Reliability-Based Processing of Markov Chains for Modeling Pavement Network Deterioration

Abstract

Accurate prediction of pavement deterioration is the most important factor in the determination of pavement repair years and optimization programming of highway network maintenance. The Nonhomogeneous Markov Probabilistic Modeling Program, developed to determine pavement deterioration rates in different stages, is described. In this program the transition probability matrices (TPMs) are considered as a time-related transition process. Each element of the TPMs is determined on the basis of a reliability analysis and a Monte Carlo simulation technique. This avoids the use of the existing conventional methods, which involve taking an average subjective opinion of pavement engineers or observing a large number of multiyear pavement performance data and conducting a number of statistical calculations. As a result a series of TPMs for an individual pavement section for different stages can be determined by running the program. Furthermore, the pavement condition state in terms of a probability vector at each stage (year) is calculated. In applying the models both the predicted actual traffic (in terms of equivalent single axle loads) at each stage and the maximum traffic that the pavement can withstand at each defined pavement condition state interval are considered to be random variables. In addition, the sensitivities of pavement deterioration rates to pavement design parameters, such as traffic growth rate, subgrade strength, and material properties, are studied. Finally, an example of calculating the TPMs for a pavement section located in southeastern Ontario, Canada, is demonstrated. It shows that the sensitivities of the TPMs to traffic growth rate, subgrade deflection, and pavement thickness are significant.

Analysis of ground penetrating radar data using hierarchical Markov Chain Monte Carlo simulation

“…The stochastic model that has successfully been used in modeling pavement performance is the Markov model (Butt et al 1987;Li et al 1996). In the Markov model, pavement-condition ratings are first transformed into discrete condition states.…”

Hierarchical asphalt pavement deterioration model for climate impact studies

“…The most common form of probabilistic models is based on Markov chains (Lytton 1987, Li et al 1996, Costello et al 2006, Yang et al 2006, Seyedshohadaie et al 2010. Markov models are based on transition matrices, which are normally derived in a structured way from past observations.…”

Reliability-based initial pavement performance deterioration modelling

“…Given the current asset condition (state i), the MC technique predicts the future condition of the asset (state j) as probability distribution (Micevski et al 2002, Baik et al 2006, Ortiz-Garcia et al 2006. Bayesian regression modelling was proposed at the end of the 1990s (Li andHaas 1996, C-SHRP 1997). The key advantage of the Bayesian regression model formulation is the power to incorporate uncertainty -which is a reality in all design and planning processes for transportation infrastructure.…”

“…The characteristics of interest here are those that have an effect on the causal variables for the performance model such as initial structure, as-built quality, environmental exposure, traffic loading and International Journal of Pavement Engineering 179 maintenance practice. The concept of similar families of pavements is not new; it has been extensively used by others (Pedigo et al 1981, Li and Haas 1996, Mauch and Madanat 2001 to analyse large databases and enhance reliability of the performance models. The next step is to decide on the causal factors that affect the deterioration process that can realistically be included in the performance model.…”