STP1375: Prediction of Remaining Life of Flexible Pavements with Artificial Neural Networks Models

    Abdallah, I
    Research Engineer, Center for Highway Materials Research, The University of Texas at El Paso, El Paso, TX

    Ferregut, C
    Professor, Center for Highway Materials Research, The University of Texas at El Paso, El Paso, TX

    Nazarian, S
    Professor, Center for Highway Materials Research, The University of Texas at El Paso, El Paso, TX

    Melchor Lucero, O
    Research Engineer, Center for Highway Materials Research, The University of Texas at El Paso, El Paso, TX

    Pages: 15    Published: Jan 2000


    Abstract

    A software program has been developed to predict the remaining life of flexible pavements using artificial neural network (ANN) technology. The remaining life due to either rutting or fatigue cracking can be predicted. The inputs to the software are the best estimate of the thickness of the layers, the deflection basin measured with a falling weight deflectometer (FWD), and optionally, the extent of damage at the time of the FWD test. The outputs are the best estimate of the remaining life and the pavement performance curve. If uncertainty in the thicknesses, FWD measurements and traffic exists, a probabilistic description of the remaining life is also provided. The main benefit of the proposed approach is that the backcalculation process for determining moduli is not necessary. The remaining lives or alternatively the critical stresses needed to calculate them are directly estimated. As such, the results seem to be more robust. In this paper, the overall procedure and the details of the methodology followed in developing the software are described. A case study is included to demonstrate the application of the methodology.

    Keywords:

    neural networks, pavements, remaining life, nondestructive testing, uncertainty, probability


    Paper ID: STP14786S

    Committee/Subcommittee: D04.39

    DOI: 10.1520/STP14786S


    CrossRef ASTM International is a member of CrossRef.