Skip to Main Content
The inaccuracy of cost estimation in feasibility study and design stage is one of main reasons that make investment of railway construction projects out of control in China. In those two stages cost estimations are affected by many uncertain factors, and the relationship among them is nonlinear and traditional model is hard to solve them. This paper applies Cost-Significant (CS) theorem to simplify the cost estimation, and sets up the forecasting methods for Cost-significant Items (CSIs) of Whole life Costing (WIC). Then the Back-propagation Neural Network (BPNN) model is made up according to BP (Back-propagation) algorithm to "distill" CSIs and cost-significant factor (csf) from the data and information of performed projects, which provides a practical solution for those problems according to the nonlinear theory. The basic theories of BPNN, CS and WIC are introduced and their applications are illustrated with an example in this paper. From the example, we can see that the relative errors are small enough for accuracy demand of cost estimations after simulation, and test result shows that the model based on CSIs, WIC and BPNN theorems is accurate and successful.