Research on nonlinear estimating methods of whole life-cycle cost for China high-speed railway project

The present investment appraisal and budget estimation methods for the high and new technique investment projects of China high-speed railways are as follows: Comparing with similar completed projects, the project to be built is divided into similar parts and unsimilar parts; the similar parts are estimated by use of the linear analysis model or according to the budget quotas; the unsimilar parts are estimated by empirical analogy or by three time estimation; therefore, great errors, poor reliability and much workload are caused. In this paper, on the basis of the theories of whole life-cycle costs(WLC) and costs-significant(CS), the nonlinear investment estimation model for prediction of a whole life-cycle high-speed railway project was established to effectively fit the highly nonlinear relationships between costs and mutiple influencing factors. The cost-significant items(CSIs) of the scheduled project were decomposed into the CSIs with similar and unsimilar data by the CS method and experts'experiences. Chaotic time series were used to estimate the similiar CSIs with the plethora of the time series data of similar and finished items. The back-propagation neural network(BPNN) was used to estimate the similiar CSIs with the plethora of the non-time series data of finished items. The fuzzy clustering(FC) method was applied to estimate the similiar CSIs with a certain amount of the data of finished items. When there was no similar finished items, the fuzzy inference system(FIS) was used to make full uas of experts experiences and calculate the costs of the CSIs and the csf data of the whole project. Through programming computer software, the proposed methods can greatly reduce the workload of calculation, simplify the estimation procedure and effectively improve the accuracy of estimation by fitting the best nonlinear trend curve of the finished historic data.