The era of extensive digitalization marked by the fourth industrial revolution has ushered in significant advancements in technologies like automation, artificial intelligence (AI), and the Internet of Things (IoT). These innovations are revolutionizing smart industries like manufacturing, smart energy systems (SESs), and the automotive industry. Industry 4.0 (I4.0) and the subsequent Industry 5.0 (I5.0) emerged as comprehensive representations of the physical world in the information world, with goals to establish smart factories and promote human-machine coexistence. However, the implementation of I4.0 and I5.0 applications faces challenges related to engineering efficiency, interoperability, and efficient service discovery and binding.

This thesis seeks to address these challenges by exploring potential strategies to develop an efficient System of Systems (SoS) that comprises individual, autonomous systems collaborating to achieve a shared goal. This research examines methods to enhance the efficacy of SoS by refining its engineering procedures, promoting interoperability between standardized protocols and heterogeneous systems, and employing dynamic adaptation mechanisms. It aims to achieve automatic service discovery and interoperability between diverse industrial standards and systems across the different domains within the smart industry by integrating the Eclipse Arrowhead Framework. This IoT framework facilitates secure and seamless communication and collaboration among devices, machines, and systems.

Moreover, this work delves into saving energy consumption in distributed SoS environments. This is achieved through the Demand Response (DR) mechanism in SESs combined with the Eclipse Arrowhead framework. In addition, the thesis examines challenges in automotive testing, specifically in Vehicle-in-the-Loop (VIL) testing environments, which are distributed SoS systems requiring efficient communication, diverse hardware and system interoperability, realistic simulation, and scalable system integration with minimal cost and resource demand. The research also explores flexible methods for integrating heterogeneous environment models into VIL frameworks and proposes a standardized service-oriented vehicle data communication framework to improve interoperability, operational efficiency, and scalability.

The overarching objective is to pave the way for flexible production processes characterized by minimal resource waste, optimized energy consumption, and sustainable solutions. Through this endeavor, the thesis contributes to shaping a more efficient, interoperable, and sustainable smart industrial landscape in the context of Industry 4.0 and beyond.

Per- and polyfluoroalkyl substances (PFAS) are persistent environmental contaminants whose transport, retention, and removal are governed by complex interactions with natural and engineered surfaces. Despite extensive research, key uncertainties remain regarding how PFAS molecular structure, sorbent chemistry, and environmental conditions jointly control adsorption behavior. This thesis addresses these challenges by systematically investigating PFAS interactions across a range of systems, from soil components to engineered sorbents and dynamic treatment processes. A stepwise experimental approach was applied, beginning with fundamental interactions with soil organic matter and iron (hydr)oxide, followed by competitive adsorption on granular activated carbon and ion exchange resin, and culminating in evaluation of PFAS removal under flow conditions.

The results demonstrate that PFAS sorption in soils is governed primarily by the chemical composition of soil organic matter rather than its total content. In addition, PFAS were shown to influence soil processes by mobilizing dissolved organic matter (DOM), particularly under conditions of weaker sorption. Adsorption onto ferrihydrite was strongly pH-dependent and exhibited non-linear behavior, indicating the formation of multilayer structures at higher concentrations. In engineered systems, adsorption behavior was controlled by both PFAS molecular structure and solution chemistry. Ion exchange resins showed high removal efficiency, particularly for short-chain PFAS, but were sensitive to competition from co-existing ions like phosphate, while granular activated carbon exhibited more variable performance depending on PFAS structure and DOM composition. Dynamic PFAS removal experiments further revealed that system design plays a critical role, with differences in kinetics and breakthrough behavior observed between rotating bed reactors and column systems.

Together, these findings provide a coherent framework linking molecular-scale interactions to system-scale performance. The work highlights that PFAS behavior cannot be understood or predicted based on single factors alone, but rather emerges from the interplay between sorbent properties, PFAS chemistry, and environmental conditions. This has important implications for both environmental risk assessment and the design of remediation strategies, emphasizing the need for mechanistic understanding and matrix-specific evaluation when addressing PFAS contamination.

Source-separated blackwater contains most of the nutrients present in domestic wastewater and therefore represents an important resource for nutrient recovery. Blackwater digestate contains nitrogen, phosphorus, potassium, and micronutrients that could be recovered and reused in agriculture, contributing to more circular sanitation systems. However, the digestate is typically too dilute for efficient transport and agricultural reuse, requiring nitrogen stabilisation and volume reduction to produce a practical fertiliser product and reduce ammonia losses during thermal treatment.

This thesis investigates the pathway towards producing a fertiliser product from blackwater digestate through two sequential treatment steps: nitrogen stabilisation and low-grade heat concentration. Nitrogen stabilisation was investigated using both biological and chemical approaches. Biological stabilisation was achieved through nitrification in pH-controlled moving bed biofilm reactors (MBBRs), where the effects of startup strategy, temperature, and pH on reactor performance were evaluated. Startup strategy and temperature strongly influenced nitrification establishment, while nitrifier abundance alone did not explain reactor performance. Stable nitrification rates up to 1.8 g N m-² d-¹ were achieved at pH 6 and 30 °C. Furthermore, complete nitrification of approximately half of the ammonium was maintained even under acidic disturbance conditions at pH 2.3 and temperatures of 20 °C and 30 °C. Chemical stabilisation was achieved through acid dosing prior to thermal concentration, using phosphoric acid in membrane distillation experiments and nitric acid in low-temperature evaporation experiments to suppress ammonia volatilisation and maintain nutrients in soluble form.

Low-grade heat concentration of blackwater digestate was evaluated using laboratory-scale air-gap membrane distillation and pilot-scale low-temperature evaporation operated at moderate temperatures (40–70 °C), suitable for waste heat utilisation. The processes were assessed with respect to nutrient retention, volume reduction, flux behaviour, and energy demand. Membrane distillation achieved up to 15-fold concentration, although permeate flux declined to approximately 15% of its initial value due to fouling and membrane wetting. Pilot-scale evaporation achieved volume reduction factors up to 85 while producing nutrient-rich concentrates. Lower pH improved nutrient retention by reducing volatilisation and precipitation losses, while potassium and sodium concentrations increased under acidic conditions.

A dynamic model describing condensate generation during evaporation was calibrated against experimental data, yielding coefficients of determination (R²) between 0.961 and 0.997. The model showed that progressive solute enrichment increased the apparent enthalpy of evaporation at high volume reduction factors. Exergy analysis demonstrated that only about 12% of the supplied thermal energy was converted into useful exergy in the produced vapor, highlighting the importance of heat recovery and waste heat integration.

The development of edge computing, Internet-of-Things (IoT), and AI-driven services involves increasing volumes of sensitive data—health records, industrial sensor readings, personal information—at the network edge. Regulatory frameworks such as the GDPR, the European Health Data Space, and the EU AI Act impose strict requirements on such processing. However, privacy-enhancing technologies face trade-offs between utility, privacy guarantees, and computational cost that limit their practical use and uptake.

This thesis investigates how these trade-offs can be addressed through encrypted machine learning (ML) inference, architecture co-design for efficient encrypted computation, and privacy-preserving model training. The six appended papers, complemented by additional publications, address these trade-offs across health data infrastructure, edge data center management, and privacy regulation.

First, fully homomorphic encryption (FHE) is applied to data processing and control pipelines to investigate practically feasible encrypted computation on health data and encrypted remote monitoring and control of edge data center systems for lightweight ML tasks. Two leading encryption schemes—CKKS and TFHE—are evaluated, showing that both data and algorithm confidentiality can be achieved simultaneously, with encrypted execution times ranging from milliseconds to seconds.

Second, the computational bottleneck of neural network inference under FHE is addressed through architecture co-design. A novel mechanism is proposed—the Inhibitor—which replaces the multiplications and softmax activations of conventional gated RNNs and Transformer attention with encryption-friendly addition and ReLU. Experiments demonstrate 3–6× speedup for encrypted inference, and knowledge distillation produces a compact Inhibitor language model competitive on corresponding NLP benchmarks.

Third, methods are developed for training on distributed or sensitive data without compromising privacy. A local conditioning approach for heterogeneous federated learning is introduced, where locally computed statistics replace cross-client coordination, enabling scalable personalization without revealing distributional information or adding communication overhead. A proactive defense against training data memorization is also proposed, reducing the log-likelihood ratio for membership identification by an order of magnitude at the chosen audit precision, while maintaining model accuracy.

Taken together, the results suggest that privacy-preserving ML can be made more feasible for resource-constrained settings, including latency-sensitive edge deployments, helping to narrow the gap between regulatory requirements and computational cost. The evaluation relies on established benchmarks and controlled numerical experiments; validation in production deployments remains future work.

Biochar demonstrates strong decarbonization potential in concrete but significantly affects workability. To address this and to better understand the effects of biochar on concrete, as well as its role in decarbonizing it, this study applied a volumetric water-to-binder ratio (w/b), replacing cement with biochar powder at 5%, 10%, and 20% by volume, and compared the results with a 5% weight-based replacement. Additionally, prewetted biochar aggregate was used to replace sand at 30%, 60%, and 100% by volume. Various properties were assessed, including rheology, mechanical strength, hydration products, fire resistance, and carbon emissions of the concrete.

For biochar powder used as a cement replacement, the results indicate that the volumetric w/b ratio effectively improved workability and enhanced internal curing, leading to increased hydration products and improved mechanical performance compared to weight-based replacement. Despite a lower cement content, the 10% volumetric biochar sample achieved higher strength than the 5% weight-based sample. Biochar used as a sand replacement in concrete can better enhance internal curing and increase the degree of hydration due to improved moisture retention from the larger replacement volumes compared with its use as a cement replacement. Furthermore, replacing sand with biochar aggregate effectively reduces total shrinkage due to improved internal curing. Total shrinkage at 28 days was reduced by 41%, 55%, and 65% for 30%, 60%, and 100% biochar aggregate replacement levels, respectively.

Regarding fire resistance, both types of biochar increased the temperature gradients in concrete during heating. However, biochar aggregate contributed to a greater increase due to its higher volume and sand replacement ratio than biochar powder, which intensified thermal damage and, at 400 °C, even offset the benefits of accelerated cement hydration.

Due to energy recovery and carbon sequestration, wood biochar replacing 20% of cement reduced concrete carbon emissions by 42%, while fruit biochar replacing 100% of crushed sand reduced emissions by 167%, indicating that the concrete can become an effective carbon sink. However, carbon emissions should not be the sole consideration. Wood biochar used as a cement replacement reduced 56-day compressive strength by 3%, 6%, and 13% at 5 vol%, 10 vol%, and 20 vol%, respectively. Similarly, when fruit biochar replaced sand, the 56-day compressive strength decreased by 7%, 21%, and 47.4% at 30 vol%, 60 vol%, and 100 vol%, respectively.