Masters Degrees (Chemical Engineering)

Permanent URI for this collection

https://scholar.sun.ac.za/handle/10019.1/782

Browse

Now showing 1 - 5 of 5

A feasibility study of elementary reinforcement learning-based process control
(Stellenbosch : Stellenbosch University, 2022-04) Bras, Edward Hendrik; Louw, Tobias Muller; Bradshaw, Steven Martin; Stellenbosch University. Faculty of Engineering. Dept. of Process Engineering.
ENGLISH SUMMARY: The classical control paradigm is widely used in industry, has well-understood theoretical guarantees, and forms part of the foundational knowledge of chemical engineers. Challenging non-linear dynamics prevent its successful application in certain cases, while classical controllers cannot automatically accommodate changing closed-loop dynamics. Advances in computational capabilities have led to a significant research interest in the application of Reinforcement Learning (RL) to control processes. In RL, a computational agent interacts with an environment to maximise the cumulative scalar rewards received. It may be viewed as an alternative paradigm for control, as is done in this thesis, or as an approach to potentially enhancing the performance of classical controllers. This simulation-based study’s purpose is to investigate the feasibility of elementary RL techniques to automatically determine the final element adjustments in a single-loop RL-based control scheme. It places into context what the strengths and limitations are of using elementary RL to control processes and highlights nuances of RL-based control without trying to outperform classical control. The control of a self-regulatory water tank model and the Van de Vusse reaction scheme model (used for benchmarking and requires advanced control solutions) were studied by applying two algorithms – Q-learning and SARSA – in a control scheme synthesized purely for theoretical study. Subsequently, these algorithms and the One-Step Actor-Critic algorithm were applied to the control of particle size in a qualitatively accurate grinding circuit model. All simulations leveraged the simplest possible RL design to allow interpretable and clear accounts of how these systems behave. The results show that the use of elementary RL techniques to obtain interpretable RL-based controllers for simulation-based study worked well for the water tank and Van de Vusse reaction scheme models. This was not the case for the grinding circuit case study. Replacing the classical control paradigm is not likely using elementary RL. Significant safety concerns arise since large amounts of operational data may be required and insufficient training in certain regions of the state-action space leads to unpredictable control behaviour. The strengths and weaknesses of the algorithms studied were investigated. It is unlikely that a reduction of control loop specific tuning parameters in comparison to classical control will be realised in practical control problems by applying RL-based control. Where applicable, classical control outperformed the elementary RL-based controllers which stresses that algorithmic adjustments are required, as is recognised in state-of-the-art RL-based control approaches. To conclude, the most practically feasible RL-based control solutions are likely to lie in the enhancement of existing control solutions by incorporating RL principles. The studied elementary RL-based control methods are not feasible for practical robust control. The control engineer must not be removed completely from the loop, and existing domain knowledge must be reconciled with computational thinking instead.
Investigation of submerged trommel screen
(Stellenbosch : Stellenbosch University, 2022-04) Laker, Chris James; Akdogan, Guven; Bradshaw, Steven Martin; Stellenbosch University. Faculty of Engineering. Dept. of Process Engineering.
ENGLISH SUMMARY: Trommels are investigated as an alternative to drain and rinse vibrating screens for the purpose of dense medium recovery. Iron ore and cyclone 40 atomized ferrosilicon were used to test screen performance on a submerged washing trommel. This trommel consists of two types of chambers, one for medium drainage and two submerged wash chambers to wash adhering medium. The performance of the drain chamber was evaluated with overflow properties such as moisture content, FeSi content and FeSi carryover, by varying medium relative density (RD) between 2.7 and 3.6. The results were compared to past studies on a vibrating drain and rinse screen by Kabondo (2018). The test work was performed on two separate trommel designs, Submerged DMS Trommel (Mk1) and Trommel Mk2. Due to design limitations in the design of Mk1, there was great uncertainty in the results from the first investigation. Mk1’s performance was however highly promising compared to vibrating drain screens. Percentage moisture in the overflow of Mk1 had a maximum of 3.60 % at a medium RD of 3.6, while vibrating screen results ranged between 10.29 % and 27.91 % at lower medium RDs below 2.3. % FeSi in the overflow of the drain chamber of Mk1 ranged between 11.05 % and 18.76 %, while the vibrating screen ranged between 9.71 % and 36.01 %. % FeSi carryover on Mk1 reached a maximum of 4.96 % while the vibrating screen ranged between 3.12 % and 10.46 %. In addition to competitive drainage of trommel Mk1, batch tests performed in the submerged wash chambers of Mk1, combined with bench tests for submerged washing and rinsing, concluded that the most effective washing method was motion of particles within a submerged bath. Efficiency for submerged washing ranged from 84.82 % to 99.7 %, compared to 74.24 % for rinsing of medium. It was justified to design a new test trommel, Mk2, from the learnings of the first campaign. The second trommel (Mk2) was designed with nine different underflow discharges. For the first time in open literature, trommel Mk2 provided insight into material distribution and utilised screen area of a trommel by evaluating these discharges. It was found that 98 % of drainage occurs through the middle and towards the direction of rotation of the trommel. The work on trommel Mk2 was performed at similar operating conditions as Mk1. The results of Mk2 had higher repeatability and statistical significance. Confidence was provided in the results and it was concluded that trommels is a competitive alternative to vibrating drain and rinse screens. Future work will include test work on all remaining operating parameters to develop a complete understanding how these parameters govern the operation of trommels.
Optimality assessment with optimality recovery for multi-modal process operations
(Stellenbosch : Stellenbosch University, 2023-03) Meyer, Tanya; Louw, Tobias Muller; Bradshaw, Steven Martin; Stellenbosch University. Faculty of Engineering. Dept. of Process Engineering.
ENGLISH ABSTRACT: The field of optimality assessment (OA) is a recent development within data-driven process monitoring. OA is a plant-wide approach to real-time optimisation that aims to minimise nonoptimal online operation caused by (1) disturbances that cannot be rejected by the regulatory control system, or (2) inevitable controlled variable setpoint drift. The distinguishing design factor of OA, as opposed to fault- and quality-related process monitoring, is the incorporation of the comprehensive economic index to quantify overall plant optimality or performance. Since optimality is only available in retrospect, the estimation of optimality during real-time operation allows for prompt intervention when nonoptimal conditions arise, so as to prevent prolonged conditions of deteriorated Performance. This work proposes an alternative to the conventional latent variable model-based OA workflows, which employ monitoring charts founded on Shewhart- or similarity-based statistics. The proposed OA workflow is designed to account for continuous and multimodal industrial process data without transition states. The workflow is developed under the framework of a novel optimality landscape which captures various stable modes in the historical process dataset as well as their associated optimality grade. In addition, the proposed optimality landscape captures the cause for the historical operating point shifting from one mode to another, which is termed a modal shift. Two types of modal shifts are captured, namely those that are caused by disturbances, or by SP change(s) that are implemented by the control system or operational team. The offline phase of the proposed OA workflow constructs a holistic reference tool called the optimality graph. The nodes of the optimality graph are discovered by 𝑘-means clustering in the latent variable space, whereas the edges are discovered by the proposed TASLA (time-based alignment of modal shifts and plant log algorithm) technique. Furthermore, metrics are developed for selecting hyperparameters that result in an optimality graph which best reflects the optimality landscape. The online phase of the proposed workflow essentially projects online conditions onto the historical optimality graph using latent variable techniques, such that the closest reference mode is identified. Consequently, real-time conditions are assigned the optimality grade of the identified reference mode. The online optimality graph reveals which actions can be implemented to shift the online operating point toward modes of differing optimality. A key outcome of this work is the utility of the online optimality graph as an optimality advisory that provides the operational team with historically substantiated decision support for implementing optimality recovery. The performance of the proposed OA workflow is tested using a simulated Tennessee-Eastman Process dataset. The proposed workflow captures the historical optimality landscape of the pseudo-industrial dataset and estimates real-time optimality well upon comparison to the ground truth. The performance of each latent variable extraction technique – namely PCA, PLS, ICA – is evaluated by considering the spread of optimality within each mode of the related optimality graph. PLS is deemed the most suited to extracting features that are reflective of the optimality landscape. The holistic nature of the proposed PLS-based OA workflow offers a good alternative to existing latent variable model-based OA workflows.
Optimizing a method for leaching PGMs from simulated spent autocatalyst material using ozone & hydrochloric acid
(Stellenbosch : Stellenbosch University, 2024-03) Knight, Marcus Alexander; Bradshaw, Steven Martin; Akdogan, Guven; van Wyk, Andries Pieter; Stellenbosch University. Faculty of Engineering. Dept. of Chemical Engineering. Process Engineering.
ENGLISH ABSTRACT:Catalytic converters are present in all modern vehicles and contain the PGMs Pt, Pd, and Rh. These PGMs act as catalysts for the oxidation of C, H, and O to CO2 and H2O; and the reduction of NOx to N2, H2O, CO2, and NO. Mining is the primary source of these metals, but much research and development has been conducted into recovery from secondary sources such as e-waste and scrap. Pyrometallurgical methods are most popular for large scale recycling but produce toxic fumes such as sulphur oxides and have a significant electrical requirement. Hydrometallurgical methods show good extraction efficiencies, especially on a smaller scale, but trade the high energy requirements and toxic emissions of pyrometallurgical processes for increased reagent requirements. A need is present for continued investigation into optimizing processes for recovering these PGMs from spent auto-catalysts using more environmentally benign hydrometallurgical methods. The use of ozone as an oxidant in chloride leaching is one option. The gas has a high oxidising potential and is safer and less aggressive when compared to other popular oxidants such as chlorine gas. Therefore, the aim of this investigation was to develop a process for leaching Pt, Pd, and Rh from simulated spent autocatalyst material using ozone and hydrochloric acid. Development of the chosen process was carried out through kinetic and statistical analyses. Experimentation was divided into two stages: Leach 1 and Leach 2. Both leaches were carried out using a Box-Behnken experimental design. The O3 mass flow, initial HCl concentration, and temperature were varied at factor levels of 3.34, 5.01, and 6.68 g/h; 1.0, 3.0, and 5.0 M; and 30, 60, and 90oC respectively. Fifteen runs were conducted with a centrepoint triplicate for each leach. From the results and a statistical analysis, it was concluded that when maximizing the overall PGM extraction the leach was optimized at 5.01 g/h O3, 5.0 M HCl, and 90oC with Pt, Pd, and Rh extractions of approximately 80%, 85%, and 42%. However, at these conditions a significant degree of impurity extraction was observed at 67% and 68% for Al and Mg. Due to the high dependency of Rh and these two impurities on temperature, a desirability analysis determined that adjusting the factor setpoints could potentially facilitate the selective extraction of Pt and Pd (at values >80%) while minimizing the extraction of Rh and impurities. This was the basis for Leach 2. The second leach was conducted at the same O3 mass flows, HCl initial concentrations of 4.0, 5.0, and 6.0 M, and temperatures of 25, 30, and 35oC. The analysis from this leach indicated optimum conditions of 3.34 g/h, 5.0 M, and 25oC. The extractions at these conditions were 64% for Pt and Pd, and Rh, Si, Al, and Mg extractions were 1.3%, 2.6%, 2.8%, and 4.1%. Therefore, the minimization of the Rh and impurity extraction was successful. However, throughout experimentation a decrease in the extractability of the PGMs was observed over time. This is highlighted by the difference in the optimum results for Pt and Pd between Leach 1 and 2, and the similar conditions used to facilitate these extractions. A possible cause of this decrease is the gradual hydroxylation of the PGM oxides as a result of their exposure to humid air. However, further investigation is required to establish exactly what decreased the PGM extractability over time.
State estimation and model-based fault detection in a submerged arc furnace
(Stellenbosch : Stellenbosch University, 2023-12) Kristensen, Isabella; Louw, Tobias Muller; Bradshaw, Steven Martin; Stellenbosch University. Faculty of Engineering. Dept. of Chemical Engineering. Process Engineering.
ENGLISH ABSTRACT: Model-based state estimators use noisy plant measurements and a process model to calculate accurate and timely estimates of the state variables for process monitoring, model-based fault detection, and model predictive control. The aim of this project was to perform model-based fault detection using state estimation in a complex chemical unit operation and compare the model-based fault detection to a datadriven technique under plant-model mismatch. A system observability analysis and fault detectability analysis was first conducted. The performance of the various nonlinear state estimation techniques, namely the extended Kalman filter (EKF), the unscented Kalman filter (UKF), the particle filter (PF), and the moving horizon estimator (MHE), was then assessed, enabling the selection of appropriate state estimation techniques for model-based fault detection. Model-based fault detection was employed using the residuals generated from the state estimators followed by residual evaluation using PCA. The modelbased fault detection was compared to data-driven fault detection using PCA on the measurements and the effect of plant-model mismatch on the performance of model-based fault detection was investigated. A submerged arc furnace (SAF) for platinum group metal smelting was used as a case study to apply these techniques. The state observability analysis found the SAF system to be locally observable and the measured states to have a higher degree of observability than the unmeasured states. Upon implementation of the state estimation algorithms, the least observable states corresponded to states estimates with the largest estimation error. The fault detectability analysis identified all faults investigated to be structurally detectable. Upon implementation of model-based fault detection, it was concluded that the more structurally detectable a fault is, the better the fault detection performance. The investigation into state estimation in the SAF showed that the EKF, UKF, and PF display good estimation accuracy and fast computation times. The PF showed superior estimation accuracy under low process noise conditions and was selected for model-based fault detection. The EKF, being the most popular algorithm in literature and displaying fairly good estimation accuracy, was selected as the second method. The computational requirements of the MHE proved to be its greatest limitation. Investigations were carried out into reducing the computational load of the method using alternative singular perturbation SAF model with larger integration steps which halved the computational requirements. However, the computation times remained inappropriate for application in model-based fault detection. Lastly, this study found that the model-based fault detection using the PF residuals outperformed the model-based fault detection using the EKF residuals and the data-driven PCA method for detection of faulty conditions within the SAF process. Due to the sensitivity of the PF residuals resulting from the nature of the algorithm, this method showed exceptionally poor robustness to plant-model mismatch. The investigation then demonstrated that residual evaluation of the PF and EKF residuals in a reduceddimensional space using PCA improved the classification performance of the method when plant-model mismatch was present. However, when no modelling error is present, the classification of PF and EKF residuals showed the best performance in the original dimension space.

Browse

Browsing Masters Degrees (Chemical Engineering) by browse.metadata.advisor "Bradshaw, Steven Martin"

Results Per Page

Sort Options