Optimization assisted probabilistic damage localization of cracks in weld connections of a steel portal frame using electromechanical impedance technique

Abstract Steel structure members are typically connected through welded connections, that are highly vulnerable to crack development. Therefore, the continuous monitoring of these connections is necessary for early defect detection and to ensure structural stability. Although monitoring approaches such as electromechanical impedance (EMI) have found extensive application in qualitative damage assessment, there is a need for more rigorous EMI-based structural health monitoring (SHM) analysis for sizing and localization of weld damages. In this study a novel methodology is developed that utilizes swarm optimization based on wavelet damage indicators derived from piezo-coupled admittance signatures to experimentally investigate and quantify the amount of crack damages. Initially, a single-story portal steel frame is instrumented with PZTs to gather the admittance signatures in pristine and weld crack damage states. The obtained signatures are then decomposed using discrete wavelet transform (DWT) functions to extract various damage parameters. An objective function has been formulated using the extracted damage parameters like root mean square deviation and DWT parameters such as mean, variance, energy, and entropy. Further, various swarm optimization algorithms like particle swarm optimization (PSO), ant colony optimization, and artificial bee colony optimization (ABC) have been utilized for determining the size and location of the weld crack damage. It has been observed that PSO-based optimization using wavelet coefficients mean achieved 99.904% accuracy, and outperformed other damage parameters in identifying the location of weld crack damages. The proposed methodology demonstrates high efficiency with fewer sensors and has potential real-world uses as a low-cost SHM solution for welded joints in steel structures.