http://hdl.handle.net/10453/19979 2026-07-13T17:08:39Z http://hdl.handle.net/10453/195485 Title: Binder jet 3D printing with recycled glass fines for circular design applications Authors: Williams, Ella L. Abstract: As linear production practices continue to devastate the environment, the need for industrial designers to prioritise circularity in products and services has become increasingly urgent. This research demonstrates how designers can integrate hands-on material and production knowledge into the development of circular outcomes. Specifically, it explores how 3D printing can be harnessed to transform the waste stream of glass fines. Glass is inherently suited to circular systems, as it can be endlessly remelted and reformed without loss of material quality. However, current recycling systems generate large quantities of glass fines: small, mixed-colour particles that cannot be used in traditional glass manufacturing. As a result, these fines are currently downcycled into low-value applications such as road base filler or pipe embedment material, terminal uses that fail to capitalise on the material’s intrinsic properties. This research presents a novel material formulation, production process, and application that transform glass fines through binder jetting, an additive manufacturing process in which a liquid binder is selectively deposited onto a powder bed. Binder jetting enables cost-effective customisation and the fabrication of complex geometries without support structures. A four-step production process is developed in which glass fines are combined with a binder, printed, dried, and then fired in a kiln. During firing, the binder burns out while the glass particles fuse, resulting in a solid glass object that retains full recyclability. By tailoring the firing schedule, parts can be produced that are both waterproof and water-absorbent, with a porosity of approximately 14.5%. This approach maintains the inherent circularity of glass while introducing a new method for fabricating complex glass forms that are not achievable through conventional glass manufacturing. To demonstrate the application potential, the research presents an evaporative cooling screen inspired by traditional porous ceramic cooling systems. The porosity created during binder jetting allows water to permeate the surface, forming a thin film that evaporates and cools the surrounding air. A 1:1 scale demonstrator composed of sixteen 3D-printed tiles illustrates how the material and process can be leveraged to create products with industrial relevance. This research contributes to discourse at the intersection of design, circularity, additive manufacturing, and materials. It shows how these domains can be holistically integrated, with each informing and strengthening the others. By embedding circular strategies within the material, production, and application, the research expands the scope of industrial design and underscores the need for designers capable of operating across traditional disciplinary boundaries. Description: University of Technology Sydney. Faculty of Design, Architecture and Building. 2026-01-01T00:00:00Z http://hdl.handle.net/10453/195484 Title: Cross-Domain Image Classification in Complex Real-World Scenarios: With Test-Time Label or Continual Domain Shift Authors: Ma, Tianyi Abstract: Cross-domain image classification is proposed to address the distribution discrepancy between the source and target domains, commonly referred to as the domain gap. Although it has been proven effective in mitigating the domain gap between the source and target domains, cross-domain image classification still faces significant challenges in complex real-world scenarios. In this thesis, two specific challenges that commonly happen, namely, label shift and continual domain shift, for cross-domain image classification are investigated and addressed. For the label shift, this thesis focuses on cross-domain few-shot image classification (CDFSIC). A novel prompt-to-disentangle method is proposed to combine the benefits of domain generalisation and adaptation by disentangling source and target knowledge. For the continual domain shift, this thesis focuses on continual test-time adaptation (CoTTA). Based on the findings in CDFSIC, we design a similar strategy called the Source and Target Disentangle Transformer to explicitly disentangle source and target knowledge, thereby facilitating both the preservation of source knowledge and the extraction of target knowledge. Then, based on the observation that the recent CoTTA method is unstable in a small-batch setting, a novel task named single-sample CoTTA is proposed. A novel strategy, named effective buffer and resetting, is designed to increase adaptation stability. Moreover, we apply this method to zero-shot models, solving the label shift and continual domain shift simultaneously. Finally, we highlighted several future directions, including active CoTTA to address larger domain gaps with human calibrations and zero-shot CoTTA to tackle both label shift and continual domain shift. Description: University of Technology Sydney. Faculty of Engineering and Information Technology. 2025-01-01T00:00:00Z http://hdl.handle.net/10453/195483 Title: Towards Robust Clinical Segmentation of Paediatric Brain Tumours in Magnetic Resonance Images Using Weakly-supervised Deep Neural Networks Authors: Loesch, Nico Abstract: The delineation of paediatric brain tumours in magnetic resonance imaging (MRI) remains a major clinical challenge due to complex tumour presentations, and the heightened sensitivity of developing brains to treatment. Manual annotation is time-consuming and subjective, with inconsistencies arising from differences in imaging quality and tumour morphology. Automated approaches show promise in addressing these issues, yet state-of-the-art supervised deep learning (DL) methods depend on extensive, pixel-level annotations that are costly and scarce in paediatric populations. To overcome these limitations, this thesis investigates weakly-supervised anomaly detection based on denoising diffusion probabilistic models (DDPMs) as an alternative for delineating paediatric brain tumours in MRI. The first contribution introduces a 3D-latent diffusion model (LDM) with a novel patch-based training strategy that enables efficient learning on volumetric data while reducing computational demand. This strategy also facilitates the extraction of pseudo-healthy anatomy from diseased individuals, mitigating data collection requirements. The applicability of a novel encoding mechanism, adapted from natural images, is further assessed for medical imaging. The approach surpasses existing weakly-supervised baselines across several benchmarks. However, the generation of artefacts raises important questions regarding performance on small lesion detection. To address these limitations, the second contribution exploits the generative capacity of LDMs to synthesise datasets with precisely controlled lesion sizes. Various conditioning strategies are systematically compared to balance fidelity and dataset consistency. Building on this foundation, the third contribution explores spatial resolution enhancement via super-resolution (SR) using a conditional LDM to improve the detection of small lesions. The results demonstrate clear gains in lesion sensitivity and resolution-aware segmentation. The fourth contribution assesses the generalisability of the proposed framework to paediatric tumours, an underexplored domain limited by scarce annotations. Experimental results show that LDMs trained solely on adult data generalise effectively to paediatric cases, while fine-tuning yields negligible gains. Additional evaluation on a private multi-institutional cohort encompassing diverse tumour types and acquisition conditions further supports the framework’s robustness. These findings demonstrate that anomaly detection can extend beyond its original training domain and underscore the framework’s relevance in low-annotation regimes. Together, these contributions advance the use of LDMs for weakly-supervised anomaly detection in medical imaging, unifying lesion detection, spatial resolution enhancement, and synthetic data generation. The framework reduces dependence on large annotated datasets and demonstrates robustness across adult and paediatric cohorts. As a result, this thesis outlines a pathway towards scalable and clinically applicable paediatric tumour segmentation beyond conventional supervised paradigms. Description: University of Technology Sydney. Faculty of Engineering and Information Technology. 2026-01-01T00:00:00Z http://hdl.handle.net/10453/195482 Title: Collaborative Robot Technology Adoption in Australian Manufacturing SMEs Authors: Haddas, Mashael Ali Abstract: Collaborative robots (cobots) represent an emerging category of technological innovation that is transforming the industrial landscape by revolutionising the interaction between machines and humans. Cobots are lightweight, cost-effective and flexible industrial solutions that have become increasingly suited to the growing shift toward mass customisation in modern manufacturing environments. Despite their potential and promise, especially for small and medium-sized enterprises (SMEs), adoption among SMEs remains relatively limited, with several challenges related to their adoption still unexplored. Driven by 1) the strategic importance of advanced technologies for developing manufacturing SMEs in Australia, and 2) the limited research on cobot adoption in this context, the present study aims to develop the Holistic Collaborative Robot Adoption Model (HCRAM) for Australian manufacturing SMEs. Guided by the design science approach, this thesis makes the following contributions: 1) it develops a conceptual framework grounded in both empirical and theoretical research on cobot adoption in the manufacturing sector; 2) it presents findings from the analysis and refinement of the conceptual framework based on perspectives from senior and mid-level managerial and technical specialists; and 3) it demonstrates the evaluation outcomes of the framework in manufacturing SMEs, drawing on data gathered from a large sample using the questionnaire instrument. The HCRAM, developed in this thesis, is, to the best of the researcher’s knowledge, the first to address this issue in the context of Australian manufacturing SMEs. It encompasses five key contexts relevant to cobot adoption: environment, human, technology, organisation, and barriers. The findings show that 11 of the 15 factors demonstrated a statistically significant association with cobot adoption, thereby supporting the validity of the findings. As a result, HCRAM can be used as a practical tool for industrial decision-makers to formulate adoption strategies for cobots in Australian manufacturing SMEs. It is supported by an online questionnaire tool designed to identify both enablers and barriers to adoption across a wide range of Australian manufacturing SMEs. It further establishes a basis for future work on cobot adoption within manufacturing SMEs and related contexts. Although this thesis employs a rigorous approach, there are several limitations relating to data collection, sampling methods, and geographic context. Therefore, future research could adopt different methodologies to further validate HCRAM and explore its applicability across various industries and national contexts. Description: University of Technology Sydney. Faculty of Engineering and Information Technology. 2025-01-01T00:00:00Z