Enhanced Crack Detection in Composite Plates: Integrating Haar Wavelet Transform with Convolutional Neural Networks

Department of Mechanical and Industrial Engineering. Tallinn University of Technology. Ehitajate tee 5, 19086, Tallinn, Estonia

^a) Corresponding author: marmar.mehrparvar@taltech.ee
^b) juri.majak@taltech.ee
^c) kristo.karjust@taltech.ee

Abstract

In order to ensure structural integrity, detecting cracks, as a common structural flaw, is crucial. The current study presents a method for crack detection and prediction in plates under free vibration using the Convolutional Neural Network (CNN) and the Haar wavelet transformation. The Haar wavelet method is employed to preprocess vibration data, extracting key features that improve CNN's ability to identify and localize cracks. The proposed approach establishes high accuracy in detecting crack locations and intensities, showcasing its potential for real-time structural health monitoring.