Damage detection approach based on Cross Modal StrainEnergy (CMSE) and Neural Network

Structural health monitoring is increasingly being used to reduce maintenance costs and prevent sudden collapsesby detecting damage as early as possible. Various methods, such as those based on vibration characteristics, areemployed for structural health monitoring. However, these methods cannot be used in isolation to localize structuraldamage. One method that uses vibration-based characteristics to find the location of damage is Cross Modal StrainEnergy (CMSE). The CMSE approach identifies damage to each element by integrating the structure's modalproperties and stiffness matrix. However, this method encounters difficulties if the mode shape of all nodes is notavailable. Additionally, data correlation issues can arise, leading to an ill-posed problem. Neural networks are nowwidely used as suitable function approximators. In this study, a neural network is employed to assist the CMSEapproach in addressing the challenges of incomplete measurements and data correlation. The neural network istrained using data from the CMSE method. The proposed neural network is capable of detecting damage in theelements with an accuracy of 96%. Furthermore, the neural network can identify minor damages, such as a 1%reduction in an element's modulus of elasticity. This finding underscores the potential of using neural networks todetermine the location of structural damage based on CMSE data.

Damage detection approach based on Cross Modal StrainEnergy (CMSE) and Neural Network Keywords:

Structural health monitoring , Cross modal strain energy , Neural network , Damage localization , Classification