http://lutpub.lut.fi:80/handle/10024/158302 2026-07-11T01:40:46Z http://lutpub.lut.fi:80/handle/10024/172618 Design and control of compact mechatronic systems based on magnetic shape memory alloys Kulagin, Ivan Magnetic shape memory (MSM) alloys offer unique actuation capabilities, combining large reversible strain, high power density and a fast response enabled by magnetically induced motion. These properties make MSM-based actuators attractive for compact and high-performance mechatronic systems. However, practical deployment remains limited due to intrinsic material hysteresis, restricted stroke, and sensitivity to operating conditions. Moreover, most existing MSM actuators operate at relatively low frequencies and employ minimal or no mechanical transmission, leaving the potential of more complex actuator architectures and advanced control strategies largely unexplored. This dissertation presents a novel strain wave gearing MSM (SWG-MSM) actuator concept that addresses these limitations by combining multiple MSM elements with strain wave gearing mechanism. A comprehensive model of the SWG-MSM actuator is developed, capturing the mechanical interactions and MSM dynamics governing actuator behaviour. The model is used to analyse actuator performance, identify key design parameters, and derive force-velocity characteristics. A laboratory prototype is designed and experimentally characterised, including measurements of the magnetic circuit, rack displacement, velocity, and output force. The experimental results confirm the technical feasibility of the proposed concept and show good agreement with the simulation. The observed deviations are primarily attributed to non-stabilised MSM elements, mechanical backlash, and simplifying assumptions in the MSM material model. Beyond open-loop characterization, this work presents the first investigation of closed-loop control for SWG-MSM actuators. Classical PID control and several reinforcement learning–based controllers are evaluated in simulation for velocity and position control under varying external loads and internal disturbances. PID control demonstrates reliable and accurate tracking, making it well suited for initial practical implementation. Reinforcement learning-based controllers, particularly the proximal policy optimization (PPO), achieve superior performance and robustness to non-linearities and discrete actuation effects, highlighting their potential for advanced control of SWG-MSM actuators. 2026-08-14T00:00:00Z http://lutpub.lut.fi:80/handle/10024/172467 Trilogy of financial technology (fintech) integration for sustained usage, performance, and growth Addae, John Agyekum This doctoral thesis examines how the integration of Financial Technology (FinTech), encompassing mobile money, multichannel banking, and central bank digital currencies (CBDCs), collectively affects sustained usage, business performance, and economic growth. To investigate these interrelated outcomes, this thesis is organised into three multilevel analytical dimensions of Trilogy of FinTech, spanning micro-, meso-, and macro-level perspectives to examine how finTech shapes sustained usage, firm performance, and economic development outcomes. This thesis makes several scientific contributions through seven publications that employ diverse mixed-methodological approaches. At the micro-level, the thesis shows that multichannel integration quality improves user value, service experience, and electronic word-of-mouth, while both enabling and inhibiting factors influence continued FinTech usage. At the meso-level, the thesis's empirical evidence shows that mobile money adoption positively influences both innovative and financial performance among SMEs. At the macro level, the thesis demonstrated that FinTech contribute to economic growth, particularly when supported by strong institutional quality. The macro-level results also highlighted that learning from discontinued central bank digital currency initiatives contributes to ongoing debates about the implementation challenges of digital currencies and public trust. From a managerial perspective, the thesis offers actionable insights into how financial service providers can enhance user experience, build trust, and sustain adoption through multichannel strategies. From a policy perspective, the study underscores the role of institutional quality, regulatory clarity, and user-centric design in ensuring the resilience and scalability of digital financial systems. This thesis’ findings support Sustainable Development Goals (SDGs), particularly Goals 8 (Decent Work and Economic Growth), 9 (Industry, Innovation and Infrastructure), 10 (Reduced Inequalities), and 11 (Sustainable Cities and Communities). 2026-06-30T00:00:00Z http://lutpub.lut.fi:80/handle/10024/172407 Emotional artificial intelligence via identity-free body language recognition and understanding Li, Deng Emotional Artificial Intelligence (EAI) refers to an Artificial Intelligence (AI) system’s ability to perceive, interpret, and respond to human emotions. EAI is essential for humancentered applications such as healthcare, education, and human–computer interaction. However, mainstream emotion recognition systems depend heavily on identity-sensitive signals including facial expressions, raw speech, and biometric data, raising privacy concerns in practice. This dissertation investigates an alternative paradigm: privacy-preserving emotional artificial intelligence through Identity-Free Body Language (IFBL). Specifically, it pursues four objectives to validate this paradigm: effective IFBL perception, efficient IFBL perception, de-identified emotion dataset construction, and Multimodal Large Language Model (MLLM)-based emotion understanding. First, this dissertation develops an effective IFBL recognition method based on visual-text contrastive learning. The proposed method achieves 66.12% and 65.08% top-1 accuracy on the iMiGUE and SMG benchmarks, respectively, outperforming all previous methods. Ablation studies show that visual-text contrastive learning contributes to a 26.43% improvement over a vision-only baseline. Second, this dissertation proposes the Motion-aware State Fusion Mamba (MSF-Mamba) for efficient IFBL recognition. MSF-Mamba enhances state-space models with a Multiscale Central Frame Difference State Fusion Module that injects motion cues into the state transition. MSF-Mamba achieves state-of-the-art (SoTA) performance on iMiGUE and SMG, surpassing all CNN-, Transformer-, and SSM-based baselines while maintaining high computational efficiency. Third, this dissertation introduces the task of De-identified Multimodal Emotion Recognition and Reasoning (D-MERR) and constructs the corresponding DEEMO dataset. Fourth, this dissertation develops DEEMO-LLaMA, an MLLM that integrates de-identified video, de-identified audio, and transcriptions for emotion recognition and reasoning. DEEMOLLaMA achieves 74.49% accuracy, outperforming the best baseline by 9.75% in accuracy. In summary, these contributions establish a complete research pipeline. This dissertation demonstrates that privacy-preserving emotional artificial intelligence grounded in identity-free body language is both feasible and effective. 2026-08-04T00:00:00Z http://lutpub.lut.fi:80/handle/10024/172343 Effects of electrodes, current density, and dynamic operation on solid carbon production in molten carbonate electrolysis Nur’aini, Anafi The urgency to reduce CO2 emissions has accelerated research into electrochemical conversion of CO2 into valuable products, such as solid carbon, carbon monoxide gas, and hydrocarbons. Investigations into each of these target products are ongoing, with the goals of achieving high production efficiency, high purity, and low production cost. The CO2 electrolysis for solid carbon production involves a cathode and an anode that are immersed in molten salt electrolytes. When a potential difference is applied, the reduction reaction at the cathode produces solid carbon, while oxidation at the anode generates oxygen gas. The carbon product from CO2 electrolysis in molten salt, commonly referred to as electrolytic carbon, exhibits diverse morphologies including graphite, carbon nanotubes (CNTs), spherical-like onion carbon, and amorphous carbon. Metal elements or compounds originating from the cathode and the anode are commonly detected through scanning electron microscopy (SEM) and X-ray diffraction (XRD). Some of these metals are separated from the carbon particles, while others remain on the surface or are incorporated within the carbon particles. Electrolytic carbon has numerous potential applications, such as a lithium-ion battery anode, an adsorbent, and bucky paper. Despite its promising properties, commercialization remains challenging because production costs are significantly higher compared with carbon obtained through conventional methods. In this doctoral dissertation, experimental studies were conducted to investigate the effects of electrode pair selection, applied current density, and dynamic operating conditions using various waveforms during electrolysis on the properties of the solid carbon product. Additionally, the influence of dynamic operation on electrical power consumption during electrolysis was examined. Titanium, nickel, and nickel-based superalloy (Alloy X) were tested as cathode and anode materials in a molten lithium carbonate. Among the electrode pairs evaluated, a titanium cathode with a nickel anode (Ti–Ni) demonstrated the most stable voltage performance, achieving the highest Faraday efficiency of 62% and voltage efficiency of 48%. The SEM analysis of carbon produced using a Ti–Ni pair revealed predominant spherical-like onion structures, whereas tubular and amorphous carbon morphologies were observed when an Alloy X cathode and a titanium anode (Ax–Ti) were employed. Subsequently, using the Ti–Ni electrode pair, current densities ranging from 0.1 to 0.4 Acm−2 were applied. The highest Faraday efficiency of 67% and voltage efficiency of 52% were achieved at 0.1 Acm−2. Notably, the trend of Faraday efficiency did not correlate with the magnitude of applied current density. The SEM analysis indicated that the applied current density had no significant effect on the morphology of the electrolytic carbon, although it slightly influenced the diameter of the spherical onion-like shape of the carbon structure. Furthermore, dynamic operations using triangular, square, sine, ramp-up, and ramp-down waveforms were applied during electrolysis. The SEM analysis revealed that waveform variation had a minimum impact on the carbon morphology, with most products exhibiting spherical-like onion structures, and a small fraction of tubular and hollow sphere structures. An analysis of electrical power consumption during electrolysis under dynamic operation indicated higher electrical power consumption compared with constant DC, due to additional losses introduced by the alternating current component. Ripple loss under dynamic operation increased with amplitude, and among the tested waveforms, the square waveform produced the highest ripple loss. 2026-08-14T00:00:00Z