21793 research outputs found
Sort by
Pressure-Dependent Characterizations and Design Optimization of a Two Phase Fluidic Suspension Strut Using AI-Based Modeling Technique
No full textThis thesis presents a study of two-phase fluidic suspension strut for predicting its pressure-dependent properties. The methods explored range from fundamental constitutive relations to AI-based simulation techniques. The design of two-phase fluid strut is greatly simplified by letting gas-oil mixture, also denoted as gas-oil emulsion strut (GOES). The design addressed many limitations of conventional compact hydro-pneumatic struts, such as, requirement of a floating piston to separate gas and oil media, reduced friction, relatively lower temperature sensitivity, and larger effective area. The design also offered added flexibility in view of number and sizes of bleed orifices and blow-off valves. In the first stage, an analytical model is formulated considering fundamental constitutive pressure and flow relations, LuGre friction model, and pressure-dependent flow coefficient. In addition to polytropic van der Waals real-gas law, the properties of the emulsion are thoroughly investigated using analytical formulations and available experimental data. The validation of the model is demonstrated using available experimental data under different operating and excitation conditions. It is shown that consideration of pressure-dependent relations could help and enhance prediction effectiveness of the model, irrespectively of operating and excitation conditions.
The validated model is used to investigate pressure-dependent nonlinear stiffness and damping properties of GOES under various operating conditions and the results are discussed to highlight design guidance. The influences of charge pressure, gas volume fraction, and gas-to-oil volume ratio on the strut’s performance were studied over the broad frequency and velocity ranges. The results revealed that the stiffness is primarily influenced by strut deflection, while the damping characteristics are strongly dependent on strut velocity, excitation frequency, and deflection. In the third stage, an optimized supervised artificial neural network (ANN) model is developed. The non-dominated sorting genetic algorithm II (NSGA-II) is applied to seek to tune strut design to address the limitations of the pressure-dependent analytical model. The ANN model provided accurate predictions of the highly nonlinear behavior of GOES under various uncertainties such as those arising from the effect of fluid inertia, deformations of the seals, and nonlinear dependence on gas volume fraction and emulsion characteristics. The optimally tuned ANN model is applied in a quarter-car model simulation platform to evaluate its effectiveness under random road excitations and varying operating conditions. The model provided an accurate analysis of ride comfort performance. A reinforcement learning-based (RL-based) semi-active control strategy is subsequently conceived to dynamically regulate the strut force using an adjustable solenoid valve. A quarter-car model was further used to evaluate the performance potentials of the proposed strategy under random road excitations. The RL-based controller was trained to regulate valve opening based on the vertical acceleration and velocity of the sprung-mass. The proposed semi-active scheme provided improved ride comfort compared to the conventional and optimally tuned passive GOES. The findings of this thesis provided a sound basis toward advancement of intelligent and tunable suspensions for vehicles by integrating more accurate nonlinear analytical models and machine learning-based approaches for real-world vehicle applications
Understanding fear of depression recurrence among individuals who are remitted from major depressive disorder: A mixed-methods study
No full textUnderstanding fear of depression recurrence among individuals who are remitted from major depressive disorder: A mixed-methods study
Stephanie Gumuchian, PhD
Concordia University, 2025
Major depressive disorder (MDD) is a prevalent psychiatric condition that significantly impacts functioning and quality of life. A key factor contributing to its high personal, social, and economic burden is the frequent recurrence of depressive episodes, which occurs in 25-40% of individuals diagnosed with MDD. Understanding the cognitive and behavioural changes that occur after a major depressive episode (MDE) and contribute to recurrence risk is crucial for enhancing treatment strategies and preventing future episodes. This mixed-methods thesis investigates the concept of fear of depression recurrence (FoDR), defined as fears that one’s depression may return, among individuals in remission from MDD. Research examining fear of illness recurrence among mental health conditions, including MDD, is limited. Thus, studying FoDR may lead to a better understanding of the changes that occur following an MDE that increase one’s recurrence risk.
In study one, we conducted semi-structured interviews with 30 individuals who are remitted from MDD to explore their experiences with FoDR. We examined the content, triggers, and severity of participants’ fears and evaluated the perceived impact of these fears on daily functioning. Most of the sample (73%) reported experiencing FoDR, with variations in the frequency, severity, and duration of fears. The content and triggers of participants’ FoDR mirrored the experiences that occurred during past MDEs. The impact of FoDR on daily functioning was nuanced, with some reporting a positive, negative, or minimal influence.
In study two, we used the information from our qualitative study to inform the development of a self-report questionnaire designed to capture the multidimensional concept of FoDR. We used exploratory and confirmatory factor analyses to identify the initial factor structure and inform item selection. Our findings supported the retention of a 24-item scale with three factors evaluating the severity, content, and triggers of fears. Our scale demonstrated excellent psychometric properties.
Our results offer initial support for the presence and clinical relevance of FoDR among individuals remitted from MDD. The Fear of Depression Recurrence Questionnaire (FoDRQ) can now be used to explore associations between FoDR and MDD recurrence risk, clinical outcomes, and coping strategies and identify when preventative interventions are needed
Deep Generative Models and Their Inversions For Bidirectional Transformation Between Data and Latent Distributions
No full textDeep Generative Models and Their Inversions For Bidirectional Transformation Between Data and Latent Distributions
Jeongik Cho, Ph.D.
Concordia University, 2025
Generative models aim to transform a simple latent distribution into a complex data distribution, enabling the synthesis of high-dimensional, realistic data. In contrast, generative model inversion addresses the reverse process, mapping a complex data distribution back into a simple latent representation. In this thesis, we introduce several novel contributions to architecture-agnostic algorithms of generative models and their inversions, as well as applications utilizing these methods.
First, we show that using multiple adversarial losses improves the performance and requires fewer hyperparameters than using an auxiliary classifier. Then, we introduce a novel encoder-based GAN inversion method for better convergence than a simple mean squared error by dynamically adjusting the scale of each element of the latent random variable. Furthermore, we propose an out-of-distribution detection method that leverages the log probability of the latent vector predicted by the encoder-based GAN inversion framework. Next, we introduce a novel method that combines the perceptual VAE and the GAN inversion technique from the second contribution to improve the GAN inversion performance. Finally, we introduce a novel GAN that allows the model to perform self-supervised class-conditional data generation and clustering using a classifier gradient penalty loss
Deep Learning for Quantitative Ultrasound and Multimodal Analysis: Liver Steatosis Diagnosis, Uncertainty Decomposition, and Diagnosis of Breast Cancer-Related Lymphedema
No full textPoint-of-Care Ultrasound (POCUS) is a portable, cost-effective imaging modality with strong potential to expand access to diagnostic tools in remote and underserved settings. However, its interpretation still depends heavily on expert knowledge, which limits broader clinical adoption. This thesis aims to enhance the interpretability, reliability, and accessibility of POCUS by leveraging deep learning techniques.
The first part of this work presents a Bayesian deep learning framework for the classification of Non-alcoholic fatty liver disease using QUS features extracted from pre-clinical duck experiments. The model not only achieves accurate classification but also provides meaningful uncertainty estimates, helping assess prediction confidence. In the second part, we propose a method to decompose predictive uncertainty into epistemic and aleatoric components in the estimation of Homodyned-K distribution QUS parameters and investigate their relationship with prediction error. The final part introduces a multimodal dataset for the diagnosis of breast cancer-related lymphedema (BCRL) using POCUS. A deep learning pipeline is developed that integrates ultrasound images and clinical features to improve diagnostic performance.
Together, these contributions apply deep learning methods to enhance quantitative tissue characterization, uncertainty estimation, and diagnosis, making ultrasound more practical and accessible in everyday healthcare
Multistage stress classification and cognitive capacity analysis using EEG
No full textStress is a physiological and psychological strain caused by mental workload. It is better to detect stress at early stages, which can help in stress management, compared to later stages, which may develop into psychiatric disorders such as anxiety and depression. In this study, we employed Muse S, a four-channel EEG headband, to record participants’ EEG data under stressed and control conditions. We utilized mental arithmetic tasks and the Stroop color-word test as stressors to induce stress among our participants. We conducted subject-dependent and subject-independent evaluations by employing 10-fold and LOSO cross-validation strategies, respectively, and analyzed the difference between the two evaluation strategies. We proposed a two-stage deep learning model that comprises a fully connected autoencoder and a bidirectional LSTM model with an attention mechanism to improve the classification metrics for subject-independent evaluation using LOSO cross-validation strategy. We employed our proposed deep learning model to perform both binary and three-stage stress classification, achieving an accuracy of 83% for binary classification, while for three stage stress classification our model reported an accuracy of 66%. We compared the cognitive capacity of our best and worst performers by employing statistical tools such as line graphs and the Mann-Whitney U test. We implemented a regression model using random forest to predict the participants’ scores by employing brain waves and response time
Engineering Droplet Microfluidic Platforms for Microbial Strain Improvement
No full textAdvancing microbial strain improvement is essential for industrial biotechnology, enabling organisms with enhanced productivity, stress tolerance, and metabolic efficiency. Traditional improvement methods—both genetic and non-GMO approaches like UV mutagenesis or adaptive laboratory evolution—generate high phenotypic diversity but require the screening of vast microbial libraries. Existing screening tools such as FACS and microtiter assays struggle with cost, resolution, and compatibility, especially for secreted or label-free phenotypes and complex morphologies like filamentous fungi.
This thesis presents a novel microfluidic electrostatic droplet sorting (EDS) platform designed to overcome these limitations. The EDS system encapsulates individual cells in nano- to picoliter droplets, creating isolated microreactors for high-throughput phenotypic screening. Unlike conventional dielectrophoretic sorting, EDS operates at lower voltages, enhancing biocompatibility and versatility, and is better suited to heterogeneous droplet populations and morphologically complex organisms.
The thesis details device design, fabrication, and integration with optical detection systems, supporting both binary and multiplexed sorting. The platform was validated with various industrial microbes, including Clonostachys rosea, Aspergillus oryzae, Trichoderma reesei, and Saccharomyces cerevisiae var. diastaticus, targeting enzymes like chitinase, amylase, and cellulase. Furthermore, a novel method was developed to mimic solid-state fermentation in droplets using colloidal suspensions, enabling screening under industry-relevant conditions.
These integrated EDS platforms address critical screening challenges, offering scalable, low-cost, and high-throughput solutions that support the development of next-generation strains for sustainable biomanufacturing
TENSILE
No full textIn 2020, Erica McAlpine authored a book-length study of mistakes in poetry called The Poet’s Mistake. The poet who wrote TENSILE, by their own assessment, has made a lot of them. Not merely an exercise in self-flagellation, these poems explore how the poet’s particular strategies for structuring their reality lead mistakes to proliferate wildly in their perceptual field like an invasive weed. What is happening, at the level of the body, affect, language, interpretation, family, cognition, and behaviour, to open up such vast horizons of humiliation? And is there even some beauty to be found in this patterning? What other vistas could be carved out of the landscape? What of work, for example—where might there be playfulness within its restrictions? Robert Frost says, “You don’t want a freedom from the tennis court, you want the freedom of the tennis court, with the net just so high and the court just so large, always.” Perhaps, then, play is the way out. Not just the desultory play of signifiers, but the play of children whose moves in every moment of the spontaneous game are contingent, intrinsically motivated acts of discovery
Curve Shortening Flow of Closed Curves in the Plane and on Curved Surfaces
No full textThe curve shortening flow of smooth curves, also referred to as flow by curvature, has seen detailed study in the last four decades. In this paper, we go over the main existence theorems of the theory, with an focus on making the presentation clear and self-contained.
Starting with convex curves in the plane, we examine the Gage-Hamilton theorem, which proves the flow of these curves exists and converges to a point. We inspect the original proof by Gage and Hamilton. We next look at Grayson's Theorem which proves the same for non-convex curves. Instead of Grayson's original proof, we study a proof by Bryan and Andrews.
The aim is to highlight the core arguments and make the proofs more accessible. We seek to emphasize the intuition underlying the key ideas.
Finally, we turn to the flow of curves on Riemannian surfaces. We present some of the necessary differential geometry material which is otherwise typically assumed. We compare results on the surface to their analogues in the plane and examine how the curvature affects the respective proofs
Automated File-Level Logging Generation for Machine Learning Applications using LLMs: A Case Study using GPT-4o Mini
No full textLogging is essential in software development, helping developers monitor system behavior and aiding in debugging applications. Given the ability of large language models (LLMs) to generate natural language and code, researchers are exploring their potential to generate log statements. However, prior work focuses on evaluating logs introduced in code functions, leaving file-level log generation underexplored---especially in machine learning (ML) applications, where comprehensive logging can enhance reliability.
In this study, we evaluate the capacity of GPT-4o mini as a case study to generate log statements for ML projects at file level. We gathered a set of 171 ML repositories containing 4,073 Python files with at least one log statement. We identified and removed the original logs from the files, prompted the LLM to generate logs for them, and evaluated both the position of the logs and log level, variables, and text quality of the generated logs compared to human-written logs. In addition, we manually analyzed a representative sample of generated logs to identify common patterns and challenges.
We find that the LLM introduces logs in the same place as humans in 63.91% of cases, but at the cost of a high overlogging rate of 82.66%. Furthermore, our manual analysis reveals challenges for file-level logging, which shows overlogging at the beginning or end of a function, difficulty logging within large code blocks, and misalignment with project-specific logging conventions. While the LLM shows promise for generating logs for complete files, these limitations remain to be addressed for practical implementation
Picturing “Topsy-Turvy Land”: Photographic Representations of Egypt and Japan
No full textThis dissertation investigates patterns that find expression in late nineteenth- and early twentieth-century photographs of Egypt and Japan. It is argued that, with the intensification of imperial activity and the rise of mass tourism, these countries became twin beacons for the Western world’s fascination with the Orient.
The period of study ranges from the 1850s to the outbreak of the First World War. Drawing on a variety of photographic materials from this era—including commercial prints, snapshots, stereographs, book illustrations, postcards, etc.—a selection of visual tropes is examined to shed light on cultural stereotypes and common image-making practices. Because photographs of Egypt and Japan were profoundly influenced by the literary tradition of the travelogue, the images under investigation are interpreted through the writing that framed them in a manner that recognizes text and image as two components of the same discursive whole.
Adopting an iconographic approach, each of the dissertation’s thematic chapters is dedicated to a prevalent photographic motif. The first is concerned with representations of monumental sculpture, the second with images of marketplaces and commodities, and the third with portraits of travellers dressed in local costume. Attending to the repetitiveness of this formulaic imagery is crucial as it reveals the self-sustaining structure of the Orientalist imagination. This is how countries as dissimilar as Egypt and Japan—culturally, politically, and geographically—could be conflated in photographic representations that positioned the Orient as the topsy-turvy counterpart of an ordered, modern, and moral Occidental civilization