Session: 02-01: Digital Thread/Digital Twin/NDE Big Data

Paper Number: 138485

138485 - Smart Edge Computing Framework for In-Line Signal Detection and Classification 

Abstract: 

The multilayered, and anisotropic nature of composite structures presents challenges in monitoring damage initiation and progression. Non-intrusive inspection methods are vital to continuous monitoring regimes and facilitate the transition to a predictive maintenance regime for these structures. The effectiveness of this transition relies on the efficacy of the onboard monitoring system to acquire signals and extract essential features on the edge. These signal features, when correlated with parameterized damage metrics, enable the characterization of damages, and help to realize an automated framework for assessment of structural integrity and determining optimal maintenance intervention points. In real-time, in-service condition monitoring, handling large volumes of continuously acquired data is challenging due to limited computing ability and memory constraints. 

This paper presents a cyber-physical architecture designed to efficiently acquire ultrasonic guided wave signals and classify them as either essential acoustic event representatives or non-essential signals. The experimental setup consists of a sparse array of transducers and an edge device hosting the signal generation and reception systems. This setup was utilized to excite a 12-layered carbon fiber reinforced composite panel with tone-burst sinusoidal signals across a range of actuation frequencies. The responses to these excitations were captured and subjected to soft-threshold-based wavelet denoising to extract the structural acoustic response in the ultrasonic frequency band. Subsequently, the conditioned signals were transformed into time-frequency scalograms, which were then employed to train a multi-class classification algorithm operating on the edge device for effective real-time signal classification in an in-service setting. 

Presenting Author: Anirudh Gullapalli Cardiff University

Presenting Author Biography: Third year PhD student at Cardiff University school of engineering under the supervision of Dr.Abhishek Kundu, Prof. Carol Featherston and Prof.Rhys Pullin.
Working on developing a cyberphysical framework for structural health monitoring of thin-walled structures.

Authors:

Anirudh Gullapalli Cardiff University
Taha Aburakhis Cardiff University School of Engineering
Carol Featherston Cardiff University School of Engineering
Rhys Pullin Cardiff University School of Engineering
Abhishek Kundu Cardiff University School of Engineering