Capital Cost Estimates and Sensitivity Analysis by the Monte Carlo Approach

Successful capital cost estimates in early project stages base 1) on a reliable statistical estimation method, 2) the use of an appropriate cost index, and 3) a powerful tool for sensitivity analysis.

Cost estimation methods such as “percentage of delivered-equipment cost“ or “Lang Factors“ are used in the process industry in the context of product and process development primarily in early project phases such as “order of magnitude estimate“ and “study estimate“. The results of these estimates are characteristic numbers such as fixed capital investment, production costs, return on investment (ROI) or pay back period. Depending on the project‘s degree of maturity or depending on the money spent at a certain point of time of a project‘s life, respectively, the accuracy of the calculated data is located in the asymmetric interval between -20/+50% and -10/+20% or in the symmetric interval between +/-30% and +/-20%, depending on the way, how contingencies were put into account.

Cost indices are specific cost escalation factors. They are based on statistical market data and they are specifically estimated for different branches of the process industry, e.g. for the chemical process industry or the refinery industry. Cost indices can be used to up-date fixed capital investment data from 1) former projects, 2) proposals or 3) company in-house data bases. They are used for the transfer of cost of equipment, plant sections or complete plants to the present. Cost Indices are valid for the estimation of cost over a time span of approximately 10 years, depending upon changes in the state of the art of the respective equipment.

Cost indices are estimated as overall parameters for 1) an average process plant using specific weighting factors and 2) for different sub-groups such as "heat exchangers and tanks“, "pipes vessels and fittings“, "buildings“ or "engineering and supervision“, respectively.

The direction and the sensitivity of the variation of input variables of a cost estimate on the output variables is typically analyzed by a sequential variation of one relevant input variable after the other whereas all remaining parameters are fixed to constant values. The main disadvantages of this sequential analysis are 1) only a limited set of input variables can be analyzed simultaneously, 2) no information on the correlation between the input variables is gained and 3)no information on the confidence interval of the output variables is gained.

A sensitivity analysis by a Monte Carlo approach is a strong tool to overcome this drawbacks.