Keywords: mathematical modelling, simulation, optimal design, optimization model, evolutionary algorithms.

During the 1980s, and particularly after the First North American Pavement Management Conference, held in Toronto, Canada, in 1985, pavement management systems (PMS) were recognized to be a major tool to aid the road engineer. Since then, they have been extensively utilized by national, state, regional, and municipal road administrations in many countries of the world to define maintenance and rehabilitation (M&R) strategies for the pavements of the road networks within their jurisdiction.

Two of the main components of a PMS are: (1) the models used to predict pavement performance; and (2) the approach used to select the best M&R strategy (taking into account the expected evolution of pavement performance). The pavement performance models employed within a PMS may be deterministic or probabilistic [1]. Deterministic models use regression equations to describe the evolution of pavement condition over time, whereas probabilistic models use Markov chains for the same purpose. At first sight, probabilistic models may look preferable because they take into account the uncertainty inherent in pavement degradation processes. However, unlike (modern) deterministic models, they do not rely on solid theoretical foundations (i.e., probabilistic models are purely empirical). Moreover, deterministic models may (and generally do) describe pavement performance in detailed, quantitative terms, with reference to the different types of distresses (cracking, rutting, etc.) that characterize pavement condition. This is practically impossible when probabilistic models are used, pavement condition being, in this case, always assessed through qualitative, aggregate measures (such as good, fair, and poor, or 1, 2, ..., 9). These reasons possibly explain why road administrations often prefer deterministic PMS (i.e., PMS based on deterministic pavement performance models) instead of probabilistic PMS.

The approach used to select the M&R strategy within a deterministic PMS typically relies on the analysis of a limited number of alternatives. Examples of PMS that follow this approach are the HDM PMS [2], possibly the PMS more widely used in the world, and the Nevada PMS [3]. The main weakness of this approach is that it does not guarantee the selection of the best possible M&R strategy. If the alternatives considered are poor, it cannot lead to a good strategy. This can only be avoided if the approach followed to select the M&R strategy is based on optimization techniques. Probabilistic PMS, such as the Arizona PMS [4,5], follow this kind of approach. But, as stated before, they are unable to describe pavement performance in detailed, quantitative terms.

This paper presents a segment-linked optimization model for deterministic PMS developed within an R&D project that is being developed at the University of Coimbra (Portugal). The model, which is used by the Decision-Aid Tool (DAT) of the PMS, is aimed at determining the least-cost M&R strategy to be implemented in a road network, taking into account the applicable technical and budgetary constraints. The designation of segment-linked is applied to the model because M&R operations are directly assigned to road segments and not to road categories, as the models normally used within probabilistic PMS do. This ultimately signifies that, unlike these models, the model presented here automatically prioritizes M&R operations (across road segments). This DAT works with the deterministic pavement performance model used in the AASHTO flexible pavement design method. The paper also presents a genetic-algorithm heuristic, called GENEPAV-D, for solving deterministic optimization models. The DAT is applied to a problem involving the road network of Oliveira do Hospital, which is one of the Portuguese municipalities. The results obtained for this problem clearly indicate that the model is a valuable addition to the road engineer's toolbox and GENEPAV-D is an efficient heuristic method for solving this type of optimization model.

1

Ferreira, A., Picado-Santos, L., Antunes, A., "Pavement performance modelling: State of the art", In Topping, B. and Kumar, B. (Eds), Computer Techniques for Civil and Structural Engineering, Civil-Comp Press, pp. 157-164, Edinburgh, UK, 1999. doi:10.4203/ccp.58.8.1

2

ISOHDM, "International Study of Highway Development and Management Tools - Modelling Road Deterioration and Maintenance Effects in HDM-4", Asian Development Bank, Manila, Philippines, 1995.

3

Sebaaly, P., Hand, A., Epps, J., Bosch, C., "Nevada's approach to pavement management", Transportation Research Record, 1524, pp. 109-117, 1996. doi:10.3141/1524-13

4

Golabi, K., Kulkarni, R. B., Way, G. B., "A State-wide Pavement Management System", Interfaces 12(6), pp. 5-21, 1982. doi:10.1287/inte.12.6.5

5

Wang, K., Zaniewski, J., "20/30 Hindsight: the new pavement optimization in the Arizona state highway network", Interfaces, 26 (3), pp. 77-89, 1996. doi:10.1287/inte.26.3.77

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £105 +P&P)