Damage Location by Maximum Entropy Method on a Civil Structure

Abstract

This article presents the application of a new damage assessment method based on a supervised learning algorithm that uses the principle of maximum entropy. The proposed algorithm employs, as reference information, the natural frequencies and mode shapes obtained with a numerical model of the structure in an undamaged condition and under different damage scenarios. The algorithm is validated with experimental tests on a laboratory six-levels-structure that is 2 m tall. The structure was anchored to a horizontal vibrating shaker table and it was excited at the support level by different seismic records and colored noise. The change in the structure from a normal condition to a damaged one is made by three consecutive reductions in the cross-section of a column. Additionally, a perturbation is introduced as a 0.5 % increase in the total mass of the structure. The experimental modal properties are identified by the Stochastic Subspace and Multivariable Output Error State sPace (MOESP) methods and they are compared with those obtained with a numerical model. The identification, location and quantification of damage that has been obtained with the proposed algorithm in the different test conditions, agrees well with the experimental damage, even for the conditions were few sensors are located in the structure.