Keywords: pavement performance, pavement distress, deterioration model, pavement management system, fuzzy system, rule-based system, non-linear regression.

Road pavement performance measures and deterioration prediction models are essential elements of pavement management systems. Existing models have been developed based on extensive data sets from field observations over long time periods. However, such models may not be directly transferable from country to country due to differences in characteristics such as pavement materials, construction quality and maintenance methods, environmental conditions, etc. In those cases that actual data are not available, pavement performance models may be derived from expert knowledge. The resulting data, generally qualitative and imprecise, can be converted to numerical data though the employment of fuzzy systems.

In this work, deterioration prediction models are developed for asphalt pavements using fuzzy systems. The development has considered several parameters such as, pavement distress type (cracking, disintegration, surface distortions, surface defects), traffic loads, environmental conditions, soil type, pavement material properties and design features, construction quality, and, most importantly, pavement age. The development includes the following three phases.

In the first, (qualitative) data were collected from expert responses regarding the factors and the way that they affect distress initiation and propagation. In addition, the literature was reviewed to investigate general equation forms used elsewhere. The procedure led to the determination of 12 distress types and 15 influencing factors which were grouped into four generalized variables, i.e., pavement strength, traffic load, pavement construction quality, and pavement age [1].

Next, a fuzzy rule-based system was developed to represent the expert knowledge using fuzzy rules if-then [2]. The fuzzy method that was used is the linguistic fuzzy model (Mamdani method). The system parameters are represented by linguistic fuzzy sets (e.g., weak pavement strength) and corresponding rules of the form if structural strength is weak and traffic load is high and ...then severity is high are developed which represent the impact of the influencing factors on distress severity. With this procedure, qualitative information is converted to quantitative data.

In the third phase, several sets of random crisp data for the independent and the dependent variables were generated by the fuzzy system. A least square non-linear regression [3] was applied to these data to determine the coefficients of the prediction model. These coefficients are different for each distress (leading thus to separate models) and determine the shape and the coordinates of the corresponding curves. In addition to individual distress models, a composite index was developed that represents the general condition of a road section based on the most severe existing distress or on a weighted average of individual distress indices.

The methodology was implemented considering the characteristics of the road network in Greece and the expertise of local pavement maintenance engineers. The implementation led to models that are considered quite realistic by pavement construction and maintenance experts. It appears that these models can be effectively employed in pavement management systems so as to provide the necessary input to optimise maintenance decisions regarding the appropriate rehabilitation type and time of application.

1

George K.P., "MDOT Pavement Management System: Prediction models and feedback system", Research Report, Department of Civil Engineering, The University of Mississippi, 2000.

2

Babuska R., "Fuzzy logic for engineering applications", Control Systems Engineering, Faculty of Information Technology and Systems, Delft University of Technology, Delft, The Netherlands, 1999.

3

Salem O., A. El-Assaly, and S. AboutRizk, "Performance prediction models of pavement highway network in Alberta", TRB 2003 Annual Meeting CD-ROM, 2003.

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 £95 +P&P)