Structural Health Monitoring of Lattice Structure Using Artificial Neural Network

Abstract

Buildings need periodical health monitoring for assessing their strength and performance in the future. However, Structural Health Monitoring (SHM) plays a vital role in safeguard against failure at the member and structural level. Vibration-based structural health monitoring is the technique used to identify the structural changes using vibration measurements which causes the difference in the damage sensitive features in structures. Artificial Neural network (ANN) is an efficient Machine Learning (ML) tool widely used in many fields due to its high degree of robustness and fault tolerance. Feature selection and proper training of a network by adjusting parameters and hyperparameters are essential to get the output desired for a set of input data. The present study based on developing ANN to predict the damages in the lattice structures. Physical changes in the structure will alter the vibration parameters of the system. These vibration parameters include natural frequency, damping ratio, and mode shapes, etc. Thus, variations associated with these properties can be interpreted as damage caused to the structure. This is the idea behind the damage detection strategy of structures. Finite element analysis of lattice structure in the damaged and undamaged state is carried out to obtain various vibration parameters. Extracted vibration data is used to train neural network models to distinguish damaged conditions of the structure and thereby generalise the behaviour of the structure. The developed neural network is tested with unknown vibration data from the structure and the location of the damage is predicted.