Structural damage location and evaluation model inspired by memory and causal reasoning of the human brain

Abstract

Structural health monitoring system should handle massive sensor information correctly and complete the damage location and evaluation. This study proposes a structural damage location and evaluation model inspired by the memory and causal reasoning of the human brain. Mass-monitoring data filtering is conducted in the short-term memory area. The long-term memory area stores useful information that contains the characteristics of structural damage and has two functions. First, the support degree index is deduced to locate the damage source of the basement. Second, the historical acoustic emission data and sparse autoencoder classifier are utilized to identify moisture content and then evaluate the damage level of the basement. Experiments show the functions of the model, such as monitoring data reduction, locating the damage source, and evaluating the level of damage, without prior knowledge of mechanics.