UWSpace

UWSpace is the University of Waterloo’s institutional repository for the free, secure, and long-term home of research produced by faculty, students, and staff.

Depositing Theses/Dissertations or Research to UWSpace

Are you a Graduate Student depositing your thesis to UWSpace? See our Thesis Deposit Help and UWSpace Thesis FAQ pages to learn more.

Are you a Faculty or Staff member depositing research to UWSpace? See our Waterloo Research Deposit Help and Self-Archiving pages to learn more.

Photo by Waterloo staff

Communities in UWSpace

Select a community to browse its collections.

Recent Submissions

  • Thumbnail Image
    Item type: Item ,
    Topology Optimization for Additive Manufacturing: Towards efficient and manufacturable structures for multi-physics applications
    (University of Waterloo, 2026-04-28) Orakwe, Joseph Nonso
    The Additive Manufacturing (AM) paradigm has continued to revolutionize the way products and components are conceptualized, given its procedure of building parts from the ground up, in a layerwise manner, which brings benefits such as part consolidation, performance-driven products, and unparalleled design freedom to build complex geometries, albeit within some constraints. To fully exploit the capabilities of additive manufacturing (AM), Design for AM (DfAM) integrates process constraints early in the design stage, leveraging topology optimization (TopOpt), lattice design, and simulation-based engineering (SbE) as core tools. TopOpt is a gradient-based method that determines optimal material layouts within a domain under governing physics, objectives, and constraints, while multiphysics topology optimization (MTO) - such as thermal-fluid topology optimization (TF-TO) - extends this framework to coupled flow and heat transfer problems. Lattices are architected cellular structures used to tailor effective material properties, enabling lightweighting & heat dissipation, while SbE drives engineering design and decisions through computational simulation. A review of the MTO literature revealed two key opportunities for novelty. First, components subjected to severe thermo-mechanical loading (e.g., gas turbine hardware) often require the simultaneous design of coolant flow channels and lightweight load-bearing regions, which can be addressed using MTO with coupled fluid, thermal, and structural physics; however, challenges remain in handling multiple material phases while ensuring manufacturability. Second, thermal-fluid topology optimization (TF-TO) is increasingly applied to heat-sink design for high-power-density electronics, but remains accessible only through limited commercial tools, while full three-dimensional (3-D) implementations demand substantial computational expertise, resources, and manufacturability integration. Integrating MTO with lattice architecture offers further potential through SbE-based functional grading and AM compatibility, yet existing studies underutilize advanced design methodologies that could yield additional performance gains while lowering the barrier to adoption. In this research, two major TopOpt-based methodologies have been proposed. To address the first opportunity, an FTS-TO framework was created, which integrates load-bearing and fluid-delivery requirements within a unified Fluid-Thermal-Structural topology optimization using a cascaded two-stage formulation. Optimization Stage 1 (OS1) addresses the thermal-fluidic optimization, while Optimization Stage 2 (OS2) performs a design-independent thermo-structural TopOpt initialized from OS1, yielding structures satisfying flow, cooling, and structural constraints. Development focused on OS2, introducing a robust multiobjective TopOpt for stiff, conductive, lightweight designs with manufactured-feature-size uncertainty resilience, later extended from 2-D to 3-D with explicit minimum & maximum size control, and self-supporting constraints. Validating simulations indicate promise in designing structures that include flow, cooling, and structural needs, while incorporating various constraints, including AM. For the second research prospect, hybrid field-driven design techniques were developed, which integrate multi-scale Triply Periodic Minimal Surface (TPMS) lattices and airfoil-type fins with TopOpt-predicted flow fields. The first hybrid technique Lattice-TopOpt (LTO) method, conforms heat-dissipative lattices to optimal flow paths to obtain AM-suitable high-performance heat sinks. 3D printability was validated via pure copper on an EOS M290 LPBF machine, and experimental validation backed up predicted superior thermo-hydraulic performance, including a real-world application to an award-winning electrochemically printed cold plate. The Field-Oriented Fin Placement (FOFP) approach orients low-drag airfoil fins along TopOpt velocity vectors, yielding lightweight heat sinks with significantly enhanced mass-based thermo-hydraulic performance. Overall, this work advances the integration of manufacturability constraints into MTO for additively manufactured heat-sink and cooling applications, with a focus on low user complexity, and is expected to contribute to more efficient thermal management solutions in the energy sector.
  • Thumbnail Image
    Item type: Item ,
    Advances in similarity-based prediction modeling
    (University of Waterloo, 2026-04-28) Kim, Minzee
    Personalized predictive modeling has been growing rapidly in recent years, especially with the availability of Electronic Health Records (EHRs). This approach aims to improve a model's predictive performance by fitting a unique model to each individual. We train the model on a subset of the training data consisting of individuals that are similar to the individual we are predicting for, identified through some patient similarity metric. Studies have shown that using a personalized model trained on a customized subset of the data leads to better prediction than using a global model trained on all the available data in the training data. In this thesis, we discuss advancements in similarity-based prediction modeling through extensive simulation studies and data analyses. Longitudinal and time-to-event data are often analyzed in biomarker research to study the association between the longitudinal biomarker measurements and the event-time outcome, in which the longitudinal information contributes to the probability of the outcome of interest. An attractive feature of fitting a joint model on this type of data is that we can dynamically predict the survival probability as additional longitudinal information becomes available. In Chapter 2, we propose a new similarity-based method for the dynamic prediction of joint models where we consider training the model on only a targeted subset of the data to obtain an improved outcome prediction. Through a comprehensive simulation study and an application to intensive care unit data on patients diagnosed with sepsis, we demonstrate that the predictive performance of the dynamic prediction of joint models can be improved with our proposed similarity-based approach. Next, we develop a new patient similarity metric designed to improve the predictive performance of a personalized model for binary response data. Specifically, we introduce a weighted cosine similarity metric in Chapter 3 that extends the standard cosine similarity metric by assigning predictor-specific weights when computing similarity between participants. These weights are estimated using the relaxed adaptive group lasso. Results from our simulation study and an analysis of intensive care unit data involving patients with circulatory system disease show that although the proposed similarity metric leads to a slight deterioration in calibration, it produces substantial gains in discrimination. Overall predictive performance measured by the Brier Score improves because the increase in discrimination outweighs the loss in calibration; therefore, our proposed similarity metric more effectively identifies clinically similar patients, resulting in improved predictive accuracy. Finally, in Chapter 4, we conduct a comprehensive comparison of several similarity metrics to investigate how the choice of similarity metric influences predictive performance in personalized modeling, again in the context of binary response data. By fitting models using only a subset of training participants who are most similar to the individual of interest, prediction accuracy for that individual can be improved. Consequently, selecting an appropriate similarity metric that identifies the most relevant subset of data is critical. We compare a range of distance-based and cosine similarity measures alongside clustering-based approaches, an area that is not well explored in the existing literature. In addition, we perform an extensive simulation study to examine how different data-generating mechanisms and underlying dataset characteristics affect the relative effectiveness of each similarity metric. Finally, we end with a discussion chapter that summarizes the key contributions of the thesis along with highlighting some key areas of future work.
  • Thumbnail Image
    Item type: Item ,
    Rethinking filtration performance assessment for public health protection
    (University of Waterloo, 2026-04-28) Batista, Elyse
    Minimizing acute health risks from waterborne pathogens such as Cryptosporidium is the paramount goal of drinking water treatment. Because oocysts of the protozoan Cryptosporidium resist chlorine-based disinfectants and require complex and expensive detection methods, removal by physico-chemical filtration (CAF) is critical. Regulatory frameworks such as the US EPA suite of Surface Water Treatment Rules (SWTRs) and their Canadian analogs prescribe treatment credits based on filtered water turbidity, assuming the achievement of turbidity targets must reflect well-operated treatment and ≥3-log oocyst removal. However, in systems reliant on low turbidity (<2 NTU), low dissolved organic carbon (DOC; <2 mg/L) source water, raw water turbidity often meets or exceeds filtered water targets, obscuring whether coagulant dosing is resulting in sufficient particle destabilisation. As a result, treatment plants may unknowingly operate at coagulant doses that fail to achieve the particle destabilization required for pathogen removal. This challenge is further compounded because low turbidity, low DOC waters have fewer interactions between particles (as particle concentrations are low) resulting in the need for higher doses of coagulant to increase contact opportunities by precipitating additional particles. Here, pilot-scale CAF investigations confirmed that oocyst removal was dependent on sufficient particle destabilization. Filter challenge studies using Cryptosporidium oocysts were run using low turbidity, low DOC source water and various coagulant doses. The experiments were replicated at similar operational conditions over several years and consistently demonstrated an increased risk of oocyst passage when insufficient coagulant was added and inadequate particle destabilization occurred. The results incontrovertibly demonstrated that turbidity is inadequate as a sole indicator of particle destabilization necessary for ensuring sufficient oocyst removal by CAF, particularly for low turbidity, low DOC source waters. Zeta potential analysis proved to be a useful tool for indicating sufficiency of particle destabilization. Zeta potential values within ±5 mV of zero were required to consistently achieve ≥3-log removal of Cryptosporidium oocysts by filtration of low turbidity, low DOC Lake Ontario source water; in these cases, ≥4-log removal of oocysts was often achieved. As expected, higher coagulant doses than those typically practiced for low turbidity, low DOC source waters—specifically, coagulant doses that led to aluminum hydroxide solid precipitation—were required to achieve these levels of Cryptosporidium removal by CAF. These findings highlight the dangers of sole reliance on turbidity as an indicator of post-filtration water quality and treatment performance and underscore the need for complimentary monitoring tools to ensure protection of public health in systems reliant on higher quality source waters. Integrating zeta potential monitoring into routine coagulation control could provide operators—especially those dealing with the challenge of determining coagulant dose in the absence of substantial turbidity—with an indication of sub-optimal particle destabilization and associated poor filtration performance. These insights also point to broader implications for regulatory policy, as current turbidity-based treatment credits may not always adequately reflect true pathogen removal performance. Further operational guidance for zeta potential operationalization alongside turbidity analysis is needed to help ensure sufficient Cryptosporidium removal by CAF.
  • Thumbnail Image
    Item type: Item ,
    Post-Training Large Language Models as Software Engineering Agents
    (University of Waterloo, 2026-04-28) LYU, Zhiheng
    Large language models (LLMs) have demonstrated remarkable capabilities in code un- derstanding and generation, yet a significant gap remains between static code generation and interactive software engineering. This thesis investigates the post-training of LLMs as software engineering agents, focusing on three interconnected challenges: infrastructure, data, and training methodology. First, we contribute to VerlTool, a unified framework for agentic reinforcement learn- ing with tool integration (ARLT). The author’s contributions center on the training orches- tration layer — the stateful environment protocol, environment server architecture, and SWE agent post-training pipeline — which make tool-augmented RL training practical and accessible for researchers. Second, we address the critical bottleneck of training data and evaluation infrastructure. SWE-Next provides a scalable, Ray-native pipeline for synthesizing verifiable software engineering tasks from open-source repositories (ongoing work with intermediate results reported). For SWE-QA-Pro, a representative benchmark for code question answering, the author contributes the data sourcing and synthesis pipeline. Third, we investigate the post-training design space for software engineering agents, spanning supervised fine-tuning (SFT), rejection fine-tuning (RFT), RL from AI feed- back (RLAIF), and RL with verifiable rewards (RLVR). Through three complementary case studies—code question answering (SFT + RLAIF), web-based information retrieval (SFT + RFT), and repository-level bug fixing (RLVR)—we demonstrate that the opti- mal training recipe depends on task characteristics such as reward verifiability, exploration complexity, and data availability. Our experiments show that task-specific post-training of smaller open-weight models can be competitive with larger proprietary models, and that matching the training method to the task structure is more important than uniformly applying all stages.
  • Thumbnail Image
    Item type: Item ,
    Audio-Visual Feature Fusion through Transformers for Automated Depression Screening in Social Media Content
    (University of Waterloo, 2026-04-28) Haque, Md Rezwanul
    Depression has become a critical public health concern, with the World Health Organization reporting that over 280 million people worldwide are affected by it. The rapid growth of social media, particularly video blogs, has drawn research attention toward analyzing user-generated audiovisual content for signs of depression. These videos capture natural facial expressions, voice characteristics, and speech patterns that may reveal more about a person's emotional state than verbal self-reports alone. However, extracting useful features from such noisy, unstructured data and combining audio and visual information in a way that preserves their complementary nature remain open problems in this domain. The thesis is organized into two main contributions. In the first part, we propose MDD-Net, a multimodal depression detection network that uses a mutual transformer to fuse acoustic and visual features. The acoustic branch employs a global self-attention network to process 25 low-level descriptors including loudness, Mel-Frequency Cepstral Coefficients, and spectral flux, capturing both content-based and positional relationships. The visual branch applies hierarchical multi-head self-attention on 68 facial landmarks extracted from each video frame. The mutual transformer then operates bidirectionally: audio queries attend to visual keys and values, and visual queries attend to audio keys and values. We also design a composite loss function that combines binary cross-entropy, focal loss, and L2 regularization to handle the noisy labels and class imbalance that are common in social media datasets. In the second part, we introduce MMFformer, a multimodal fusion transformer network that takes a different approach to the same problem. For video, a pre-trained vision transformer augmented with residual connections extracts high-level spatial patterns from facial data. For audio, a transformer encoder built on the audio spectrogram transformer paradigm models temporal dynamics in speech signals through patch and positional embeddings. On the fusion side, we propose and compare three distinct strategies: late transformer fusion, intermediate transformer fusion, and intermediate attention fusion, each operating at a different level of the processing pipeline. We evaluate both architectures on the D-Vlog dataset, a publicly available collection of 961 YouTube vlogs from 816 individuals annotated for depression. MMFformer is additionally tested on the LMVD dataset, a larger corpus of 1,823 vlogs collected from four different social media platforms. MDD-Net reaches an F1-Score of 77.07% on D-Vlog, which is an improvement ranging from 1.82% to 17.37% over previously reported methods. MMFformer achieves 90.92% on D-Vlog and 90.48% on LMVD, surpassing the best existing results by 13.92% and 7.74% respectively. Cross-corpus validation between D-Vlog and LMVD further confirms that the developed architectures generalize across different platforms and populations.