Operational based stochastic cluster regression-based modeling for predicting condition rating of highway tunnels

Abstract

Despite the higher capital costs of tunnels and supplementary cost of maintenance, fewer deterioration models have been built compared with other highway components. Most of these models were limited to structural defects, not operational conditions. Moreover, there are inherent subjectivity and inaccuracy of the developed models, which may affect the maintenance process. The aim of this research and proposed contribution are investigating and modeling the impact of explanatory variables and non-periodic maintenance effect on highway tunnel condition. The stochastic regression analysis has been conducted to come out with a realistic tunnel condition through the Monte Carlo simulation methodology. The research methodology consists of three phases: cluster analysis, regression modeling, and stochastic analysis. A dataset of 473 highway tunnels along 41 American states from the National Tunnels Inventory (NTI) has been used. Nine models have been developed with a high coefficient of determination (R2 = 90.8%). The obtained results and the models could help advance the development of tunnel deterioration models from a management perspective, not only from the structural view. The developed models help highway authorities to prioritize the maintenance and objectively make informed investment decisions based on the historical data. This research has come as a response to the significant problems facing the highway authorities regarding tunnels and asset management.