THE APPLICATION OF FUZZY LOGIC AND ARTIFICIAL NEURAL NETWORKS TO PARAMETRIC COST ESTIMATING OF FLOATING OFFSHORE STRUCTURES

摘要

Estimating the development costs for floating offshore structures a number of years before actual construction is a complex multi-disciplinary task that requires appropriate methods and techniques. Given that the average cost of an offshore oil and gas development comprising a floating offshore structure is approximately USD1 billion, the impact of wrong decisions based on poorly determined front-end cost estimates are far reaching through the asset lifecycle. Typically, during the concept selection and front -end of an offshore development project, a variety of data focused approaches have been employed in the development of cost estimates, a majority of which are based on individual judgement, experience, extrapolation, analogies and parametric cost modeling using regression analysis. However, in a number of situations like frontier development and the use of new technology, the requisite data for these cost models are not readily available. The use of approximate parametric information and the existence of uncertainty in the front -end decision making makes the ability to predict construction cost for offshore structures with limited data necessary. Not all sectors of floating offshore structure construction have taken advantage of other types of data focused modeling techniques such as neural networks and fuzzy logic. Neural networks can be used to develop parametric cost estimates of construction projects or specific hull construction operations for floating structures like FPSOs, SPARs, Semisubmersibles, Tension Leg Platforms or other floating structures. Due to their versatility and ability to handle fuzziness, neural networks have performed well in estimating specific construction operations for which cost is dependent on specific parameters. This paper presents methods of developing parametric cost estimates for floating structures using artificial neural networks (ANN) which is in line with the author's ongoing research into the development of a probabilistic cost engineering model for offshore structures at the University of Western Australia. Fuzzy set theory is used to enumerate the parameters and cost drivers while linguistic rules are developed to provide a measure of the resulting cost. The model is expected to be a very useful tool, which improves front-end engineering while increasing the level of confidence in engineered costs.