PRELIMINARY COST MODELLING FOR PROCESS VESSELS

The aim of the project was to generate cost estimation formulae using the minimum amount of design information. Capital cost data for pressure vessels were analysed with a range of techniques including regression analysis, neural networks, rational polynomials, non-linear equations and fuzzy matching. Two sets of real industrial cost information were studied. The results showed that regression analysis using a power-law (or multiplicative function) gave convenient and reasonable cost correlations. The best overall technique was a neural network having a particularly simple structure (i.e., a single hidden layer with just one or two elements within it).