A combined modal strain energy and particle swarm optimization for health monitoring of structures

Abstract

Main complexities of classical health-monitoring techniques, such as modal strain energy (MSE) and modal curvature (MCR), are noise sensitivity and false alarming. On the other hand, the modern techniques, such as model updating, become very complex for a large number of variables and large search space. In this present study, a new health-monitoring technique based on the combination of MSE and model updating is presented for fast and accurate identification of structural damage. In this technique, the probable locations of an unknown damage are determined using MSE-based damage index (MSEDI). The effect of noise is reduced by passing the vibration data through Morlet wavelet filter. In this context, the peaks of the MSEDI in the wavelet domain are considered as the suspicious locations of damages. The exact location and the severity of the damage are then determined by applying model updating technique. To achieve this, reduction of stiffnesses at the probable locations is considered as the updating variable. The objective function is developed using MCR in wavelet domain and minimized by particle swarm optimization technique. The final updated values of the stiffnesses at those locations represent the actual locations and severity of the damage. The technique is applied on a simulated single-storied plane steel frame structure and a similar experimental model with welded joints for single and multiple damage scenarios. The damage was introduced either near to or away from the joints by making saw-cut grooves representing loss of stiffness. The results depict the effectiveness of the technique for potential application in real-life structures.