems. ... Read more

The thesis, titled ‘Towards the Integration of Additive Manufacturing for Freeform Steel and Glass Façade Construction,’ explores the transformative role of additive manufacturing (AM) in enhancing the design and construction of Freeform Steel and Glass Façades (FFSGF). These façades involve intricate components, necessitating collaborative efforts among designers, engineers, and fabricators to achieve complex façade geometries with good performance while ensuring efficiency in design, material usage, fabrication, and cost. Recent advancements in AM technology have made it a potentially viable and increasingly accessible fabrication strategy, particularly for mass-customized components such as those used in freeform construction. The research aims to answer the central question of how AM can effectively contribute to developing node solutions supporting Freeform Steel & Glass Façade (FFSGF) construction.The study systematically examines opportunities for improvement in existing solutions. Chapters 2 and 3 provide a critical overview of FFSGF construction typologies and explore the current landscape of AM in façade applications. Chapters 4 and 5 delve into the design and development of crucial components—structural nodes and gasket nodes—utilizing different AM technologies. Prototypes are manufactured and evaluated in comparison to existing solutions. Chapter 6 provides a case study of integrating AM product development into a larger construction project, emphasizing interdisciplinary collaboration. The thesis underlines the advantages of the systemization of AM node design throughout the design and fabrication of AM nodes. This research contributes to the evolving knowledge at the intersection of design, engineering, construction, and AM, aiming to provide a valuable resource for building industry professionals navigating the complexities of incorporating AM into the fabrication of freeform steel and glass façades. ... Read more

Across the world, countries are facing housing shortages and the Netherlands is no different. The increasing demand for new housing exceeds the growth rate of the architecture, engineering, and construction industry. Current solutions remain small in scale and therefore unsustainable. Multi-family housing is the optimal typology to address the housing shortage but the industry cannot design and build these projects fast enough. Automation can help. In 2016, Kevin Kelly[2016] reframed the conversation about automation, by stating "our most important mechanical inventions are not machines that do what humans do better, but machines that can do things we can't do at all." Humans cannot address the housing crisis alone, but automation through the application of deep learning models can bring well-designed spaces to everyone.

Since the introduction of computers, architects have looked for ways to automate menial tasks. Some researchers even imagine a future where the machine becomes a partner to architects and designers, contributing to the design process. This ambition requires that the algorithms train themselves. Innovations in the field of deep learning have made this possible by allowing algorithms to train themselves through the use of artificial neural networks. When it comes to applying deep learning to generative design tasks, however, there is little research. The studies that have been done generate geometry that is small in size (64 x 64 x 64 voxels) and focuses on objects like chairs, not on buildings. 3D generative adversarial networks show promise for generating building geometry. By automating design, it is possible to apply expert knowledge on good design to all projects so everyone has access to well-designed buildings.

This research aims to develop the architecture for a generative adversarial network that produces feasible building geometry. An important first step was identifying and pre-processing a data set that could be used for this purpose. The data set is released with the publication of this thesis so it can be used for further research. Through this thesis, the Improved 3D Wasserstein Generative Adversarial Network architecture has also been developed and documented. The research found that using a combination of Wasserstein loss with gradient penalty, Leaky ReLU activation functions in the generator and the critic, and the RMS Prop optimizer results in an architecture with stable training and outputs that are similar in size, shape, and proportion to the training data with minimal noise in the output. Performance of 3D Wasserstein generative adversarial networks with these hyperparameters was improved even further when using ten layers and a larger number of channels. The experiments concluded that generative adversarial networks can be used to generate building geometry and can be an area of continued research to improve generative design tools and support the automation of architectural design. ... Read more

Current building operations can be improved through smart predictive operation based on weather and use patterns in order to save energy with minimal impact on the building fabric and daily use. The existing literature has investigated implementations, and potential savings through combining with variable tariffs, however, this thesis addresses the issue of how different buildings differ in their suitability for such smart control.

In this thesis, a digital twin of a school is created and adjusted to test differences in building fabric factors. These are combined with multiple Deep Reinforcement Learning (DRL) agents, which are trained to operate the schools more efficiently by controlling the heating set-point as well as natural ventilation of the buildings in order to save energy while maintaining comfort. The DRL agents vary in their ability to observe future weather as well as their internal network model architecture.

The results show a high energy saving compared to a simple baseline, despite the few building controls available to the agents. In addition, some algorithms out-compete even rule-based controllers, which were tested as a stricter baseline. The results also confirm a theory revealed through the literature review, that buildings with higher energy input, storage and control have a larger potential for energy savings. Additionally, the types of DRL models used also greatly influences the agents’ ability to perform well, and generally more advanced models performed better. The findings can be used to access a building’s suitability for and potential benefits from such predictive smart control. ... Read more