- ItemDiscrete element method (DEM) calibration of wet materials for bulk handling.(Stellenbosch : Stellenbosch University, 2023-03) Scheffler, OC; Coetzee, CJ; Stellenbosch University. Faculty of Engineering. Dept. of Mechanical and Mechatronic Engineering.ENGLISH ABSTRACT: The discrete element method (DEM) has demonstrated potential as a design tool for assessing cohesionless (dry) materials when accurate input parameter values are specified. However, the calibration processes establishing precise contact model parameters and their values for bulk materials, with cohesion present, remain somewhat novel. Accordingly, the behaviour of cohesive materials was examined at the bulk level. To this end, moisture was employed to induce material cohesion in three different sand grades. Concurrently, the DEM approaches for simulating the latter were investigated, with two strategies emerging for calibrating the cohesive parameters depending on the level of cohesion in the material. To this end, the prevailing non-cohesive contact models were acquainted with their bulk calibration procedures. Moreover, the most favourable cohesive contact models were analysed, and their potential to replicate the bulk cohesive response of the moistened sand was examined. These included the Full Johnson, Kendall & Roberts (JKR), Simplified Johnson, Kendall & Roberts (SJKR), liquid-bridge and generic linear cohesive models. Subsequently, the linear cohesive model’s non-cohesive contact parameters were calibrated for six numerical particle upscaling factors, utilising the established calibration techniques. The particle-wall friction coefficient, damping coefficients and, significantly, the particle-particle friction coefficient remained constant as the particles were upscaled. However, the calibrated particle densities and contact stiffnesses increased with particle upscaling. Additionally, the linear cohesive model’s cohesive contact parameters were calibrated for the sands at four degrees of pore saturation, namely 0 % (dry), 5 %, 10 % and 15 %. The latter established the maximum tensile/rupture force (F) and rupture distance (D) at each upscaling factor. Moreover, the calibrated non-cohesive parameters could be kept constant for the test material in dry and wet conditions. The upscaling factors translated to ratios of 1.0 to 4.0 for the largest (. – . mm, "rough") sand grade, 2.5 to 9.9 for the middle (. – . mm, "medium") sand grade and 16.5 to 60.0 for the smallest (:. – . mm, "fine") sand grade. Unique F and D combinations for the "rough" and "medium" sands were obtained by combining (superimposing) the numerical replications of a vertical displacement angle of repose test, a draw-down test’s shear angle and the centroid elevation of the rotating drum’s cohesive particle bed, resulting in the first calibration strategy for establishing the cohesive contact parameters for mildly cohesive materials. In addition, a centrifuge was also utilised to analyse the materials’ slope angle at elevated lateral g-forces. A combination of the latter with the vertical displacement angle of repose was used to obtain unique F and D parameter values for the "fine" sand (through the superimposition of the F and D response surfaces) and resulted in the second calibration strategy for highly cohesive materials. Consequently, the F was found to scale cubically with numerical particle size, whilst the D was scale invariant. A minor increase of D with material pore saturation was also found. However, the bulk cohesive response was insensitive to the magnitude of D, but its facilitation of an attractive tensile branch greatly enhanced the bulk cohesive response of the wet material. The validity of the calibrated parameter sets was evaluated by comparing the bulk cohesive flow of the sands across a conveyor transfer point. To this end, an impact plate was placed in the path of the feeding stream, flowing off the conveyor. Accordingly, the maximum (peak) forces exerted on the material boundary during bulk flow and the residual forces (material weight) after bulk flow were analysed. Additionally, qualitative comparisons of the impact plate’sphysical and predicted cohesive pile formations were made. Only the three smallest scale factors provided enough resolution to accurately replicate the largest sand grade’s peak and residual forces. Furthermore, for the "medium" sand, the three smallest scale factors accurately replicated the peak and residual forces for the wet cases. For the "medium" sand’s non-cohesive dry case, the three smallest scale factors were accurate for the peak forces, whereas only the smallest scale factor was accurate for the residual forces. The "fine" sand grade’s forces were accurate for the five smallest upscaling factors when moisture-induced cohesion was present, whilst only the peak forces were accurately replicated with the three smallest scale factors for the non-cohesive case. Consequently, the degree to which particles may be upscaled increased as cohesion increased the characteristic lengths of the bulk material region being replicated. The latter translated to a minimum numerical resolution of four particle diameters for simulating the peak forces of particle-boundary interactions, whilst a resolution of eight particle diameters was required to replicate material build-up. Also, the qualitative replication of the material’s pile formation was sensitive to accurately representing the physical material’s particle size distribution, whilst quantitative measurements appeared insensitive to the latter. Moreover, employing the JKR, SJKR, or liquid-bridge contact models did not improve the modelling of the bulk cohesive behaviour.
- ItemModel-based exploration of the inter-relationships between diesel fuel properties and engine performance and exhaust emissions(Stellenbosch : Stellenbosch University, 2023-03) Schaberg, PW; Harms, TM; Groenwold, A; Stellenbosch University. Faculty of Engineering. Dept. of Mechanical and Mechatronic Engineering.ENGLISH ABSTRACT: This work utilises artificial neural network models to enable examination of the relationships between diesel fuel properties and engine performance and exhaust emissions in novel ways. The models were trained with experimental test data to accurately predict engine performance and exhaust emission parameters using duty-cycle, engine control, and fuel property parameters as inputs. The training data were collected during an engine test campaign conducted at the Sasol Fuels Application Centre in Cape Town, using a 15 litre, 373 kW heavy-duty diesel engine equipped with a common-rail fuel injection system and variable geometry turbocharger. The test fuels were formulated by blending market diesel fuels, refinery components, and biodiesel, to provide variations in pre-selected fuel properties, namely hydrogen to carbon ratio (H/C), oxygen to carbon ratio (O/C), derived cetane number (DCN), viscosity, and mid- and end-point distillation parameters. Care was taken to ensure that correlation between these fuel properties in the test fuel matrix was minimised, to avoid confounding model input variables. The test engine was exercised over a wide variety of transient test cycles during which fuel injection pressure, injection timing, air flow, and recirculated exhaust gas flow were systematically varied. The transient test data were used for training dynamic, recurrent neural network models so that transient engine operation could be accurately simulated. Modelling was performed in the MATLAB programming environment and the cluster supercomputer facilities at the national Centre for High Performance Computing were used for model training. The resulting models could predict the transient engine torque and fuel consumption, and nitrogen oxide (NOx), soot, carbon monoxide (CO), total hydrocarbon (THC), and carbon dioxide (CO2) exhaust emissions with good accuracy, which provided assurance that the characterisation of the test fuels using the selected fuel property parameters was sufficient to capture the fuel-related effects. The model inputs can be varied independently within the limits of the training dataset, enabling model-based parametric studies to be performed to quantify the relative impacts that the input variables have on engine performance and emissions. The newly-developed tool therefore allowed the effect of fuel properties to be examined through a new lens. NOx emissions were found to be primarily determined by the H/C and O/C ratios of the fuel, while soot was additionally impacted by DCN and viscosity. CO emissions showed the same trends as soot emissions, except that with DCN an opposite trend was observed. THC emissions were impacted by all fuel parameters but showed very little sensitivity to variations in engine control parameters. The models were also incorporated into a numerical optimisation routine which allowed synergies between fuel properties and engine control parameters to be identified to improve engine brake thermal efficiency (BTE). The H/C ratio was found to offer the greatest potential for improving the trade-off between BTE and NOx emissions. Besides providing a powerful new way to examine engine-fuel interactions, the tool can also be very useful for predicting the impact that new fuel formulations or fuel specification changes may have on engine performance and emissions.
- ItemSolar thermal treatment of manganese ores(Stellenbosch : Stellenbosch University, 2023-03) Hockaday, Aletta Carolina; Dinter, Frank; McGregor, Craig; Reynolds, Quinn; Stellenbosch University. Faculty of Engineering. Dept. of Mechanical and Mechatronic Engineering.ENGLISH SUMMARY: The use of sunlight as an alternative energy source is well established in the electricity sector, but mineral processing has lagged in adopting renewable energy sources. Specifically, the combustion of fossil fuels for process heat and using coal as a reductant are still the norm. It is possible to heat minerals directly with concentrating solar technology. Simultaneously, thermal decomposition reactions can cause the removal of oxygen or the decomposition of carbonate minerals without reductants. The use of concentrating solar technology for the pre-treatment of manganese ores was investigated in this light. Pre-heating, calcination and pre-reduction were identified as processes within the temperature range achievable by solar particle receivers. Manganese ores differ in the amount of carbonate minerals present and in the degree of oxidation of manganese. After chemical and mineralogical assaying of three South African manganese ores, their behaviour was investigated in thermogravimetric laboratory experiments. Tests were also done using a rooftop flat mirror parabolic dish concentrator to confirm that the laboratory studies may be used to inform the ores’ behaviour in on-sun conditions. Comparing the mass loss achieved during experiments to expected mass loss under equilibrium conditions, it was clear that mass loss was kinetically limited at temperatures below 950 °C. A dynamic reaction rate model was formulated from published literature to explain the ore behaviour in this temperature range. The model was validated against the measured mass loss data from all the experiments. Scaling of the technology was investigated for a 2.5 MWt solar plant. A radiation view factor model was formulated describing the rotary solar receiver energy flows. The reaction rate model was incorporated into this dynamic process model, enabling the calculation of the material composition in each receiver zone and for the products. The model was evaluated for daily and monthly periods in minute timesteps. The products were shown to produce energy savings and reduce greenhouse gas emissions when used in high-carbon ferromanganese production. The solar thermal plant process model showed that a 2.5 MWt solar plant can treat 5800 to 10600 metric tons of manganese ore per year at a levelised cost of heat of 380 R/MWh for 20-year project life. At the same time, while avoiding greenhouse gas emissions, this energy cost is lower than that of electric heating and heating with diesel combustion.
- ItemAdaptive digital image correlation using neural networks(Stellenbosch : Stellenbosch University, 2023-03) Atkinson, Devan James; Becker, Thorsten Hermann; Neaves, Melody; Stellenbosch University. Faculty of Engineering. Dept. of Mechanical and Mechatronic Engineering.ENGLISH SUMMARY: Subset size selection is crucial to the accuracy and precision of digital image correlation (DIC) measured displacements. Increasing the subset size improves noise suppression (reducing random errors) at the cost of spatial resolution (ability to accurately measure complex displacement fields). The tradition of global correlation parameter assignment is suboptimal because the speckle pattern quality and displacement field complexity can vary spatially. Dynamic subset selection (DSS), which assigns location specific optimal subset sizes, is challenging because the metrological performance of correlation is dictated by complex interactions between correlation parameters (subset size and shape function) and image set properties (noise, speckle pattern and displacement field complexity). This dissertation uses an open-source DIC framework to investigate the potential of artificial neural networks (ANNs) for error prediction and DSS, prior to the DIC process, from purely image information. ANNs are capable of modelling complex relationships within noisy, incomplete data without imposing fixed relationships, inspiring their recent resurgence for DIC applications. Despite the plethora of open-source DIC algorithms available, none offer spatially and temporally independent assignment of correlation parameters. Subsequently, a modular, open-source DIC framework capable of such flexibility is developed. This framework is predominantly consistent with current state-of-the-art practices and performs on par with well-established open-source and commercial DIC algorithms. Drawing direct links between the well-documented theory of DIC and its nuanced practical implementation, bridges this gap in literature which has acted as a barrier to newcomers intending to develop the capabilities of DIC. This framework, implemented in 117 and 202 lines of MATLAB code for 2D and stereo DIC, respectively, is attractive as a starting point to further the capabilities of DIC. The feed-forward ANN developed using this DIC framework, predicts random errors based on the speckle pattern quality (contained within a subset) and standard deviation of image noise more accurately and precisely than established theoretical derivations. A DSS framework is developed which uses this ANN to appoint subset sizes, based on the local speckle pattern, that offer random errors consistent with a stipulated threshold value. Appropriate selection of the random error threshold offers a favourable compromise between noise suppression and spatial resolution for up to moderate displacement gradients. Consequently, in the presence of varying speckle pattern quality this framework outperforms the traditional approach of trialand- error global subset size selection for the same mean subset size. Speckle pattern characteristics outside the training scope reveal the generalisability limitations of the DSS method, and associated ANN, as it performs on par with the traditional global subset size approach, motivating the need to broaden its training scope. Investigation of convolutional neural networks for dynamic shape function selection is initiated, showing they are capable of quantifying displacement field complexity between image pairs to guide spatially and temporally independent shape function assignment. The dissertation reveals that ANNs are an attractive approach to model the correlation parameter assignment. Furthermore, such models facilitate dynamic correlation parameter assignment from purely image information such that they can operate as a pre-process to DIC.
- ItemA design framework for aggregation in a system of digital twins(Stellenbosch : Stellenbosch University, 2022-04) Human, Carlo; Basson, AH; Kruger, K; Stellenbosch University. Faculty of Engineering. Dept. of Mechanical and Mechatronic Engineering.ENGLISH SUMMARY: The digital twin (DT) concept has become a popular means of capturing and utilising data related to physical systems and has been applied in many domains. The data provided within DTs allow for the integration of services and models to improve understanding and decision-making related to the physical system. Through aggregation, multiple DTs can be combined to represent larger, more complex system, while maintaining the separation of concerns. The design framework presented in this dissertation aims to enable systematic, effective decisions when designing a system of DTs to represent a complex physical system. In particular, this framework adopts hierarchical aggregation as one of its primary enablers and it considers the use of a services network, such as a service-oriented architecture, as well. The design framework is intended to be broadly applicable, by remaining vendor-neutral, and it enables traceability of design choices. The approach starts with an analysis of physical system complexity to identify key needs related to managing complexity. A suitable requirements classification is then introduced to help translate the needs into requirements that the system of DTs should satisfy. Hierarchical aggregation is also introduced as a primary architectural approach to manage complexity. Hierarchical aggregation allows for the separation of concerns, computational load distribution, incremental development and modular software design. The design framework is arranged in six steps: 1) needs and constraints analysis, 2) physical system decomposition, 3) services allocation, 4) performance and quality considerations, 5) implementation considerations and 6) verification and validation. The dissertation then introduces a general reference architecture that combines a system of DTs (which follows hierarchical aggregation principles) with a services network to allow for reliable and adaptable service provisioning. The design framework is then discussed in the context of the general reference architecture. The design steps of the design framework are then moulded into six design patterns, which simplify the design process by focussing of key quality attributes. The quality attributes considered for the respective design patterns are performance efficiency, reliability, maintainability, compatibility, portability and security. The use of the design framework and design patterns are then demonstrated and validated through three case studies, two high-level case studies and one detailed case study. The high-level case studies consider a water distribution system and a smart city, respectively. The detailed case study considers a heliostat field. The dissertation concludes that the design framework, as well as the design patterns, enable a systematic approach to designing a system of DTs. The design framework can also be applied to numerous and varying domains, such as the case studies considered.