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.
Communities in UWSpace
Select a community to browse its collections.
- The University of Waterloo institution-wide UWSpace community.
Recent Submissions
Microarray Image Denoising Leveraging Autoencoders and Attention-Based Architectures with Synthetic Training Data
(University of Waterloo, 2024-09-16) Czarnecki, Chris
Microarray technology has for many years remained a golden standard in transcriptomics. However, preparation of physical slides in wet labs involves procedures which tend to introduce occasional dirt and noise into the slide. Having to repeat experiments due to environmental noise present in the scanned images leads to increased reagent and labor costs. Motivated by the high costs of repeated wet lab procedures we explore denoising methods in the narrow subfield of microarray image analysis. We propose SADGE, a domain-relevant metric to quantify the denoising power of methods considered. We introduce a synthetic data generation protocol which permits the creation of very large microarray image datasets programmatically and provides noise-free ground truth useful for objective quantification of denoising. We also train several deep learning architectures for the denoising task, with several of them beating the current state-of-the-art method on both PSNR and SADGE metrics. We propose a new training modality leveraging EATME module to condition the image reconstruction on ground-truth expression values and we introduce an additional loss term (DEL) which further enhances the denoising capabilities of the model while ensuring minimal information loss. Collectively, innovations outlined in our work constitute a significant contribution to the field of microarray image denoising, influencing the cost-effectiveness of microarray experiments and thus impacting a wide range of clinical and biotechnological applications.
Adaptive Model Predictive Control for Microstructure Control in Laser Material Processing
(University of Waterloo, 2024-09-16) van Blitterswijk, Richard Hendrik
Laser material processing (LMP) has revolutionized traditional fabrication methods across industries, evolving from laser cutting to encompassing advanced techniques like laser heat treatment (LHT), laser welding, and laser additive manufacturing, enabling precise alteration of material properties and unprecedented design freedom across industries. However, achieving consistent material characteristics remains a significant challenge, particularly in advanced additive manufacturing processes such as laser-directed energy deposition (LDED), where the complex interplay between process parameters and material properties hinders uniform product quality, emphasizing the need for advanced process control strategies.
Conventional control methods like proportional-integral-derivative controllers struggle to anticipate the intricate interactions inherent in LMP processes, making it difficult to control multiple parameters simultaneously. Model-based control strategies, leveraging numerical models, offer promise in providing a comprehensive understanding of process dynamics. However, their practical implementation in real-time control applications is impeded by the computational challenges of numerical models. Overcoming these obstacles is crucial to harnessing the full potential of numerical models for enhanced process control and ensuring consistent, reproducible material characteristics.
In this research, a novel adaptive model predictive control (AMPC) algorithm was developed to address the challenges of ensuring consistent material characteristics in LMP processes. Initially, a two-dimensional (2D) adaptive thermal model was designed for real-time prediction of thermal dynamics during the LHT process, focusing on parameters like peak temperature and spatial cooling rate. Subsequently, a one-dimensional (1D) adaptive thermal model was developed with improvements on efficiency, accuracy, and suitability for control applications compared to the 2D counterpart, focusing on real-time prediction of the temperature distribution and spatial cooling rate. Additionally, a model predictive control (MPC) algorithm utilizing a 2D thermal model was developed for single-input single-output (SISO) peak temperature control during LHT to improve the consistency of hardness and hardening depth. Finally, an AMPC algorithm was designed using the 1D adaptive thermal model for multi-input multi-output (MIMO) temperature and spatial cooling rate control during LDED to achieve consistent material characteristics throughout the process.
A series of LHT and LDED experiments were designed to assess the real-time thermal dynamic prediction capabilities of the models and the real-time control capabilities of the MPC algorithms in LMP. These experiments encompass open-loop LHT and LDED scenarios, targeting the validation of adaptive 1D and 2D thermal models, respectively. Additionally, closed-loop LHT and LDED experiments were designed to investigate the efficacy of the MPC algorithms in controlling one or multiple process parameters to achieve consistent hardness values. The 2D adaptive thermal model effectively adjusted to the thermal dynamic changes in real-time, yielding precise predictions of peak temperature and spatial cooling rates during LHT. Similarly, validations of the 1D adaptive thermal model showcased near-perfect temperature and cooling rate predictions during LDED, along with impressive computational efficiency. Utilizing the SISO MPC algorithm ensured consistent hardness and hardening depth through closed-loop peak temperature control during LHT. Meanwhile, deploying the MIMO AMPC algorithm enabled consistent hardness across the entire deposition process. This was achieved by simultaneously controlling the temperature and spatial cooling rates during the LDED experiments.
In conclusion, this research marks significant advancements in real-time process control within LMP applications. Through the integration of adaptive thermal models and MPC algorithms, the study achieves the crucial objective of ensuring consistent material characteristics in LMP-manufactured parts. The developed AMPC algorithm demonstrates unprecedented levels of control, stability, and reliability. Moreover, its versatility and simplicity extend its applicability beyond LMP processes, enabling adoption in various advanced manufacturing processes utilizing concentrated energy sources. Thus, the AMPC methodology holds the potential to address the crucial need in advanced manufacturing by ensuring consistent and reproducible material characteristics in manufactured parts across the entire industry.
Troll Patrol: Detecting Blocked Tor Bridges
(University of Waterloo, 2024-09-16) Vecna, Ivy
Tor is an important tool for protecting people against Internet surveillance and censorship. Therefore, some governments that wish to monitor or restrict their people’s use of the Internet attempt to block access to Tor. Bridges are circumvention proxies that provide routes around this censorship, enabling people to access Tor, even in countries that ordinarily censor it. However, a motivated censor may work to identify these bridges and block access to them.
To impede the censor’s attempts at identifying and blocking bridges, reputation-based systems for bridge distribution such as Hyphae, Salmon, and Lox have been proposed. These systems place greater trust in users when the bridges they know remain uncensored and reduced trust in users when bridges they know become censored. In order to enact these changes in trust, it is necessary to know which bridges have been blocked and which have not, but Tor does not currently have a systematic way to detect blocked bridges.
In this work, we present Troll Patrol, a system for automatically detecting censorship of Tor bridges. This system infers bridge reachability based on already-existing bridge usage statistics and novel anonymous user reports that we design for this purpose. We evaluate our system using a simulation and demonstrate that user reports improve our ability to detect bridge censorship, compared to using statistics on bridge use alone. We describe an attack that allows the censor to evade detection if classification of bridge blockage relies on bridge statistics alone, and we demonstrate that user reports allow us to defend against this attack.
Fabrication of Proton Conducting Electrochemical Half-cell Based on Perovskite Structure Material
(University of Waterloo, 2024-09-16) Zhao, Pengcheng
The rising concerns about 𝐶𝑂2 emissions from industrial processes and fossil fuel combustion are driving the development of clean energy sources. Among these, hydrogen energy stands out as an efficient carrier with high storage capacity and minimal environmental impact. This thesis focuses on the fabrication of a solid oxide electrochemical half-cell (SOC) based on proton-conducting materials in a perovskite structure, which can be used for hydrogen generation or utilization. The primary material used is Barium Zirconium Cerium Yttrium Oxide (BZCY) due to its proton conductivity, chemical stability, and mechanical strength under varying conditions.
In this work, several nanomaterials synthesis methods were utilized, including sol-gel and combustion processes, to achieve high-purity BZCY172 material with the desired particle size and composition. A variety of membrane fabrication techniques, such as screen printing, dry pressing, and manual blade coating were employed to construct the bi-layer electrolyte membranes, aiming for uniformity and high-density. Through extensive experimentation, the optimal sintering temperature for the bi-layer membrane was determined, which successfully produced a dense electrolyte layer with a thickness of 20-30μm. Furthermore, the maximal diffusion coefficient (Do) and activation energy for diffusion (Ea) values for barium ion diffusion within the BZCY172 material were determined using Fick’s second law model based on experimental data, offering new insights into material performance under high-temperature conditions.
This thesis also tackled key challenges in proton-conducting SOC fabrication, such as optimizing the sintering process to enhance densification, controlling barium evaporation during high-temperature sintering, and incorporating suitable additives to promote grain growth and reduce porosity. Characterization techniques, including X-ray diffraction (XRD) and scanning electron microscopy (SEM), were employed to analyze the microstructure and chemical composition of the synthesized materials and fabricated membranes, further advancing the understanding of their performance in electrochemical applications.
Overall, this research contributed to the field of hydrogen energy and proton-conducting SOCs by providing a detailed investigation into the fabrication and optimization of BZCY-based electrochemical systems.
Lossless Basis Expansion for Gradient-Domain Rendering
(University of Waterloo, 2024-09-16) Fang, Qiqin
Gradient-domain rendering utilizes difference estimates with shift mapping to reduce variance in Monte Carlo rendering. Such difference estimates are effective under the assumption that pixels for difference estimates have similar integrands. This assumption is often violated because it is common to have spatially varying BSDFs with material maps, which potentially result in a very different integrand per pixel. We introduce an extension of gradient-domain rendering that effectively supports such per-pixel variation in BSDFs based on basis expansion. Basis expansion for BSDFs has been used extensively in other problems in rendering, where the goal is to approximate a given BSDF by a weighted sum of predefined basis functions. We instead utilize lossless basis expansion, representing a BSDF without any approximation by adding the remaining difference in the original basis expansion. This lossless basis expansion allows us to cancel more terms via shift mapping, resulting in low variance difference estimates even with per-pixel BSDF variation. We also extend the Poisson reconstruction process to support this basis expansion. Regular gradient-domain rendering can be expressed as a special case of our extension, where the basis is simply the BSDF per pixel (i.e., no basis expansion). We provide proof-of-concept experiments and showcase the effectiveness of our method for scenes with highly varying material maps. Our results show noticeable improvement over regular gradient-domain rendering under both L1 and L2 reconstructions. The resulting formulation via basis expansion essentially serves as a new way of path reuse among pixels in the presence of per-pixel variation.