Early Cost Estimating of Road Tunnel Construction Using Neural Networks

• Published in: Journal of construction engineering and management Vol. 138; no. 6; pp. 679 - 687
• Main Authors: Petroutsatou, K, Georgopoulos, E, Lambropoulos, S, Pantouvakis, J. P
• Format: Journal Article
• Language: English
• Published: Reston, VA American Society of Civil Engineers 01.06.2012
• Subjects: Applied sciences; Building economics. Cost; Buildings. Public works; Computation methods. Tables. Charts; Construction costs; Cost engineering; Cost estimates; Exact sciences and technology; Mathematical models; Neural networks; Roads; Structural analysis. Stresses; Technical Papers; Tunnel construction; Tunnels (transportation); Tunnels, galleries; Greece; Europe; Construction cost; Data analysis; Tunnel construction; Estimation; Neural network; Modeling; Construction costs; Road tunnel; Neural networks; Cost estimation; Comparative study; Data gathering; Tunnels
• ISSN: 0733-9364; 1943-7862; 1943-7862
• DOI: 10.1061/(ASCE)CO.1943-7862.0000479

AbstractRoad tunnel construction is subject to underground uncertainties and risks, and as such it is difficult to predict the final construction cost, especially at the conception phase where issues are evaluated and important design decisions are made. A system assisting in the early cost estimation of road tunnels would therefore be of great value as it would allow the quick costing of alternative and more economical solutions. The development of such an early cost estimation system is discussed in this paper. First, the basic parameters (geological, geometrical, and work quantities-related) affecting temporary and permanent support and final construction cost are determined. After that, appropriate real-world data derived from the analysis of 33 twin tunnels of 46 km total length constructed for the Egnatia Motorway in northern Greece from 1998 to 2004 and related to work quantities is collected and normalized. Appropriate price lists are then applied to calculate the costs; subsequently, cost-estimating models are developed using two types of neural networks: (1) the multilayer feed-forward network; and (2) the general regression neural network. Finally, these models are compared against real quantities and costs for accuracy and robustness. The main conclusion is that the models developed are fit for their purpose and may lead to fairly accurate work quantities and cost estimates of road tunnels.