Browsing by Author "Lindner, Berndt Gerald"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Bi-objective generator maintenance scheduling for a national power utility
    (Stellenbosch : Stellenbosch University, 2017-03) Lindner, Berndt Gerald; Van Vuuren, J. H.; Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering.
    ENGLISH ABSTRACT: One of the key focus areas for the management of a power utility in a regulated energy market is planned preventative maintenance of the power generating units in its power system. The so- called generator maintenance scheduling (GMS) problem refers to finding a schedule according to which the planned maintenance can be performed on the generating units in a power system. A novel bi-objective optimisation model is proposed in this dissertation for the GMS problem in which demand satisfaction reliability is maximised and electricity production cost is minimised. The first scheduling objective is one of the most common objectives in GMS problems in the literature, namely minimising the sum of squared net reserve levels. This objective serves to create an even (reliable) margin of generating capacity over expected demand. The second scheduling objective is the (linear) production cost associated with a maintenance plan of all the generating units in a system. The latter objective is aimed at exploiting the following correlation: planning maintenance on a cost-efficient power station during a high-demand period incurs a higher fuel cost. Production cost is simply taken as fuel cost in this dissertation since it is the most prominent production cost component of power generation. Dominance-based multi-objective simulated annealing is adopted as model solution technique. Solving the aforementioned model clearly demonstrates that maintenance schedules which min- imise the sum of squared reserves are typically also associated with low production costs, but that the lowest sum of squared reserves maintenance schedule does not necessarily achieve the lowest production cost (a sentiment also reported in the literature). Hence there is a need for adopting a multi-objective modelling approach in the context of GMS problems in search of trade-off solutions rather than adopting a standard single-objective modelling paradigm. A sensitivity analysis is performed in respect of model constraint relaxations and the degree of constraint violations. In the process, certain soft constraints which sensitively influence the model objectives are identified. A decision support system, whose working is based on the bi-objective optimisation model described above, is designed and a concept demonstrator of this system is implemented on a personal computer. This concept demonstrator may be used to find and analyse trade-off solutions to instances of the GMS model and offers interactive features which facilitate sensitivity analyses in a very natural way. The viability and practical use of the concept demonstrator is finally illustrated by applying it to two realistic GMS case studies. It is found that the decision system is capable of producing high-quality sets of trade-off maintenance schedules in each case.
  • Thumbnail Image
    Item
    Determining optimal primary sawing and ripping machine settings in the wood manufacturing chain
    (Stellenbosch : Stellenbosch University, 2014-04) Lindner, Berndt Gerald; Vlok, P. J.; Wessels, C. B.; Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering.
    ENGLISH ABSTRACT: For wood manufacturers around the world, the single biggest cost factor is known to be its raw material. Thus maximum utilisation, specifically volume recovery of this raw material, is of key importance for the industry. The wood products industry consists of several interrelated manufacturing steps for converting trees into logs and logs into finished lumber. At most primary and secondary wood processors the different manufacturing steps are optimised in isolation or based on operator experience. This can lead to suboptimal decisions and a substantial waste of raw material. The objective of this study was to determine the optimal machine settings for two interrelated operations, namely the sawing and ripping operations which have traditionally been optimised individually. A model, having two decision variables, was developed which aims to satisfy market demand at a minimal cost. The first decision was how to saw the log supply into different thicknesses by choosing specific sawing patterns. The second was to decide on a rip saw’s settings, namely part priority values, which determines how the products from the primary sawing operation are ripped into products of a certain thickness and width. The techniques used to determine the machine settings included static simulation with the SIMSAW software to represent the sawing operation and mixed integer programming to model the ripping operation. A metaheuristic, namely the Population Based Incremental Learning algorithm, was the link between the two operations and determined the optimal settings for the combined process. The model’s objective function was formulated to minimise the cost of production. This cost included the raw material waste cost and the over or under production cost. The over production cost was estimated to include the stock keeping costs. The under production cost was estimated as the buy-in cost of purchasing the under supplied products from another wood supplier. The model performed well against current decision software available in South Africa, namely the Sawmill Production Planning System package, which combines simulation (SIMSAW) and mixed integer programming techniques to maximise profit. The model added further value in modelling and determining the ripping priority settings in addition to the primary sawing patterns.