The use of value functions in multi-attribute decision making in the presence of uncertainty.
Engineering Honours Degree 2008
University of Adelaide
Many decisions are made in circumstances where uncertainty exists, both within the petroleum industry, and in life in general.
The project aimed to explore the effects of uncertainty on the use of value functions in multi-objective decision-making.
Initially, a review of a number of aspects of quantitative decision-making was undertaken.
A methodology was developed to account for the presence of uncertainty by using probability distributions for attribute scores, instead of discrete values, and applying these distributions to value functions. A modified form of Monte Carlo Simulation was then used upon a standard decision matrix model, to generate probability distributions for the final net-weighted value of the decision alternative.
The problem of selecting a dynamometer for a rod-pumped oil well was selected to be used as a test-case decision. 5 designs of dynamometer were used as the decision alternatives, and probability distributions for each of the 7 decision attributes were assigned to each of the alternatives.
The methodology was applied to the dynamometer decision, and probability distributions were successfully generated for the final net-weighted values of each of the 5 alternative dynamometers. A clear ranking of the alternatives was able to be developed.
Thus, probability distributions that incorporate uncertainty are able to be applied, in lieu of known discrete values, to value functions as part of a multi-objective decision-making process. Monte Carlo Simulation can then be applied to generate output distributions for each alternative, in order to account for the uncertainty in the input distributions in the output.