Mosquitoes of the species Aedes Aegypti exhibit a particularly distinctive figure-eight shaped wingtip trajectory, which includes a vertical component of motion (characterized by deviation angle) that is hypothesized to influence aerodynamic force generation through interactions with the wing’s own wake, referred to as wake capture. While standard quasi-steady (QS) aerodynamic numerical models can effectively predict the forces generated by translational, rotational, coupling and added mass effects during wing motion, they do not capture the additional forces arising from wake interaction.

This study aims to leverage this known limitation of quasi-steady models, to investigate the influence of deviation angle on wake capture forces during hovering flight. Aerodynamic forces were measured using a robotic flapping-wing apparatus programmed to replicate biologically plausible kinematics of a mosquito, across a range of deviation angles (0° - 6°) (including an outlier at 7.5°). By subtracting forces predicted by a validated quasi-steady model from the experimentally measured forces, the residual unsteady component associated with wake capture was isolated and quantitatively analyzed.

The results of this study demonstrated that wake capture forces play a beneficial aerodynamic role in the mid-to-high deviation angle range, specifically between approximately 4° and 6°. Within this interval, the wake capture lift and drag forces exhibit relatively elevated mean values, and the experimentally observed lift-to-drag ratio surpasses quasi-steady predictions, indicating enhanced aerodynamic efficiency. Correspondingly, power utilization increases, and flapping efficiency peaks near 5° to 6°, collectively highlighting an optimal regime where unsteady wake interactions benefit aerodynamic performance. This favorable trend could be attributed to the pronounced out-of-plane wing motion at these deviation angles, which intensifies three-dimensional wake structures and possibly promotes more effective wing-wake vortex interactions that augment lift and overall flapping efficiency.

In spite of the high uncertainty in the force sensor readings, this study successfully established a kinematically accurate, high-fidelity setup for examining the influence of deviation angle on wake capture phenomena. The elevated noise from the inbuilt force sensor and structural vibration in the wing assembly were identified as the primary limitations, which could only be fully resolved through hardware replacement beyond the present scope. Nevertheless, the findings provide novel insights into the role of deviation angle in shaping wake capture forces, with both scientific implications for advancing flapping-wing micro air vehicle (FWMAV) design and societal relevance given the connection between mosquito flight and public health.

This thesis presents a novel method for non-destructive plant health monitoring using ultrasound to detect acoustic radial resonance modes in plant stems. By transmitting ultrasound through the stem and analysing the resulting frequency spectrum, structural and physiological changes can be inferred. Analytical models and finite element simulations (COMSOL) were used to predict resonance behavi0uor, and a custom-built measurement setup was developed to test both plant-proxies and live cucumber stems. Experimental results show that radial mode frequencies shift consistently as the plant dries, supporting the hypothesis that acoustic resonance contains measurable signatures of hydration state. The method proved sensitive to stem geometry, internal structure, and material properties. The study demonstrates the feasibility of acoustic resonance as a tool for plant health assessment. Focusing on the radial mode offers a simplified yet informative signal compared to more complex acoustic approaches. Future work should aim to improve the hardware, automate spectral analysis, and validate the method across different species and growing conditions, paving the way for practical in situ monitoring applications in agriculture.

Skid-steer mobile robots present a unique challenge for reinforcement learning (RL) due to their nonholonomic constraints, dynamic wheel coupling, and high susceptibility to slip. Standard RL methods often struggle to achieve stable torque control in such settings, particularly under long-horizon tasks with sparse or delayed rewards. This thesis introduces a knowledge-assisted RL framework that systematically integrates expert demonstrations and curriculum learning to improve sample efficiency, training stability, and final policy performance.

The foundation of the method is a torque-based Deep Deterministic Policy Gradient (DDPG) agent, augmented through two key innovations: (1) KAMMA (Knowledge-Assisted Mixed Mode Actioning), a probabilistic switching mechanism that alternates between expert and learned actions to avoid interference artifacts and accelerate early-stage convergence; and (2) Curriculum-Driven System Identification, where the task complexity is gradually increased via staged velocity profiles to reveal underlying terrain-robot dynamics in a structured manner.

Experiments conducted in Isaac Sim demonstrate that this integrated KAMMA + Curriculum approach outperforms both baseline KA-DDPG and imitation-only variants across key metrics, including trajectory tracking error, policy smoothness, and convergence speed. The results confirm that combining staged learning with adaptive knowledge infusion enables robust torque-level control and offers a scalable template for learning-driven system identification in robotics.

Until recently, multi-stable mechanical metamaterials have been primarily used in passive energy absorption systems. However, the ability to actively program these structures has gained significant interest, expanding their functionality to enable on-demand adaptive deformation. While existing active programming methods effectively induce global state transitions, localized actuation remains largely unexplored. This study introduces a novel approach to actively programming multi-stable metamaterials via local thermal stiffness modulation at boundary conditions. Using a polymer bi-material design with distinct glass transition temperatures between the beam and boundary supports, the system can transition from a bi-stable to a mono-stable state, enabling controlled snap-back behaviour after deformation. An analytical model is developed to characterize the snap-through behaviour of the unit cell, providing insight into the geometric interactions and sensitivities associated with various design parameters. Experimental implementation, using multiple additive manufacturing techniques, revealed key limitations and design considerations. In particular, the importance of constraining the second buckling mode and careful material selection emerged as fundamental design requirements for ensuring functionality. This work contributes to the growing field of actively programmable mechanical metamaterials, with implications for compact motion systems in future work.

In industries such as oil and gas, power generation, and chemical processing, the structural integrity of piping systems is crucial for operational safety and reliability. Various factors such as corrosion, erosion, and manufacturing defects can result in a loss of wall thickness, potentially leading to catastrophic failures. Detecting and assessing the integrity of piping systems without causing disruption to operations or compromising their structural integrity is a challenge in asset management and maintenance.
Guided Wave Tomography (GWT) is a promising non-destructive testing (NDT) technique for assessing the integrity of a pipeline by mapping its wall thickness. Using tomographic algorithms that use forward wavefield extrapolators, based on the gradient of the misfit a defect reconstruction can be made. Although the forward model is computationally inexpensive, the currently implemented gradient calculation is costly due to the large number of simulations required for each inversion parameter. This study explores the feasibility of using the adjoint state method for gradient computation in GWT, requiring only a single forward simulation independent of the number of model parameters.
A nonlinear conjugate gradient optimization method is implemented to iteratively update the wall thickness based on the misfit between simulated and observed data. The inversion process makes use of the dispersion curves to establish the relationship between phase velocity, frequency, and wall thickness. The study specifically focuses on the $S_0$ mode, analysing its suitability for complex defect reconstruction. Numerical experiments demonstrate that the adjoint method effectively localizes defects in both noise-free and noise environments, provided that there is a sufficient contrast in the sensitivity kernel between the defect and its surroundings. The technique achieves a resolution of approximately $1$ to $1.5\lambda$, with frequency continuation and multimodal inversions improving reconstruction accuracy. However, challenges arise in handling noise and wall thickness interchanges, which can be mitigated using adaptive regularization strategies.
The results suggest that the adjoint method is a viable alternative to traditional finite-difference sensitivity calculations. Future research should focus on experimental validation with real-world or as noise-free alternative 3D elastodynamic data evaluating the method on complex defect geometries and varying pipe configurations. Finally, the algorithm can be enhanced by integrating Hessian-based optimization techniques, thereby implementing a second-order adjoint method.

This research critically analysed the manufacturing process of a loader bucket with grapple. Such buckets are commonly used to unload dry bulk materials such as wood chips. In order to do this as quickly as possible, these buckets transport tonnes of material each cycle, the buckets themselves have a mass of tonnes too and can be as tall as 4 metres. These buckets are built to order as one-offs or in small batches, optimized for the needs and wishes of the customer. The goal of this research was to reduce the manufacturing lead time of such a bucket from 14 days to ten days or less. The bucket design was changed to accommodate a solution to the problem, but such a change should not increase the module’s weight by more than five percent, nor reduce its functionality in other ways. Each bucket has to remain specific to each customer and their requirements. This research first analysed these requirements and available variations in the bucket design. Following this, the operations in the manufacturing process were identified and analysed with regards to their criticality in the total lead time. Each operation was evaluated regarding its impact on the overall lead time, as well as the dependence on resources available in the factory, such as machinery like a milling machine. Furthermore, the dependency of other operations was also taken into account. According to these factors, the operations with the highest impact on the lead time were identified. From this, various lead time reduction methods were proposed which reduce the impact of these critical operations on the overall manufacturing process. By parallelizing the part with which the manufacturing of the bucket starts, the lead time could be reduced to just more than ten days. To reach the ten day goal, this parallelized part also had to be manufactured more efficiently. For this, six redesigns of the part were devised which do not require machining. These were evaluated in a survey among six employees from the engineering, planning and manufacturing departments. The most promising designs could reduce the expected lead time of the part itself by two days. By combining the suggested parallelization, along with the shortened manufacturing of the mounting plate without machining, the goal of a lead time of ten days could be achieved.

This research successfully establishes a proof of concept for the hot forming of PEEK sheets using resin-printed polymer molds.
Mechanical metamaterials, through their architected inner structures, offer unique properties that enable unparalleled functionalities. In the MECOMOS project, a novel manufacturing method is proposed to fabricate a multi-stable 3-DoF PEEK metamaterial that enables tip, tilt, and z-translation. The method involves forming PEEK layers to create the geometries of the unit cells, followed by joining these formed layers to produce a monolithic metamaterial.
In this research, a new forming method, hot forming, is developed and analyzed for shaping these layers. This method creates microscale features in PEEK sheets by implementing the process steps of hot embossing into matched die thermoforming.
Polymer molds with various geometries, including rounded cylinders, blocks and trapezoids, are successfully resin printed. Key variables influencing the mold output in resin printing include the cleaning procedure, post-curing, and layer thickness. The smallest printable features measure 400x400x100 µm, and the printing errors range from 40 to 150 µm and are predominantly independent of feature size.
The proof of concept for the hot forming process reveals its strong potential as a manufacturing method.The forming parameters are highly forgiving, with a broad range of process parameters still yielding high replication quality. The suitable forming temperature spans from 100 to 150°C, and pressing forces between 750 and 3000 N prove effective.

The current global environmental crisis has resulted in increased efforts towards more efficient and sustainable industrial processes. High-temperature heat between 150 ◦C & 400 ◦C accounts for a major part of the energy demand in industrial processes. At the same time, large quantities of waste heat are unutilised at temperatures up to 200 ◦C. A heat pump could upgrade this waste heat to cover a part of the demand, resulting in considerable savings both for the planet and the operator. Nevertheless, heat pumps have seen little development in heat delivery temperatures above about 150 ◦C. This research aims to assess the current technological potential and limitations of high-temperature heat pumps. Subsequently, solutions and recommendations are developed.
Focussing on mechanical vapour compression heat pumps, a thorough understanding of such cycles is gained first. The performance of a heat pump is highly dependent on the choice of cycle setup and working fluid, with the compressor posing the largest limitations for higher temperatures. To assess these, this project develops a heat pump model which simulates many different working fluids for different component configurations. The model was subjected to two temperature domains, covering waste heats of 100 ◦C & 200 ◦C and process heat temperatures in the range of 150 ◦C to 400 ◦C.
Results were obtained for all fluids incorporated in RefProp 9.0 and showed that multistage compression with intercooling and superheating considerably improved the performance of nearly all fluids. By comparing fluids based on efficiencies, capacities, temperatures and pressures, benzene and propylcyclohexane showed the best performance for the lower and higher part of process heat temperatures, respectively. The results however also showed the potential superiority of water as it has the best efficiencies and the largest applicability range, which combines with the hazard-free & environmentally friendly nature, low cost and wide availability. The main downside of water appeared to be the persistent, unacceptably high compression temperatures, combined with large pressures and pressure ratios. It was subsequently investigated how the disadvantages of water could be handled. A solution was found in the usage of Liquid Piston Gas Compression (LPGC), in which a rising liquid column, supplied by a pump, acts as a reciprocating piston in a compression chamber. This setup conveniently allows for liquid spray injection to cool the steam upon compression and alleviates limitations on the pressure ratio. By using the same water as the liquid in the LPGC, any temperature rise is compensated by the evaporation of liquid, resulting in more steam with a lower temperature. A numerical model of this type of compressor was made in which dynamics were modelled down to individual droplets. This simplified approach provided insight into the compression path with such liquid injection and allowed the approximate determination of the required amount of spray. Results showed that the injection could cool the vapour adequately even for high temperature lifts. The LPGC was subsequently incorporated into a single-stage heat pump cycle and compared the results for other fluids using ordinary compressors. These results showed large CoP improvements of 15-25 % CoP and low discharge temperature. With that, it was shown that an environmentally friendly fluid could be used in a simple single-stage configuration and still provide the best performance compared to any other fluid.

Drones are increasingly used nowadays, primarily for visual inspection tasks facilitated by onboard cameras. The field of aerial manipulation tries to expand the capabilities of drones by attaching a manipulator, enabling physical interaction. Unfortunately, the usability of aerial manipulators is hindered by disturbances resulting from the movements of the manipulator. These disturbances, including reaction forces and a shifting centre of mass, not only affect manipulation accuracy but also pose safety risks by potentially destabilizing the drone. In this thesis, a design is presented that addresses this challenge by leveraging the theory of dynamic balance.
A new design approach of making a manipulator fly, instead of the common approach of mounting a manipulator arm to a drone was used. This new approach avoids interference with the drone's components, allowing to focus on the design of the manipulator arm. Furthermore, it made it possible to create a manipulator which can manipulate above, to the side and underneath itself. This makes the presented manipulator arm more versatile than common aerial manipulators whose workspace is mostly located only above or below the drone. The kinematics, workspace and balance conditions of the manipulator arm are presented. Furthermore, the design's workspace is optimised while the mass of the manipulator is minimized in a bilevel optimisation. Finally, the design is validated both by simulation and measurements performed with the built prototype.
The design presented is the first inherently fully dynamically balanced manipulator with omnidirectional workspace which can be used for aerial manipulation.

This study explores the design and implementation of an electrowetting-on-dielectric (EWOD) system for precise die positioning in a chip assembly environment, utilizing a millimeter-scale ”chip-on-droplet” approach. Two methodologies were investigated: coarse positioning using digital signals, a method well-documented in prior research, and fine positioning using analog signals, which, to the best of our knowledge, had not been applied to droplet positioning at the time of writing. The primary advantage of the fine positioning approach lies in its independence from feature size, allowing for significantly smaller step sizes. Experimental results demonstrated that the analog system achieved step sizes as small as 3.5 μm in a ”droplet only” setup, outperforming the ”traditional” coarse positioning system. Performance limitations were found to arise primarily from the tracking setup rather than the EWOD system itself.
The study also evaluated various surface layers, confirming prior findings that a smooth, thin, and homogeneous surface layer is critical for achieving consistent and precise movement. Attempts to en- hance droplet guidance through modified surface structures were unsuccessful due to pinning forces exceeding the generated EWOD forces. These findings highlight the potential of analog signal-driven EWOD systems for fine positioning applications, with important implications for chip assembly tech- nologies.