Optimality in sewer network design

De Villiers, Nico-Ben (2018-03)

Thesis (PhD)--Stellenbosch University, 2018.

Thesis

ENGLISH ABSTRACT: Sewer networks form a vital part of urban infrastructure and represent a capital intensive expense in the public sector. In South Africa, a recent estimation stated that R44.75 billion is required to provide basic sanitation services to those in need, while the budget allocation for sanitation was approximately R3.2 billion nation wide. This state of affairs is not limited to South Africa alone. It is clear that funds allocated to sanitation have to be spent as effectively as possible. It is within this space that this dissertation makes a contribution, by exploiting modern digital computing power and optimization techniques to find more cost effective solutions to sewer network design and analysis problems. This investigation proposes and evaluates new solutions that address critical shortcomings in existing sewer network optimization algorithms. The distinct lack of adequate benchmark problems to test and compare the performance of new sewer network optimization algorithms is addressed and a thorough investigation into the nature of optimal solutions is made by capturing trial solutions of optimization algorithms at regular intervals. The characteristics of the solutions are quantified through proposed network characteristic parameters. The evolution of trial solution characteristics allow insight and understanding into how the network changed over the course of the algorithm's life cycle. This information is then exploited to propose heuristic influence factors which guide the optimization procedures toward decisions favouring characteristics which have demonstrated good solution fitness. In this way the proposed sewer network optimization algorithms, which are shown to perform better than the algorithms it took inspiration from, are further improved through the use of heuristics. The heuristics, and their formulation, reveal characteristics of optimal sewer network designs which can be used by future researchers as well as design engineers, to improve their solutions.

AFRIKAANSE OPSOMMING: Rioolnetwerke vorm 'n belangrike deel van stedelike infrastruktuur en verteenwoordig 'n kapitale intensiewe koste in die openbare sektor. In Suid-Afrika, verklaar 'n onlangse beraming dat R44.75 miljard nodig is om basiese sanitasiedienste te verskaf aan diegene in nood, terwyl die begrotingstoewysing vir sanitasie nagenoeg R3.2 miljard landwyd was. Hierdie stand van sake is nie net beperk tot Suid-Afrika nie. Dit is duidelik dat fondse toegewys aan sanitasie so effektief as moontlik bestee moet word. Dit is binne hierdie spasie dat hierdie proefskrif 'n bydrae lewer, deur die gebruik van moderne digitale rekenkrag en optimaliseringstegnieke om meer kostee effektiewe oplossings vir rioolnetwerkontwerp en ontledingsprobleme te vind. Hierdie ondersoek stel en evalueer nuwe oplossings wat kritieke tekortkominge aanspreek in bestaande rioolnetwerk optimalisering algoritmes. Die duidelike gebrek aan voldoende maatstafprobleme om die prestasie van nuwe rioolnetwerk optimalisering algoritmes te toets en te vergelyk word aangespreek. 'n Deeglike ondersoek na die aard van optimale oplossings word gemaak deur proefoplossings van optimaliseringsalgoritmes vas te lê met gereelde tussenposes. Die eienskappe van die oplossings word gekwantifiseer deur die voorgestelde netwerk eienskap parameters. Die evolusie van proefoplossing eienskappe laat insig en begrip toe in hoe die netwerk verander in die loop van die algoritme se lewensiklus. Hierdie inligting is dan uitgebuit om heuristiese invloedsfaktore voor te stel wat die optimalisering prosedures dryf na besluite wat gedemonstreerde eienskappe bevoordeel. Op hierdie manier word die voorgestelde rioolnetwerk optimalisering algoritmes, wat getoon word om beter te presteer as die algoritmes wat dit uit inspirasie geneem het, verder verbeter deur die gebruik van heuristiek. Die heuristiek, en hul formulering, onthul eienskappe van optimale rioolnetwerk ontwerpe wat deur toekomstige navorsers, sowel as ontwerp ingenieurs, gebruik kan word om hul oplossings te verbeter.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/103463
This item appears in the following collections: