Evaluation of Computational Intelligence Methods using Statistical Analysis to Detect Structural Damage

Abstract

Structural damage detection using dynamic measurements has led to the development of several approaches in the last decades. Most of conventional methods associate modal variations of the structure to damage, like methods based on strain energy deviation, methods based on changes in curvature mode shapes, flexibility matrix analysis, etc. However, the technique used to extract the modal parameters can considerably affect the results of damage identification methods, introducing additional uncertainties. Thus, approaches involving computational intelligence and statistics to detect structural changes directly from raw dynamic measurements are being investigated in recent researches. The present work aims to compare several computational intelligence algorithms to identify structural damage using statistical parameters of structural time histories. The proposed algorithms are analyzed in a numerical model of a simply supported beam. The good results encourage the development of computational tools using statistical analysis for structural damage assessment.