Thesis (MSc)--Stellenbosch University, 2000. / ENGLISH ABSTRACT: A severely polluted industrial effluent is generated by the local gelatinmanufacturing industry. Due to increasingly stringent restrictions on discharge qualities enforced by the National Water Act of 1998 and National Environmental Management Act of 1998, as well as increasing trade-effluent charges implemented via the Local Municipal Bylaws, the industry is compelled to consider a system to pre-treat the polluted effluent. A study was undertaken to examine the viability of anaerobic treatment of the gelatin-manufacturing effluent, since the anaerobic digestion technology is well recognised for the high success rate in the treatment of high-strength, complex wastewaters. Various laboratory and pilot-scale studies were done, using different hybrid Upflow Anaerobic Sludge Blanket (UASB) and contact designs. Two mesophilic laboratory-scale hybrid UASB digester designs, fitted with polyethylene (AD-1) and polyurethane (AD-2), performed well at a hydraulic retention time (HRT) of 1.0 d. Chemical oxygen demand (COD) removal efficiencies of up to 90% (avg. 53%) for AD-1 and 83% (avg. 60%) for AD-2 at organic loading rates (OLR) of 9.56 and 4.62 kg COD.m-3.d-1, respectively, were obtained. High sulphate (S04) removal efficiencies of up to 96% (avg. 86%) for AD-1 and 98% (avg. 82%) for AD-2 were also achieved, respectively. A maximum total solid (TS) removal of 65% (avg. 25%) for AD-1 and 62% (avg. 28%) for AD-2 was reported. An average methane content of 80% (AD-1) and 79% (AD-2) with average methane yields per COD removed of 2.19 and 1.86 m3. kg CODremoved.df-o1r AD-1 and AD-2 were found, respectively. When the same digesters (AD-1 and AD-2) were combined in a muItiphase series configuration, a total COD removal efficiency of up to 97% (avg. 80%) at an OLR of 8.32 kg COD.m-3.d-1,was achieved. Excellent total S04 removals of 96% (avg. 69%) were accomplished. Up to 82% TS (avg. 29%) was also removed during this study and the biogas consisted of 89% methane (avg. 79%). For this multi-phase combination up to 92% volatile fatty acids (VFA) (avg. 48%) were removed, indicating possible selective phase separation of the respective fatty acid producing/utilising bacterial populations. The use of a laboratory-scale UASB bioreactor with recirculation, resulted in COD removal efficiencies of up to 96% (avg. 51%) at an HRT of 3.0 d, and 95% (avg. 54%) at a HRT of 1.0 d. Low performances were generally found, with average S04 and TS removals of 59% (max. 97%) and 26% (max. 67%), respectively at an HRT of 1.0 d. The biogas production was very low throughout the study (0.05 - 0.63 I,d-1 ). A pilot-scale UASB reactor (300 I) was constructed and performed satisfactory with a 58% average COD removal and maximum of 96%. S04 and TS removals up to 96% (avg. 44%) and 93% (avg. 63%), respectively, were obtained. The methane content of the biogas was 85%. The pilot-scale studies were conducted under actual field conditions, where various shock and organic loads had to be absorbed by the system. The pilot-scale contact configuration (300 I) did not perform satisfactory as a result of continuous blockages experienced in the feed and recirculation lines. Maximum COD, S04, VFA and TS removal efficiencies of 41% (avg. 27%), 62% (avg. 41%), 64% (avg. 27%) and 39% (avg. 21%), respectively, were obtained. The results of all the studies indicated acceptable COD removals with increasing OLR's. Indications of the presence of active methanogenic and sulphate-reducing bacterial populations were apparent throughout the studies. One possibility for the successful start-up and commissioning of the anaerobic reactors was the use of a well-adjusted biomass, which consisted of highly selected and adapted microbial consortium for the specific gelatinmanufacturing effluent. It was clear from this study that gelatin-manufacturing effluent can be treated successfully, especially with the use of the UASB design. A welldefined data base was constructed which could be of great value for further upscaling to a full-scale digester. / AFRIKAANSE OPSOMMING: 'n Hoogs besoedelde industriele uitvloeisel word gegenereer deur die plaaslike gelatien-vervaardigings industrie. As gevolg van toenemende streng beperkings op die kwaliteit van uitvloeisels wat bepaal word deur die Nasionale Water Wet van 1998 en Nasionale Omgewings Bestuurs Wet van 1998, asook toenemende munisipale heffings wat geimplementeer word via Plaaslike Munisipale Wette, word die industrie verplig om die uitvloeisel vooraf te behandel. 'n Studie is onderneem om die lewensvatbaarheid van anaërobe behandeling van gelatien-vervaardigings uitvloeisel te ondersoek, aangesien anaërobe verterings tegnologie alombekend is vir die goeie sukses behaal in die behandeling van hoë-sterkte, komplekse uitvloeisels. Verskeie laboratorium- en loods-skaal studies is gedoen, met verskillende hibried Opvloei Anaërobe Slykkombers (OAS) en kontak ontwerpe. Goeie werksverrigting was verkry by 'n hidroliese retensie tyd (HRT) van 1.0 d met twee mesofiliese laboratorium-skaal hibried OAS verteerder ontwerpe wat uitgevoer was met poli-etileen (AD-1) en poli-uretaan (AD-2) materiaal. Chemiese suurstof behoefte (CSB) verwyderings van so hoog as 90% (gem. 53%) vir AD-1 en 83% (gem. 60%) vir AD-2 by organiese ladingstempo's (OLT) van 9.56 en 4.62 kg CSB.m-3.d-1,was onderskeidelik verkry. Hoë sulfaat (S04) verwyderings van tot 96% (gem. 86%) vir AD-1 en 98% (gem. 82%) vir AD-2 was ook onderskeidelik verkry. 'n Maksimum totale vaste stof (TVS) verwydering van 65% (gem. 25%) vir AD-1 en 62% (gem. 28%) vir AD-2 is gerapporteer. 'n Gemiddelde metaan inhoud van 80% (AD-1) en 79% (AD-2) met 'n gemiddelde metaan opbrengs per CSB verwyder van 2.19 en 1.86 m3.kg CSBverwyder.dv-i1r AD-1 en AD-2, was onderskeidelik gevind. Met die aanwending van dieselfde twee verteerders (AD-1 en AD-2) in 'n series gekoppelde multi-fase konfigurasie, is 'n totale CSB verwydering so hoog as 97% (gem. 80%) verkry by 'n OLT van 8.32 kg CSB.m-3.d-1. Uitstekende totale S04 verwydering van 96% (gem. 69%) is behaal. Tot 82% TVS (gem. 29%) was vewyder gedurende die studie en die biogas het uit 89% metaan (gem. 79%) bestaan. Vir die multi-fase kombinasie is 'n maksimum van 92% vlugtige vetsure (WS) (gem. 48%) verwyder, wat dui op die moontlike skeiding van selektiewe fases van die onderskeie vetsuur produserende/verbruiker bakteriële populasies. CSB verwydering van tot 96% (gem. 51%) by 'n HRT van 3.0 d en 95% (gem. 54%) met 'n HRT van 1.0 d was verkry, tydens die gebruik van In laboratorium-skaal OAS bioreaktor met hersirkulasie. Lae werksverrigting was oor die algemeen waargeneem, met gemiddelde S04 en TVS verwyderings van 59% (maks. 97%) en 26% (maks. 67%) by In HRT van 1.0 d. Die biogas produksie was baie laag gedurende die studie (0.05 - 0.63 I,d-\ In Loods-skaal OAS verteerder was opgerig en bevredigende resultate was verkry met In gemiddeld van 58% CSB verwydering en maksimum van 96%. S04 en TVS verwyderings so hoog as 96% (gem. 44%) en 93% (gem. 63%) is onderskeidelik verkry. Die metaan inhoud van die biogas was 85%. Die loods-skaal studie was uitgevoer gedurende ware veld kondisies, waartydens verskeie skok en organiese ladings deur die sisteem geabsorbeer is. Die loods-skaal kontak konfigurasie (300 I) het nie bevredigende resultate getoon nie, as gevolg van voortdurende blokkasies wat ondervind is in die toevoer en hersirkulasie pype. Maksimum CSB, S04, WS en TVS verwyderings van 41% (gem. 27%), 62% (gem. 41%), 64% (gem. 27%) en 39% (gem. 21%) was onderskeidelik verkry. Die resultate van al die studies het aanvaarbare CSB verwydering aangedui by toenemende OLT's. Indikasies van aktiewe metanogene en sulfaat-reduserende bakteriële populasies was ook teenwoordig gedurende die studies. Die suksesvolle aansit-prosedure en begin van die anaërobe verteerders kan toegeskryf word aan die gebruik van In goed aangepaste biomassa, wat uit hoogs selektiewe en aangepaste mikrobiese populasies vir die spesifieke uitvloeisel bestaan. Hierdie studie het getoon dat gelatien-vervaardigings uitvloeisel suksesvol met die OAS ontwerp behandel kan word. In Goed gedefinieerde data basis kan voorsien word, wat van groot waarde sal wees vir verdere opgradering na In volskaalse verteerder.

published_or_final_version / Environmental Management / Master / Master of Science in Environmental Management

published_or_final_version / Environmental Management / Master / Master of Science in Environmental Management

A large number of machine learning algorithms are critically dependent on the underlying distance/metric/similarity function. Learning an appropriate distance function is therefore crucial to the success of many methods. The class of distance functions that can be learned accurately is characterized by the amount and type of supervision available to the particular application. In this thesis, we explore a variety of such distance learning problems using different amounts/types of supervision and provide efficient and scalable algorithms to learn appropriate distance functions for each of these problems. First, we propose a generic regularized framework for Mahalanobis metric learning and prove that for a wide variety of regularization functions, metric learning can be used for efficiently learning a kernel function incorporating the available side-information. Furthermore, we provide a method for fast nearest neighbor search using the learned distance/kernel function. We show that a variety of existing metric learning methods are special cases of our general framework. Hence, our framework also provides a kernelization scheme and fast similarity search scheme for such methods. Second, we consider a variation of our standard metric learning framework where the side-information is incremental, streaming and cannot be stored. For this problem, we provide an efficient online metric learning algorithm that compares favorably to existing methods both theoretically and empirically. Next, we consider a contrasting scenario where the amount of supervision being provided is extremely small compared to the number of training points. For this problem, we consider two different modeling assumptions: 1) data lies on a low-dimensional linear subspace, 2) data lies on a low-dimensional non-linear manifold. The first assumption, in particular, leads to the problem of matrix rank minimization over polyhedral sets, which is a problem of immense interest in numerous fields including optimization, machine learning, computer vision, and control theory. We propose a novel online learning based optimization method for the rank minimization problem and provide provable approximation guarantees for it. The second assumption leads to our geometry-aware metric/kernel learning formulation, where we jointly model the metric/kernel over the data along with the underlying manifold. We provide an efficient alternating minimization algorithm for this problem and demonstrate its wide applicability and effectiveness by applying it to various machine learning tasks such as semi-supervised classification, colored dimensionality reduction, manifold alignment etc. Finally, we consider the task of learning distance functions under no supervision, which we cast as a problem of learning disparate clusterings of the data. To this end, we propose a discriminative approach and a generative model based approach and we provide efficient algorithms with convergence guarantees for both the approaches. / text

<p>The thesis addresses time-domain inverse electromagneticscattering for determining unknown characteristics of an objectfrom observations of the scattered .eld. Applications includenon-destructive characterization of media and optimization ofmaterial properties, for example the design of radar absorbingmaterials.A nother interesting application is the parameteroptimization of subcell models to avoid detailed modeling ofcomplex geometries.</p><p>The inverse problem is formulated as an optimal controlproblem where the cost function to be minimized is thedi.erence between the estimated and observed .elds, and thecontrol parameters are the unknown object characteristics. Theproblem is solved in a deterministic gradient-basedoptimization algorithm using a parallel 2D FDTD scheme for thedirect problem.This approach is computationally intensive sincethe direct problem needs to be solved in every optimizationiteration in order to compute an estimated .eld.H ighlyaccurate analytical gradients are computed from the adjointformulation.In addition to giving better accuracy than .nitedi.erences, the analytical gradients also have the advantage ofonly requiring one direct and one adjoint problem to be solvedregardless of the number of parameters.</p><p>When absorbing boundary conditions are used to truncate thecomputational domain, the equations are non-reversible and theentire time-history of the direct solution needs to be storedfor the gradient computation.Ho wever, using an additionaldirect simulation and a restart procedure it is possible tokeep the storage at an acceptable level.</p><p>The inverse method has been successfully applied to a widerange of industrial problems within the European project,IMPACT (Inverse Methods for Wave Propagation Applications inTime-Domain).T he results presented here includecharacterization of layered dispersive media, determination ofparameters in subcell models for thin sheets and narrow slotsand optimization problems where the observed .eld is given bydesign objectives.</p>

Theoretical waste minimisation opportunities and options for electroplating were sought from the literature. Their suitability under the specific site conditions of a chromium electroplating plant were evaluated using the results of a waste minimisation audit (audit). The audit showed that many waste minimisation practices were already in place. These included counter current flowing rinse systems, multiple use of rinses and recycling of the drag-out solution back into the plating solution. Two types of information were collected during the audit, namely new chemical monitoring (concentration levels of sodium, iron, zinc, copper, lead, chromium and nickel and conductivity, total dissolved solids and pH) and flow rate data and existing data (composition of the process solutions, products and waste outputs, and raw materials, workpieces and utility inputs). The data were analysed using four established waste minimisation techniques. The Scoping Audit and the Water Economy Assessment results were determined using empirically derived models while the Mass Balancing and the True Cost of Waste results were obtained through more detailed calculations. The results of the audit showed that the three most important areas for waste minimisation were water usage, effluent from rinse water waste streams and nickel consumption. Water usage has the highest waste minimisation potential followed by nickel. Dragged-out process chemicals and rinse water consumption contribute to ranking the effluent stream the most important waste minimisation opportunity identified by the True Cost of Waste Analysis. Potential financial savings were roughly estimated to be in the order of R 19949 and R 126603 for water and nickel respectively. Intervention using only "low cost-no-cost" waste minimisation measures was recommended as a first step before contemplating further focus areas or technical or economical feasibility. / Thesis (M.Sc.)-University of KwaZulu-Natal, Pietermaritzburg, 2004.

This study involved an assessment of the Pietermaritzburg Waste Minimisation Club (PWMC) during 2001, and a waste minimisation audit conducted at two coil coating lines. Waste minimisation is the reduction or elimination of waste at source4 • It is often achieved through waste minimisation clubs which comprise a small number of companies, generally in the same geographicaI area, that are interested in reducing waste 1"711 . The success of the PMWC was evaluated in terms of the degree to which members implemented waste minimisation principles. Two questionnaires were used to assess the success of the club. These indicated that although the training material gave the members a good understanding of the basic principles of waste minimisation, the material has not given the members enough practical information to implement a waste minimisation programme in their companies. The main barriers to waste minimisation identified include production pressure, operational constraints, lack of human resources and a lack of management time. The drivers for waste minimisation were financial savings, improved plant utilisation and improved environmental performance. Coil coating is a continuous process where a coiled sheet of aluminium is cleaned, pretreated and coated with paint. The flow rates, compositions and costs of all input and output streams to the cleaning and pretreatment sections were gathered from operators or measured. The data were collected over a three-month period to obtain a representative sample, and then analysed to determine waste minimisation opportunities using mass balances, monitoring and targeting, a scoping audit and a true cost of waste assessment. The scoping audit was found to be the most useful technique because it accurately prioritised the waste minimisation opportunities but required a relatively small amount of data for its application. However, the scoping audit underestimated the savings that could be achieved at the coil coating department and therefore the 'scope to save' percentages, which were developed for United Kingdom industries, need modification to better reflect South African industry. Opportunities for waste minimisation on Coil Coating Line 1 (CCL1) included reducing the water consumption, reducing the acid and chromium raw materials consumption, and finding a cheaper energy source for heating the process and rinse tanks. Potential fmancial savings of R116 000 and environmental savings of 18 200 kL of water or effluent per year were calculated for CCL1. The chromium and acid effluent treatment and solid waste disposal are the main areas for waste minimisation on Coil Coating Line 2 (CCL2). Savings could be achieved in these areas by using roller application ofthe chromium pretreatment rather than spray application, and by preventing a leak of chromium pretreatment into the acid process and rinse tanks. Other savings can also be achieved by operating the chromium process tank as a fed-batch process, and operating the alkali and acid process tanks as continuous processes at the specified chemical concentrations and with recycle of the rinse water (dragout). The total financial savings that can be achieved on CCL2 are R5.3 million, and potential environmental savings are 31 600 kL ofeffluent per year. / Thesis (M.Sc.)- University of Natal, Pietermaritzburg, 2002.

Our ability and capacity to generate, record and store multi-dimensional, apparently unstructured data is increasing rapidly, while the cost of data storage is going down. The data recorded is not perfect, as noise gets introduced in it from different sources. Some of the basic forms of noise are incorrect recording of values and missing values. The formal study of discovering useful hidden information in the data is called Data Mining. Because of the size, and complexity of the problem, practical data mining problems are best attempted using automatic means. Data Mining can be categorized into two types i.e. supervised learning or classification and unsupervised learning or clustering. Clustering only the records in a database (or data matrix) gives a global view of the data and is called one-way clustering. For a detailed analysis or a local view, biclustering or co-clustering or two-way clustering is required involving the simultaneous clustering of the records and the attributes. In this dissertation, a novel fast and white noise tolerant data mining solution is proposed based on the Crossing Minimization (CM) paradigm; the solution works for one-way as well as two-way clustering for discovering overlapping biclusters. For decades the CM paradigm has traditionally been used for graph drawing and VLSI (Very Large Scale Integration) circuit design for reducing wire length and congestion. The utility of the proposed technique is demonstrated by comparing it with other biclustering techniques using simulated noisy, as well as real data from Agriculture, Biology and other domains. Two other interesting and hard problems also addressed in this dissertation are (i) the Minimum Attribute Subset Selection (MASS) problem and (ii) Bandwidth Minimization (BWM) problem of sparse matrices. The proposed CM technique is demonstrated to provide very convincing results while attempting to solve the said problems using real public domain data. Pakistan is the fourth largest supplier of cotton in the world. An apparent anomaly has been observed during 1989-97 between cotton yield and pesticide consumption in Pakistan showing unexpected periods of negative correlation. By applying the indigenous CM technique for one-way clustering to real Agro-Met data (2001-2002), a possible explanation of the anomaly has been presented in this thesis.

When we interpret the input vector of a Boolean function as a binary number, we define interval Boolean function fn [a,b] so that fn [a,b](x) = 1 if and only if a ≤ x ≤ b. Disjunctive normal form is a common way of representing Boolean functions. Minimization of DNF representation of an interval Boolean function can be per- formed in linear time. The natural generalization to k-interval functions seems to be significantly harder to tackle. In this thesis, we discuss the difficulties with finding an optimal solution and introduce a 2k-approximation algorithm.

This paper presents a two-stage model of optimization as a quick method to choose the best potential links for implementing urban travel demand management (UTDM) strategy like road pricing. The model is optimized by minimizing the hidden cost of congestion based on user equilibrium (MHCCUE). It forecasts the exact amount of flows and tolls for links in user equilibrium condition to determine the hidden cost for each link to optimize the link selection based on the network congestion priority. The results show that not only the amount of total cost is decreased, but also the number of selected links for pricing is reduced as compared with the previous toll minimization methods. Moreover, as this model just uses the traffic assignment data for calculation, it could be considered as a quick and optimum solution for choosing the potential links.