Renewable energy, particularly wind power, plays a vital role in combating global warming and air pollution. However, the inherent variability of wind complicates turbine performance predictions, especially for offshore wind farms. Standardised power curve models often fail to account for site-specific conditions, leading to performance deviations.

This thesis addresses these limitations by developing a new method to correct power curve predictions using detailed, site-specific data. Nacelle-mounted LiDAR and SCADA data from RWE’s offshore wind farms are utilised to enhance performance assessments during the development phase. Existing methods only rely on turbulence intensity (TI) to explain performance deviations, but these methods are based on smaller, onshore turbines and do not adequately account for the complexities of larger offshore turbines. The research introduces power performance matrices that incorporate additional wind characteristics such as shear, veer, and yaw misalignment, rather than relying solely on TI. The study finds that while LiDAR is less effective in measuring TI, it excels at capturing shear and veer, both of which show consistent patterns in relation to turbine performance. The findings reveal that extreme shear values negatively affect power production, while neutral or low shear has positive effects. Similarly, high veer correlates with decreased performance. 

Based on these insights, new power performance matrices were developed, using shear, veer, and TI, along with normalised wind speed, to adjust the power curve. These matrices can be dynamically created using site-specific data, making them flexible and adaptable for different turbines and wind conditions. The method enables selecting “inner range” parameters, providing a more accurate reference for power deviations. Additionally, this approach is particularly advantageous because shear and veer are well-captured by both floating and nacelle LiDAR, making it applicable across various stages of wind farm development. The primary limitation of this method is its reduced effectiveness in regions with significant mechanical atmospheric mixing, such as near coastal or mountainous terrain. In such areas, local topography can create complex wind flows, requiring more parameters than shear and veer to fully capture the wind dynamics. Another limitation is wake analysis; despite directional filtering to remove wake effects, some residual wake impact persists, suggesting further refinement is needed. 

Although more data is necessary to fully validate the approach, the current methodology shows promise in enhancing energy yield estimations and improving long-term predictions. The use of site-specific power matrices, particularly for offshore turbines, could significantly increase the accuracy of performance predictions, benefiting wind farm developers and manufacturers. As more data from LiDAR measurements and power performance tests become available, the matrices can be adapted to different turbine types and conditions by normalising wind speeds and creating dynamic “inner ranges” for site-specific use. Future research should focus on refining wake filtering methods and understanding the full correlation between shear, veer, and TI to further enhance prediction accuracy. In conclusion, this thesis proposed a novel method for adjusting turbine power predictions using site-specific wind characteristics, with potential to greatly improve the reliability of performance estimates. By integrating shear and veer into the power curve adjustments, this approach allows more reliable performance estimations. With further data collection, this methodology holds the potential to become a key tool for optimising wind energy production in the future.

To detect cyber-physical attacks targeted at a water storage unit part of the Water Distribution Network (WDN), this thesis investigates the use of process data provided by Dunea, a Dutch drinking water production company, for this purpose.

In order to achieve this, a model-based anomaly detection method is employed. This consists of deriving an accurate model that represents the nominal dynamics of the system, computing the residual between the estimation and the measurement and evaluating the residual using the Non-Parametric CUSUM (NP-CUSUM). Additionally, cyber-physical attacks are designed to expose the water storage unit, during its draining and replenishing, to the most extreme attacks that could potentially impact the WDN.

Results show that all designed attacks initiated during the reservoir's replenishing are detected within this mode of operation. Regarding the attacks initiated during the reservoir's draining, various attacks are detected within this mode, and some attacks are detected afterwards when the system operates in the replenishing mode. The model-based anomaly detection method implemented using the process data can detect all the designed and simulated cyber-physical attacks against the water storage unit.

This thesis's contributions include developing a benchmark for implementing potential future anomaly detection methods for Dunea's and other drinking water production companies' water storage units. Additionally, given the provided process data, a nonlinear hybrid automaton is created to describe the nominal behaviour of the water storage unit. Lastly, multiple attack profiles are studied, including their impact on the physical system and effectiveness in evading detection.

This thesis presents the design and implementation of a process monitoring system for the
HIsarna pilot plant aimed at the early detection of anomalies, specifically foamers. Though
rare, foamers represent severe process faults that can significantly impact the plant’s safety
and reliability. Early detection of these events is crucial for maintaining operational stability
and steady iron production.

Given the ambiguous definitions of nominal and non-nominal operations in HIsarna, this thesis constructs a precise definition based on foamer characteristics. Foamers are non-nominal operations, while the absence of these foamer characteristics indicates nominal operations. The monitoring system utilizes data from a gas analysis sensor in the plant’s topside outlet, the dogleg. The dogleg gas analyser is chosen for its reliability and the absence of monitoring methods based on this sensor. To ensure the accuracy of the sensor data, an optimizationbased method is employed to eliminate air ingress disturbances.

An anomaly detection method based on distance measures is developed, incorporating the
HIsarna Operation Model (HIOM) to reduce sensitivity to operating point changes. Historical
process data is analyzed to establish a distance measure that reflects the correspondence of
current measurements to nominal operations. The performance of the anomaly detector is
evaluated based on the number of foamers detected, hours per false alarm, and detection
time.

The primary contribution of this thesis is the development of an anomaly detection algorithm
using Mahalanobis distance for the HIsarna pilot plant. This metric indicates how closely
current measurements match historical nominal operation data. Additional contributions include a rigorous definition of nominal and foaming operations, an optimization-based method for air ingress removal, and an observer for dogleg gas analysis data, which can estimate carbon conversion and oxygen flow prediction bias of HIOM.

This thesis addresses the detection and prevention of adversarial attacks on self-triggered control (STC) systems and demonstrates how manipulating the sampling times can be used to provide a secure control framework. Secure control is vital to guarantee both the safety and security of critical infrastructure, which has been the target of malicious attacks over the past decade.

First, a novel watermarking scheme is proposed based on the early triggering of a well-designed STC policy, such that stability is guaranteed to be preserved. We show that this watermarking scheme, together with an event-triggered χ² detector we design, is able to detect replay attacks. An online heuristic for obtaining an optimal early triggering policy is provided. If certain assumptions hold we show that the early triggering policy is a discrete uniform one. Through an illustrative example, both a quantitative and qualitative comparison between two other watermarking schemes are provided. We conclude that none of the watermarking schemes can claim absolute superiority, and trade-offs between all considered schemes exist.

Next, we propose a new type of attack called a switched zero dynamic attack (ZDA), and provide an algorithm on how to construct these switched ZDA. We show that certain STC systems are susceptible to such attacks, and demonstrate that by tuning the triggering parameters there exist sufficient conditions such that these attacks are no longer disruptive. The effect of additive perturbations and a non-zero initial condition, as well as the proposed tuning method, are shown in a numerical example. We provide a qualitative comparison between several other countermeasures in the literature, which we extend for aperiodic sampling when needed. Finally, shortcomings and future directions are discussed.

Wind energy is growing to be an essential part of the transition towards sustainable energy sources. To facilitate this, it is crucial that wind farm operators can offer competitive energy prices compared to fossil fuel sources; effective and efficient wind farm operations are therefore imperative. However, many full-scale wind farms deal with wake effects, e.g. the disturbed air that travels downstream of a turbine and potentially ends up in other neighbouring turbines. The result of this lower-velocity, higher-turbulence flow field is both decreased power production and increased fatigue loads. One proposed solution for this problem comes in the form of 'wake steering': the yawing (rotating) of upstream turbines' nacelles to facilitate a degree of control over the deflection of wakes. By doing so, the problematic disturbed flow fields can be strategically guided between downstream turbines to maximise collective power production. However, the yawing action itself brings some adverse effects: the upstream turbines, now no longer directly facing the wind, feel decreased power production and increased fatigue loads themselves. These fatigue effects, accumulating over time, can eventually cause increased maintenance costs and nullification of any revenue gains through power optimisation. The problem thus becomes a farm-wide collective revenue optimisation task. This thesis investigates how long-term revenue in wind farms can be maximised, considering both profits through power optimisation and maintenance costs through fatigue-induced component failures due to wake steering control. First, a realistic wind farm simulation environment is constructed, based on a Graph Neural Network (GNN) surrogate wind farm simulation model, to facilitate efficient reinforcement learning training. Next, the environment is used to train fully centralised reinforcement learning agents based on a GNN architecture, resulting in agents that can generalise across all wind conditions and unseen wind farm layouts. Ultimately, the results show that an 'informed' agent that considers all profits and costs involved manages to significantly reduce the cost of energy compared to 'greedy' (power optimisation only), 'risk-averse' (damage minimisation only) and 'baseline' (zero-yaw) policies, furthermore considerably maximising long-term wind farm revenue by as much as 20%. Altogether, this thesis shows promising results in using graph-based reinforcement learning to train maintenance-conditioned, inflow-agnostic, and layout-agnostic wake steering controllers for wind farm revenue optimisation.

Linear Parameter-Varying (LPV) systems can be used as a bridge to extend the well studied model based control methods of Linear Time-Invariant systems to certain nonlinear systems. Despite significant attention in literature over the last two decades, finding an efficient global state space identification algorithm remains an open problem. Furthermore, a common assumption has been that the scheduling signals governing the time-varying dynamics are known or measured exactly. These drawbacks are found to inhibit the number of use cases for model based LPV control. This thesis explores new ways of identifying LPV systems for more general nonlinear systems with limited information on the optimal scheduling variable and scheduling dimension.

To this end, two novel rank constrained least squares problems are presented to identify system matrices and scheduling signals from input and full state measurements, and a dictionary of possible scheduling signals. The use of the methods is demonstrated in a simulation experiment, and the results are compared to existing full state measurement methods.

At last, the full state measurement methods are coupled with a state sequence estimation method from literature in order to obtain a method for identification of quasi-LPV systems from Input-Output measurements. Here "quasi" indicates that the scheduling signal is dependent on the system state and input. The scheme is flexible, and a proof of concept is given on a small nonlinear system.

Task and Motion Planning (TAMP) has progressed significantly in solving intricate manipulation tasks in recent years, but the robust execution of these plans remains less touched. Particularly, generalizing to diverse geometric scenarios is still challenging during execution. In this work, we propose a reactive TAMP method to deal with disturbances and geometric ambiguities by combining an active inference planner (AIP) for online action selection with a proposed multi-modal model predictive path integral controller (M3P2I) for low-level control. The AIP generates online alternative plans, each of which is translated into a cost function to be sampled for the proposed method. The proposed M3P2I then uses a parallelizable physics simulator for throwing different rollouts, leading to a coherent optimal solution by averaging the weighted samples based on their costs. Our method empowers real-time adaptation of action sequences to rectify failed plans, while also computing low-level motions to address dynamic obstacles or disturbances that could potentially invalidate the existing plan.

Theoretical findings are validated in simulation and in the real world. We show that the framework exhibits reactiveness in different scenarios, including battery charging, push-pull among obstacles, and pick-place with disturbances. We show that our framework outperforms an off-the-shelf RL method in the reactive pick-place task in terms of position error and orientation error. We also show that M3P2I is generalizable in combining different constraints, such as generating hybrid motions of push and pull for the mobile robot, and grasping objects with different grasping poses. The real-world experiments show that the system exhibits reactiveness and robustness against human disturbances in a variety of manipulation tasks.

The supporting videos can be found at https://sites.google.com/view/m3p2i-aip.

The Monitoring technique plays a vital role in ensuring the proper functioning of modern industrial systems that are highly sophisticated and automated. On two different applications, this thesis investigates two major categories of information redundancy monitoring techniques, model-based and data-driven.

The first application focuses on ground fault detection in microgrid systems. Leveraging the model information of the system, we propose a design approach for the fault detection filter by creating a linear programming problem. This design ensures the complete decoupling of the disturbance and guarantees fault sensitivity. Recognizing that decoupling is not always feasible, we create a new optimization problem by exploiting available disturbance patterns, so that the filter suppresses the impact of the disturbances while ensuring the fault sensitivity. Simulation studies validate the effectiveness of the proposed designs. The second application deals with non-intrusive load monitoring (NILM) in building systems. Our approach involves a two-stage process that utilizes data to perform NILM. In the first stage, events are identified from the aggregate load measurement. In the second stage, an integer programming problem is formulated to estimate the load for each appliance. The effectiveness of our method is evaluated on a real-world dataset and compared with several other NILM approaches, demonstrating competitive performance in terms of accuracy and computational complexity.

A significant decrease in greenhouse gas emissions can be achieved by including a prediction of future power consumption in the control of ships’ power plants during transits. Moreover, including a prediction of power consumption in the Energy Management System (EMS) during Dynamic Positioning (DP) operations can also contribute to a reduction in greenhouse gas emissions. However, predictions of future power consumption for DP operations, as well as its implementation in an EMS, have not been investigated widely, since this is a complex task for a relatively specific application. This research aims to investigate how power consumption during DP operations can most accurately be predicted for near-future and far-future, within reduced computational requirements for real-time applications.

In this research, six different approaches for power consumption prediction are investigated. These approaches can be categorized into projection models, which determine the power consumption independent of time based on the environmental conditions, and forecast models, which rely on the recent past behaviour of the system.
The four projection models consist of two Physical Models (PMs), a Data-Driven Model (DDM), and a hybrid model. The first PM is the static model, in which the environmental loads are directly compensated by the thrusters, hence the ship is assumed not to move. The second PM is the dynamic model, in which the motions of the ship are also modelled. Then, the DDM projection is based on Kernel Regularized Least Squares (KRLS) to capture the nonlinear relation between environmental conditions and power consumption based on historical data. Lastly, the hybrid projection model is an integration of the dynamic model in the DDM.
Afterwards, two forecast models are developed, being one DDM and one hybrid model. The DDM forecast is based on a combination between KRLS and Time Series (TS) forecasting. The hybrid forecast is an integration of the dynamic model in the DDM forecast.

Simulations are performed for each model using a data set provided by RH Marine. The DDMs show the most accurate results, taking into account the required computational effort. The results show that the forecast model, using the combination between KRLS and TS, predicts the near-future at 1 s in the future with 2.8 % error, increasing to 6.9 % at 120 s in the future. Afterwards, the projection model, using merely KRLS, predicts the far-future up to the length of the operation horizon with an accuracy of 11.2 %, based on the weather forecast. This multi-horizon prediction of power consumption can enable the EMS to make accurate short-term decisions in the control of the power plant and to define optimal scheduling of the power plant components over the
complete horizon of the DP operation.

Steel production is a critical index to measure the infrastructural growth and development of a nation. The steel production capacity of a nation has a significant impact on its GDP. Even though steel rolling has been in the industry since the early 1700s, there have not been significant advancements in its technology space. The safety hazards to personnel and failure of process components have persistently posed daunting challenges for the steel industry. The advent of automation in the late 1900s helped the industry manage some of these challenges to a certain extent. Despite these technological improvements, the realm of steel rolling is still not explored thoroughly because steel rolling is a highly integrated and complex system with numerous process parameters impacting the quality of the finished product. As a result, the study of the dynamics of steel rolling is still under active research targeted toward improving the complex processes involved in the industry.

External factors play a significant role in the complexity of the steel rolling process. The scope of the work herein attempts to identify and model some of these significant external factors, also known as "disturbances" in control terminology. A Finite-Element Method (FEM) based simulator for the rolling process simulation incorporating the external disturbances is explored. The outcomes from this simulation will enable establishing regression models that facilitate quantifying external disturbances’ effects on the technological/process parameters in steel rolling.

The thesis is focused on studying the external disturbance of speed mismatch and proposing a quantitative model for evaluating the effect of speed mismatch on the process parameters of rolling. Subsequent to the development of the proposed disturbance quantitative model, three controller systems - Proportional Integral Derivative (PID) controller, Linear Quadratic Regulator (LQR) controller, and Model Predictive Control (MPC) controller will be implemented to evaluate comparisons between the controllers for disturbance rejection and reference tracking.

The scope of the work presented in this thesis is significant as it focuses on developing a quantitative model for the process disturbances that prevail in the steel rolling industry.

The energy transition towards carbon neutrality requires a rapid electrification of all energy sectors by 2050. In the EU-27, the REPowerEU strategy, initiated in March 2022, is pursuing this target even more ambitiously. However, in order to achieve the aforementioned goals, the increasing demand for green hydrogen will cause an increase of the renewable energy needs as early as 2030. In this way, offshore wind can be a solid supplier of the renewable energy for green hydrogen production. However, since nearly 80% of the worldwide wind energy potential is situated in deep waters, floating offshore wind turbines (FOWT) can be used to cover those energy needs. In addition, the literature review showed that when green hydrogen production via FOWT is considered, the most economically and energy efficient layout is the in situ topology, where hydrogen is locally produced on the FOWT. Although FOWT and green hydrogen production via FOWT have been lately examined in literature, a literature gap was found. More specifically, no attention was given on the performance change of a turbine when it is adapted from a bottom fixed (BF) application to a FOWT. Also, the effect of the in situ hydrogen plant to the FOWT performance was not considered. Thus, the aim of this project is to highlight if FOWT for in situ hydrogen production should be aerodynamically redesigned to improve their performance or to tackle possible energy losses. In this direction, several sub-questions should be answered, including the selection of turbine, floater type and site to be investigated, the effect of the floater design on the FOWT performance, the performance change of a turbine due to its adaptation as FOWT and the effect of the added mass of the hydrogen plant on the FOWT performance. Lastly, in case of a proven FOWT performance deterioration, solutions should be provided so as to regain performance.  Aligned with the previous goals, the IEA 15MW reference wind turbine on top on the UMaine Volturn US-S semi-submersible floater is simulated in OpenFAST under steady and turbulent wind fields, according to wind & wave conditions of a typical US East Coast site. In addition, the in situ hydrogen plant is incorporated in the model using a simplified approach. The results suggest that in the whole partial load region, the FOWT experiences power losses due to the static floater pitch angle, which reduces the inflow wind speed seen by the rotor. However, between 9 and 12 m/s, a peak shaving routine is incorporated in the FOWT controller, which results in early power shedding and contributes, together with the static floater pitch, to considerable power losses. Furthermore, the simulations conducted using the OpenFAST FOWT model, which incorporated the hydrogen plant, suggest that it has a negligible effect on the FOWT performance and can be omitted from the model.  Finally, the comparison of the FOWT and of the BF turbine under turbulence, highlights the fact that the FOWT exhibits a spanwise aerodynamic torque reduction and a spanwise airfoil aerodynamic inefficiency in terms of angle of attack, that can be solved via an aerodynamic redesign. As a result, a variety of blade twist angle and airfoil chord length profiles are developed, tested and evaluated using the FOWT annual energy production (AEP) as an indicator. The results point out that all solutions result in a slight FOWT AEP gain, compared to the original FOWT design, at the expense of increased rotor loading, which effectively increases the static floater pitch. Thus, the aerodynamic redesign approach requires a more cautious approach.

Due to the continuously increasing volume of road freight transportation, there is a need to aid or automate truck-trailer driving. Truck docking, the process of parking a truck-trailer combination at a loading dock, is one of the most difficult manoeuvres for professional drivers. In this work, we propose a Model Predictive Control (MPC)-based truck docking driver assistance system. The objective of the system is to support the truck driver while parking the vehicle by means of visual instructions. MPC is an advanced control method that can be used to control Multiple-Input and Multiple-Output (MIMO) systems based on a finite horizon optimization. The control structure allows one to formulate a multi-objective control strategy with explicit constraints. This work has been conducted within the scope of the VIsion Supported Truck docking Assistant (VISTA) project and the practical application of the proposed system is to experiment with an MPC-based approach for the VISTA system currently in development. To determine the optimal control action, the proposed MPC-based driver assistant makes use of a kinematic model describing the motion of the vehicle as well as a second-order linear time-invariant model representing the driver’s behaviour. The system is examined by testing four performance categories: path tracking error, robustness, computational speed, and driver acceptance. The performance of the system is tested with professional truck drivers and people without a truck driver’s licence in a Virtual Reality (VR) simulation. The results suggest that the proposed MPC-based driver assistant is able to perform well regarding reference path tracking and is robust against driver errors, with satisfactory computational speed. The feedback and comments from the professional drivers during testing also indicate definite possible advantages of using the system. As of now, the MPC-based solution is preferred over previous concepts of the VISTA system.

Job scheduling is the process where jobs are arranged in a specific sequence. This process has always been a crucial subject for companies in numerous industries. A company could significantly improve its business performance if the scheduling process is not performed optimally. As companies and products become more and more data-driven, new tools to improve processes are emerging. Robust job scheduling methods optimize job schedules with job duration uncertainties. Typical robust scheduling methods only use straightforward statistical metrics such as the job duration and variance. Data relevant to the process' job duration, stored as observational data, is not typically used in robust job scheduling methods. A field of research that is concerned with the use of observational data is causal inference. It can be applied to observational data to explain causal variable behaviour and therefore also the statistical metrics used in the scheduling process. In addition, it can be used to predict variable behaviour after certain variable modifications. This motivates our search for an approach that uses the tools of causal inference on available process data to identify the causal relations within these job processes. Based on this information, it predicts the effects of parameter modifications. These predictions are incorporated in the scheduling optimization to potentially further improve the job scheduling performance. In this thesis, we will therefore study research on causal inference and a method of robust job scheduling and propose an algorithm that extends the robust job scheduling method. The algorithm includes the tools of causal inference and additionally solution sensitivity analysis.

With a growing number of citizens and tourists, the scarce public space, roads, and public transport in Amsterdam are experiencing rising pressure. Instead of utilizing the conventional transportation routes, Autonomous Surface Vessels (ASVs) could transport goods and people via the 165 canals present in Amsterdam.
However, urban canals are challenging for motion planning since the space can be narrow and contain other human-operated boats.
Unfortunately, there are limited existing works regarding motion planning for ASVs in urban canals with dynamic obstacles. Additionally, motion planners for other applications can not be applied directly because of the lacking traffic structure in the canals, the specific dynamics of ASVs, and the regulations applying to canals.
Therefore, the purpose of this thesis is to develop a motion planning framework that can plan dynamically feasible motions for ASVs in urban canals complying with regulations. These rules are not only mandatory but adhering to them makes the ASV's motion socially compliant, and therefore, more predictable by other canal users.
We build upon local model predictive contouring control and extend it with regulation compliance. Adherence to rules is achieved by constructing a new cost function for the optimization problem that allocates a higher cost on specific sides of the obstacle vessel. With this new cost function, the Roboat can behave according to the regulations in takeovers, head-on encounters, and crossings with boats from port and starboard.
Furthermore, the effect of predicting the obstacle vessel trajectories with a Social variational recurrent neural network is compared to the constant velocity model.
The entire motion planning framework is compared in simulation with LMPCC and the current motion planning and control framework of the Roboat, breadth first search combined with Nonlinear Model Predictive tracking Control (NMPC).
Additionally, the motion planning framework is implemented on a quarter-scale Roboat and is tested in an outdoor environment with disturbances.

Online video completion aims to complete corrupted frames of a video in an online fashion. Consider a surveillance camera that suddenly outputs corrupted data, where up to 95% of the pixels per frame are corrupted. Real time video completion and correction is often desirable in such scenarios. Therefore, this thesis improves the Tensor-Networked Kalman Filter (TNKF) as presented in [12] by developing the Tensor-Networked Square-Root Kalman Filter (TNSRKF). The TNSRKF is a Square-Root Kalman Filter (SRKF) realized in a Tensor Networks (TN) structure. The square-root Kalman filter is an inherently stable algorithm, and indirectly allows for more information retention throughout the algorithm. Thereby, the filter aims to improve the performance of the TNKF. Furthermore, implementing the filter in TN results in an intuitive implementation of the filter while allowing for larger video dimensions than the widely used matrix format. This thesis concludes that the TNSRKF is too computationally burdensome to complete a video in an online fashion due to the implementation of the Modified Gram-Schmidt (MGS) algorithm in Tensor-Train (TT)-format. In addition, results demonstrate that a low-rank orthogonality matrix is not realizable, making it impractical to update the state covariance matrix via any method that uses an orthonormal matrix.

Wind turbines offer an appealing method for power generation. In the fight against global warming wind energy helps realise green house gas emission reduction targets. Moreover, given a globally increasing demand for energy, the role of wind energy is set to become even more important. Historically, the cost of wind energy has decreased significantly, making it more competitive with respect to fossil fuels. This trend is set to continue thanks in part to a driving factor, which is the increasing size of wind turbines. However, there are several impediments to the upscaling of wind turbine sizes. The larger, more flexible wind turbine components experience greater loads. To mitigate increasing fatigue loading, which limits the lifetime of components, thereby increasing the cost of wind energy, smart control methodologies are needed. The existing literature indicates that individual pitch control (IPC) offers good perspectives for fatigue load mitigation. An interesting approach to performing IPC is by means of model predictive control (MPC). Unlike conventional IPC implementations, MPC offers the designer a more intuitive way to evaluate trade-offs between, e.g., power generation and load mitigation. In addition, an important advantage of MPC is its ability to handle constraints explicitly. A particularly interesting application of MPC for wind turbine operation concerns convex economic model predictive control (CEMPC). By means of a simple change of variables, the normally troublesome nonlinear dynamics are transformed into linear dynamics. As a result, relevant optimization problems can be reformulated in a convex fashion. The formulation of such a convex optimization problem is useful because such problems enable a global optimum to be found, and can be solved relatively efficiently. Unfortunately, the existing CEMPC framework is limited to collective pitch control (CPC). To facilitate further load reduction, it would be useful to extend the CEMPC framework to the domain of IPC. This thesis therefore seeks to investigate how the existing CEMPC framework can be extended to the domain of IPC for the purpose of wind turbine load mitigation. It is shown that an improved CPC implementation of CEMPC can be extended to IPC by considering a blade-effective wind speed and CPC-equivalent aerodynamic power for each blade. With regards to load reduction, a case study of out-of-plane blade root bending moment fatigue load mitigation indicates several important difficulties that require further investigation. Arguably the most important impediment for future load reduction prospects within CEMPC using IPC is the inability to consider the pitch, which is an important variable to minimize various sorts of loading, as a free optimization variable.

As systems become more lightweight, to satisfy inertia and size requirements, vibration becomes a prominent factor in their dynamics. This vibration is undesirable and various suppression methods exist such as passive, semi-active, and active. In this thesis, active control methods are explored for this purpose. The present technology utilizes under-actuation for suppression of multiple modes, which has a sub-optimal performance. This work provides a comparative study on the different vibration suppression algorithms, which would aid in developing a distributed placement of actuators (over-actuation) and sensors. The main aim is to achieve multi-mode suppression systems and improve collocation for higher-order modes which facilitates accurate control of the end-effector of a system. A comparison is drawn between point actuation and over-actuation in terms of energy consumption, amount of damping, and precision. Also, a new control strategy is developed to circumvent the limitations posed by the present control strategies such as low frequency spillover and steady-state error. The benefits in terms of damping and shortcomings are presented based on the conclusion drawn

In this thesis we consider multiple Automated Guided Vehicles (AGVs) navigating a common workspace to fulfill intralogistics tasks, typically formulated as the Multi-Agent Path Finding (MAPF) problem. To keep plan execution deadlock-free, one approach is to construct an Action Dependency Graph (ADG) which encodes the ordering of AGVs as they proceed along their routes. Using this method, delayed AGVs occasionally require others to wait for them at intersections, thereby affecting the plan execution efficiency. If the workspace is shared by dynamic obstacles such as humans or third party robots, AGVs can experience large delays. A common mitigation approach is to re-solve the MAPF using the current, delayed AGV positions. However, due to its inherent complexity, solving the MAPF is time-consuming, making this approach inefficient, especially for large AGV teams. To address this challenge, we present a novel concept called a Switchable Action Dependency Graph (SADG) which is used as the basis for a shrinking and receding horizon control scheme to repeatedly modify an acyclic ADG to minimize route completion times of each AGV using an optimization based approach. Our control strategies persistently maintain an acyclic ADG, necessary for deadlock-free plan execution. The proposed control strategies are evaluated in a simulation environment and show a reduc- tion in route completion times when a fleet of AGVs is subjected to random delays. Finally, the methods are also implemented using ROS and validated in the Gazebo simulation envi- ronment to illustrate practical feasibility when applied to real systems.

The increased implementation of digitalisation all over the world has led to an exponential growth of available data across various industries. Consequently, there is a large growth of Machine Learning (ML) techniques being applied to process data. Predictive maintenance is a digital strategy using condition-based monitoring techniques to track the performance of equipment to detect possible defects in advance. In oil refineries, various types of process equipment are used, of which flow control valves are essential to regulate the throughput of heavy, possibly dangerous material. Control valve failures can lead to production loss and increase maintenance costs. This paper addresses the use of time-series data of the valve controller for automated failure diagnosis of flow control valves. Statistical features are extracted from the time series and the significant predictors for the output are adopted in the model using the ANOVA test. The classification and prediction of the failure behaviour are performed using Random Forest (RF) classification. The performance of the diagnosis is measured using the indicators: accuracy and log loss. Findings show that five failure behaviour categories can be predicted, using new and unseen data for the model, with high accuracy(81%).

The introduction of control systems in the automotive industry has significantly increased safety. Improved control over the vehicle dynamics has been shown to contribute to substantial reductions in the number of deaths and serious injuries resulting from road traffic crashes. The introduction of Electronic Stability Control yielded impressive improvement in vehicle stability. A meta-analysis revealed that a 49% reduction of single-vehicle accidents is realized. Recent research continues the development of a fully autonomously operating vehicle. The vehicle requires the ability to operate in all situations safely, in order to reach the highest level of automation. Vehicle performance should be guaranteed within, at and beyond the limits of friction. Previous research revealed that the unstable drift motion could enlarge the operating envelope of a vehicle. An extensive amount of research is dedicated to controlling a drift. The results show that control systems are increasingly capable of stabilizing a steady-state cornering scenario. The main limitation of these studies is that only a portion of the vehicle motion that is observed in reality can be considered to be steady-state motion.

This thesis presents a multi-objective trajectory optimization which extends the steady-state analysis to a dynamic driving scenario. Based on experimental data obtained with a 1:10 scaled vehicle, accurate vehicle and tire models are derived. It is validated that the models closely mimic the dynamics of the scaled vehicle. In order to justify the use of drifting, the differences between stable and unstable driving equilibria are studied. The stability and controllability are assessed through the construction of the phase portraits and the computation of the Controllability Grammian. The findings, obtained under the assumption of steady-state conditions, are then validated in the dynamic driving scenario. A two-step optimization approach is presented. Spline optimization based on a simplified model is used to obtain initial conditions for a high fidelity model-based optimization. The scope is limited to a single corner, which is optimized under varying velocities and friction conditions.

Under the assumption of steady-state conditions, it is found that the drift motion imposes various benefits over normal driving. Higher cornering velocities and therewith yaw rates can be achieved in a drift. Besides, the principles of tire saturation and force coupling allow for controlling the lateral and yaw dynamics of the vehicle through the rear longitudinal tire force. This increases the maneuverability of the vehicle. The results of the dynamic optimization extend the findings of the steady-state analysis. In the dynamic maneuvers, drifting is found to improve vehicle maneuverability at high velocities and in scenarios of low friction. The approach presented in this work forms a basis for studying the effects that drifting could have on vehicle motion in reality. The relevant aspects of vehicle motion are translated into a multi-objective optimization. The methods that are developed in this work release the simplifying assumption of steady-state driving conditions. As a result, the drift motion can be studied in a more realistic driving scenario. It is expected that through further improving the optimization algorithm, the full operation envelope of the vehicle can be explored.