|
View Paper
A GENERALIZED GLOBAL OPTIMIZATION APPLICATION AND ITS USE IN DETERMINING THE HYDRAULIC PROPERTIES OF UNSATURATED SAND
Session: Laboratory Testing II / Essais en laboratoire II
Marc Lebeau, Laval University (Canada)
Jean-Marie Konrad, Laval University (Canada)
Global optimization consists in finding the best solution to a problem that may contain multiple sub-optimal solutions. Although it dates back to the origins of geometry, it is now extensively applied in various fields of engineering. This paper focuses on the deterministic algorithms that have been implemented in a generalized global optimization application. The application integrates the algorithms into a comprehensive modeling environment that uses the JUPITER API to communicate with any process model with batch processing capabilities. After a brief description of its structure and algorithms, the capabilities of the application are illustrated by finding the hydraulic property functions of unsaturated silica sand through inverse modeling of a multistep outflow experiment with both suction and cumulative outflow measurements. The canonical differential evolution algorithm is shown to be particularly well suited to this optimization problem in which the objective function is topographically complex. The estimated hydraulic property functions are also shown to be in excellent agreement with those independently determined using steady-state methods.
|
