DIPLOBACK: Neural-network-based backcalculation program for composite pavements

Lev Khazanovich, Jeffery Roesler

Research output: Contribution to journalArticlepeer-review

24 Scopus citations

Abstract

A neural-network-based backcalculation procedure is developed for multilayer composite pavement systems. The constructed layers are modeled as compressible elastic layers, whereas the subgrade is modeled as a Winkler foundation. The neural networks are trained to find moduli of elasticity of the constructed layers and a coefficient of sub-grade reaction to accurately match a measured deflection profile. The method was verified by theoretically generated deflection profiles and falling weight deflectometer data measurements conducted at Edmonton Municipal Airport, Canada. For the theoretical deflection basins, the results of backcalculation were compared with actual elastic parameters, and excellent agreement was observed. The results of backcalculation using field test data were compared with the results obtained using WESDEF. Similar trends were observed for elastic parameters of all the pavement layers. The backcalculation procedure is implemented in a computer program called DIPLOBACK.

Original languageEnglish (US)
Pages (from-to)143-150
Number of pages8
JournalTransportation Research Record
Issue number1570
DOIs
StatePublished - Jan 1 1997

Fingerprint Dive into the research topics of 'DIPLOBACK: Neural-network-based backcalculation program for composite pavements'. Together they form a unique fingerprint.

Cite this