Abstract
The role of a falling weight deflectometer (FWD) test lies in the capacity to measure pavement system responses in relation to transient loads. These loads are applied on surfaces of the pavements. A study by ARA Inc. & ERES Consultants Division (2004) indicated that the backcalculation of pavement layer moduli via FWD has continually gained adoption or wide employment, yet it has not gone without limitations. As concurred by Ceylan, Guclu, Tutumluer and Thompson (2005), FWD results are dependent on pavement response static analyses. Indeed, most of the previous studies highlight that the discrepancies between FWD test dynamic nature and the static assumption yield significant errors in moduli (Chatti, Ji & Harichandran, 2004). It has also been established that a number of dynamic solutions exist in relation to pavement response but computational complexities with which these approaches are associated imply that their application in programs of conventional backcalculation is impractical (Goel & Das, 2008). Indeed, it is the limitation of FWD results and the impractical nature of adopting dynamic solutions that have paved way for the application of artificial neural network technologies while seeking to address the backcalculation dilemma with precision. Artificial neural networks constitute simple processing element collections that are highly interconnected and, upon training, could aid in the approximation of inverse functions (Goktepe, Agar & Lav, 2006). Goktepe, Agar and Lav (2005) observed that the approximation is achieved via repeated shows of forward problem solutions. Gopalakrishnan (2010) avowed that the leading advantage associated with artificial neural networks (ANN) concerns the aspect of speed.
Export citation and abstract BibTeX RIS
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.