Masters Degrees (Civil Engineering)

Permanent URI for this collection

Browse

Recent Submissions

Now showing 1 - 5 of 604
  • Item
    Adaptive conversational systems harnessing human expertise in modern chatbots
    (Stellenbosch : Stellenbosch University, 2024-03) Van Eeden, Christiaan; Du Preez, Johan; Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering.
    ENGLISH ABSTRACT: Chatbots are a rapidly evolving technology integral to the automation of customer service. Traditionally designed around complex if-then-else statements, the technology has seen a paradigm shift towards deep learning techniques for improved flexibility and user interaction. This research proposes a novel approach for training chatbots to learn more effectively from human support agents within automated customer service. Capitalising on the extensive data generated by these agents, we develop a framework to train chatbot models to emulate human-like responses, encapsulating nuances of unique language and individual mannerisms. Using state-of-the-art machine learning and natural language processing techniques, we train these models, achieving more contextually appropriate and authentic responses that accommodate the subtle complexities of human interaction. While commercial large language models (LLMs) like ChatGPT have demonstrated proficiency in customer service automation, they present limitations that could hinder their practical use. Such models, being proprietary, are not available for modification, impeding a company’s capacity to customise them to their specific needs. The costs, both financial and technical, associated with training a bespoke LLM can also be prohibitive for many organisations. Furthermore, these models, with their billions of parameters, require substantial hardware resources and may struggle to manage high-volume, swift interactions typical in a customer service environment. Additionally, due to their general-purpose nature, these models can occasionally produce unpredictable or undesirable responses as they lack specific domain knowledge. Our proposed model, developed on the more manageable GPT-2, offers a tailored, cost-effective, and adaptive solution to these challenges. Although our research is limited in its scope, the findings indicate an improvement in the usability and effectiveness of chatbots trained with our proposed method. This study contributes to the broader field of AI-driven customer service by augmenting the development of more sophisticated and user-friendly chatbot systems.
  • Item
    Evaluating five mine residue disposal facility case studies using a limit equilibrium method of slices.
    (Stellenbosch : Stellenbosch University, 2024-03) Greyling, Deon; MacRobert, Charles; Stellenbosch University. Faculty of Engineering. Dept. of Civil Engineering.
    ENGLISH ABSTRACT: In this study, five slope failure case studies, associated with mine waste facilities, such as Tailings Storage Facilities and Waste Rock Dumps, are evaluated using the Morgenstern-Price Limit Equilibrium Method of Slices. The available information of each study has been studied and key information about its investigation, soil test work, pre-failure slope geometry, and the piezometric surface have been extracted to develop a set of limit equilibrium slope stability models. The models were used to search for the critical slip surface and to back-calculate the shear strength of the critical soil region that satisfies a computed factor of safety of unity. The critical slip surfaces, determined from the limit equilibrium models, closely resemble the observed performance of the studied slopes. The back-calculated shear strength parameters of the critical soil region are compared to shear strength parameters derived from laboratory test results and typical values published in the literature. The back-calculated values compared well to the shear strength values published in the literature for similar soils. However, in two cases the back-calculated values did not compare well to values derived from laboratory test results. In both cases, direct shear tests were completed on fine-grained soil, which may have led to erroneous results. In two case studies, the back-calculated values compared well to the test results, and in another, the back-calculated shear strength is between the peak and residual shear strengths derived from laboratory test results. For three case studies, an additional analysis was done where two empirical correlations was used to specify the fully softened shear strength and residual shear strength of finegrained soil. The results of the slope stability models showed that the stability of the slopes is sensitive to the porewater pressure regime, emphasising that it should always be appropriately accounted for in the analysis of a slope that has a consequence of failure. Furthermore, the results of the analyses highlight that the engineering properties of the soils comprising the slope and its foundation should be sufficiently characterised and verified through appropriate geotechnical laboratory and in situ test work. The analyses affirm that the Morgenstern-Price Limit Equilibrium Method of Slices is a suitable tool to investigate slope stability provided that the piezometric surface, soil parameters and slope geometry are accurately defined.
  • Item
    The influence of pristine graphene on conventional concrete
    (Stellenbosch : Stellenbosch University, 2024-02) Van Wyk, Abraham Erasmus; Combrinck, Riaan; Stellenbosch University. Faculty of Engineering. Dept. of Civil Engineering.
    ENGLISH ABSTRACT: Population growth and rapid urbanisation have brought forth the demand for infrastructure development and the need for durable, low cost and accessible materials. Material compositions such as concrete are commonly used in the built environment. Cement is the primary binding agent in concrete. To produce one metric ton of cement, 0.59 metric ton of CO2 is emitted into the atmosphere (IEA, 2022). The Global Cement & Concrete Association (GCCA) implemented the Net Zero Emissions vision 2050 to reduce the industry’s CO2 emissions. Significantly reducing the amount of cement used in concrete is a viable strategy to lower CO2 emissions during this transitional period towards Net Zero Emissions 2050. This can be achieved by increasing the mechanical and durability properties of concrete whilst reducing the cement content. This research study aims to substantially improve the mechanical and durability properties of conventional concrete by using pristine graphene (PG), which has the potential to ultimately reduce the carbon footprint of cement-based materials (CBMs). PG used in concrete has shown the potential to improve concrete’s mechanical and durability properties significantly, but to unlock the full potential of PG has been one of the greatest challenges. Thus, to unlock the potential of PG and improve the mechanical and durability properties of conventional concrete, different PG application techniques (PGATs) were developed. These PGATs are known as pre-dispersion methods and have two functions namely, 1) reduce the number of layered graphene (exfoliate) and 2) facilitate the distribution of graphene throughout the cementitious matrix. This is done by using three different pre-dispersion methods namely, 1) wet pre-dispersion, 2) dry pre-dispersion and 3) combined (wet and dry) pre-dispersion. All three pre-dispersion methods were used in this study. The PG was obtained from First Graphene Ltd Australia. Two PG forms namely, flake (PUREGRAPH) and agglomerate (PUREGRAPH AQUA) were used in this study. The PG in flake and agglomerate form that was used has a lateral platelet size of 56.9 μm and 52.1 μm, respectively. Additionally, the PGATs aims to resolve some of conventional concrete’s oldest problems. Concrete is described as a heterogeneous material. Therefore, concrete is not uniform in composition due to the difference in stiffness of each traditional constituent (sand, cement and coarse aggregate). The difference in stiffness causes predominantly smaller sized particles to be situated in a region near the surface or aggregates when concrete is formed. This causes porous crystalline structures to form in a region around aggregates which are susceptible to mechanical failure and durability issues. This region is known as the interfacial transition zone (ITZ). Thus, this study strategically positioned PG in the ITZ to improve the bond between the mortar matrix and the surface of coarse aggregate to improve the mechanical and durability properties of conventional concrete. The compressive and in-direct tensile splitting strength of conventional concrete were improved by 18.8% and 7.6%, respectively after 56 days of water curing. Additionally, the amount of crystalline growth on the surface of the coarse aggregate were investigated to determine the bond between the mortar matrix and the surface of the coarse aggregate. The bond strength between the mortar matrix and the surface of the coarse aggregate was quantified by using a direct tensile test referred to as the ITZ strength test. PG improved the bond strength by 22.1% after 28 days of water curing. Furthermore, the oxygen permeability and water absorption were reduced by 42.8% and 17.4%, respectively. Therefore, PG applied to conventional concrete have shown the potential to improve the ITZ which directly influences the mechanical and durability properties. Although significant research still needs to be conducted, this study shows that PG could facilitate greener, stronger and longer lasting concrete as well as provide a partial solution to decrease the global cement consumption which will lower carbon emissions of the cement industry.
  • Item
    Development and testing of an automated time-motion detector for construction activities
    (Stellenbosch : Stellenbosch University, 2024-02) Meyer, Ewald Mauritz; Wium, Jan; Jurgens, Chris; Stellenbosch University. Faculty of Engineering. Dept. of Civil Engineering.
    ENGLISH ABSTRACT: This study looked into the intricacies of construction rework and the methods that have been used to identify, quantify and manage rework. From this research, a novel process of combining equipment, software and methodology is presented in the development of a concept of an automated time-motion detector system, with a proposed method to implement the system to quantify rework of structural elements on construction sites. The aim of this study is to develop and test a technological solution to detect time spent on construction activities and ultimately uncover un-announced rework on construction sites. This is achieved by completing the following objectives: (i) Determine the most feasible technological solution for the time-motion monitoring of construction activities that could lead to quality failures. (ii) Develop a time-motion detector that can be used to monitor construction activities remotely. (iii) Implement and test the system by performing simplified tests in a controlled environment to determine the accuracy of the system. (iv) Discuss results and future implementations. The motivation for this study was based on the challenge of efficient construction management and the inability to learn from quality failures on construction sites resulting from a lack of suitable data and documentation relating to the execution of construction activities. This lack of execution information hinders construction companies to become more efficient and profitable. In research, a shift is seen towards the gathering and analysis of data to improve the efficiency and profitability of construction companies after very successful implementation of data analysis and management in other fields of study. In this study, a proof of concept of an automated time-motion detector was developed in Python to record and analyse construction activities to determine the human capital required for these activities. The system was tested with several trial tests, where it was found that the system has an average accuracy of 84% while valuable operational data, limitations and recommendations for the use of this system were obtained.
  • Item
    Development and automation of an electronic H-TRIS fire testing system
    (Stellenbosch : Stellenbosch University, 2024-02) van Wyk, Ryno; Walls, Richard Shaun; Devine, Courtney; Streicher, Dirk; Stellenbosch University. Faculty of Engineering. Dept. of Civil Engineering.
    ENGLISH ABSTRACT: In this research a fire testing method, named the Heat-Transfer Rate Inducing System (HTRIS) using electric Radiant Panel Arrays (RPAs), is developed, automated and characterised. An investigation is conducted to study the suitability and benefits of the testing method in the field of fire engineering, as it is not currently a standard testing method. An existing H-TRIS has been developed and consists of propane-fired radiant panels mounted to a frame with a linear stepper motor controlling the emitting heat of the system. In contrast, the electronic H-TRIS (eH-TRIS) developed in this study comprises electric radiant panels mounted to a frame with the temperature of the radiant panels controlling the emitting heat of the system. The system is developed by designing and building (1) the experimental setup and (2) the control box of the testing method. The experimental setup consists of the radiant panels mounted to a trolley, and the control box comprises the hardware and electrical components used to control the system. The system is automated by developing a feedback control system which controls the temperature of the radiant panels based on the power supply. A Graphical User Interface (GUI) is developed to specify the system inputs and configuration parameters. The system is characterised by measuring the heat flux received at various positions at specified separation distances from the radiant panels. Based on operational tests conducted, an empirical model is developed to predict the heat flux received on a surface at a specified distance from the radiant panels. Similar to the cone calorimeter, the system is calibrated to ensure accuracy over time. The eH-TRIS enables the simulation of the behaviour of materials or systems experiencing different heat sources - a real fire, furnace test, or specified thermal boundary conditions of a fire model. This eH-TRIS ensures reproducible results at a relatively low cost compared to furnace tests. The contribution of this study is (1) the development and automation of an H-TRIS using electric RPAs (eH-TRIS), and (2) the quantification of the spatial heat flux variation on a receiving surface tested with the eH-TRIS. The advantage of the system is that the heat fluxes emitted can be controlled more accurately, more consistent spatial variations in heat flux can be produced, laboratory safety is improved and the system is adaptable to different configurations and setups. The primary disadvantage of the system is that it reaches lower maximum heat fluxes relative to gas burner setups. The affordability, the quick setup of tests and the variety of materials that can be tested, presents the H-TRIS as one of the more successful testing methods for modelling information.