Dissertation Defense Announcements

Behavioral and Social Science (BSS) classes in higher education provide students with understandings of human behaviors, motivations, and actions that are crucial to confronting both social and personal problems. Moreover, most community college degrees require that students take at least one BSS class—anthropology, economics, political science, psychology, and/or sociology. While BSS classes are important—both from a philosophical as well as a degree-requirement standpoint—without effective student engagement, that importance may be lost. Oftentimes, BSS classes are still taught largely through didactic instruction. Yet, active and collaborative learning has proven to be a more effective instructional approach. Moreover, the need for active and collaborative learning may be even more crucial in community college BSS classes due to the unique demographics of these institutions. Building on findings that active and collaborative learning in BSS classes is more effective than didactic instruction, the purpose of this study is to better understand BSS instructors’ knowledge of active and collaborative learning and to identify the factors that foster this instructional approach and those that present hurdles. By determining these factors, recommendations can be made for replicating effective active and collaborative learning in BSS classrooms and/or working to minimize the roadblocks to this instructional approach.

Additive manufacturing (AM), particularly laser powder bed fusion (LPBF), is of great interest to the aerospace community as it can be used to manufacture parts with complex internal and external geometries. This is in contrast to conventional manufacturing methods that limit the complexity of part designs. Of particular interest are the manufacture of parts with cooling channels consisting of complex surface topographies designed to improve thermal performance of the channels. While conventional machining can reduce the roughness of external surfaces, most surface treatment processes cannot be applied to internal channels, especially when the dimensions are a millimeter or submillimeter scale. Additive manufacturing offers an alternative that has the potential to overcome this limitation.
For a successful industrial adoption of AM for parts requiring complex cooling channels, an understanding of the relationship between the as-built surface finish and heat transfer is needed. In LPBF, there are numerous build parameters, such as part orientation during the build, that affect the final part surface topography and hence heat transfer. The classic literature on the impact of surface roughness on heat transfer (Moody's diagram) uses a simplified treatment of surface roughness, while powder bed fusion processes generate complex surfaces with strong anisotropic features, spatter, and surface and subsurface defects, all of which may affect heat transfer and fluid flow.
The primary focus of this work is to study the effects of AM roughness characteristics (build orientations, density of spatter deposits and their sizes, amplitudes/wavelengths, etc.) on heat transfer from the corresponding AM surfaces and pressure drop across cooling channels. Both numerical and experimental investigations are carried out for this purpose. Computational Fluid Dynamics (CFD) models for mini-channels using StarCCM+ (a commercial CFD code) were developed by acquiring the roughness data from real AM surfaces with various roughness parameters. To explore the correlation between sand-grain modeled roughness and AM surface roughness with 90° build orientation, CFD simulations for the entire system model (experimental setup) were employed. Further, CFD modeling of mini-channels with different wavy surfaces helped in determining the suitable dimensions for the mini-channel experimental set-up and the range of Reynolds numbers necessary for carrying out relevant experiments. Based on CFD findings, an exchangeable experimental setup was developed and on the basis of roughness characterizations, the AM parts with three critical orientations (0°, 45°, and 90°) were fabricated and then machined to the required shape and size to fit into the set-up. In addition, an Inconel part with a smooth surface has been machined from a forged Inconel-625 circular bar to serve as the baseline control condition.
Both CFD and experimental results were compared for different Reynolds numbers. The experimental results validated the CFD findings. Significant differences in the Nusselt numbers and pressure drops were observed across the different AM surfaces, with the surface with 90° build orientation performing best in terms of heat transfer. Based on these results, further investigation on the effects of 90° weld tracked surfaces in a circular form was also carried out. For this exploration, two aluminum (Al-6061) channels - one with a smooth surface and the other with internal threads serving as artificial waviness, similar to an AM surface with a 90° build orientation to the fluid flow direction - were conventionally manufactured and both CFD and experimental investigations were carried out for different mass flow rates. Both CFD and experimental results show that the artificial waviness (structured surfaces) has a significant impact on heat transfer and leads to a high cooling efficiency with a Nusselt number approximately 3x larger for various flow conditions compared to the smooth channel. However, the intentional structured surface also leads to larger pressure drops and may require extra pumping power, depending upon the application.

Group purchasing (GP) is a procurement strategy by which the retailers can negotiate better prices by increasing their negotiation power through collaboration with each other. GP problem can be modeled as a generalized newsvendor problem, although it is more realistic to model this problem with stochastic demand, current literature on GP is mostly focused on problems with deterministic demand. Comparing the single retailer newsvendor vs. a newsvendor problem with multiple retailers, there has been more attention paid to the newsvendor problem with single retailer. When there are multiple retailers, competition would be another important aspect to consider, which is lacking in parts of the literature and will be considered in this research. Different contracting scenarios such as revenue-sharing and buyback contracts are other aspects which can be considered in the GP problem which has not been studied so far. Given that; four research questions are defined to investigate in this study: 1) the first question investigates the newsvendor problem with quantity discount pricing from supplier by exploring an analytical approach to solve this problem building on existing solutions from the literature; next a second novel solution approach is proposed which solves the problem in fewer steps; answering this question makes the foundation for our subsequent research questions. 2) the second research question studies the GP problem with multiple symmetric retailers; this research question is an extension of the first research question which investigates the GP supply chain consisting of multiple symmetric retailers. 3) third research question explores the solution to GP with multiple asymmetric retailers and suppliers; since this problem is complex to solve, the GP problem is divided into two sub-problems, retailers’ problem, and suppliers’ problem which are solved separately and then brought together to provide an answer to the overall GP problem, and 4) finally, fourth research question introduces different supply chain contracts to the GP problem and investigates studying the effect of these contracts on the retailers’ profit. Mathematical results as well as managerial insights are provided for each model through sensitivity analysis and numerical experiments.

This dissertation explores the potential of natural language models, including large language models, to extract causal relations from medical texts, specifically from Clinical Practice Guidelines. The outcomes of causality extraction from Clinical Practice Guidelines on gestational diabetes are presented, marking a first in the field. We also release, the first of its kind, an annotated corpus of causal statements in the Clinical Practice Guidelines.
We address the challenge of classifying causal sentences with a small amount of annotated data at the inter-sentence level by treating it as a cross-domain transfer learning problem. Obtaining these classified sentences is the first step in extracting causality. Furthermore, we delve into the importance of modal verbs and the degree of influence from cause to effect. We show the capability of three models (BERT, DistilBERT, and BioBERT) to identify the degree of influence in the text.
Lastly, we tackle the challenge of sparse annotated data for the causality extraction from Clinical Practice Guidelines by, again, using transfer learning. We investigate the correlation between data similarity and the efficacy of transfer learning. We also investigate a zero-shot and few-shot approach to cross-domain transfer learning and quantify the link between data similarity and success rates. With the cross-domain few-shot transfer learning, we achieve an F1-score of 81%, which suggests transfer learning as a possible solution to address the limited availability of annotated data.