Files in this item



application/pdfYerkin_Abdildin.pdf (3MB)
(no description provided)PDF


Title:Multiattribute utility functions for the deep borehole filter restoration problem
Author(s):Abdildin, Yerkin
Director of Research:Abbas, Ali E.
Doctoral Committee Chair(s):Abbas, Ali E.
Doctoral Committee Member(s):Uddin, Rizwan; Sreenivas, Ramavarapu S.; Allison, James T.
Department / Program:Industrial&Enterprise Sys Eng
Discipline:Industrial Engineering
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Decision analysis
multiattribute utility functions
the World’s biggest challenge
energy problem
uranium mining
in situ leach mining
deep borehole filter restoration problem
borehole recovery problem
nuclear power
twos-complement exclusion algorithm
ternary-decimal exclusion algorithm
partial utility independence
utility assessments
preference elicitation
decision maker
evaluation of alternatives
multiple attributes
eliminating redundancy
excluding duplicate terms
experimental study
confidence intervals
Abstract:The energy problem is one of the biggest challenges facing the World in the 21st century. It is related to the issues of natural resource extraction, resource depletion, power generation, environmental degradation, and atmospheric change such as global warming. Since more than 80% of the world’s primary energy is generated from fossil fuels (coal, oil, gas), emissions of carbon dioxide (CO2) from all fossil-fuel burnings are the largest cause of climate change. Global climate disruption, in turn, impacts on human health, flora, and fauna. The global energy demand is expected to double by 2050 and that is inevitably due to global population growth, global economic growth, and continued urbanization. To meet the increasing demand for energy and to avoid catastrophic climate change, increases in energy efficiency and increases in the fraction of low carbon energy sources are required. Uranium is a good energy source, because it has high energy density and nuclear power does not contribute to carbon dioxide emissions. However, difficulties in uranium mining cause large worldwide shortages of uranium for power generation. Decision makers in uranium mining are often challenged by various uncertainties in their decision problems (financial, technological, geological) and multiple objectives (increase profits, decrease radiation hazards, improve safety of operations, preserve environments). This dissertation studies multiattribute utility functions for modeling such challenging decision problems using the example of the deep borehole filter restoration problem from the uranium extraction industry. In this problem, the filter of the production borehole (or well) is periodically contaminated or clogged, causing significant uranium output reduction. The efficient modeling of this decision-making problem is of paramount importance for uranium mining worldwide and requires normative decision analysis. Motivated by the complexity of multiattribute decision problems under uncertainty and multiple objectives, this dissertation considers a set of open research problems related to the number of attributes and their degree of ‘interdependence under uncertainty,’ formally, utility dependence and independence. This dissertation characterizes the special functional forms of multiattribute utility functions (MUFs) under the partial utility independence (PUI) condition, verifies their applicability to the deep borehole filter restoration problem, evaluates the alternatives of the decision problem by three different approaches, and introduces novel methods for excluding redundant utility assessments. In Part I of this work we present our study: (1) what are the objectives and the corresponding attributes (i.e. factors or criteria of the decision-making process) of the deep borehole filter restoration problem from the well-field manager’s point of view, (2) does the ultimate decision maker find these attributes good or not, (3) does utility independence (UI) among the attributes exist in this decision problem, (4) whether or not the canonical functional forms from the theory of decision analysis are applicable to this decision problem, (5) are the decision analysis tests easy for the experts in the uranium extraction industry to use? For this, we create new tests for assessing interdependence among the attributes of the decision problem. In the first experiment with Test 1, 105 professionals in uranium mining were requested to provide their preferences among the four most important attributes from the well-field manager’s point of view. In the second experiment with the more formal Test 2, 40 experts were asked to provide their preferences from among three of these four attributes. Based on the results of the experimental study, with 95% confidence we can conclude that the proportion of the population (thousands) of experts who assert utility independence of the attributes is at most 0.23. The results imply that the conventional approach, the assumption of utility independence, may not be valid for our decision problem. In Part II, we evaluate five alternatives of the deep borehole filter restoration problem by constructing a multiattribute utility function (MUF) utilizing two different approaches: (i) under the assumption of utility independence of attributes, and (ii) with the existing (assessed from the decision maker) partial utility independence of the attributes. The four most important attributes were selected by the ultimate decision maker, the Deputy Director General of a transnational corporation. We first assume utility independence of the attributes and utilize the corresponding multilinear form of the MUF. We then utilize the general decomposition approach for constructing the MUF under the assessed partial utility independence conditions. Finally, we compare the results of these two approaches and the results of the profit analysis. By direct assessments, (i) we find that utility independence among the attributes of the deep borehole filter restoration problem is not a valid assumption, (ii) we verify that the decision maker’s preferences assert partial utility independence, (iii) we illustrate the assessments required under partial utility independence assertions, (iv) we compare the decisions made using the assumption of utility independence and the existing partial utility independence conditions, and (v) determine that the assumption of utility independence yields recommendations, which are different from the true preferences of the decision maker. Our results also demonstrate that the assessments required for the construction of the MUF by the utility independence approach is easier for the decision maker (DM), but the DM is more comfortable with the partial utility independence conditions. To our knowledge, this is the first practical study on the comparison of profit analysis, the utility independence approach, and the partial utility independence approach in a complex real life multiattribute decision problem. In Part III, we present algorithms for excluding redundant assessments from the set of assessments required for construction of the multiattribute utility function. In complex decision problems, the number of utility assessments, and, therefore, the number of questions for the decision maker, increases dramatically. It is thus very important to check the consistency of the assessments and eliminate all redundant utility assessments. With the efficient algorithms introduced in this dissertation, the elimination of the redundant utility assessments is considerably simplified. The results of this dissertation were applied to an important and complex decision problem in the uranium extraction industry, the deep borehole filter restoration problem. The decision modeling proposed in this dissertation should also help decision makers in addressing the worldwide 14% shortage of uranium needed for nuclear power generation.
Issue Date:2014-09-16
Rights Information:Copyright 2014 Yerkin Garyshuly Abdildin
Date Available in IDEALS:2014-09-16
Date Deposited:2014-08

This item appears in the following Collection(s)

Item Statistics