Adaptive fuzzy inference system for automated pavement condition evaluation of large pavement sections from Ground Penetrating Radar (GPR) thickness data

Monitoring pavement sub-surface layer thicknesses is essential to ensure stable pavement performance under heavy traffic loading. In addition, accurate estimation of pavement subsurface layer thicknesses is required for pavement condition evaluation and remaining life analysis. Traditionally this vital information is ascertained using conventional techniques such as coring/drilling at discrete locations, which are often destructive. In contrast, ground-penetrating radar (GPR) is a non-destructive proximal sensing technique gaining popularity in pavement structural condition monitoring and thickness estimation. In this work, data collected using a 1.5 GHz ground-coupled GPR system is used to estimate asphalt layer thicknesses for a 3 km long tollway in Queensland, Australia. An automated adaptative fuzzy inference system is proposed to evaluate pavement conditions. Specific parameters need to be considered before feeding inputs to the fuzzy block. The segmentation of a large section is based on mean, standard deviation, and variation in thicknesses. The inputs to the fuzzy module are boundary limits in thickness variations and thickness counts that fall within the standard distribution curve. The fuzzy module uses Mamdani fuzzy inference with triangular and trapezoidal membership functions. The rules are designed to determine the priority of the expert system, which is input dependent. The output from the fuzzy module is a pavement condition classification rating which is a pavement performance indicator. Successful implementation of this algorithm is envisaged to benefit the pavement engineers in planning rehabilitation and maintenance of existing infrastructure.