Mechanical and Mechatronics Engineering

Permanent URI for this collectionhttps://uwspace.uwaterloo.ca/handle/10012/9912

This is the collection for the University of Waterloo's Department of Mechanical and Mechatronics Engineering.

Research outputs are organized by type (eg. Master Thesis, Article, Conference Paper).

Waterloo faculty, students, and staff can contact us or visit the UWSpace guide to learn more about depositing their research.

Browse

Recent Submissions

Now showing 1 - 20 of 1568
  • No Thumbnail Available
    Item
    Microstructure control and property enhancement of NiTi-stainless steel dissimilar joints
    (University of Waterloo, 2025-03-21) Zhang, Kaiping; Zhou, Y. Norman; Peng, Peng
    Dissimilar joining between Nickel-Titanium (NiTi) and stainless steel (SS) is of significance in many areas especially biomedical applications, however, achieving reliable NiTi-SS joints is highly challenging due to the formation of brittle intermetallic compounds (IMCs) in the fusion zone (FZ) or the interface. Two strategies can be summarized to address this issue: (1) restricting the mixing of molten metals and (2) replacing the most harmful Laves (Fe,Cr)2Ti with ductile phases. The former one poses large processing complexity and may lead to NiTi plastic deformation degrading the functional properties. The latter struggles to eliminate brittle IMCs entirely in the FZ and may introduce toxic elements. This research investigated both aspects to control the microstructure and properties of NiTi-SS joints by leveraging the flexibility of laser beam and the thermomechanical process of resistance welding. The combination of laser beam defocus and large offset enabled the laser weld-brazing of NiTi and SS wires. This approach successfully eliminated the IMCs network in the FZ, shifting the conventional and complex FZ brittleness issue to a focus on controlling the brazed interface. Additionally, laser welding mode significantly influenced the macrosegregation and porosity in the FZ of NiTi-SS joints. Low laser power density and long welding time mitigated the macrosegregation and porosity by weakening the laser keyhole effect and prolonging the molten pool duration. In NiTi-SS laser weld FZ, large pores were caused by the instability or collapse of the laser keyhole, while small pores originated from the Ni vaporization. Both IMCs control strategies were investigated in resistance spot welding (RSW) of NiTi and SS for the first time. The use of Nb interlayer resulted in a unique sandwich-structured joint, where two FZs were separated by solid-state Nb, suppressing the mixing of dissimilar molten metals. Nb-containing eutectics formed at both interfaces, enhancing the joint strength with a 38% increase in fracture load and a remarkable 460% increase in energy absorption. In another approach, increasing Ni concentration via a melted Ni interlayer effectively replaced Fe2Ti with relatively ductile Ni3Ti in the FZ. However, high Ni content also induced large pores and cracks, limiting the effectiveness of this strategy in NiTi-SS RSW. A novel processing approach leveraging interfacial liquid control was proposed, achieving a solid-state joined interface in NiTi-SS fusion welding (e.g., resistance microwelding) without any additional interlayers. The produced NiTi-SS joints showed superior strength, superelasticity and corrosion resistance compared to NiTi joints or base metal. The ultrathin reaction layer at the solid-state joined interface contributed to a strong metallurgical bonding, while Joule heating effects and interfacial reactions enhanced superelasticity and corrosion resistance of the joint. Notably, a face-centered-cubic (FCC) reorientated layer (ROL) was found between SS and IMC layer at the controlled ultrathin interface. The formation of this ROL was uncovered based on an epitaxial growth model. This ROL introduced a strong crystallographic mismatch with the textured SS, resulting in the fracture at this interface. These phenomenal findings offer valuable insights for studying material interface and controlling dissimilar-metal welding process.
  • No Thumbnail Available
    Item
    Influence of Absorbency and Additives on Performance of Battery-Free IoT Water Leak Sensors
    (University of Waterloo, 2025-03-19) MacGregor, Oluwadamilola Solomon; Zhou, Norman
    Leak detection is a reliable solution for controlling the potentially destructive outflow and wastage of water. Several types of devices are used in domestic and industrial spaces; however, most have their power sources run out, and thus require battery change. The associated costs add to overhead expenditure of the user. This necessitates the use of leak detectors that are self-powered, having no use for external sources of power. Integrating water leak detection systems with Internet of Things (IoT) technology such as Bluetooth low energy (BLE) and long-range (LoRa) protocols provides advantages such as real-time monitoring, which informs incidents and ultimately saves huge cost. The use of IoT-enabled sensors and cloud-based data analytics offers pre-emptive control mechanisms for prompt identification and containment of localized leaks. This helps reduce wastage of water and damage to property, both of which reduce costs as remote access through IoT networks guarantee instant notifications for preventative measures. Scalability fosters effortless deployment in residential, commercial, and industrial environments. In a self-powered IoT water leak device, parameters such as capillary action and electrochemical reactions directly impact power generation and beacon activation. Energy generation and harvesting happen as water interacts with active materials within the sensor device. There must be a cathode and an anode, to interact with the leaking water which would be the electrolyte. Therefore, the materials selected to play such roles in the device are crucial for the desirable chemical interactions, once in contact with the leaking water. In a water leak detector where the most crucial feature is sensitivity to water, capillary action is one of the most significant parameters to consider. Both the design of the sensor casing and channels through which the water travels, are to foster a seamless flow. Also, within the sensor chamber, each material in the stack must demonstrate capillarity. Therefore, porosity is key, as their pore sizes determine what material passes through and what might otherwise be trapped to impede the flow of the water being transmitted. Therefore, capillary action is explored for absorbent materials and the sensor casing. Both filter paper (FP) and fabric materials are examined, to ascertain which one gives optimally combined advantages for absorbency and repeatability. FP showed superior performance, due to its pore size. This advantage becomes particularly useful where additives are considered for the powder mixture. Without additives, the stacked materials have only water to interact with. While this is sufficient to power BLE, it is not enough for LoRa technologies which require higher power. To account for this, additives can be included in the materials within the sensor stack. Salts are among such additives that can provide active ions when interacting with water. Subsequently, these ions facilitate electricity generation due to increased current. Therefore, the power output of the device can be increased when additives are introduced. In previous similar works, it was shown that pure materials without any additives produce an open-circuit voltage (OCV) of 2 V and short-circuit current (SCC) of only 10 mA. This combination was able to power the sensor for beacon activation through 7 cycles of wetting-drying rounds of repeatability, but only for the BLE protocol. To solve for this limitation, NaCl was added in varied proportions. 10 wt.% NaCl was found to outperform other samples. After several rounds of repeatability, the values of current and voltage were observed to diminish. A sensor without NaCl typically lasts 7 rounds of repeatability, sensors containing NaCl last only about 3 rounds. The primary concern with the use of such additives may be an imminent trade-off between the increased power generation and possible corrosion which compromises shelf life. One of the downsides of using additives to enhance power generation is the corrosion of metallic materials in the sensor. To study the effect of NaCl on the corrosion of the metallic material, and thus the shelf life of the sensor, electrochemical corrosion tests were performed. As expected, it was observed that higher salt content resulted in higher corrosion rate. Therefore, repeatability was significantly reduced in higher salt contents, thereby limiting the overall shelf life of the sensor. Ultimately, the use of salts should be limited and be specific to the target use case.
  • No Thumbnail Available
    Item
    Data-Based Modeling of Electrochemical Energy and Thermal Systems: Fuel Cell and Lithium-Ion Battery
    (University of Waterloo, 2025-01-24) Legala, Adithya; Li, Xianguo
    As a solution to combat climate change and environmental pollution, electrochemical energy systems such as Proton Exchange Membrane Fuel Cell (PEMFC) and Lithium-Ion Battery (LIB) are being developed as the replacement for fossil fuel-powered combustion engines, especially for ground transportation and aviation applications. These electrochemical energy systems must be able to operate independently and in conjunction with each other by complementing their advantages and limitations, such as efficiency, range, thermal behavior, aging, and operating environment. This interoperability requires accurate real-time computational models to control, diagnose, and adapt according to field requirements. A typical electrochemical energy system model needs to incorporate effects related to reactant concentrations, system overpotentials, thermodynamics, porous media mechanics, membrane dynamics, gas diffusion, electrode degradation, electrolyte status, ion transport, and chemical kinetics across various operating conditions, all of which result in complex interactions affecting the accuracy and reliability of the system. Today, both PEMFC and LIB use complex computational physics-based fluid dynamics models in the product development phase, which requires enormous computational power and long lead times for iterative prototype improvements. On the other hand, both PEMFC and LIB rely on simple lookup tables and semi-empirical equations as plant models that require intensive calibration activity to determine the mode of control and diagnosis for automotive applications. However, considering the present-day automotive propulsion systems, which operate in widely varied applications and geographic locations and have short product development cycles, these approaches are not able to comprehend the complexities, hindering the ability of these systems to operate at their full potential and leading to catastrophic failures (e.g., Thermal runaway). Data-based modeling techniques are one of the potential solutions, which is quite in contrast with other empirical or physics-based models where the entire input-output relations of the model are established primarily based on the data. Data-based models use aspects of statistics, probability, and network architecture, avoiding the complexities of physics-based models and intensive calibration, providing better accuracy in most cases, primarily where the complex mechanisms can’t be modeled using specific governing equations, and fast, efficient computation with much less computational resource requirement. This thesis focuses on data acquisition (identifying and collecting the relevant data) and data-based model development by incorporating machine learning algorithms and regressors to predict the system's performance, thermal behavior, aging, and faults in real-time (on-board diagnostics). Data for these models is acquired through two approaches: experimentation by utilizing Fuel Cell and Green Energy Lab facilities such as the Automated Battery Test Station (ABTS), G20 fuel cell automated test station, and by partnering with the relevant industry. In the second approach, data is generated by simulation of physics-based models (CFD, Semi-empirical, equivalent circuit models) that are experimentally validated in the literature and developed within the research groups of UWaterloo. Development of a data-based model includes the identification of feature vectors (inputs), prediction attributes (outputs), state estimates (internal parameters), non-linearity of the systems, correlation factors of various system entities, and application of machine learning techniques such as feed-forward artificial neural network, support vector machine classifier - regressor, along with their respective adaptations and calibration processes. The primary objectives of this study are to develop data-based models for three main application areas: (i) Prediction of PEMFC performance, internal states of the membrane, cell voltage degradation, and system outputs. (ii) Prediction of LIB heat release rate during discharge and thermal dynamics of an open system during an exothermic reaction. (iii) Prediction of fuel cell battery hybrid electric vehicle’s system dynamics and thermal behavior. During this study, various data-based models were developed to tackle the problems encountered in fuel cell-battery hybrid systems, such as predicting the fuel cell performance, fuel cell voltage degradation, PEMFC membrane dynamics, lithium-ion battery thermal dynamics, thermal behavior during exothermic reactions and dynamics of fuel-cell battery hybrid system. The results presented in this study proved the data-based model’s applicability in surrogate modeling, real-time system monitoring, controls, and diagnostics of electrochemical energy systems both at the component level and system level. Additionally, the results implicate that the data-based model can serve as a complement and alternative to the traditional computational fluid dynamics models as well as complex physics-based and empirical models to predict thermal gradients and system internal states during multifaceted reactions.
  • No Thumbnail Available
    Item
    Modulation Strategies of Cu-based electrocatalysts for Enhancing Electrocatalytic CO2 Conversion
    (University of Waterloo, 2025-01-23) Wang, Lei; Wu, Yimin; Tan, Zhongchao
    Electrocatalytic CO2 reduction (ECR) into value-added chemicals and fuels using renewable energy contributes to global decarbonization, offering an elegant solution for achieving carbon neutrality and fostering sustainable development of human society. However, this strategy highly relies on the rational design of catalysts to enhance product selectivity and activity. To advance CO2 conversion technology, systematic and comprehensive studies on ECR are urgently needed to demonstrate the origins of catalytic activity, elucidate the relationship between structural defects of catalysts and catalytic activity, and reveal the dynamic evolution of active sites under ECR reaction conditions. In this thesis, mechanistic studies and functional catalyst design are extended from lab scale to large scale. The regulation of grain boundaries structures and local microenvironments is employed to stabilize oxidized copper species, thereby enhancing the selective production of desired products. Firstly, at the lab scale, we introduce oxidation and alloying strategies into grain boundaries systems. Low-loading Ag and water oxidation induce oxygen enrichment at the grain boundaries, leading to a grain boundary oxidation effect. In situ characterizations indicate that the grain boundaries and grain boundary oxidation effects contribute to strengthening resistance of the oxidative Cuδ+ species to the electrochemical reduction. Experimental and theoretical results demonstrate that in intricate grain boundaries assemblies, the oxidation state of copper plays a crucial role in the C2+ product pathway, while the nanoalloy effect tends to the formation of CH4 product. Secondly, to achieve the industrial-scale ECR to multi-carbon products with high selectivity using membrane electrode assembly (MEA) electrolyzers, we introduce activated carbon black with different functional groups to modulate the interfacial microenvironment of Cu nanoparticles, enhancing CO coverage to suppress hydrogen evolution reaction (HER). In situ multimodal characterizations consistently reveal that in situ generated strongly oxidative hydroxyl radicals can create a locally oxidative microenvironment on the catalyst surface, stabilizing the Cuδ+ species and leading to an irreversible and asynchronous change in morphology and valence, yielding high-curvature nanowhiskers. The well-stabilized Cuδ+-OH species serve as active sites during MEA testing. By comprehending this mechanism, we achieve selective ethylene production with a Faradaic efficiency (FE) of 55.6% for C2H4 at a current density of 316 mA cm-2. The insight of these reaction mechanisms bridges the gap between lab-scale studies and industrial-scale implementation, contributing to the development of sustainable and carbon-neutral industries.
  • No Thumbnail Available
    Item
    Assessing Cervical Spine Response to Head-First Impact Using Vertebral Segments, Head-Neck, and Full-Body Computational Human Models
    (University of Waterloo, 2025-01-23) Morgan, Maxwell; Cronin, Duane
    Head-first impact (HFI), which can occur in automotive rollovers and sports collisions, is associated with a high risk of cervical spine injury. Cervical spine injuries from HFI such as fracture-dislocations frequently lead to severe spinal cord injuries and in some cases death, as reported in field data and epidemiology. Experimentally, isolated motion segments have been tested in compression and bending to mimic loads incurred in HFI, while cadaveric head and necks with torso surrogate masses (TSMs) and full body (FB) cadavers have been inverted and dropped to investigate HFI. However, isolated segment tests are limited in producing the complex kinematics of HFI, and TSM response has not been quantified with respect to FB testing. Recently, computational human body models (HBMs) have been developed to simulate humans in injurious loading conditions, but have only seen limited application in HFI. In this study, computational models were applied to investigate HFI, using an isolated vertebral segment model, an isolated head and neck with a TSM, and a contemporary FB human model. First, an existing and validated cervical motion segment model was loaded in combined compression and flexion relevant to HFI, to investigate the loads and moments associated with fracture-dislocation failures. Next, TSM and FB head-first impacts were modelled using a contemporary HBM in three postures (flexed, neutral, extended) at three impact velocities. Finally, the FB model was compared with a unique set of experimental full body cadaver HFI tests. In isolated segment loading, combined compression and flexion produced hard tissue failure patterns reported in fracture-dislocations. Fracture-dislocation was achieved by simultaneously rotating and translating the superior vertebra anteriorly. Comparing the isolated head and neck TSM and FB models, the TSM condition demonstrated higher neck forces, internal energy, and a larger volume of hard tissue failure compared to the FB models under the same impact conditions. Despite similar head contact forces between TSM and FB, the compliant thorax of the FB model reduced the neck forces by half, which significantly reduced corresponding energy stored in the neck tissues. The neutral and extended neck postures predicted higher neck forces due to facet joints engaging, while neck flexion in the flexed posture reduced neck forces by misaligning the spine from the impact. Finally, it was found that the FB model had similar head impact forces and comparable T1 rotation to cadaveric HFI experiments, but a high sensitivity to initial posture was identified. This study identified forces and moments that can create a fracture-dislocation in a motion segment using prescribed boundary conditions. The TSM and FB simulations demonstrated compression loads and moments of a similar magnitude to the motion segment, but differed in timing, generating higher axial loads leading to the onset of fracture in the spine. The neck loads were higher using the TSM boundary condition compared to the FB condition. Both TSM and FB models identified the importance of neck posture on response, showing that an initially extended neck posture leads to higher neck forces compared to a flexed posture. This study identified the importance of full body boundary conditions for the simulation of HFI, the complex dependency of kinematic and kinetic response on neck posture, providing model results in agreement with the small number of full body experiments. Further experimentation was recommended to provide detailed measurements necessary for model assessment and validation. Future computational studies will integrate the motion segment and FB results to improve the understanding of fracture-dislocation.
  • Thumbnail Image
    Item
    Modeling Ablation in Al/CuO Nanothermite Pellet Combustion
    (University of Waterloo, 2025-01-22) Mondegari, Mohsen; Hickey, Jean-Pierre
    Nanothermites are reactive materials composed of metal and metal oxide nanoparticles, engineered to produce rapid exothermic reactions with high energy density. Aluminum and copper oxide (Al/CuO) are widely used due to their strong reactivity and ability to achieve efficient combustion, making them ideal for applications in energetic materials. This thesis investigates the combustion of Al/CuO nanothermite pellets, with a particular focus on ablation—mass loss due to thermal degradation and chemical reactions. A numerical model is developed to capture key combustion characteristics, including flame speed, pressure distribution, and temperature response, while accounting for both thermal and mechanical effects across varying packing densities. Leveraging the Porous-material Analysis Toolbox based on OpenFOAM (PATO), this model simulates complex reactions, heat flux, and ablation dynamics, thereby addressing existing gaps in understanding ablation effects on nanothermite combustion and enhancing predictive capabilities for these materials. In its simulation methodology, this study adapts PATO to model multiphase reactive materials, formulating governing equations for mass, momentum, and energy conservation. A two-dimensional axisymmetric model was selected to represent the cylindrical pellet structure. Key parameters, including material porosity, permeability, specific heat, and thermal conductivity, were tailored to the properties of nanothermite materials, while distinct boundary conditions were applied to simulate ignition and ablation phases. Simulations were conducted across various packing densities, with some models incorporating ablation-specific boundary conditions to capture changes in flame speed, peak pressure, and pellet stability under different conditions. Results indicate that ablation intensity significantly influences the combustion dynamics, with higher intensities leading to reduced flame speeds and peak pressures. The study's findings highlight the potential for controlled nanothermite combustion by optimizing pellet packing density and ablation characteristics, offering applications in propulsion and micro-energetics.
  • Thumbnail Image
    Item
    Learning Agent-based Model Predictive Controllers for Holistic Vehicle Control
    (University of Waterloo, 2025-01-22) Zhong, Jiaming; Khajepour, Amir; Pant, Yash Vardhan
    Holistic vehicle control (HVC) is an advanced, integrated approach that optimizes a vehicle’s mobility, stability, and safety by coordinating all available control systems. As the automotive industry progresses towards electrification and increasing intelligence, functional integration has emerged as a dominant trend in vehicle control systems. This evolution necessitates the simultaneous coordination of multiple controllers to achieve diverse objectives. The growing demand for flexibility and reliability in automotive systems has given rise to a “plug-and-play” paradigm in control system design. This approach, while beneficial, poses significant challenges for traditional “all-in-one” integrated control methods, such as integrated model predictive control (MPC). Distributed control schemes have demonstrated greater scalability and robustness compared to integrated schemes, especially in applications such as vehicle control systems, where managing complex, multi-agent dynamics is important. A recently proposed prominent approach within this framework is agent-based model predictive control (AMPC), where controllers are treated as interactive agents and are coordinated together to achieve a commonobjective iteratively, taking advantage of the distributed control structure. However, in practice, two critical pain points arise: a) Uncertain contributions from unknown controllers: The optimal control performance from the AMPC highly depends on the prediction accuracy, which requires all agents or their contributions to be accurately known. This requirement is often too idealistic for practical implementation. For example, when a third party develops a “black-box” controller, its underlying algorithm is unknown, and there is no specified interface to know its contribution to vehicle dynamics. This lack of information is likely to cause a significant error in predicting vehicle behaviour, leading to unexpected or even harmful control results. b) Limitations on controller-oriented decomposition: Decomposing from the objective’s perspective is a more practical and ideal approach in the development of function-oriented or feature-oriented automotive control systems. However, AMPC cannot decompose coupled objectives with shared agents because it is designed only to decompose the integrated system from the controller’s perspective, not from the objectives. The objective in AMPC is usually implemented as a weighted sum of multiple goals, which greatly limits the flexibility of control system design. This thesis is hence motivated to overcome these two pain points through practical solutions using data-driven machine-learning techniques and flexible distributed schemes for multi-agent-multi-objective (MAMO) control systems. For the first pain point, this thesis proposes a practical hybrid control scheme: learning agent-based MPC (L-AMPC). This scheme combines the model-based AMPC approach with data-driven learning methods to improve the control performance for multi-agent systems. The Gaussian process regression (GPR) enhanced by an online data management strategy serves as the learning core to predict unknown contributions along the prediction horizon, completing the system model in the MPC for more accurate control. Meanwhile, a stochastic framework is formulated to guarantee control safety and feasibility using soft chance constraints based on the prediction variance. The proposed hybrid control scheme is efficient for real-time implementation and is flexible to any control agent topology. For the second pain point, this thesis proposes a distributed control scheme: multiobjective AMPC (MO-AMPC). This scheme adapts the alternating direction method of multipliers (ADMM) into a general control strategy that achieves global optimization while decoupling objectives. Three formulations that can maintain convergence while addressing control regularization and inequality constraints are systematically developed. The convergence and computational efficiency of the proposed methods are verified and compared on two vehicle control scenarios with multi-objective configurations. Furthermore, this thesis proposes a data-driven distributed control scheme based on the MO-AMPC: learning multi-objective AMPC (L-MO-AMPC). To accelerate the converging process of MO-AMPC, a learning-based initialization method for iterations is proposed. The proposed scheme is compared with the MO-AMPC scheme through a path-tracking simulation using various controllers. The results show that the L-MO-AMPC scheme achieves similar control performance while significantly reducing computational time. All proposed controllers (L-AMPC, MO-AMPC and L-MO-AMPC) are verified by real-time simulations respectively. In addition, the effectiveness and performance of L-AMPC and MO-AMPC are also verified in real vehicle experiments. The results from simulations and experiments could be concluded as follows: • Compared to AMPC, the proposed L-AMPC can achieve higher tracking performance in well-learned scenarios with the learning capability and always guarantee constraint satisfaction even in less-learned scenarios. • To decompose the MAMO system from the objective’s perspective, the proposed MO-AMPC achieves the same global optimum as the corresponding integrated MPC with greater flexibility and can potentially reduce the computational cost. • Compared to model-based MO-AMPC, the proposed L-MO-AMPC has significantly higher computational efficiency while still retaining the property of converging to the global optimum, making it well-suited for real-time implementation.
  • Thumbnail Image
    Item
    Data-Driven Simulation and Optimization of Renewable Energy Systems
    (University of Waterloo, 2025-01-20) Ye, Wenrui; Wen, John; Nathwani, Jatin
    The transition to renewable energy systems is critical in mitigating climate change and reducing fossil fuel dependence. However, integrating these variable and intermittent sources into the existing grid raises challenges such as dynamic energy demand management and resource underutilization, leading to increased operational costs and hindering broader adoption. This thesis develops algorithms to optimize renewable energy systems, enhancing their integration and operational efficiency. The research makes a significant contribution to enhancing the utilization, reliability, and economic viability of renewable energy systems, supporting a smoother transition to sustainable energy practices. This thesis first enhances the energy generation of photovoltaic panels by optimizing their tilt angles to maximize solar energy capture under varying environmental conditions. The machine learning models provided accurate predictions of photovoltaic output, allowing for data-driven insights into optimal system performance. A subsequent optimization process identified the best tilt angles during the operation. The results demonstrated an increase in annual energy output by up to 9.7% compared to fixed-tilt systems. This confirms that dynamic tilt adjustment is an effective strategy for maximizing photovoltaic energy generation. The second part of the research focuses on optimizing the capacity of renewable energy system components, with a particular emphasis on energy storage systems such as batteries. This project addressed the challenge of determining the optimal capacity for each component to efficiently meet energy demands while minimizing costs. A two-stage optimization approach was applied: first, a genetic algorithm generated candidate configurations with specific capacities; second, a simulation was conducted using an energy management algorithm to evaluate the performance of these configurations. The optimized configurations led to an overall system energy independence score of 0.51 and an 18.12% higher internal rate of return, validating the effectiveness of the integrated optimization approach in capacity planning and highlighting the importance of appropriately sized energy storage in enhancing system performance. The final part introduces an advanced energy management algorithm inspired by Model Predictive Control, integrating both batteries and hydrogen storage to enhance renewable energy utilization. By employing time series data and transformer-based models, the system accurately predicts future energy demand. These predictions enable a rolling window optimization technique that utilizes machine learning for dynamic energy management. The inclusion of hydrogen storage allows excess renewable energy to be stored as hydrogen, providing a versatile energy carrier for applications beyond electricity and improving overall renewable energy utilization. This approach improved demand forecasting accuracy by 41.21% and increased the adjusted green hydrogen production rate from 29.54% to 54.3%, demonstrating that advanced predictive energy management strategies, combined with diverse energy storage solutions, significantly enhance system adaptability, efficiency, and renewable energy utilization. These studies demonstrate that optimizing component configurations and energy management strategies—while integrating advanced energy storage systems like batteries and hydrogen—substantially improves the efficiency, reliability, and economic viability of renewable energy systems. The research provides valuable insights for integrating renewable energy into existing grids and supports the transition toward more sustainable and resilient energy infrastructures.
  • Thumbnail Image
    Item
    AI-Assisted Ultrasound-Guided High-Intensity Focused Ultrasound (USgHIFU) in Non-Invasive Surgery
    (University of Waterloo, 2025-01-13) Lari, Salman; Kwon, Hyock Ju; Kim, Jong Uk
    This comprehensive study combines several innovative approaches to enhance the precision and efficacy of high-intensity focused ultrasound (HIFU) for cancer treatment. HIFU, a non-invasive therapeutic technique, uses high-frequency ultrasound to ablate tumors, but requires careful planning due to its potential for collateral damage to healthy tissues. To overcome these challenges, multiple methodologies are introduced. By employing a Physics-Informed Neural Network (PINN) integrated with a realistic breast model, breast tumors are targeted with high precision. The model utilizes a bowlshaped acoustic transducer to focus ultrasound waves directly on the tumor, achieving intense localized heating. The PINN method efficiently solves partial differential equations in a mesh-free domain, providing high accuracy with significantly lower computational demands than traditional finite element methods (FEM). This model is employed to understand the governing dynamics of the HIFU process, particularly the heat transfer mechanisms during sonication. Using machine learning techniques, the model simulates the absorption mechanism and temperature rise, validated through ex vivo experiments with bovine liver. This helps in accurately predicting and visualizing the effects of treatment, facilitating the development of personalized treatment strategies. A novel deep learning-based optimization procedure is used for preoperative treatment planning. It determines optimal focal locations and sonication times for each ablation session, ensuring minimal tissue over- or under-treatment. This algorithm has shown high potential in handling various HIFU presurgical plans and is especially effective in creating precise treatment plans based on patient-specific material properties. Addressing the limitations of manual HIFU operations, a real-time, low-cost image segmentation framework based on the Swin-Unet architecture is proposed. This system is trained and tested on B-mode imaging simulations and real images from HIFU-treated chicken breast, demonstrating high efficiency in lesion segmentation and offering potential for monitored, automated HIFU therapy. By integrating these diverse approaches, the study not only enhances the thermal effects of HIFU but also offers a framework for safer, more precise, and individually tailored cancer treatments. The combined use of advanced simulation models, deep learning approaches, and innovative technologies aims to significantly improve the therapeutic outcomes of HIFU, making it a more viable option for cancer treatment with reduced risks and enhanced efficacy.
  • Thumbnail Image
    Item
    Integrated Investigations of Lumbar Spine Biomechanics, Implant Fixation, and Design for Additive Manufacturing under Physiologic Loading
    (University of Waterloo, 2025-01-09) McGregor, Martine; McLachlin, Stewart
    Spinal implant loosening and migration are significant concerns following the surgical treatment of musculoskeletal spinal disorders because of the complex and highly variable loading experienced in the spine. During post-operative rehabilitation, the mechanical interface between the vertebral bone and metallic implant is subjected to variable, combined (multi-axial) loading modes during activities of daily living. However, most current methods for pre-clinical evaluations of spinal implants focus on application of simplified loading conditions (e.g. uniaxial toggle, static compression or pull-out tests) which do not fully represent physiologic multi-axial loading. As a result, there is a lack of biomechanical understanding in how spinal implants can fail following surgery. One of the goals of this thesis is to better understand mechanisms of clinical implant failures by improving pre-clinical spinal implant testing methodologies through the adoption of physiologically-derived loading conditions. Spinal implants are often manufactured using titanium alloys like Ti6Al4V, beneficial for their superior osteogenic qualities. However, the mismatch in mechanical moduli between titanium and bone is a known cause of problematic stress shielding and failure of the bone-implant interface. To address this challenge, laser powder bed fusion (LPBF) additive manufacturing allows for the production of unique metamaterials that can act to reduce the bulk modulus of titanium implants, bringing mechanical properties closer to that of human bone. However, there is currently a disconnect between the understanding of additive manufacturing (AM) and the mechanical properties of bone. Therefore, a second goal of this thesis was to develop design for additive manufacturing (DfAM) tools for selecting lattice properties to best match vertebral bone properties while ensuring manufacturability for LPBF. Based on the two overall goals, the thesis research was defined based on the intersection of three themes: (1) multiaxial spinal loading, (2) spinal implant evaluation, and (3) LPBF AM. Within these themes, four in vitro biomechanical human cadaver investigations were completed to develop physiologically-derived, multi-axial loading test protocols for a commercially available joint motion simulator to evaluate the impact of multi-axial loading on implant loosening and migration. The first study examined the effect of uniaxial and multiaxial cyclic loading on pedicle screw loosening in osteoporotic human bone (n=7). It was found that multiaxial loading significantly accelerated pedicle screw loosening as well as increased deformation volume at the bone-implant interface. The second (n=8) and third (n=8) biomechanical studies focused on developing six degree-of-freedom (6DOF) test protocols for spinal loading simulation. Lumbar spine stability was compared between existing pure moment test methods against physiologically-derived 6DOF loading waveforms associated with activities of daily living. It was found that the commercial joint motion simulator was reliable for applying complex, physiologically-derived lumbar spinal loads. The loads associated with in vivo spinal movements resulted in reduced range of motion compared to application of pure moments. Longitudinal gait simulation also had little impact on lumbar spine segment biomechanics. The fourth in vitro investigation (n=8) was aimed at leveraging the developed protocols to evaluate implant migration of an LPBF additively manufactured spinal implant. It was found that cage penetration into the adjacent vertebra (failure of the bone-implant interface) happened most often during simulated gait testing, while cage migration, towards the spinal cord, occurred primarily during flexion-extension. Two additional DfAM investigations were undertaken to develop tools for selecting Ti6Al4V lattices for orthopaedic applications and evaluating manufacturability of lattice designs. A comprehensive literature review was undertaken to synthesize a dataset relating lattice parameters to compressive mechanical properties. Gibson-Ashby plots were generated to relate this dataset to human bone properties for selection of lattice parameters in orthopaedic implant designs. A manufacturability evaluation was also completed to determine which lattice structures were considered manufacturable based on geometric and defect analysis. It was found that surface-based lattice structures were the most manufacturable due to the high interconnectivity of the down-skin regions. Lastly, these DfAM tools were used to develop novel lumbar interbody cage implants, which were investigated under the physiologically-derived loading (described above). Collectively, the research completed in this thesis provides new biomechanical understanding of the lumbar spine and spinal implants under multi-axial cyclic loading. With improved knowledge of how spinal implant failures can occur under multi-axial loading, pre-clinical testing standards can now be improved to help avoid these failure mechanisms, ultimately leading to the hope of improved clinical outcomes following spinal surgery. Further, new DfAM tools were created and evaluated providing new avenues for the use of emergent LPBF AM techniques for orthopaedic applications.
  • Thumbnail Image
    Item
    Detection of Cervical Spine Vertebrae on MRI towards Improving Multi-modal Image Registration
    (University of Waterloo, 2025-01-09) Chu, Jonathan Ho-Yin; McLachlin, Stewart; Wong, Alexander
    Multi-modal image registration, such as MRI to CT, is an important but often challenging aspect for clinical image analysis. It has applications in treatment planning requiring image fusion, or inter-subject, atlas-based analyses, as well as longitudinal analyses. Multi-modal image registration in the cervical spine presents extra challenges because of the variability in the field of view (FoV) of magnetic resonance imaging (MRI) of the spinal column between different image series, with cervical vertebrae having a similar appearance leading to many local registration minima. This thesis explores methods to detect, localize, and label cervical vertebrae and the spinal cord in anatomical MRI, focusing on the application of deep learning techniques. Specifically, we generated a custom annotated dataset of the cervical spine MRI, based on the Spine Generic dataset [1], resulting in 149 T1w and 100 T2w labelled images. We then successfully trained a Mask R-CNN model [2] and utilized a weighted directed acyclic graph (DAG) to leverage the sequential hierarchy of the vertebrae to filter detections for the cervical vertebrae and spinal cord detection task. This resulted in state-of-the-art performance, where the model was robust to varying cervical spine FoVs. Lastly, we integrated the detector model into a multi-sequence and multi-modal deformable image registration pipeline, where the inference results were used to crop images to an appropriate FoV and seed initial alignment, prior to deformable registration. The multi-sequence pipeline utilized the generated custom dataset and successfully demonstrated the use of a trained cervical vertebrae detector for FoV cropping prior to affine and deformable registration. The multi-modal pipeline was created, adopting from the multi-sequence pipeline, but with an additional CT vertebrae detection branch and utilized affine registration prior to FoV cropping. It was evaluated utilizing a prospectively maintained clinical dataset (SpineMets) provided by collaborators at Sunnybrook Research Institute, containing treatment planning CT and MRI scans. The pipeline demonstrated the ability to successfully perform deformable registration on multi-modal imaging. However, performance on this clinical dataset was limited by the Mask R-CNN model. The methodology developed in this thesis demonstrates the potential of deep learning object detection models combined with leveraging vertebrae hierarchy to provide robust detections of vertebrae in the cervical spine. In addition, we illustrated the application of these models to determine vertebrae FoVs towards improving multi-modal deformable image registration. By leveraging these techniques, this research can be applied to automate and increase efficiency of cervical spine image registration for clinical needs, resulting in reduced cost and burden on clinical experts.
  • Thumbnail Image
    Item
    Graph-Based Autonomous Vehicle Motion Planning Using Game Theory
    (University of Waterloo, 2025-01-07) Panahandeh, Pouya; Khajepour, Amir; Fidan, Baris
    Autonomous driving technologies promise safer, more efficient, environmentally friendly, and accessible mobility systems. To realize these benefits, advanced planning and control algorithms are crucial. Given the interactions between Autonomous Vehicles (AVs) and human-driven vehicles in mixed traffic, as well as with pedestrians and cyclists, the deci- sions and trajectories of AVs significantly impact other road users. Therefore, considering these interactions is vital for achieving safe and efficient AV driving behavior. In automated driving, the basic idea is to traverse from point A to point B au- tonomously. This state space is often represented as an occupancy grid or lattice that depicts where objects are in the environment. From the planning point of view, a path can be set by implementing graph-based algorithms that visit different states in the grid, solving the path-planning problem. The graph-based algorithms treat the static and dy- namic objects/actors detected by the perception system as static impassable areas inside their costmap and fail to capture future actions. Considering the intention of road users results in a more reliable path and control for the AV to follow. Game theory is a framework for addressing problems involving multiple decision-making agents, where each agent’s decision is influenced by the choices of others. The appropriate solution depends on the game’s structure and the relationships between players. In this work, we explore Nash, Stackelberg, and Bayesian equilibria as solutions to the interactions between AV and road users. Nash equilibrium ensures that all participants in the game are treated equally, such that no player can reduce their cost by changing their strategy unilaterally. Stackelberg equilibrium considers a hierarchical structure, where leaders and followers exist among the players. Leaders commit to a strategy first, and followers react optimally to this decision. Bayesian equilibrium incorporates uncertainty and incomplete information, where players have beliefs about the types and strategies of other players. By investigating these equilibrium concepts, we aim to develop optimal strategies for both AV and road users, ensuring effective and efficient motion planning for AV. In this research, a graph-based algorithm will be integrated with game theory to plan the motion of AV, considering the future actions and decisions of other road users. Numerical and experimental simulation results demonstrate that the proposed framework effectively manages interactions between AV and other road users, such as human-driven vehicles or pedestrians, across various scenarios.
  • Thumbnail Image
    Item
    Real-Time Stairs Detection and Terrain Classification: Evaluating LiDAR Sensor Performance
    (University of Waterloo, 2025-01-07) Saha, Shovon Sudan; Tung, James; Hashemi, Ehsan
    The growing use of wearable and autonomous systems in dynamic environments requires reliable terrain classification and obstacle detection to ensure user safety, adaptability to environmental conditions, and optimize system performance. This thesis aims to develop a complete real-time processing pipeline for terrain classification and stairs detection using LiDAR sensors, focusing on the comparative performance of both low- and high-accuracy sensors: the Cyglidar D1 and RoboSense RS-LiDAR-M1 v2. The research introduces a novel end-to-end pipeline that integrates innovative feature extraction methods with efficient classification algorithms, optimized for real-time processing of 3D point clouds. The study utilizes Support Vector Machines (SVM) with linear and non-linear kernels to classify terrain or stairs, deploying a combination of wearable and tripod-mounted configurations to collect data. These models were implemented on the Jetson Nano, an embedded platform, to evaluate real-time feasibility. Key techniques included normal vector extraction as part of the feature extraction process, noise simulation to account for lower-cost sensor conditions, and comprehensive real-time performance analysis, enabling robust classification across diverse scenarios. The results indicated that while the Cyglidar D1 sensor achieved a commendable F1 score of 0.96 in stairs detection, its performance in terrain classification was hindered by noise and overlapping features, particularly between plain and grassy terrains. On the other hand, the RoboSense RS-LiDAR-M1 v2 excelled, achieving an F1 score of 0.99 in terrain classification with minimal processing delay. The system proved robust against noise, maintaining an F1 score of 0.99 even with 10 mm of added Gaussian noise. These findings validate the effectiveness of the proposed pipeline, demonstrating its ability to balance sensor cost, noise tolerance, and algorithmic efficiency while delivering robust performance in real-time applications. The research provides valuable insights into the trade-offs between sensor accuracy and cost, contributing to advancements in LiDAR-based terrain classification and obstacle detection systems. Moreover, it underscores the importance of preprocessing and sensor selection in enhancing classification reliability in real-world scenarios.
  • Thumbnail Image
    Item
    A Study on High-Frequency Induction Welding of TRIP 690 Tubes using Mechanical Tests and Computer Simulations
    (University of Waterloo, 2025-01-02) Okoroafor, Sydney; Biro, Elliot
    The High-Frequency Induction Welding (HFIW) process is increasingly being adopted for manufacturing tubes used in hydroformed Advanced High Strength Steel (AHSS) components, specifically TRIP 690, in the automotive industry. This trend is driven by strict government climate legislation on automobiles that promotes the development of lightweight materials (AHSS) and efficient manufacturing techniques in the industry. Despite its advantages, the HFIW of TRIP 690 faces significant challenges, particularly the recurring issue of oxide inclusion defects. These defects are often undetectable by conventional tube mill inspection technologies and can only be identified through destructive mechanical testing. These defects also lead to poor mechanical properties and potential failures during complex loading scenarios like tube hydroforming. These oxide inclusion defects have not been explored in literature, leading to a critical knowledge gap in the HFIW of TRIP 690 that reduces the yield of high-quality TRIP 690 tubes during the HFIW process. This research aims to bridge the knowledge gaps associated with the HFIW of TRIP 690. It first investigates the influence of welding parameters and oxide inclusions on the mechanical properties of TRIP 690 tubes. Key findings indicate that the Ring Hoop Tensile (RHT) test yields reliable mechanical property data, revealing notable discrepancies when compared to traditional flat sheet data. The study also establishes that welding power and speed significantly affect Ultimate Tensile Strength (UTS), Uniform Elongation (UE), and fracture toughness. Optimal operating regions are identified through mechanical properties-process mapping, linking these properties to the mechanical properties-thermal process map for heat input and temperatures at the vee apex experienced in the vee region. Furthermore, the presence of oxide inclusions is shown to detrimentally impact mechanical performance, resulting in substantial reductions in UTS, elongation, and toughness. A preliminary mesoscale Finite Element Analysis (FEA) model demonstrates the potential to predict failure behavior in samples containing oxide inclusions through simulations. In addition, this research explores the thermal dynamics of the vee region during the HFIW process. A numerical model developed in COMSOL Multiphysics integrates the thermal modeling with experimental validation from tube mill trials, providing a comprehensive analysis of how critical parameters—such as heat flux, coil-to-weld point distance, and the thermophysical properties of TRIP 690—affect weld quality. Through this study useful tools have been developed that can aid in the optimization of HFIW of TRIP 690.
  • No Thumbnail Available
    Item
    Multi-Physics Smoothed Particle Hydrodynamics Implementation to Enhance Vertebral Fracture Prediction in a Finite Element Model of a Lower Cervical Spine Segment Under Compression
    (University of Waterloo, 2025-01-02) Ngan, Sophia; Cronin, Duane
    Events such as vehicle rollovers can lead to compression of the spine and vertebral fractures. Bone fragments from vertebral fracture displace, or occlude, the spinal canal, deforming the spinal cord and leading to the potential of a spinal cord injury (SCI). Finite element (FE) human body models (HBMs) provide an opportunity to predict vertebral fractures and investigate SCIs. Models such as the Global Human Body Models Consortium (GHBMC) model use strain-based element erosion to model hard tissue fracture by removing elements from the simulation upon reaching threshold strains. While strain-based element erosion allows for the prediction of fracture initiation, the method results in the loss of hard tissue material. Under compression, the loss of hard tissue material limits the post-fracture predictive ability of the model due to the loss of structural support and absence of fractured material that may occlude the spinal canal. The objective of the work in this thesis was the implementation of a multi-physics modelling approach to combine strain-based element erosion with smoothed particle hydrodynamics (SPH) to preserve hard tissue material and simulate the movement of fractured material under central compression in a C5-C6-C7 cervical segment FE model. The implementation of SPH was then assessed by comparing the response of the segment model to experimental results and by evaluating SPH particle dependency. Finally, a parametric study was conducted using the model with the SPH implementation to investigate the response of the FE segment model under varied impact severities, aged hard tissue material parameters, and eccentric loading. The model with the SPH implementation was numerically stable and was found to improve the prediction of the trend and magnitude of the force-displacement response, with the area under the curve compared to the experimental response improving from a 34% difference to a 4% difference. Additionally, the implementation of SPH allowed for modelling the flow of hard tissue material and, consequently, occlusion of the spinal canal. The prediction of maximum occlusion in the model compared to the experiments improved from a 137% difference to a 5% difference. Increasing the number of SPH particles generated for each solid element showed numerical instability, illustrating a need for compatibility between the size of the solid element and the number of SPH particles. Varying the impact severity of the central compression load showed that the occlusion in the FE segment model appeared to have a greater dependency on the maximum displacement applied in compression rather than the maximum velocity of the impact due to the amount of fractured material in the simulation. Applying hard tissue material parameters representative of an older age group resulted in higher occlusion and a lower force-displacement response, in agreement with experimental data. The resulting multi-physics approach improved the model predictive capabilities in all cases. Future research will include a spinal cord in the FE segment model to more accurately assess changes in the spinal canal geometry and potential for SCI.
  • No Thumbnail Available
    Item
    Neck Injury Risk in Non-Neutral Position Rear Impact Scenarios Using a Reference Geometry-Based Repositioning
    (University of Waterloo, 2024-12-19) Seif Reis, Matheus; Cronin, Duane
    Rear-end vehicle collisions may lead to whiplash-associated disorders (WADs), comprising a variety of neck and head pain responses. The source of WADs is an active area of research, but the facet joints, cervical spine ligaments, neck muscles, nerve roots, vertebral arteries, and intervertebral discs have been postulated to be potential injury sources. While most investigations focus on whiplash injuries in neutral position impacts, front-seat vehicle occupants tend to spend a significant amount of time outside the neutral axial head position. Multiple cervical tissues can experience increased mechanical loads and, consequently, higher injury risks during head-rotated impacts. Given the limited experimental data for non-neutral position rear impacts, computational human body models (HBMs) can inform the potential for tissue-level injury. Previous studies have considered external boundary conditions to reposition the head axially but were limited in reproducing a biofidelic movement for a detailed finite element (FE) head and neck model. The objectives of this study were to implement a novel head repositioning method using soft tissue stress initialization to achieve targeted axial head rotations, validate the model intervertebral motion against experimental data, and simulate neutral and non-neutral rear impacts to assess the model tissue-level response and quantify the injury risk. The head and neck model was extracted from the detailed commercial Global Human Body Models Consortium (GHBMC) M50-O FE model, representing a 50th percentile male occupant. The repositioning used a novel method to achieve equilibrium in the target axial rotations of 24.5°, 33.5°, 42.5°, 51.5°, and 60.5°, by pre-stretching the neck soft tissues. The head-rotated models underwent verification tests to ensure the target positions and equilibrium were achieved, and the method was validated by comparing spinal rotations to experimental results. A neutral position and five head-rotated models were evaluated at three rear impact severities (4g, 7g, and 10g) with and without muscle activation. Head motion, ligament distractions, muscle strains, nerve root compressions, and annulus fibrosus (AF) strains were assessed and compared with physiological limits to estimate the injury risks. The repositioning method was able to rotate the head to the five target positions, showing general agreement with reported intervertebral rotations. Under the rear impact scenarios, higher axial head rotations increased the combined three-plane rotations experienced by the head. Increasing head rotations led to higher ligament distractions and muscle strains during rear impacts, increasing their potential for injury. The trends for AF strains varied depending on the muscle activation, with peak strains being observed for initial rotations of 0° and 60.5° without muscle contraction and only for 0° with muscle contraction. Overall, muscle activation decreased ligament displacements, muscle strains, and AF strains, while higher impact severity increased the injury risk for these same tissues. Nerve root compression was minimally affected by head rotation, muscle activation, and impact severity. The models predicted injury potential for the ligaments and muscles starting at a 4g rear impact acceleration and for the AF starting at 7g, particularly under high head rotations. Therefore, the proposed repositioning method introduced in this study enabled the model to achieve steady head rotations with realistic cervical spine movements, increasing the biofidelity of non-neutral rear impact simulations. The tissue-level assessment revealed the effect of varying head rotations, muscle activation, and impact severities on cervical tissues, allowing for a comprehensive estimation of injury thresholds across ligaments, muscles, nerve roots, and AFs. The results of this study can be applied to future assessments and design of head restraints and protection systems.
  • Thumbnail Image
    Item
    The Role of the Laminar Separation Bubble in Wind Turbine Aeroacoustics and Dynamic Stall
    (University of Waterloo, 2024-12-18) Zilstra, Alison; Johnson, David A.
    The implementation of airfoils in low Reynolds number (Re) conditions requires careful consideration of the natural boundary layer (BL) transition as it covers a significant portion of the airfoil surface and often includes the formation of a laminar separation bubble (LSB). This study focuses on the role of the LSB in two design challenges facing low Re applications including small wind turbines (SWT): the generation of tonal airfoil self-noise and the blade loading fluctuations that occur during dynamic stall. Computational fluid dynamic (CFD) and computational aeroacoustic (CAA) methods were applied to two low Re airfoils to first validate the ability of the chosen methods to predict the complex LSB behaviour and second investigate the role of the LSB in these processes. Multiple detailed experimental data sets were combined to form a comprehensive validation of the simulated aerodynamic and aeroacoustic behaviours. The investigation of tonal airfoil self-noise was completed using the SD 7037 airfoil at a modest Re of 4.1x10⁴. The numerical methods of incompressible wall-resolved large eddy simulation (LES) and the Ffowcs-Williams and Hawkings (FW-H) acoustic analogy correctly predicted the tonal aeroacoustic noise and the Kelvin-Helmholtz (K-H) rolls that form in the LSB were identified as the acoustic source. Analysis of the transient BL development showed that the K-H rolls were amplified through a LSB pressure feedback process that altered the development of the laminar BL upstream of the LSB. Later, the analysis of the deep dynamic stall process was completed using unsteady Reynolds-Averaged Navier-Stokes (URANS) simulations of the SD 7037 airfoil at Re=4.1x10⁴ and a pitching reduced frequency of k=0.08, and the S833 airfoil at Re=1.7x10⁵ and k=0.06. The simulated timing of the dynamic stall agreed with experimental data for both airfoils which is an advancement from the early prediction of stall seen consistently in previous numerical studies. The accurate prediction of dynamic stall was found to be dependent on the correct simulation of the bursting of the LSB, which initiated the complete separation of the boundary layer and the formation of a leading edge vortex. The airfoil self-noise and dynamic stall simulations both proved the important role the LSB plays in these behaviours for low Re airfoils. All simulation methods also required strict grid refinement at the leading edge of the airfoil to correctly develop the transient BL transition behaviours for the tonal noise prediction and to capture the bursting of the LSB in the dynamic stall simulations. Grid refinements offset from the airfoil surface were also required to capture the BL transition that occurs in the separated region of the LSB. The successful application of LES and FW-H for tonal noise prediction and URANS for dynamic stall prediction opens the possibility of incorporating these methods into the aerodynamic and aeroacoustic design of airfoils for low Re applications including SWTs.
  • No Thumbnail Available
    Item
    Automated 3-DoF Control Development for Magnetically-Levitated Microrobots Using Machine Learning
    (University of Waterloo, 2024-12-17) Nofech, Joseph; Khamesee, Mir Behrad
    This study presents a novel methodology for achieving three-degree-of-freedom (3-DoF) control for an attractive-type magnetically-levitated (maglev) microrobot using machine learning. Traditional micromanipulation methods face challenges associated with friction and maintenance requirements; particularly in applications such as cell injection, where current devices tend to be limited by their high cost and maintenance requirements. The precise and low-maintenance nature of attractive-type levitation makes it a viable alternative to traditional micromanipulation methods, but a primary challenge lies in the difficulty of achieving precise 3-DoF control for such systems due to the complexity in the magnetic fields they generate. This research addresses this challenge by introducing a machine learning-based methodology that automates the learning of levitation dynamics across the workspace. Our presented approach introduces and incorporates an automated system for generating training data with minimal human intervention, enabling a machine learning model to learn how the levitated microrobot responds to system inputs. This information is then used to establish 3-DoF position control of the levitated microrobot. Our automated methodology simplifies the setup process for new and newly-modified attractive-type levitation platforms, and is demonstrated to improve performance over conventional methods by accounting for observed variations in the levitation dynamics throughout the workspace; achieving up to a 20% reduction in root mean square error during trajectory tracking and up to a 36% reduction in step response settling times. The results demonstrate the ability of our automated methodology to significantly reduce the accessibility barriers associated with establishing and modifying attractive-type maglev platforms; effectively replacing the usual methods of finite element simulation, precise magnetic field measurements, and/or analytical calculations while providing enhanced levitation control over traditional methods. This advancement contributes to the field of micromanipulation and microforce sensing by offering a more accessible and efficient approach to achieving precise control in attractive-type maglev systems.
  • Thumbnail Image
    Item
    Atmospheric pressure spatial atomic layer deposition of metal oxides for different applications
    (University of Waterloo, 2024-12-12) Durgamahanti, Poojitha; Musselman, Kevin
    Atmospheric-pressure spatial atomic layer deposition (AP-SALD) is an emerging technique for the rapid, open-air deposition of metal-oxide thin films. In this thesis, I study two metal oxides, namely silicon oxide and zinc oxide, for use in two different applications. Silicon oxide (SiOₓ) is first studied. It is a highly versatile material used in different applications. However, its conventional growth and deposition methods often require very high temperature or the use of plasma. In this thesis, I present a plasma-free, low-temperature process for depositing high quality SiOₓ thin films using AP-SALD. An aminodisilane precursor, diisopropylaminosilane (SiH₃N(C₃H₇)₂, DIPAS), was synthesized and tested with different oxidants such as ozone and 30% hydrogen peroxide aqueous solution. Initial attempts with hydrogen peroxide solution resulted in precursor condensation and the formation of nano crystallite SiOₓ contaminated with organic molecules, indicating that the deposition process is oxidant limited. In contrast, using ozone as the oxidant facilitated the deposition of high quality amorphous SiOₓ films. The microstructure was highly dependent on the deposition temperature, transitioning from nano crystallites at lower temperatures to amorphous films at temperatures of 70°C to 100°C. X-ray photoelectron spectroscopy (XPS) confirmed the deposition of continuous SiOₓ films at 70°C or above using ozone, and the growth per cycle was ~1 Å/cycle, consistent with atomic layer deposition (ALD) of SiOₓ. This work shows that high-quality SiOₓ films can be produced by AP-SALD using DIPAS and ozone, without the aid of plasma or any surface functionalization, at low growth temperatures (T >= 70°C). For the second study, I investigate the nucleation and growth mechanism of ZnO on different 2D materials. Uniform deposition of metal oxides on 2D materials is a crucial step to realize the integration of 2D materials in practical devices. In this study, I investigate the nucleation and growth of ZnO on 2D transition metal dichalcogenides, specifically MoS₂ and WS₂. The ZnO depositions were carried out using an atmospheric-pressure spatial atomic layer deposition system operated in atmospheric pressure spatial chemical vapor deposition (AP-SCVD) mode at 100°C. The nucleation of ZnO was found to be different on MoS₂ and WS₂, whereby the ZnO nuclei formed larger clusters and nanoribbon on MoS₂ as compared to WS₂. AP-SCVD led to rapid ZnO nucleation on both CVD-grown MoS₂ and WS₂ in as little as 5 AP-SCVD oscillations and complete film closure was achieved on CVD-grown WS₂ flakes in less than 60 AP-SCVD oscillations. This was attributed to the higher precursor partial pressures and uniform precursor delivery afforded by the AP-SCVD process. Raman and photoluminescence (PL) spectroscopy revealed that AP-SCVD is a benign process that doesn’t damage the underlying 2D materials and rather helps to passivate defects via oxygen/water adsorption from the air, when performed in the appropriate temperature window. Deposition of the ZnO was found to impact the optical and structural properties of CVD-grown MoS₂ and WS₂ differently. For the MoS₂ - ZnO heterostructure, electron doping and strain dominate, resulting in a reduction in the PL of MoS₂, whereas for the WS₂ - ZnO, strain and dielectric screening have a larger impact, resulting in an enhanced PL.
  • No Thumbnail Available
    Item
    Development of Conditional Source-term Estimation (CSE) Framework for Turbulent Buoyant Diffusion Flames
    (University of Waterloo, 2024-12-12) Abdalhamid, Ahmed Mohamed Khairy Abdalnaby; Devaud, Cecile
    Conditional Source-term Estimation (CSE) is a turbulent combustion model that relies on conditional averages of species mass fractions to provide closure for the mean chemical source terms. The model has the ability to account for detailed chemical kinetics while remaining computationally affordable for engineering cases. CSE has been successfully applied previously to simulate flames of different regimes and configurations. The current work aims at extending the CSE framework to model turbulent buoyant diffusion flames for the first time. Large Eddy Simulation (LES) is used to solve transport equations of mass, momentum, species and enthalpy. Chemistry is pretabulated prior to the simulations. The LES FireFOAM solver is coupled with CSE approach. The new solver is first tested against University of Maryland methane line fire using different implementations for radiation modelling, all relying on the optically thin assumption. The results are validated against experimental measurements at different locations as well as previously published predictions of the same case using different combustion models. The predicted temperatures are in good agreement with the experimental data, except very close to the fuel inlet. The predicted flame height closely matches the experimental value. The predictions are consistent with previously published results as well. Next, the inclusion of radiation absorption is considered by solving the radiative transfer equation (RTE), avoiding the optically thin approximation for the first time with CSE. Moreover, the weighted sum of grey gases (WSGG) approach is employed to estimate the absorption coefficients. The new framework is applied to the simulation of two medium and large scale methanol pool fires. The medium scale case is investigated using two angular meshes and compared to the optically thin results. The case that solves the RTE with fine angular mesh shows improved predictions. The radiative flux results show that the predictions align better with the measurements with the finer angular mesh outside the flame region. However, the results from both angular meshes are in agreement within the flame. For the large scale case, the time-averaged temperature predictions are in good agreement with the experimental data, except on the centerline downstream of the flame region where the results are overpredicted. Finally, the newly developed CSE framework is extended to include soot modelling by coupling with two of the most widely used soot models in fire analysis. The new approach is used to simulate the ethanol pool fire case, providing acceptable predictions of soot and species mole fractions. In conclusion, this study proves the capability of CSE in accurately simulating turbulent buoyant diffusion flames. Future work may address more complex fuels, cases of extinction with doubly-CSE, suppression by water mist, and more detailed soot models.