Faculty of Engineering, Built Environment and Information Technology
School of Engineering
Department of Chemical Engineering
Selected Highlights from Research Findings
Most chemical products, for example shampoos, are simply physical mixtures of different ingredients. However, a myriad of subtle interactions actually occur among the ingredients of micro-structured multiphase-materials.
For the optimum formulation of such products a continuous and analytic representation is needed to account for the variation of physical properties as the fractions are changed. Models based on fundamental theory are rarely applied, because at present they are unable to account for these complex interactions.
A neural network model for the effect of mixture composition on physical properties has been developed. It is conventional to express composition in terms of the mass fractions x1, x2… xn with xi ³ 0 and x1 + x2, …. + xn = 1.
The model is a three-layer perceptron network with the following specifications: Each of the n components in the mixture is associated with a specific input node-hidden layer neuron pair. The hidden neurons apply a transfer function to the sum of weighted inputs xi. The output neuron employs the input xi as variable weights and applies the inverse of the transform utilised by the hidden neurons.
The resulting model has up to n2 adjustable parameters with n determined by pure component behaviour. Advantages of this network architecture are that binary data sets suffice for “training” and that it is readily extended to incorporate additional species while retaining all previously determined weights.
Dr WW Focke
Institute of Applied Materials
+27 (0) 12 420 2588
walter.focke@up.ac.za
Biological techniques for treating toxic wastes have gained popularity as they offer the capability to treat pollutants under natural conditions with reduced production of unwanted by-products. Biological treatment processes are thus regarded as both environmentally friendly and cost effective in the long term. In this study, the biochemical pathway for Cr(VI) reduction was studied in two known Cr(VI) reducing microorganisms.
The hexavalent form of chromium, Cr(VI), is a known carcinogen and toxic agent to both microbial and mammalian cells. Cr(III) on the other hand, is 1000 times less toxic than Cr(VI) and is a required cofactor for metabolism of carbohydrates in mammals. For example, chromium (III) picolinate is prescribed routinely to human subjects in weight loss programmes. Additionally, Cr(III) is less mobile in the environment and is easily precipitated as Cr(OH)3.
Two mechanisms for biological Cr(VI) reduction were investigated in Escherichia coli and Bacillus sp. by blocking and uncoupling electron-flow through the membrane-associated electron-chain respiratory pathway. Energy conservation through metabolic processes was interrupted by cell growth inhibitors but Cr(VI) reduction continued under an uncoupler (2,4-dinitrophenol (2,4-DNP)).
E. coli cells grew in a coculture with the phenol-degrading species, Pseudomonas putida. In the coculture, E. coli cells utilised metabolites formed from phenol degradation by the P. putida. Phenol was used as the sole carbon source and electron donor in the P. putida/E. coli coculture whereas glucose was used in the Bacillus sp. pure culture system.
Cell behaviour under the uncoupler was determined to be equivalent to that observed under Cr(VI) overloaded conditions in a continuous-flow bioreactor system. Cr(VI) reduction activity was optimal under relatively high Cr(VI) loadings but respiratory activity was minimal. Thus, it was concluded from these results that cells may of necessity prioritise Cr(VI) reduction to create a tolerable environment before redirecting energy-yielding reducing equivalents towards cell growth and maintenance processes.
The determination of the Cr(VI) reduction pathway in bacteria will allow accurate modelling and optimisation of Cr(VI) reduction processes for many chromium polluted waters in South Africa. Sources of Cr(VI) in South Africa include ferrochrome foundries, platinum refineries, leather tanneries, electroplating plants, and paper and pulp processing plants (Mintek, 1994). South Africa is estimated to hold approximately 72% of the world’s chromium reserves (U.S.EPA, 2001). Much of this reserve is freed from ore during the manufacturing of stainless steel.
Biological removal of Cr(VI) remains a viable option for cleanup of pollution due to accidental spills as well as due to intentional discharge. One such example is the recent accidental spill of Cr(VI) at the Brits facility of the now-defunct African Chrome Smelter on 8 December 2005 which resulted in the pollution of the Crocodile River and the adjacent groundwater aquifers (DWAF Press Release, December 2005).
Prof Evans Chirwa
Chemical Engineering
+27 12 (0) 420 5894
evans.chirwa@up.ac.za
In trickle bed reactors the existence of multiple hydrodynamic states in the trickle flow regime is a well documented, but as yet poorly quantified phenomenon. The prewetting procedure is known to drastically impact the hydrodynamic state and morphology of the liquid flow. In this work, three limiting cases for the case of no gas flow were identified: Non-prewetted, Levec-prewetted, and pseudo-Kan-prewetted.
For the quiescent air-distilled water-3 mm glass spheres system, it was shown that bed-scale maldistribution exists in all three modes of operation, but is more severe in the Levec- and nonprewetted modes. The incorporation of a novel volumetric utilization coefficient into an existing momentum balance-based hold-up model was sufficient to accurately model the hold-up in all prewetted modes. This results in the first hold-up model that takes the prewetting procedure into account. Present results support the notion that Kan-type prewetting results in film flow, while Levec- and nonprewetted modes are dominated by rivulet flow.
Dr W Nicol
Chemical Engineering
+27 (0) 12 420 3796
willie.nicol@up.ac.za
Finding a suitable control structure for any process usually involves comparing the performance of different possible control structures and choosing one which best satisfies chosen criteria. It is desirable to do this performance comparison off-line, as installation of a sub-optimal controller will cost both time and money.
Monte Carlo modelling provides a well-documented method of evaluating the statistical properties of stochastic systems. Applied to control system design, Monte Carlo modelling can incorporate detailed process models and accurate estimates of input distributions to give an accurate estimate of the effect of different control strategies on the system.
In this study, Monte Carlo modelling was used to compare three candidate controllers in order to determine the best in terms of two criteria, namely variance reduction and set point tracking. The modelling technique yielded results that could be interpreted without difficulty, showing one controller to be clearly superior to the others according to these criteria. These results can be used to implement the best controller without expensive trial and error procedures.
In situ experiments on an operational digester correlated well with the simulation results, showing the best controller to reduce variance by 43% and reduce the mean error by 90% when compared to the controller currently in use. It was shown that, due to the increasing availability of powerful computing systems, Monte Carlo modelling is a viable technique for controller performance analysis on highly non-linear processes.
Prof PL de Vaal
Chemical Engineering
+27 (0) 12 420 2475
philip.devaal@up.ac.za
A mathematical formulation was developed for freshwater and wastewater minimisation in multipurpose batch plants through the exploitation of recycle and reuse opportunities. A superstructure that entails all possible recycle and reuse possibilities was used as the basis for the formulation.
The existence of an optimal production schedule, which provides starting and finishing times for water using operations, was assumed. The work addressed the specific case where the contaminant concentration in the outlet water stream from each operation was maintained at its maximum.
This condition allows the overall model to be cast as a mixed integer linear programming (MILP) problem for which global optimality is guaranteed. The effectiveness of the developed formulation was demonstrated through a published literature example and a practical case study.
Prof T Majozi
Chemical Engineering
+27 (0) 12 420 4130
thoko.majozi@up.ac.za
|
