This study is inspired by a paradox present in highly decentralised multi-business organisations: the same autonomy that drives unit-level performance systematically suppresses cross-unit innovation collaboration. Despite this, the paradox remains underexplored, particularly inorganisational settings where the corporate centre refrains from using formal authority tomandate collaboration. This thesis investigates how a regional headquarters might enable voluntary cross-unit innovation collaboration without compromising business unit autonomy. A single embedded qualitative case study was conducted at Company X, drawing on ten semistructured interviews across various levels of the organisation. Three mutually reinforcingcategories of barriers and levers have been identified: structural, motivational, and cognitive.Together, these barriers create a low-collaboration equilibrium that resists any single-leverintervention. This thesis identifies the regional headquarters as the most effective actor fordeploying a set of levers to shift this equilibrium towards voluntary collaboration without directly mandating it. The central theoretical contribution extends Chesbrough's Open Innovationframework within organisational boundaries, conceptualising the regional headquarters as anintra-organisational Open Innovation intermediary, a role the existing literature has yet tosystematically theorise. The resulting framework offers both a grounded conceptual addition tothe Open Innovation and corporate parenting literatures and a set of actionable recommendations for practitioners.

The quantum Fisher information has gained popularity due to its versatility and its multi-purposed nature. For example, in quantum metrology the quantum Fisher information is used to quantify the maximal accuracy for which a parameter may be estimated and in quantum information theory it is used as a witness of multipartite entanglement. In condensed matter physics the quantum Fisher information has been used to study phase transitions, primarily in equilibrium quantum many body systems. In this work we study the quantum Fisher information for a p+ip superfluid both in equilibrium and out of equilibrium. In equilibrium the p+ip superfluid exhibits a topological quantum phase transition between a topological weak pairing phase (BCS) and a topologically trivial strong pairing phase (BEC). After an instantaneous interaction quench, the long time asymptotic state of p+ip superfluid is characterized by three different dynamical phases and within each phase the behaviour of the asymptotic order parameter changes. In equilibrium, we find that the QFI calculated with respect to particle number using the BCS ground-state fails to probe the topological quantum phase transition. On the other hand, in our non-equilibrium treatment we find that by quenching the system out of equilibrium the quantum Fisher information can be used to extract topological information of the pre-quench state and the character of the energy dispersion of the post-quench state. Our results therefore suggest that by introducing a quantum quench the quantum Fisher information is able to extract useful information and does not rely on the choice of operator, as opposed to the equilibrium case.

Quantum photonic technologies need reliable ways to study light emission from nanoscale defects and other quantum emitters. FLIM is useful because it measures decay dynamics rather than brightness alone, making it sensitive to local optical environments and non-radiative loss channels.

This thesis develops and evaluates a confocal microscope for high-resolution FLIM using TCSPC method and a SNSPD. The work combines hardware calibration with Python software for synchronized scanning, photon-event collection, live visualization, ROI fine scanning, and post-scan lifetime analysis. The full time-resolved photon data are saved for each pixel, enabling later lifetime extraction, instrument-response correction, and multi-exponential fitting.

The platform is tested on point defects in hBN and excitonic regions in MoSe2/MoS2 heterobilayers. The hBN measurements show localized emitters with lifetimes consistent with room-temperature SPE behavior, while targeted fine scans improve spatial detail around selected defects. In the heterobilayer sample, spectral filtering and lifetime fitting help separate monolayer and bilayer contributions, with double-exponential analysis giving a clearer description of overlapping decay dynamics.

Overall, this work shows that an SNSPD-assisted confocal microscope can provide sensitive, time-resolved maps of quantum materials across both broad-area scans and focused emitter-level measurements. The platform offers a practical foundation for future studies where spatial precision and timing resolution are both essential.

Cellulose nanofibers (CNF) are among the most promising renewable building blocks for high-performance materials. The combination of TEMPO-mediated oxidized cellulose nanofibers (TCNF) with a flow-focusing channel yields the strongest bio-based filaments reported to date. This thesis follows TCNF filaments through their full processing life cycle: controlling the flowing dispersion, spinning parameters, drying temperature, recycling process, and functionalizing, and asking how each stage determines the structure and properties of the filaments.

The flowing dispersion is monitored in situ by integrating a flow-focusing geometry directly into a polarized optical microscope (POM). Spinnability is shown to be governed by the homogeneity of the fiber network rather than individual dimensions, providing a predictive criterion that replaces batch-by-batch trial-and-error testing. Post-spinning thermal drying reshapes filament structure and mechanics through two competing effects: moderate heating degrades performance, whereas high-temperature densification and co-crystallization restore strength to the single-fiber value. Circularity is addressed by redispersing dried filaments into a spinnable dispersion: blending with fresh TCNF restores colloidal stability and spinnability while retaining a large fraction of the original mechanical performance. Finally, the filaments are functionalized into carbon-fiber composites whose mechanical and thermal-transport behavior extends the flow-focusing platform from purely structural fibers toward functional, thermally conductive materials.

Together, these studies establish a coherent, resource-conscious framework for engineering the properties of bio-based cellulose filaments across their full life cycle, from the flowing dispersion to a recyclable, functional fiber.

Cellular networks have become the cornerstone of modern communications in the past decades. Starting with the first deployments by the end of 1970s until nowadays, cellular networks have connected every place on earth, offering services to billions of people. Initially planned for voice communication only, over time they have enabled support for various applications, some of which are characterized by stringent traffic requirements, ranging from high data rates, extremely low latencies, up to an almost 100% reliability. Of course, this was not an abrupt process, but it took years until this became possible, and several generations of cellular networks, each of which brought significant improvements to the user experience. 

The current generation of cellular networks, 5G, deployed first in 2019, is experiencing rapid development. 5G networks arose from the need to support data-intensive applications, such as virtual reality, augmented reality, extended reality, high-quality live video streaming, and online gaming, which was not possible with the previous generations of cellular systems, on the one hand, and providing access to a huge number of machine-type devices, on the other. 5 The research community backed up by industrial players have already started working on 6G networks, planned to be fully operational by 2030. 

With the ever-increasing number of users running these bandwidth-hungry applications, even the operation of 5G is jeopardized, given the scarcity of radio spectrum. Therefore, it is of paramount importance to properly allocate the network resources so that as many users can run cutting-edge applications on their smartphones with a satisfying level of service. This is one of the aspects considered in this habilitation. 

Further, applications related to immersive communications require, besides radio resources, computation resources as well to obtain results quickly. This fact, together with the rapid development in the area of mobile edge computing has enabled the rise of in-network computing feature in 5G, where data-intensive tasks can be executed on edge clouds, which are usually collocated with base stations. Hence, understanding the performance limits of edge clouds has become particularly important. Besides understanding the achievable performance, optimizing different aspects of edge clouds is important too. To that end, as part of this habilitation, we solve different optimization problems taking into account various resource limitations. 

The dynamic nature of cellular communications requires providing support for mobile users when leaving the area of coverage of one base station and entering the area of another base station, a process known as handover. Therefore, it is also important to investigate mobility management in cellular networks, and in particular in 5G. More importantly, considering the joint problem of performing handovers and resource allocation to users within several base stations while they are moving around is challenging. This should be done while providing performance guarantees to the users and considering different types of fairness. This is to be followed by solving optimization problems, whose results should lead to considerable improvements compared to the results obtained using the current 3GPP-based standards. This is the third aspect covered in this habilitation. 

Another way of improving the performance of wireless communication systems is by predicting correctly the channel conditions. The common way used nowadays is by transmitting pilot symbols. However, frequent pilot transmissions lead to an increase in overhead signaling, increasing the latency, making it detrimental for time- and reliability-sensitive applications. A representative example are robots in factory environments. Given the nature of the traffic in this use case, reducing the signaling overhead is of paramount importance. A way to achieve this is by using depth images as part of a machine learning system for predicting the channel state information without feedback and pilot signals. Such a system should be capable of predicting even the channel conditions in the next slots. This approach in optimizing the cellular network operation is the fourth aspect considered in this habilitation. 

Although the aforementioned topics are related more to the Radio Access Network part of the cellular network, we have also considered optimization aspects that cover the Core Network too, encompassing this way the entire cellular architecture. In addition, we have also focused on minimizing energy consumption, as a contribution to sustainability, which is one of the key pillars of next generation cellular networks. 

In this habilitation, we follow two broad approaches that lead to performance improvements of different aspects in the operation of cellular networks. In the first, we formulate constrained optimization problems, and solve them using the respective techniques, depending on the type of the optimization problem at hand (convex optimization problems, integer linear programs, mixed-integer linear programs), including the tools from Machine Learning, such as Reinforcement Learning. The second approach consists in exploiting one or more of the features of the problem to obtain an optimal solution, without using the traditional techniques in formulating and solving (constrained) optimization problems. 

In all the problems considered in this habilitation, we compare our approaches against state of the art for different scenarios, running simulations mostly on publicly-available datasets, and show the advantages our approaches offer in terms of the optimality of the results and reduced complexity. The results obtained here can be used by cellular operators to optimize various aspects of cellular network operation as well as to provide different types of fairness.

Predictive maintenance plays a central role in modern industrial production, where unplanned downtime in rotating machinery such as bearings, gearboxes, motors and pumps leads to significant economic and operational losses. Vibration data is the most established information source for condition monitoring of rotating machinery, and machine learning – particularly Convolutional Neural Networks (CNN) – has emerged as a key approach for vibration-based fault diagnosis. The research field is, however, fragmented across architectures, data representations and application contexts.

This thesis carries out a systematic literature review to examine how CNN models are used for fault diagnosis with vibration data within predictive maintenance of rotating machinery in industrial production, and how their performance compares to other machine learning models. Approximately 20 peer-reviewed publications from 2015–2026, primarily indexed in Scopus, were analysed through a qualitative thematic content analysis.

The results show that CNN is mainly used to automate feature extraction, either through 1D-CNNs applied directly to raw vibration signals or 2D-CNNs applied to time-frequency representations such as STFT and wavelet transforms. Hybrid architectures, especially CNN–LSTM, are frequently used to handle non-stationary signals. CNN performs best when datasets are large and signals are complex, whereas classical models such as SVM, KNN and Random Forest remain competitive for smaller datasets and when interpretability is prioritised. Recurring limitations include scarcity of labelled fault data, heavy dependence on benchmark datasets, weak domain transferability and limited interpretability.

The study concludes that CNN is a flexible and powerful but context-dependent component in predictive maintenance, and that the central challenge lies in bridging laboratory-based model development and real industrial applications.

The majority of the public bathhouses in Sweden were constructed during the 1960´s and 70´s and had, when they were constructed, an estimated lifespan of 50 years. Today, more than 50 years later, we can see the result of this, as many of the public bathhouses today are worn out in many ways. The bathhouse as a typology is now facing a crossroad at the same time as its benefits remain: they contribute to public health, water safety, development and social cohesion, to mention a few aspects. My project takes place in my hometown, Bollnäs in Hälsingland, Sweden. The project explores the bathhouse as a typology and investigates what role the bathhouse of the new generation can play in a small town. The central intention of the project has been to investigate how architecture can respond to the intertwining between the great complexity that a bathhouse brings with it and the relatively small programme that the municipality proposes, which is natural in relation to the size of the town. I have approached this project as an exploration of the architectural idea of the bathhouse as something to be experienced rather than a building to be visited. 

Järvafältet is the location of Stockholms only state owned agricultural operated farmland and it has been used for agriculture since at least the 1400s. However, the farmland has remained unused since 2018, and there has been ongoing debate about its future use. Several suggestions have been made for possible alternative uses for the farmland. Especially to open up parts of the farmland for local food production or community gardens. The surrounding areas include Tensta, Kista and Husby, which are lower income areas with few social activities and meeting places. I propose transforming a portion of the farmland into a public farmland, where individuals can grow their own crops and plants, similar to the concept of kolonilott (allotment gardens). To complement this I will design a agricultural building that serves the needs of the farm users and park visitors. This building will include utility storage, workstations, rest areas, bathrooms and a café. It will provide the necessary support to keep the area more accessible for the ones using the farmland. Additionally, it can function as a gathering space for meetings, workshops, or community events. It can also be a possible resting stop for people who will visit the rest of the green area and nature reserve.

The digitalization of Swedish industry has placed digital environment, health, and safety (EHS) management systems at the center of companies' daily operations. This study evaluates the IA system (Information for Work Environment), the market's dominant and cost-free tool for systematic work environment management (SAM) in accordance with the Swedish Work Environment Authority's regulations AFS 2023:1, focusing on its user interface, usability, and cognitive ergonomics. The problem formulation is based on the hypothesis that a conflict exists between technical usability and legal compliance. The study was conducted as a thesis project on behalf of the consulting firm Vaara EHS.

To answer the research questions, a mixed-methods approach with methodological triangulation is applied. The implementation comprises objective system testing of the desktop and mobile platforms, five semi-structured interviews with managers and work environment engineers, and a survey based on the validated System Usability Scale (SUS) measurement tool. The empirical material is analysed based on the Human-Technology-Organization (HTO) model, the ISO 9241-11 usability standard, Nielsen's design heuristics, and cognitive load theory.

The results demonstrate a functional dual nature, where the system's overall usability achieves a high SUS score of 82.75, indicating strong system acceptance. Conversely, the desktop interface burdens administrators and managers through inflexible modules, rigid drop-down lists that obscure specific data under the catch-all category "other", and a total lack of integration with regular work order systems. Furthermore, the system's inadequate dashboards force managers to regularly export raw data to Excel to manually create customized decision support for management. The conclusion is that the IA system offers a strong reactive database for long-term traceability according to Section 14 of AFS 2023:1, but its proactive support falls short due to insufficient automated communication flows. This forces organizations into informal, manual routines outside the system to fulfil the requirements for written action plans according to Section 13 of AFS 2023:1. For future development, integration of order systems and the implementation of artificial intelligence are recommended to reduce cognitive friction and proactively improve data quality.

University students around the world are organising gaming communities on their own initiative, yet little research has examined why they do so or how these communities function in practice. This study investigates what motivates university students to participate in and contribute to grassroots gaming communities, and how digital and physical interactions shape the way these communities are organised and sustained. Using a qualitative case study approach, thirteen semi-structured interviews were conducted with university students involved in communities organised around the online geography game GeoGuessr.

The findings show that motivation is rarely driven by a single factor. Social connection is the most consistent primary motivator. Students most commonly start these communities through existing friendships, with the shared activity serving as a natural catalyst, though communities themselves consistently become spaces where new connections form. Competition and learning function as secondary motivations that reinforce each other over time, and among more experienced participants, teaching others emerges as a distinct and self-sustaining motivation.

Regarding organisation, physical and digital interactions serve complementary but distinct roles. Among participants with access to physical co-presence, in-person interaction functions as the more natural catalyst for grassroots community formation, sparking the initial bonds and shared experiences that are harder to ignite from scratch online, while digital platforms provide the organisational backbone that keeps communities connected, coordinated, and active across time and distance in ways that physical meetings alone cannot always support. Digital infrastructure, however, concentrates organisational labour in ways that create structural fragility when key contributors disengage. Engagement is cyclical, and re-engagement after inactivity is driven by the interplay of both digital and physical cues.

The findings are discussed in relation to existing literature on gaming communities, motivation, and digital habitats, and assessed against Jung's (2020) Gamers' Communicative Ecology Model, which provides a valuable descriptive starting point despite contextual limitations. The study contributes to a broader understanding of informal organisation and collective participation in digital environments.

In industrial production environments, the attention of production personnel constitutes a safety critical resource. The use of personal audio equipment, such as headphones, has become increasingly established, while there is a lack of industry wide guidelines that account for the potential risks associated with such use. Existing research addresses auditory isolation, situational awareness, cognitive load and noise exposure as separate phenomena. This creates uncertainty regarding the combined risks that the use entails in an industrial context.

The purpose of the report is to investigate whether the use of personal audio equipment constitutes a safety risk in industrial production environments, and how findings from existing research can be applied in an industrial context. Volvo Powertrain AB's operations in Skövde and Köping serve as the empirical context of the report. These production facilities maintain differing approaches to personal audio equipment, which enables comparative analysis. The method consists of a systematic literature review, an analysis of reported safety incidents, as well as site visits and interviews with production workers at both facilities. The theoretical framework is based on Endsley's model of situational awareness and Sweller's findings on cognitive load. The legal framework is based on a selection of relevant provisions from the Swedish Work Environment Act (1977:1160).

The results indicate that personal audio equipment potentially impairs the ability to perceive safety critical acoustic information. Level dependent hearing protection enhances the perception of sound but may cause localization errors when identifying hazards. The cognitive load induced by listening to music is conditional and depends on the character, tempo and volume of the music. Experiences from Köping show that an established ban on personal audio equipment is practically feasible and tends to become normalized over time. Compliance with safety regulations is also considered to improve when leaders actively emphasize the importance of following workplace safety rules.

The report concludes that regulating the use of personal audio equipment is warranted and motivated by the employer's preventive responsibility under Chapter 3, Section 2 of the Swedish Work Environment Act. This responsibility entails an obligation to reduce elements that constitute a potential safety risk.

Balancing high customer service levels with cost efficiency remains a central challenge in modern logistics management. Although route optimization is well established mathematically, its operational implementation is often constrained by information asymmetry. In distributed transport networks, variable costs such as transit freight and terminal fees are frequently subject to delayed documentation (backbilling), rendering them unavailable to planners during execution. As a result, exclusive reliance on traditional Least Cost Routing models introduces a structural information gap, increasing the likelihood of sub-optimal routing decisions in which apparently efficient routes result in unforeseen financial deficits after invoicing.

This study addresses this gap by examining methods for integrating and visualizing lagging variable costs within a Decision Support System (DSS) for operational route optimization, and by evaluating the impact of this integration on the quality of routing decisions. Employing a case study of DACHSER’s European transport network, the research develops a conceptual decision model and quantitatively assesses it through back-testing with historical transport data.

The theoretical framework integrates Activity-Based Costing (ABC) to map cost drivers with Multi Criteria Decision Making (MCDM) to systematically balance financial trade-offs against strict two-day transit Service Level Agreements (SLAs). Furthermore, the study explores principles of Visual Management and behavioral nudging to reduce cognitive load for operators working under time constraints. However, a theoretical tension is identified regarding the risk of automation bias; over designed interfaces that aggressively nudge operators toward financially optimized routes may inadvertently cause them to overlook qualitative operational risks.

Ultimately, this research bridges the gap between retrospective financial auditing and proactive operational planning, contributing to a deeper understanding of how integrated data modeling and visual cost governance can enhance decision quality in complex logistical networks.

Complex construction projects rely on well-structured internal logistics to avoid delays. However, the goods reception at the studied site suffered from an unstructured layout and a broken information flow. This caused severe traceability gaps and forced workers into erratic search paths. To solve this, the thesis aimed to develop and evaluate an analog visual management framework. The research followed a single case study approach, gathering empirical data through site observations, spaghetti diagrams, value stream mapping, and semi-structured interviews.

The results show that dedicated 5S drop-off zones and an analog Poka-Yoke tracking board effectively eliminate disorganized search patterns. This intervention saves an estimated 30 minutes per worker daily. Scaled across a typical site workforce of 20 personnel, this results in a projected 10 labour hours saved per day. The framework also closes the administrative information gap. By using personalized magnetic tokens, site management can secure the chain of custody during material retrieval. The study concludes that construction sites should apply the ”Lean first, digital second” principle. Before investing in complex Warehouse Management Systems (WMS), major operation and sustainability gains (SDG 8 and 12) can be achieved through low-cost analog tools.

This study evaluates the implementation of an automated robotic cell with a Haeger machine for screw assembly at AB Linde Maskiner. The purpose is to investigate its economic viability, technical performance and flexibility. Furthermore, the social and environmental sustainability of the automated cell is compared to that of the current manual assembly process.

The results are based on empirical data collection, operational observations, and investment analyses. The robotic cell reduces the annual setup time by 25 percent and downtime by 47 percent. The weighted cycle time is reduced to 16.9 seconds. For more complex components, the cycle time is reduced by up to 53 percent, while it increases slightly for simpler components. Economically, the initial investment of SEK 4.2 million yields a positive net present value of approximately SEK 2.9 million over a 10-year period at a 7 percent discount rate. However, the payback period is 4.2 years, which exceeds the company's requirement of three years. Sensitivity analyses show that the net present value remains positive, but that the payback requirement is not met in any scenario. Furthermore, for the calculation to hold, it is required that operator presence is limited to 25 percent.

From a social perspective, the number of job opportunities decreases, but the work environment improves as monotonous tasks are replaced by technical duties. Environmentally, the vision system minimizes scrap during operation, but product-specific pallets risk generating material waste during future product revisions. It is recommended that the company request a quote guaranteeing a maximum of 25 percent operator presence, explore the market for more cost-effective alternatives, and investigate advanced vision systems.

This study was conducted in the MIA warehouse at AstraZeneca's production facility in Gärtuna, Södertälje. The purpose of the study was to analyze the internal material flow for consumables and identify inefficiencies in the current work processes. The supply of consumables is a critical secondary process for pharmaceutical manufacturing and the study examines how standardization and digitalization can contribute to increased operational stability.

The methodology consists of a literature review combined with empirical data collection through semi-structured interviews, direct observations, Gemba walks, and indicative time studies of frequent work steps. The collected data has been analyzed in relation to established theories within Lean Production, industrial logistics, and internal and external efficiency.

The results show that the material flow in the MIA warehouse is characterized by significant process variation, manual and analog registration routines and unclear organizational ownership. Several work steps are performed based on experience and personal knowledge rather than documented working methods, which creates variation in work performance and makes the business vulnerable to personnel changes. Information management is largely dependent on paper-based documentation, manual information transfers and individual assessments. The lack of systematic notifications also contributes to a reactive working method and hidden flow losses, including in the form of speculative control rounds to the CNC lock. The study further shows that standardization is the most important prerequisite for long-term improvements and that digital solutions can contribute to increased traceability, more efficient information flows and reduced administrative workload when implemented as support for standardized processes.

This study explores how founder–angel investor relationships evolve over time in early-stage ventures, focusing on expectations, investor involvement, governance, and misalignment. Using a qualitative research design, semi-structured interviews were conducted with founders and angel investors and analyzed through thematic analysis. The findings show that these relationships are dynamic and continuously renegotiated as ventures develop. While early interactions are often characterized by trust, shared expectations, and informal coordination, relationships become more complex over time. Investor involvement is valued when it aligns with venture needs, while communication, trust, and mutual adaptation remain central despiteformal governance structures. The study also highlights recurring misalignment regarding expectations, strategic direction, and decision-making authority. Overall, the study contributesto a deeper understanding of how founder–angel investor relationships are formed, maintained, and managed in early-stage ventures.

Industrial production systems, despite various interventions, produce persistent and unsustainable consequences (overproduction, resource intensity, chemical waste). This study argues that these outcomes are not operational failures but structural outcomes: they are produced by how production systems are configured, and those configurations persist because they are self-reinforcing.

To diagnose this problem, the study draws on how nature works as a system. Grounded in ecological succession theory and complex adaptive systems (CAS) theory, the study develops a stage-based diagnostic framework that positions industrial production systems across four structural configurations. Ecological succession provides a developmental logic for understanding how configurations consolidate over time; CAS theory explains why they resist incremental change.

The framework is operationalized through five diagnostic dimensions, each grounded in succession mechanisms and observable from secondary data. It is applied to upstream denim manufacturing, one of the most resource-intensive segments of the global textile industry.

The analysis reveals an asymmetric configuration profile: the system occupies a Stage 1-2 position overall, with feedback loop relevance advancing faster than configuration lock-in or structural modularity. This asymmetry helps explain why partial improvements do not consolidate into system-level reconfiguration. Expert validation confirms the framework's characterization and identifies economic incentive structures and measurement fragmentation as boundary conditions outside its current scope.

The study contributes a mechanism-based diagnostic framework applicable to industrial production systems beyond the textile industry. By re-examining sustainability issues in production as symptoms of a single structural configuration rather than isolated inefficiencies, it makes visible why interventions that do not address the underlying architecture are absorbed without producing structural change. The structural visibility, it argues, is a necessary prerequisite for designing interventions that can change this architecture.

Premium brands are under growing pressure to extend the life of their products without weakening the meanings that justify their prices. In B2B hospitality, this is harder than it sounds: durable products stay visible to guests for years, but hotels, not the brand, are the ones maintaining, using, and eventually replacing them. Research on luxury brand management explains how authority is built at the point of sale. Research on lifecycle extension explains how to prolong product use. Neither explains how a premium brand governs what happens after the sale, across multiple stakeholders it does not directly control.

This thesis asks under what conditions lifecycle control is compatible with a luxury strategy in B2B markets for durable goods. It is a qualitative single-case study of a premium sleep products manufacturer operating in the hospitality segment. The empirical material consists of 14 semi-structured interviews with internal company respondents, hotel and hospitality actors, and an external luxury-brand expert, supplemented by internal documents and secondary sources. The analysis draws on a framework integrating luxury brand management, product lifecycle extension, B2B distribution governance, material constraints, and digital traceability.

The findings show that compatibility depends on five conditions: brand-governed intervention, perceptual invisibility or prestige framing, material feasibility, partner compliance, and selective traceability. Five lifecycle-control mechanisms are identified, with digital product passports serving as an enabling traceability infrastructure: hotel lifecycle stewardship contracts, replaceable comfort-and-hygiene layer subscriptions, atelier recrafting and recertification, closed-loop take-back credit, and third-party-financed operating leases. Stewardship contracts and comfort-layer subscriptions are the most immediately viable because they extend brand authority through routines, training, and replaceable components without requiring asset ownership or reverse logistics. Recrafting has a strong symbolic fit but belongs only to flagship applications. Leasing and broad take-back models involve higher financial and operational risk and are better deferred.

The thesis reframes lifecycle extension in luxury as a brand governance problem rather than an environmental or operational one. It develops the Brand-Controlled Lifecycle Governance Framework, which maps how decision rights, information flows, and quality-control structures can sustain brand authority through the post-sale phase. It also introduces selective traceability as a governance principle, distinguishing internal visibility for control from external disclosure for brand narrative. The practical implication is that lifecycle control should be built gradually, starting with care routines, then replaceable components, then internal data infrastructure, before moving toward leasing or large-scale take-back.

  Increasing sustainability reporting requirements under frameworks such as the Greenhouse Gas Protocol (GHG Protocol) and the Corporate Sustainability Reporting Directive (CSRD) have increased the importance of accurate Scope 3 emissions reporting. For many industrial manufacturing firms, Scope 3 Category 11: Use of Sold Products represents a significant share of total emissions. However, current reporting practices are often based on generalized assumptions and secondary data due to limited access to downstream customer information.

This study investigates the challenges, opportunities, and enabling practices associated with collecting primary downstream data for Scope 3 Category 11 reporting within industrial manufacturing firms. The study adopts an abductive qualitative study approach conducted in collaboration with a global industrial manufacturing firm, anonymized as Alpha. Empirical data was collected through semi-structured interviews, focus groups, and document analysis involving sustainability, sales, and strategic functions.

The findings show that transitioning toward more primary-data-based Scope 3 Category 11 reporting is constrained by organizational fragmentation, manual reporting practices, methodological uncertainty, and the commercially sensitive nature of operational customer data. At the same time, the study identifies opportunities for more representative downstream emissions reporting through selective and scalable approaches to primary data collection. In particular, site-specific energy mix data emerges as a more feasible alternative to detailed operational usage data, as it is perceived as less commercially sensitive and is often publicly disclosed by large customers.

The findings further highlight that successful implementation depends on trust-based customer relationships, clear organizational ownership, cross-functional collaboration between sustainability and customer-facing roles, and increased automation supported by integrated digital systems and standardized reporting processes.

The study contributes to the understanding of how industrial manufacturing firms can transition toward more representative and scalable Scope 3 Category 11 reporting practices while balancing reporting accuracy, organizational readiness, and inter-organizational collaboration.

Pharmaceutical formulation still includes manual and semi-manual work that can expose operators to physical strain. At the same time, automation decisions are limited by Good Manufacturing Practice (GMP), validation requirements, space constraints, and product quality demands. This thesis examines how task-level automation opportunities in pharmaceutical formulation can be identified and prioritized by combining ergonomic workload, human-automation interaction (HAI), and constraints in a regulated production environment.

An exploratory single case study was conducted in a pharmaceutical formulation production line. Data were collected through process observations, operator input, task exposure measurements, Rapid Entire Body Assessment (REBA), document review, and stakeholder interviews. Based on literature in process improvement, ergonomics, and HAI, a staged framework was developed and applied in three rounds. Round 1 screened 31 task-level pain points using operator input, frequency, and safety, health and environment (SHE)-criticality. Round 2 evaluated the 12 remaining tasks using task frequency, task duration, and REBA scores. Round 3 assessed the three highest-ranked tasks based on technical difficulty, space limitations, regulatory difficulty, cost, implementation effort, and operator input.

The results indicate that ergonomic risk affected task prioritization, but physical strain alone was not enough to select final automation candidates. The strongest candidates were tasks where ergonomic burden, clear task boundaries, and practical feasibility overlapped. The thesis presents an early-stage framework for screening automation opportunities at task level before detailed technical design, investment analysis, or implementation planning.

This thesis presents the development of a discrete event simulation tool for BakeMyDay's bread production line. The simulation is built in MATLAB SimEvents and models the entire production flow from raw material handling in the silo to finished packaged products. The purpose is to create a planning tool that can be used to identify bottlenecks, test what-if scenarios, and optimize the production schedule without disrupting real production.

The simulation models bread products flowing through seven production stages: silo, dough mixers, forming stations, yeast ovens, egg wash, topping, baking ovens, and packing stations. Each production resource is modeled with realistic capacity constraints and product-specific process times.

Results show that the Vetelina station, which handles manual forming of a specific product, is the bottleneck in production with a utilization rate of 96.5%. An analysis shows that an additional worker at the station reduces forming time from 69.7 to 55.76 minutes and yields three additional packed wagons per shift. An exhaustive search method based on 1000 simulation iterations resulted in a 6.1% higher production output compared to the original schedule.

The thesis demonstrates that discrete event simulation is an effective tool for production planning in the food industry and that simulation-based decision support systems can provide concrete and measurable improvements in production efficiency.

To manage their exposure to emerging risks and opportunities, organisations monitor external developments, such as new regulations, market shifts, and technological innovations. This activity, known as environmental scanning, presents several challenges for analysts performing it while also proving difficult to automate due to its judgment-intensive nature.

The study investigates how analysts perceive the value and limitations of agentic AI, meaning AI that can independently execute multi-step tasks, in automating this type of judgment-intensive business process.

The empirical case explored in the study focuses on the automation of regulatory environmental scanning within the sustainability function of a global heavy manufacturing company. Data is collected from 18 interviews, the development of five AI prototypes, and four workshops where analysts tested AI-supported versions of scanning tasks.

The study finds that value varies across process steps, being strongest for information processing tasks but still requiring significant human oversight for tasks needing accountability or depending on interpersonal relationships.

Secondly, the study finds that, in addition to increasing productivity, agentic AI created value through mechanisms such as process formalisation and reduced dependencies. The implication is that much value is realised through process transformation rather than the technology itself.

Finally, the study finds that agentic AI may fall short when the underlying process bottleneck is misunderstood, when users do not trust the system, or when the agent lacks sufficient internal context or customisation. This suggests that successful implementation requires early end-user involvement and gradual adoption with meaningful human control.

Spain’s utility-scale solar market is undergoing a period of revenue compression. Average midday clearing prices fell from € 109/MWh in 2021 to € 41/MWh in 2025, while cumulative installed capacity nearly tripled. Non-compensated curtailment more than doubled in 2025, peaking at 1.2 TWh in July alone. Price cannibalization and curtailment reinforce each other and realized revenues are now falling well below the original financial projections of these assets.

This thesis evaluates whether co-locating Bitcoin mining as a flexible demand-side load at a utility-scale solar plant can recover value from the energy that is either lost or undervalued at the asset boundary. The analysis is conducted from the perspective of the asset owner, on a 100 MW reference plant. A 15-minute dispatch simulation covers the period May 2024 to March 2026, using historical Iberian Electricity Market (MIBEL) spot prices, plant generation and curtailment records, and Bitcoin network data. Hardware selection and fleet sizing are performed on Equivalent Annual Annuity (EAA) across 13 candidate ASIC models. The discounted cash flow uses a 10% discount rate, reflecting the industry partner’s internal stress-test hurdle for renewable investments.

Four operating strategies are compared: curtailment-only dispatch (Technical Scenario (TS)1), curtailment plus local PV energy use (TS2), profit-gated grid import (TS3), and continuous data-centre-like operation (TS4). The dispatch logic is structured as a merit-order stack across three energy layers, with each scenario activating a different subset of layers.

The profit-gated dispatch strategy (TS3) is the recommended configuration. A fleet of 47.94 MW of Antminer S23 Hydro generates an EAA of € 12.61M per year, with an NPV of € 47.8M, IRR of 25.3%, and discounted payback of 3.34 years within the 5-year hardware lifetime. Utilisation reaches 97.9%, with grid import supplying 61.3% of consumed energy, local PV energy use 28.2%, and curtailment 10.5%.

It is also concluded that higher utilisation is not a sufficient condition for higher value. TS4 reaches 100% utilisation by removing the profitability gate on grid import, but generates € 0.21M per year less than TS3 because the additional operating hours are loss-making on a variable cost basis. Moreover, optimal hardware selection depends on the energy mix: under low-cost energy regimes the optimiser selects low-CAPEX hardware, while under high-cost energy regimes it switches to high-efficiency hardware whose unit CAPEX is over ten times higher, as it remains profitable across a wider range of scenarios, on an operational basis. Curtailed energy and local PV energy contribute significantly more to the net margin than to the share of consumed energy, due to their low-cost characteristics.

The sensitivity analysis stress-tests the model across five different parameters: Bitcoin price, mining network difficulty, MIBEL spot price, curtailment volume and discount rate. NPV barely changes with curtailment volume, moving by under € 2M across a −40% to +40% range, meaning that a future reduction in Spanish grid congestion would not destroy the financial case. The project is exposed to BTC price declines below 0.8 times the period mean, where NPV turns negative, and to MIBEL spot price increases, which compress the margin on the import layer. The 50 MW import cap imposed by Royal Decree 244/2019 binds before the EAA curve peaks under upside BTC price scenarios, so the project’s upside is regulatory rather than economic.

Bitcoin mining co-location in utility-scale solar parks can materially improve returns for Independent Power Producers (IPPs) under current market conditions and regulatory framework. The results are conditional on the modelling assumptions and future evolution in Bitcoin price and mining network difficulty, which can heavily swing the project’s financials. The dispatch logic is technology-agnostic and can be extended to other flexible loads with minimal changes.

The weaponization and securitization of semiconductor supply chains have transformed semiconductors into strategic geopolitical assets, creating new challenges for industrial firms dependent on global supply networks. This study explores how industrial firms are affected by these developments and investigates the adaptation mechanisms implemented to mitigate their effects. An exploratory multiple-case study based on eleven semi-structured interviews was conducted with supply chain managers, procurement specialists, and industry experts from Europe and Asia. The findings show that firms are affected by both regulatory and physical disruptions, resulting in increased uncertainty, supply chain fragmentation, and reduced predictability.

To address these challenges, firms implement five primary adaptation mechanisms: supply chain integration, diversification, logistics diplomacy, supply chain mapping, and supply and demand planning. While these mechanisms are implemented differently between case companies, with respective trade-offs, they collectively aim to improve resilience, control, flexibility, visibility, and preparedness. The study contributes to the literature by presenting a new explanatory framework, linking the effects of weaponization and securitization in a semiconductor supply chain context to firm-level adaptation strategies, which extends the research beyond macro-level geopolitical discussions. From a managerial perspective, the findings suggest that resilience increasingly depends on firms' ability to balance the benefits and trade-offs between cost efficiency, organizational flexibility, and control, while tailoring their strategies to specific geopolitical risks and organizational contexts.

Demand planningis an important process within manufacturing companies and their supplychains.Within demand planning, demand forecasting is considered one of the most critical processes, and literature shows that artificial intelligencehas potential within this area. Despiteits expected potential to improve forecast accuracy, research regarding how it can be implementedfrom an organizational viewis limited.Therefore, this study seeks to investigate how artificial intelligence can be adopted to create value in demand forecasting.

To investigate this, a case study was conducted in the context of thecase company, a large global manufacturing firm.The study adopted a qualitative approach and primarily consisted of interviews and site visits. Through these qualitative methods, barriers related to the individual and organizational levels regarding the use and adoption of artificial intelligence within demand forecasting were identified. The barriers were identified with the combined theoretical framework:Technology Acceptance Model and Technology-Organization-Environment framework. Based on these barriers, recommendations for the adoption ofartificial intelligence within demand forecasting were developed.

The results show that there are several central barriers. These can be summarized as lacking data management, limited artificial intelligence and organizational readiness, resource constraints, trust and risk perception, unclear human involvement and poor system and processintegration. The results also indicatea lack of change management, complex technology, limited readiness among partners and stakeholders, as well as constraints related to regulations.

Through the identification of these barriers, the development of recommendations that can support the case company’sadoption of artificial intelligence within demand planning is enabled. However, these recommendations may also be applied to manufacturing companies operating in similar contexts and this study highlights the currently underrepresented research area concerning organizational perspectives on AI adoption within demand planning. Furtherresearch is, however, recommended to make the results more generalizable.

Sweden’s 2045 climate framework relies on bio-energy carbon capture and storage (BECCS), and the Stockholm region concentrates ten combined heat-and-power and waste-to-energy emitters generating about 2.78 MtCO2 per year. This thesis determines the least-cost design and rollout of the regional CO2 transport network from these emitters to the Northern Lights export terminals at Värtahamnen and Igelstahamn. A GIS least-cost-path analysis on a regulation-weighted friction surface identifies the feasible pipeline corridors, and a mixed-integer linear programming model then co-optimises build years, transport mode, pipeline diameters, compressor and booster stations, buffer storage, and weekly-to-monthly dispatch over 2026 to 2050, with fourteen scenario solves across the two corridors. The optimum activates eight of the ten emitters, anchored by Värtaverket and Högdalenverket from 2032, chooses for each emitter between a direct pipeline, a feeder onto a shared trunk, and road tankers, serving the high-volume emitters by pipeline and the low-volume emitters by truck, and converges all routes onto a single regulated corridor into each hub. The delivered system cost is 102 EUR/t in the Northern corridor and 105 EUR/t in the Southern, with the Northern Lights tariff alone above 60 % of system NPV and the regional engineering cost of 22 to 50 EUR/t in agreement with independent benchmarks. The topology is stable in five of seven scenarios per corridor, and steady-state operation requires about 328 ship voyages per year with a winter road tanker peak near 780 trips per month. The results give Stockholm stakeholders a quantitative reference for the late-2020s BECCS investment decisions.

Industrial water is systematically underpriced relative to its economic value. Utility tariffs typically reflect only the cost of abstraction, treatment, and distribution, omitting the costs of scarcity, supply disruption, regulatory tightening, and environmental externalities. The result is a pricing failure that distorts corporate capital allocation. Investments in water resilience and efficiency that are economically rational at the true cost of water often fail conventional return-on-investment scrutiny at the posted tariff.

This thesis develops a location-specific shadow pricing framework that calculates the per-cubic-metre economic cost of industrial water by decomposing it into four components: the utility tariff, a scarcity premium derived from WRI Aqueduct baseline water stress data and the cost of alternative supply, a disruption premium derived from regional supply-failure probabilities and sector-specific resiliency factors, and a regulatory trajectory premium reflecting anticipated future cost increases. The framework is operationalised through three case studies spanning three industries, three continents, and three distinct water stress profiles: TSMC’s semiconductor fabrication facility in Phoenix, Carlsberg’s brewery in Fredericia, Denmark, and BHP’s Escondida copper mine in the Atacama Desert, Chile.

The framework produces shadow prices ranging from $9.43 per cubic metre at Escondida to $11.45 per cubic metre at Fredericia, with shadow-to-tariff ratios as high as 23.6 at sites with binding scarcity constraints. Applied to the matched-pair Levelized Cost of Water for the resilience investment each facility has actually deployed, the framework surfaces a cost-effectiveness flip on the LCOW comparison for TSMC Arizona and BHP Escondida: at the utility tariff each system’s LCOW is unviable, but at the shadow price it is cost-effective. Net present value analysis at the sector weighted average cost of capital sharpens this result. TSMC’s NPV rises from a marginal $1.1 million at tariff to $429.4 million at shadow price, and BHP’s flips from -$4.76 billion at tariff to +$2.37 billion at shadow price. Carlsberg’s recycling investment is cost-effective at the existing Danish full-cost-recovery tariff, illustrating how mature regulatory pricing already approaches the shadow price floor in well-governed jurisdictions.

The thesis contributes a transparent, traceable analytical structure for evaluating industrial water investments under uncertainty, with documented input derivations and full sensitivity analysis. The contribution is the framework’s analytical structure and the cost-effectiveness flip it surfaces, not a claim to reproduce internal corporate accounting or to predict any specific firm’s investment decision.

Smålandshamnar AB, the operator of Oskarshamn Harbour, is planning the construction of new ferry and cargo terminals as part of its long-term strategy to support the Fossil-Free Sweden initiative, which aims to achieve net-zero domestic shipping emissions by 2045. A key component of this transition is the deployment of Onshore Power Supply (OPS) to berthed vessels, complemented by local renewable electricity generation from rooftop photovoltaic (PV) systems. However, large and intermittent OPS demand presents significant challenges for both the port operator and the local Distribution System Operator (DSO), Oskarshamn Energi, whose available grid capacity is limited. Consequently, expanding the electricity supply through conventional grid reinforcement may be both technically constrained and economically expensive.

This thesis investigates the techno-economic feasibility of integrating rooftop PV and Battery Energy Storage Systems (BESS) at the ferry and cargo terminals to reduce grid dependency, mitigate peak demand, and lower electricity costs under realistic Swedish SE3 market conditions. A two-level optimization framework is developed in which Mixed-Integer Linear Programming (MILP) determines the optimal daily dispatch strategy. At the same time, the Non-dominated Sorting Genetic Algorithm II (NSGA-II) identifies the optimal sizing of the PV and BESS systems based on technical and economic trade-offs. In addition, three battery technologies, lithium iron phosphate (LFP), sodium-ion batteries (SIB), and vanadium redox flow batteries (VRFB), are evaluated using a consistent modeling framework.

The optimization results show that the optimal rooftop PV capacities are 900.9 kWp for the ferry terminal and 2,503.8 kWp for the cargo terminal. Without battery storage, the PV systems achieve self-consumption rates of 48.0% and 84.7% for the ferry and cargo terminals, respectively, demonstrating substantial potential to reduce grid electricity consumption. Economic analysis further indicates that standalone PV systems are financially attractive, yielding positive Net Present Values (NPVs) under current market conditions. In contrast, none of the evaluated BESS technologies is economically viable. Although battery storage effectively reduces peak demand and enables energy arbitrage, the associated capital investment outweighs the achievable economic benefits under the prevailing SE3 electricity market, resulting in negative NPVs and undefined Internal Rates of Return (IRRs).

Overall, the results demonstrate that rooftop PV provides a cost-effective solution for supporting port electrification at Oskarshamn Harbour. In contrast, BESS deployment requires either significant reductions in investment costs or additional revenue streams, such as participation in ancillary service markets, to become economically feasible. The proposed optimization framework offers a practical decision-support tool for planning sustainable port electrification under constrained grid conditions.

This thesis examines downstream startup funding as a two-margin outcome: whether a venture receives external funding after a demo-day pitch, and how much it raises conditional on becoming funded. The study analyzes 217 public Y Combinator and Antler pitch transcripts, matched with downstream funding data and macro-financial indicators. Pitch transcripts are used to measure team, market, and technology signals, while clustering identifies transcript-based venture archetypes. The results show that market-related pitch language is associated with funding incidence, whereas funding amount is more strongly associated with macro-financial conditions and bundled YC–Antler regime context. The findings suggest that startup funding should be studied as a staged, context-dependent evaluative process.

Core-collapse supernova (CCSN) explosions mark the final stage in the lives of massive stars, leaving behind compact remnants while also launching powerful shock waves into their surroundings. Long before the final explosion, the massive progenitor stars continuously shed material through stellar winds and binary interactions, sculpting the circumstellar environment into which the supernova (SN) eventually explodes. The properties of this material therefore carries information about the progenitor’s mass-loss history prior to explosion. When the ejecta collide with this circumstellar material (CSM), forward and reverse shocks are formed, heating both the ejecta and the CSM to X-ray emitting temperatures.

Hidden at the centre of the expanding ejecta is the newly formed compact remnant. Some of these remnants are rapidly rotating, highly magnetised neutron stars known as pulsars. As the pulsar spins down, its rotational energy is converted into electromagnetic radiation and relativistic particles, powering an associated pulsar wind nebula (PWN). This emission extends into the X-ray band.

Some CCSNe are considerably more luminous than the rest. At the extreme end of the luminosity distribution lie the superluminous supernovae (SLSNe). Due to the higher energy explosions, the conventional powering mechanisms of CCSNe are not enough, and they are believed to require an additional energy source. Two favoured models are shock interaction with the CSM and a highly magnetised millisecond pulsar, often referred to as a magnetar. However, it remains uncertain whether either mechanism alone can account for the observed emission, whether both contribute, or whether additional powering mechanisms are involved.

Many open questions remain regarding how the properties and evolution of massive progenitor stars shape the resulting SN, their circumstellar environment, and their compact remnants. This thesis addresses these questions through systematic X-ray studies of large CCSN samples, using the X-ray emission to probe the compact-remnant properties and progenitor mass-loss history. In addition, it expands the very limited sample of X-ray detected SLSNe, providing new constraints on their powering mechanisms.

Late-time X-ray observations of nearby CCSNe provide constraints on the birth properties of compact remnants. The results favour pulsar populations with relatively long initial spin periods, while simultaneously demonstrating that late-time X-ray emission is predominantly produced by CSM interaction. We also find that the hard X-ray emission of SN 1979C is consistent with emission from a young pulsar, under the assumption of little-to no ejecta absorption.

We expand the very limited sample of X-ray detected SLSNe through a multi-epoch study of SN 2018bsz, one of the nearest hydrogen-poor SLSNe discovered to date. The observed X-ray evolution is difficult to reconcile with standard magnetar models, particularly when ejecta absorption is taken into account, whereas interaction with CSM naturally reproduces both the luminosity and temporal evolution. These results provide new evidence that CSM interaction may contribute significantly to the powering of at least a subset of hydrogen-poor SLSNe and highlight the diversity of physical mechanisms within this class.

Finally, through a comprehensive X-ray study of Type II SNe, we add to the X-ray detected SN sample, investigate early CSM absorption, and map the mass-loss history of the progenitor stars. We find that for some SNe the inferred mass loss evolves with time before explosion in a way that deviates from the expectations of a steady progenitor wind.

The overall increase in wind and solar generation makes electricity production increasingly dependent on weather forecasts, which are mainly produced by Numerical Weather Prediction (NWP) models. Errors in these forecasts feed through into power forecast errors, which in turn drive activity in the intraday market. This thesis investigates how NWP forecast errors transmit into intraday price behaviour in the SE3 bidding zone of the Nordic electricity market, and whether this information can support trading strategies for wind and solar producers. In contrast to existing literature, forecasts are not treated as a single point forecast from a single model: different NWP models are considered, and their regime-dependent behaviour and disagreement are examined to obtain additional signals.

Power forecasts for wind and solar are produced through a two-layer architecture built on four NWP models: the global models ECMWF IFS and ICON Global, and the regional models MET Nordic and ICON EU. In the first layer, a base model is trained for each; in the second layer, an ensemble is trained that receives base model predictions, the predicted weather regime, and disagreement features. Analysing the results shows that the base models make similar errors, and more importantly, 80 % of the time err in the same direction. This limits the potential gains for an ensemble and the explanatory power of individual models for the intraday market. However, the ensemble is still able to learn systematic and regime-dependent behaviour of the base models, as it improves by 9.1 % and 6.9 % over the best individual Mean Absolute Error (MAE), for solar and wind forecasting in the SE3 bidding zone, respectively. Inter-model disagreement is then tested as a predictor of the ensemble’s forecast uncertainty. For solar forecasts, the spread between models is moderately correlated with the absolute ensemble error (Pearson correlation 0.542) and increases monotonically across spread bins, whereas for wind, the relationship is weak (Pearson correlation 0.208). Disagreement is therefore a useful uncertainty signal for solar, but not for wind forecasting.

The DA/ID spread in SE3 is analysed to identify its drivers by combining a LightGBM model withSHAP attribution and an OLS regression with Hansen threshold testing. The spread is shown tobe only weakly predictable, due to the high noise in the relatively illiquid Nordic intraday marketand a limited dataset resulting from recent structural changes in the Nordic electricity markets.Several consistent and systematic drivers are nonetheless identified. The strongest is the totalcumulative error, the sum of all base-model errors for both wind and solar: when generationis underpredicted, the system faces a shortage and intraday prices rise above the day-aheadbaseline, and when it is overpredicted, a surplus pushes them below. Including all base-modelpredictions carries value, even if the additional variance explained is small. Furthermore, theeffect is asymmetric: negative errors (shortages) matter more than positive ones (surpluses), andwind errors have a far larger impact than solar errors. This mechanism is also visible at theweather-regime level: clusters in which the forecast models carry a systematic bias tend to showa correspondingly signed mean DA/ID spread.

The second-strongest signal is residual load: when it is high, supply is tight, and intraday prices tend to rise above the day-ahead baseline. The threshold regression confirms a structural break in how forecast errors drive the spread, with the impact of a negative forecast error roughly doubling above a residual load of 2326 MWh. Finally, cross-zonal variables also have a significant effect, enabled by the Cross-Border Intraday (XBID) project, which links the European intraday markets. While SE2 and SE4 are the largest interconnectors for SE3, their dominant north-to-south flow means they vary little and carry little explanatory power; the smaller borders with DK1, NO1, and FI, whose prices and intraday capacities vary far more, show the strongest and most consistent effects in both the SHAP analysis and the OLS regression. Overall, cheaper power in a neighbouring zone dampens SE3 intraday prices, provided enough capacity is available on the relevant interconnector.

Finally, the signals identified in the intraday analysis are translated into a qualitative discussion of how they can improve trading strategies for wind and solar producers. Two key decisions are considered: the Day-Ahead (DA) bid quantity and the timing of intraday position corrections. At the time of the DA bid, the following signals are available: the forecasted residual load level, the forecasted weather cluster, and the inter-model disagreement. A high residual load level, a cluster with a negative bias (overprediction), or high disagreement between models all point to bidding at a lower quantile of the probabilistic forecast, thereby reducing the risk of being short in an expensive regime. During intraday trading, the cumulative forecast updates become the primary signal, mainly driven by wind power. When a zonal update signals a system-level shortage, intraday prices are expected to rise, and a producer with a short position should seek to correct as early as possible, while one with a surplus can afford to wait and try to sell at a higher price. This highlights the importance of receiving frequent forecast updates so that one can act on them before the market reacts. Cross-zonal variables can also provide additional context in a trading strategy: when cheap power is available in a neighbouring zone and interconnector capacity is unconstrained, a shortage in SE3 is more easily absorbed, thus reducing the risk of waiting longer to correct a position. It is important to note that, due to the limited availability of out-of-sample data and the structurally illiquid intraday market in SE3, no quantitative improvement in trading revenue is claimed.

Building occupancy is of great importance for the operation of building HVAC systems and is directly related to energy efficiency, thermal comfort, and indoor air quality. Commonly, occupancy is modelled with a static schedule, which does not capture the highly dynamic, varied behaviour of occupants. To improve occupancy modelling methods, machine-learning-based occupancy models are tested in this thesis and integrated directly into IDA ICE via Python within a co-simulation framework.

This study used occupancy and indoor temperature data from four residential apartments in the KTH Live-In Lab; the findings should be read within that building type and occupancy regime. The models trained on this data were Long Short-Term Memory and Random Forest. The models were trained to predict short-term (1h), day-ahead (24h), and week-ahead (168h) occupancy schedules and were evaluated using metrics such as AUC, F1-score, accuracy, and transition accuracy. The predicted schedules were then implemented in IDA ICE via co-simulation and compared with a static schedule that assumed unoccupied periods occurred on weekdays from 9 am to 5 pm.

The results showed that model performance decreased as the prediction horizon increased, with the short-term models (1h) performing best overall. Random Forest and Long Short-Term Memory showed similar overall prediction performance. However, Random Forest was selected for the final co-simulation because it provided representative predicted schedules while reducing software complexity in the Python–IDA ICE workflow. The co-simulation results showed that ML-predicted schedules did not overall reduce the apartments' energy consumption compared with the static schedule. The effect was highly apartment-dependent. However, in terms of load behaviour, the ML-predicted schedules were more accurate at predicting peak loads and showed some load shaving. Furthermore, thermal comfort was maintained across all apartments, while indoor air quality results were mixed and apartment-dependent.

The main contribution of this thesis is a reproducible Python–IDA ICE co-simulation workflow for evaluating ML-based occupancy schedules in building energy simulation. The workflow connects offline prediction metrics, such as AUC, F1-score, accuracy, and transition accuracy, to simulated building performance outcomes, including heating energy use, load behaviour, thermal comfort, and indoor air quality. The results show that good offline prediction performance does not necessarily lead to improved building performance. Therefore, the thesis demonstrates the need to evaluate ML-based occupancy schedules within a building energy simulation tool, rather than relying only on offline metrics.

This study aims to analyze and evaluate the choice of cement type and its impact on the climate in different building scenarios for a four-storey building, the M&K building. Through life cycle assessments of a reference project, six different cement alternatives have been compared, including the construction and operational phases and the effect of carbonation in the calculations. Both manual calculations and the Configo calculation software have been used to determine the climate impact, gain an overview, and compare the results.The results showed that cement with a reduced clinker content contributes to significantly lower carbon dioxide emissions than traditional cement, and that scenarios with CCS technology have the lowest emissions. It was also found that the construction phase accounts for the largest share of emissions, while carbon dioxide uptake through carbonation had a smaller effect and cannot compensate for the emissions from cement production. Factors such as waterproofing layers and exposure affect the degree of carbon dioxide uptake, but the choice of cement type is crucial.The work highlights how certain limitations, such as assumptions regarding material choices, construction practices, and future energy use, influence the results. In summary, the study shows that the choice of cement type and material optimization during the design phase are important decisions for achieving the climate goals of the construction sector.

This study examines knowledge sharing in boundary processes between a Community of Practice (CoP) and other employees within a Swedish technology company. The research was conducted as a qualitative within-site case study using an exploratory sequential mixed methods approach consisting of semi-structured interviews and questionnaires. The purpose of the study was to identify strengths and challenges related to knowledge sharing in boundary processes between a community of practice and employees outside the community, and the overall goal was to suggest improvements to enhance knowledge sharing. 

The findings show that the CoP enables strong internal knowledge sharing through shared practices, informal interactions, and internal communication channels. At the same time, challenges arise when knowledge needs to cross CoP boundaries. Three central dimensions were identified: transparency, negotiability, and coordination. A lack of structured documentation frameworks, unclear shared processes, and ambiguous expectations played an important role in connecting professional groups and support communication between them but also created dependencies on specific individuals. Physical separation and limited opportunities for face-to-face interactions further strengthened an “us and them” dynamic between the groups. 

The study concludes that effective boundary processes require organizations to balance the strengths of CoPs with organizational structures that support communication, coordination, and mutual understanding across boundaries. Suggested improvements include standardized documentation frameworks, earlier involvement from different groups in projects, clearer expectations, and increased opportunities for both formal and informal (social) interaction between groups. 

AI coding assistants promise faster software development, but they also move a critical part of quality assurance into a less visible place: the moment when developers decide whether generated code is trustworthy enough to use. This thesis examines how professional software developers verify AI-generated code in real work situations, and how developer experience and organizational conditions shape those decisions.

The study is based on qualitative interviews with twelve professional software developers in Sweden who use chat-based or agentic AI coding assistants,including Cursor, Claude Code, Codex, Kiro, and Claude. The interviews were structured using the Critical Incident Technique, where each participant reconstructed one recent episode in which they had to decide whether and how to verify AI-generated code. The material was analyzed using the Gioia methodology, moving from first-order codes grounded in participant accounts to second-order themes and aggregate dimensions.

The findings show that developers calibrate trust in AI-generated code through situated verification and reliance decisions. Trust was not mainly expressed as a general attitude toward an AI tool. It became visible in what developers did:which cues they noticed, which checks they performed, when they stopped checking, and whether they accepted, modified, refined, stepped back from,or abandoned the AI output. Output cues such as large or irrelevant changes, functional success, and repeated non-convergence shaped the first judgment. Perceived task risk changed how much confidence was needed, while time pressure, fatigue, long AI outputs, slow feedback loops, and data sensitivity limited what developers could realistically verify. Developer experience shaped whether participants could challenge the AI output independently, especially when tasks involved architecture, security, or system-level reasoning.

The thesis contributes a grounded model of trust-calibrated verification in AI-assisted software development. It argues that verification is not only an individual developer skill, but also an organizational quality mechanism shaped by work design, staffing, infrastructure, and review practices. For organizations adopting AI coding assistants, the central challenge is therefore not only to generate code faster, but to design development work so that appropriate verification remains possible.

The pursuit of sustainable bio-based materials has driven interest in using nanocellulose as a building block for high-performance materials. To translate the advantageous nanoscale properties of nanocellulose into macroscopic materials, it is important to control the nanostructure. Controlling nanoscale alignment creates anisotropy, providing structural and functional advantages. While several reviews cover alignment of nanocellulosic materials, research on fabrication routes that successfully align cellulose nanofibrils for industrially scalable concepts remains limited. This study aims to investigate and design fabrication route(s) that yield high in-plane alignment of cellulose nanofibrils in two-dimensional formats. In the established fabrication route, cellulose nanofibrils were aligned via shear-induced flow by subjecting cellulose nanofibril dispersions to a cylindrical vial rotating about its own axis, producing films along the vial walls. To address flowability challenges in such experimental set-up, arising from highly entangled networks of the elongated fibrils, surface modification was introduced to impart steric stabilisation. Specifically, TEMPO-oxidised cellulose nanofibrils were ionically grafted with polyetheramines, JEFFAMINE® M-2070 and M-600. Furthermore, for the optimisation of the fabrication route, two types of dispersion media were evaluated: water and acetone, as well as different weight percentages of cellulose nanofibrils. Characterisation measurements of the fabricated films included wide-angle X-ray diffraction, angle-polarized Fourier-transform infrared spectroscopy, and optical birefringence. The degree of orientation was evaluated using Herman’s orientation parameter (S). The use of steric stabilisation was shown to be advantageous for shear-induced alignment, increasing the inter-fibrillar distance and lowering viscosity. Among the dispersion systems evaluated, Jeffamine M-600, water, and cellulose nanofibrils content of 0.2 wt% were the most favourable, achieving an S-value of 0.32. The established fabrication route is a promising alternative for alignment of cellulose nanofibrils and is suitable for large-scale production; with further optimisation, higher degrees of orientation may be achievable.

Shock waves provide an effective means of dynamically compressing fluids and generating intense, localized thermodynamic and mechanical states involving high temperatures and pressures. Such fluid states are relevant to applications including inertial confinement fusion, shock-wave lithotripsy, targeted drug delivery, materials research, and studies of fuel ignition and detonation. This thesis investigates two related flow scenarios in which shock dynamics and shock-induced interactions govern strong compression, energy concentration, and the resulting flow evolution in fluids.

The first part of the thesis examines shock convergence in gases, where a curved shock strengthens as it propagates toward its focus, producing high temperatures upon focusing and subsequent reflection. The experiments are performed in a shock tube, which enables controlled generation and study of the converging shocks. High-speed shadowgraphy, optical emission spectroscopy, and photometry are used to measure the shock trajectory and focal temperature, while numerical simulations employing ideal- and real-gas equations of state complement the experiments. The results show that the convergence process is influenced by the initial shock strength, upstream nonuniformities, and real-gas effects, which together govern the focal temperatures that can realistically be achieved. 

A generalized scaling for the focal temperature is developed, incorporating the effects of shock symmetry, convergence ratio, and initial gas pressure. Demonstrated for argon, the scaling enables prediction of the focal temperature from accessible initial parameters and provides a basis for inverse design, allowing the conditions required to reach a target temperature to be estimated. Subsequent to focusing, the shock reflects and propagates upstream, where it interacts with the incoming boundary layer. These interactions are found to depend strongly on the specific heat ratio of the gas and can lead to bifurcation of the reflected shock. Such non-ideal flow effects alter the thermodynamic state behind the reflected shock, which is important to characterize in studies of chemical kinetics and detonation employing shock-focusing geometries.

The second part of the thesis investigates shock--bubble interactions in a water--gelatine solution used as a water surrogate. These studies address how shocks interacting with light inhomogeneities in liquids can generate localized pressure peaks, high-speed jets, and coherent vortical structures. Experiments in the water--air limit remain relatively scarce because generating strong shocks in water is challenging, the interactions involve gigapascal-scale pressures, and strong density gradients limit the use of conventional optical visualization techniques. Planar Mach 1.4 and Mach 1.5 blast waves are generated by exploding copper foils using a pulsed-power driver, and their interaction with cylindrical air bubbles is visualized using X-ray phase-contrast imaging at a synchrotron. The measurements resolve the interaction from incident-wave impact through jet formation, water-hammer shock emission, remnant-lobe collapse, and late-time vortex-pair formation. Compared with the corresponding idealized step-like shock interaction, the finite-width blast-wave driver, combined with the cylindrical geometry of the bubble, produces thinner and slower jets, weaker water-hammer shocks, larger remnant lobes, and a sequence of events that is moreclearly separated in space and time. This separation allows the mechanism responsible for a second upstream pressure peak and the subsequent multi-stage vorticity evolution to be identified.

Finally, the effect of shock curvature is investigated through the interaction of cylindrical and semi-cylindrical converging shocks with a coaxial cylindrical air bubble. In the cylindrical configuration, the bubble undergoes inward collapse accompanied by Richtmyer--Meshkov instability spikes, while the transmitted shock focuses within the bubble and reaches high Mach numbers, leading to elevated temperatures in the bubble gas. In the semi-cylindrical configuration, the asymmetric interaction concentrates momentum into a single supersonic jet and gives rise to localized high-pressure regions, internal Mach reflection, and coherent vortical structures.

Overall, the thesis shows how shock waves can generate intense thermodynamic and mechanical states through geometric convergence and interactions with gas--liquid interfaces. It further advances understanding of how the interactions and flow features that develop during these phenomena influence the resulting pressure and temperature states and the subsequent evolution of the flow. This understanding is important for interpreting these phenomena, both to better harness shock-driven compression and energy concentration and to anticipate and limit their potentially damaging effects in natural and engineered systems.

The KTH seminar on 28 August 2026 connected two sides of the same development: artificial intelligence can strengthen water intelligence, while AI and digital infrastructure themselves create new water demand and local water risk. Across the presentations and roundtable discussion, a central conclusion was that Water–AI becomes useful only when physical process understanding, monitoring, trustworthy data, institutional responsibility and implementation mechanisms are connected.Prosun Bhattacharya established the scientific foundation by framing groundwater as hidden infrastructure and stressing that more data do not automatically produce better understanding. He argued that AI should strengthen hydrogeological interpretation and conceptualisation from local to regional and, where appropriate, potentially global scales. Experiences from Bangladesh illustrated a progression from field-based safe-aquifer targeting and local knowledge towards digital groundwater intelligence and real-time monitoring. Joydeep Dash complemented this perspective by reversing the lens from “AI for water” to “water for AI”, highlighting the rapid expansion of data centres in water-stressed Indian cities and the need to consider local water risk, transparency, circular water use, investment and opportunities for Sweden–India collaboration. Sanjeev Sharma presented a physics-informed neural-network approach to anthropogenic groundwater stress in Bangladesh, showing how groundwater-level observations can be combined with aquifer mass-balance constraints while also making clear that the three analysed sites were dominated by agriculture-related water use and that industrial attribution remains to be tested in dedicated industrial settings. Per Kinnander presented WATERFORCE as a knowledge-to-action pathway linking research, public authorities, utilities, industry and communities through shared evidence, scenario exploration, coordinated action and continuous learning. Md Zarif Oeishik presented a scalable screening concept for rooftop rainwater harvesting in the textile and ready-made garment sector, aimed at identifying where rainwater harvesting could meaningfully reduce groundwater dependence before more detailed feasibility studies are undertaken.

The panel and roundtable discussion connected these contributions to monitoring-system design, attribution of groundwater-level changes, the need for lithological information, data sharing and privacy, industrial and mining applications, community legitimacy, assurance and the development of reusable and scalable Water–AI demonstrators rather than isolated pilot projects. AI needs water, and water increasingly needs intelligence. The objective is not AI for its own sake, but process-informed, locally contextualised and accountable water intelligence that shortens the path from observation and understanding to well-founded decisions and responsible action.

Choir sound emerges from individual singers, yet each singer also hears and responds to the collective sound of the ensemble from within it. Studying this relationship under controlled conditions is challenging in live choir experiments, where the singer’s own voice, the surrounding voices, the room response, and the ongoing adaptations of performers are all present in the sound field. Consequently, the singer’s own voice (Self) and the surrounding ensemble (Other) are difficult to isolate and control independently under repeatable conditions. This thesis aimed to establish a controlled experimental framework for investigating choir acoustics at the level of the individual singer and to examine how such a framework can be used to study singers’ perception, Self–Other balance, and vocal adaptation.

Across the included studies, controlled interaction with reproduced Other, spatial rendering, and the representation of Self informed an integrated virtual choir framework that kept the singer actively performing. The framework was applied in a within-singer choir-spacing experiment with 39 experienced choir singers performing one at a time with spatialised virtual-choir playback under three spacing conditions. Perceptual responses were analysed alongside separated, calibrated, system-referenced measures of Self level, Other level, and Self-to-Other Ratio (SOR).

Wider spacing increased perceived audibility of Self and was associated with greater perceived ease of tuning, whereas the audibility of different parts of Other showed no uniform improvement. Most singers favoured wider spacing for concert performance. Other changed substantially more than Self, so the Self–Other balance shifted across conditions rather than being restored. The Self–Other relationship became more apparent at phrase level, and perceived self-hearing was more clearly associated with SOR than with Self level alone. These findings demonstrate the feasibility of investigating singer–ensemble interaction with controlled virtual acoustics during active singing.

The construction and civil engineering industry is facing increasing demands to reduce greenhouse gas emissions and transition towards fossil-free operations. Within the industry, heavy machinery accounts for a significant share of fuel consumption and climate impact, making it a key component of this transition. Although technologicaldevelopment is advancing, implementation is slower than the available technical potential, indicating that the challenges are not only technical but also economic and organizational. The purpose of this study is to analyze how decision-making processesfor heavy machinery in the industry can be developed to enable more accurate comparisons between conventional diesel-powered machines and fossil-free alternatives, ensuring that the decision-making process does not hinder an evidencebasedtransition. The study was conducted as a case study at Skanska.

The results show that the decision-making process is characterized by increasing complexity, making it more difficult to evaluate different investment options. Economic and climate-related aspects are currently handled by different units and individuals within the organization, with some exceptions among managers with a profit and loss responsibility. This separation makes it difficult to develop a holistic understanding of how economic and environmental key performance indicators interact and which investment is most beneficial in terms of lifespan, efficiency, and operational costs. The study also identifies organizational challenges related to knowledge sharing, leadership,and implementation, which further affect the conditions for well-informed decisionmaking. Finally, the study highlights a need to better integrate sustainability and economic perspectives into shared performance indicators and into clearer and more accessible decision-making tools. This is important for creating a shared understanding and for engaging employees, managers, and other organizational functions in the transition process.

The European Union is transitioning towards a fossil-free electricity system. Growing shares of variable, non-dispatchable renewable sources reduce supply-side flexibility, increasing the need for balancing resources. Energy-intensive industrial consumers can contribute to the stability of the system through demand response programs. Power-to-heat technologies are particularly promising due to fast-regulating equipment, while thermal storage and complementary heat-producing units provide a stable heat supply to the industrial processes. However, industrial power-to-heat flexibility in balancing markets is still underexplored, particularly in the Swedish mFRR market.

The thesis develops a two-stage MILP model and applies it to an industrial steam generation system, represented by Fiskeby Board AB. The objective is to assess the economic potential, the operational scheduling and the technical feasibility for the mill’s participation in the Swedish day-ahead and mFRR markets. Stage 1 optimizes the hourly baseline operation and determines upward and downward commitments to the mFRR capacity market under three operational scenarios. The resulting baseline is used as input for Stage 2, which defines the redispatch of the system at a 15-minute resolution to assess delivery feasibility under an energy activation signal.

The results show potential for both upward and downward activations, with downward commitment being more prevalent. The highest committed capacities are 11 MW for upward and 12 MW for downward. The Hybrid scenario has the highest economic potential, with a reduction of the net operating cost from 5.11 MEUR to 4.51 MEUR, while the specific cost of steam drops from 23.98 EUR/MWh to 21.18 EUR/MWh. Stage 2 is tested only for the Hybrid scenario and shows no unmet activation. However, prolonged downward activations show an expected increase in steam venting, indicating that Stage 1 partially overestimates downward deliverability.

Overall, the results show that Fiskeby’s steam generation system can potentially participate in the Swedish mFRR market, although the analyzed economic results represent a potential under the current assumptions, rather than guaranteed profitability.

The integration of Exhaust Gas Recirculation (EGR) technology into Combined Cycle Power Plants (CCPP) represents one of the most promising near-term pathways toward decarbonized gas-fired power generation. By recirculating approximately 30 to 40% of the exhaust gas back to the gas turbine inlet, EGR enriches the CO₂ content of the exhaust stream and enables carbon capture at substantially higher efficiencies than post-combustion processing of dilute flue gas. However, it increases the risk of continuous ingestion of Calcium-Magnesium-Alumino-silicate (CMAS) particulates into the gas turbine hot gas path.

CMAS particles, fine silicate-based compounds in the 1 to 10 μm aerodynamic diameter range, melt upon entering the combustion chamber at temperatures exceeding 1,240°C and deposit as a viscous glassy phase onto Thermal Barrier Coatings (TBC) of first-stage turbine nozzle guide vanes and blades. The resulting dissolution of the yttria-stabilized zirconia (YSZ) TBC matrix, pore infiltration, and crystallization-induced stress leads to TBC spallation, a failure mode that dramatically reduces component service life and threatens plant availability.This thesis establishes a comprehensive CMAS management framework for EGR-CCPP systems, that we will call EGR 1. The work identifies the primary CMAS source within the EGR circuit as the Alkaline Earth Silicate (AES) insulation lining the Heat Recovery Steam Generator (HRSG) casing. Through thermal expansion analysis based on the EGR 1 Apros thermal model, fiber release rates of 34 to 134 μg/m²/cycle are estimated across the HRSG thermal gradient, constituting a continuous internal CMAS precursor source during every start-up and shut down cycle.

Cross-sector expertise from the aviation industry, which has confronted analogous CMAS degradation from volcanic ash ingestion for decades, is systematically reviewed and applied to validate monitoring methodology. A sensor architecture based on the Dekati Electrical Low-Pressure Impactor (ELPI⁺) coupled with stage-specific filter analysis and cross-correlated with online particulate matter (PM) observation is proposed as the most stable and scientifically defensible measurement approach for the EGR flue gas stream, given the unique combination of high temperature, humidity, and compositional complexity of that environment.

A four-tier CMAS threshold classification framework, Normal, Elevated, High, and Critical, is derived from materials science literature, anchored to the Wellman-Whitman damage initiation datum of 1.8 mg/cm² and calibrated against measured PM emission benchmarks from industrial gas turbines firing natural gas (PM₁₀ = 0.03 mg/Nm³; Vierling et al., 2011). A defense-in-depth mitigation strategy spanning Prevention (insulation substitution, inlet filtration), Resistance (CMAS-tolerant TBC materials), and Removal and Monitoring (WESP, DCC, integrated sensor system) is developed and assessed.

A large number of Swedish bridges are located in the privately owned road network, and justlike the project bridge in this study, lacks documentation, which in the future will need to behandled through testing and investigation.The goal of this paper has been to verify the load carrying capacity of an older steel girderbridge that lacks documentation by comparing theoretical calculations with practicalmeasurements. By using strain gauges, data has been collected with the bridge loaded with avehicle of known weight. The data has then been compared with the theoretical calculationsthat the Swedish Transport Agency recommends, to obtain a more actual understanding of theloads that effect the bridge.In addition to the strain gauge measurements, a more practically applied method of measuringdeflection has been evaluated. The strength of the steel girders has been examined throughdestructive testing, and how the values from that testing can be implemented has beendiscussed.The paper primarily focuses on steel structures and bridge superstructures, but it also discussesthe parts on the project bridge made of wood and their load carrying capacity. The Swedishtraffic regulations' load rating classes and how they can be applied to smaller privately ownedbridges are also discussed.The bridge under investigation is over 120 years old and the thesis describes the assumptions

made regarding cross-sections and the strength of older rolled steel girders when finding up-to-date documentation is a difficulty.

The results of the strain measurements show that the measured load effect is approximately20% lower than the load effect calculated in accordance to the Swedish Transport Agency'sguidelines. Destructive testing also indicates that the yield strength of the steel is slightly higherthan the value assumed according to the same guidelines.

Human movement provides rich information about mobility, physical function and overall well-being. Accurate measurement and analysis of movements are therefore important for identifying impairments, guiding rehabilitation, optimizing sports performance, and informing assistive device design. In recent years, wearable sensing technologies such as inertial measurement units (IMUs) and pressure insoles have gained increasing attention as alternatives to conventional laboratory-based motion analysis systems. IMUs can estimate joint kinematics, while pressure insoles enable measurement of plantar loading and ground reaction forces. Compared with optical motion capture and force plate systems, wearable sensors offer greater portability and accessibility, with potential for use in clinical and at-home environment. Despite these advantages, estimation of internal biomechanical parameters, including joint torques and muscle activations, remains challenging because it requires integration of multiple sensor modalities with biomechanical models. To address this gap, an integrated biomechanical analysis system was developed by combining multi-modal wearable sensing with musculoskeletal simulation for real-time estimation of joint kinematics and kinetics. By providing biomechanical estimates during movement without extensive offline processing, the system has the potential to support immediate assessment and feedback in clinical, rehabilitation, and real-world settings. This thesis comprises two studies investigating the validity and application of the developed wearable system. In both studies, the wearable system was evaluated concurrently with a conventional laboratory-based reference system during a range of common activities, including static standing, physical performance tasks, and walking under different conditions.

In the first study, nine non-disabled participants were assessed to validate the wearable system against a laboratory-based reference. Joint angles, joint torques, and muscle activations were evaluated across of a range of tasks. The wearable system demonstrated good overall agreement for sagittal plane joint angles and ankle torques. Larger discrepancies were observed for knee and hip torque during walking, likely due to the lack of shear force measurements from the pressure insoles. Muscle activations estimated using static optimization showed overall consistent temporal patterns with electromyography across different movement conditions. A preliminary drift assessment further indicated that the system could preserve measurement consistency for at least five minutes of continuous movements.

In the second study, six individuals with transtibial amputation were assessed to further evaluate the validity of the wearable system for a clinical population. Wearable-derived joint angles, joint torques, and vertical ground reaction forces were compared with laboratory-based measurements across functional activities. The wearable system showed good agreement for joint angles and vertical forces, whereas accurate joint torque estimation was largely limited to the ankle. The validated variables were subsequently used to quantify inter-limb asymmetry, revealing pronounced differences in ankle range of motion and limb loading, particularly in the peak and impulse of vertical ground reaction force and ankle torque. Several asymmetry measures, especially those related to limb loading, were also associated with clinical mobility outcomes, such as Timed Up and Go (TUG) completion time.

These findings support the validity and potential clinical applicability of the integrated wearable system for real-time biomechanical assessment. Its ability to provide immediate estimates of joint kinematics, external loading and selected internal biomechanical variables may facilitate repeated assessment during rehabilitation and prosthetic evaluation. At the same time, the present hardware configuration constrains the accuracy of some derived variables such as proximal joint torques. An important contribution of this thesis is therefore the identification of biomechanical parameters that can be estimated with sufficient accuracy, together with the functional contexts in which these parameters are most informative. Future work will build on these validated parameters to investigate clinical and real-world applications, while alternative data-driven approaches maybe explored to improve estimation of biomechanical variables that remain difficult to estimate using the current model-based framework. 

Soil–steel composite bridges are commonly used for short-span railway applications due to their efficient construction and low material use. However, their dynamic behavior under high-speed train loading is still less established. This thesis investigates the dynamic response of the Åre soil–steel composite culvert in Sweden using field measurements and three-dimensional finite element analysis. Hammer impact testing and train-passage measurements were performed on both the culvert and the adjacent track section. The processed field data were used to calibrate finite element models in Brigade Plus, including a benchmark track–soil model and a culvert–soil model with a simplified equivalent orthotropic shell representation of the corrugated culvert. The calibrated models were validated against measured responses from an X62 Regina train passage and then used to simulate Eurocode HSLM-A high-speed train loading between 150 and 350 km.hThe monitored train passages showed that the ballast response above the culvert and in the adjacent benchmark track section approximately 20 m away were of similar magnitude, while the acceleration measured directly on the steel shell was lower. The validation showed that the models reproduced the general response patterns and dominant dynamic characteristics. However, acceleration and displacement amplitudes were underestimated by approximately a factor of two to three, indicating that the numerical models were stiffer than the real structure. The high-speed simulations showed that the adjacent track model produced significantly higher acceleration and displacement responses than the culvert model throughout the investigated speed range, likely due to uncertainties in the calibrated backfill soil properties. This was a major limitation of the comparison between the track model and culvert model when assessing acceleration limits. Therefore, the lower acceleration response predicted in the culvert model should not be interpreted as an exact representation of the real system. Instead, it should be treated as an indication that the culvert response is likely not higher than the adjacent track response, while a more accurate conclusion would require improved soil data, embankment geometry, and closer FRF agreement during calibration. No resonance response was identified from the acceleration–speed curves or the corresponding time-history results. Even when the simulated culvert displacement due to HSLM-A loading was conservatively doubled to account for the validation discrepancy, the maximum response remained below the 2 mm Trafikverket serviceability limit. Overall, the results indicate that the Åre culvert is expected to perform satisfactorily under the investigated high-speed train loading, with no clear indication of resonance or excessive displacement. 

The European transport sector faces critical sustainability challenges, with road freight accounting for 72% of freight-related GHG emissions. Rail freight offers a sustainable alternative with significantly lower carbon intensity, yet its modal share remains limited due to suboptimal terminal utilization and inefficient routing methods in large-scale multinational networks. This research addresses this by optimizing rail freight utilization in the Scandinavian - Mediterranean (Scan-Med) Corridor using advanced shortest path algorithms. More specifically, I implemented and evaluated a novel one-to-many shortest path algorithms— Standard kA*, and Simplified kA* —specifically designed for multi-destination routing scenarios (i.e., identifying optimal connections from local stations to multiple freight terminals) based on the traditional A* algorithm. The Simplified kA* algorithm demonstrates exceptional efficiency, reducing node expansions by 99.96% and runtime by 98.2% compared to traditional methods, while maintaining solution optimality. Applied to integrated European infrastructure and freight flow cluster datasets, my approach reveals critical patterns in terminal utilization and network flow patterns. The analysis of 2005–2020 data identifies the megahub Lehrte as the dominant hub (535% volume growth) and distinct terminal specializations—such as Hamburg Süderelbbrücke as a multimodal hub and Verona Quadrante Europa as an import-heavy node when the shortest path algorithm and the rail freight corridor are fully utilized within a hub-and-spoke network structure. The results provide insights to improve corridor utilization, reduce operational costs, and advance sustainable freight transportation. This work establishes a computational foundation for future large-scale rail freight network optimization and decision-support tools.

With climate change increasing the frequency and severity of extreme weather events, it becomes increasingly important to quickly assess the damages and respond appropriately. Remote sensing methods like satellite and drone observations allow us to survey these large-scale events with observations at matching scales. This thesis addresses mismatches between such applications in disaster response and standard computer vision practice: which data is used or left unused, how generalization is evaluated, and how predictive uncertainty is modeled. For building damage detection, we modify the commonly-used dataset split on the xBD dataset to measure generalization to completely unseen disaster events, instead of just unseen images within known events. We further condition the model predictions on the type of natural hazard involved, information that is always readily available in the application context and helps the model to generalize to unseen events. For wildfire spread prediction, we first introduce a new dataset which provides access to a fire's recent history instead of only the current day's observation. Finally, motivated by the limited performance of wildfire spread prediction throughout the literature, we rethink how to model the prediction targets. Instead of predicting only one segmentation mask or independent pixel-wise probabilities, we produce a set of distinct outputs that aim to capture the underlying uncertainty about possible outcomes.

The electrification of heavy-duty freight transport is accelerating under the European Union's Alternative Fuels Infrastructure Regulation (AFIR), creating growing pressure to deploy charging infrastructure along the Trans-European Transport (TEN-T) Network. While AFIR defines regulatory spacing and capacity targets, it does not specify how charging infrastructure should be prioritised spatially under operational, structural, and environmental constraints. Sweden faces a particularly challenging planning context because of its long transport corridors, sparse northern regions, uneven freight distribution, and limited publicly available infrastructure data. This thesis evaluates Sweden's heavy-duty charging readiness against the AFIR 2027 milestone and develops a planning framework for identifying candidate charging locations for AFIR 2030. A reproducible geospatial-analysis pipeline was developed by integrating Samgods freight-chain data, NOBIL charging-station records, National Land Cover Database 2023 data, and Svenska Kraftnät transmission substations. The framework combines four analytical components: network-distance regulatory auditing, network-topology and freight-flow analysis, operational state-of-charge corridor simulation, and Pareto-based candidate evaluation for new charging sites.The results show that Sweden already exceeds the AFIR 2027 minimum spacing requirement on the analysed network. However, the transition toward AFIR 2030 reveals geographically distinct bottlenecks. Northern Sweden is characterised primarily by sparse-network coverage gaps, operational risks, and limited grid accessibility, whereas southern Sweden is dominated by high-demand capacity-upgrade requirements at existing charging stations. The analysis further demonstrates that structurally critical freight corridors do not necessarily coincide with the highest-flow corridors in Sweden.36 corridor-level spacing gaps were identified for candidate Pareto-front charging sites selection. 509 candidates generated and evaluated using land-cover friction and least-cost grid-accessibility objectives. The resulting Pareto-front framework provides a transparent and weight-free candidate set for planners while preserving flexibility for downstream stakeholder prioritisation. The thesis demonstrates that future heavy-duty charging deployment under AFIR requires geographically differentiated planning strategies rather than a uniform national rollout approach.

This study investigates how increasing use of digital technologies influences the workforce development within organizations in pharmaceutical manufacturing. To investigate this, both individual (micro) and organizational (macro) dynamics are taken into consideration and then analyzed with the help of the two theoretical frameworks, Unified Theory of Acceptance and Use of Technology (UTAUT) and the Dynamic Capabilities framework.

The study uses a qualitative case study approach with semi-structured interviews conducted in a pharmaceutical manufacturing company. The findings show that digital technologies have reshaped daily work by reducing manual tasks, increasing efficiency, and improving access to information. Technology adoption is influenced by perceived value, social influence, and organizational support, while organizational success depends on the ability to sense opportunities, seize them, and reconfigure resources.

The findings also show an increased demand for competencies related to digital technology. These competencies can be divided into hard skills and soft skills and the findings indicate that data literacy and openness to change are the two competencies that are most necessary today compared to previous years in relation to digital transformation.

In conclusion, the study demonstrates that digital transformation can be successful in pharmaceutical manufacturing by relying on both technological deployment and continuous organizational adaptation. The perception of digital tools is strongly impacted by organizational support, learning initiatives and social factors. Ultimately,long-term digital transformation can be sustained through continuous development of competencies and organizational structures.

This thesis examines how carbon cost exposure under EU ETS Phase 4 affects capital expenditure (CAPEX) among manufacturing firms on the Carbon Leakage List (CLL), distinguishing between net-buyers of allowances and firms whose free allocation covers their verified emissions. Using installation-level compliance data from the European Union Transaction Log (EUTL) matched to firm-level financial data from Orbis, a difference-in-differences design is applied to a balanced panel of EU27 CLL manufacturing firms over the 2017–2024 period. An exposure score measuring the share of verified emissions not covered by free allocation is interacted with annual average EUA prices to identify the investment effect of rising compliance costs. No significant aggregate effect was found, suggesting that benchmark tightening alone does not alter physical investment behavior. However, isolating firms that must purchase allowances reveals a significant negative effect at baseline and one-year lag, although sectoral analysis indicates this is concentrated in Mineral Products rather than reflecting a broad cross-sectoral pattern. To illustrate the magnitude, a firm moving from the 25th to the 75th percentile of the net-buyer exposure distribution is associated with a CAPEX ratio approximately 0.83 percentage points lower at the 2022 average EUA price. These findings suggest that the investment effects of carbon pricing depend on whether compliance costs translate into actual liquidity obligations.