Generative Adversarial Neural Networks for Guided Wave Signal Synthesis

Heesch, Mateusz; Dworakowski, Ziemowit; Mendrok, Krzysztof

doi:10.1007/978-3-030-64908-1_2

Cited by 1 publication

(1 citation statement)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This work tackles the application of GANs to generate synthetic signals of a quality sufficient for usage in designing and training damage detection algorithms, extending the previous work on GW-GAN [ 22 ] (guided wave—generative adversarial network) a generative adversarial neural network model for GW based on the state-of-the-art architecture for image synthesis StyleGAN2 [ 23 , 24 ]. The motivation behind this work is the difficulty of obtaining guided waves data of sufficient quality, quantity, and distribution [ 2 ] to adequately explore applications of some of the newer machine learning (ML) advances in GW analysis.…”

Section: Introductionmentioning

confidence: 99%

Diagnostic-Quality Guided Wave Signals Synthesized Using Generative Adversarial Neural Networks

Heesch

Dziendzikowski

Mendrok

et al. 2022

Sensors

Self Cite

View full text Add to dashboard Cite

Guided waves are a potent tool in structural health monitoring, with promising machine learning algorithm applications due to the complexity of their signals. However, these algorithms usually require copious amounts of data to be trained. Collecting the correct amount and distribution of data is costly and time-consuming, and sometimes even borderline impossible due to the necessity of introducing damage to vital machinery to collect signals for various damaged scenarios. This data scarcity problem is not unique to guided waves or structural health monitoring, and has been partly addressed in the field of computer vision using generative adversarial neural networks. These networks generate synthetic data samples based on the distribution of the data they were trained on. Though there are multiple researched methods for simulating guided wave signals, the problem is not yet solved. This work presents a generative adversarial network architecture for guided waves generation and showcases its capabilities when working with a series of pitch-catch experiments from the OpenGuidedWaves database. The network correctly generates random signals and can accurately reconstruct signals it has not seen during training. The potential of synthetic data to be used for training other algorithms was confirmed in a simple damage detection scenario, with the classifiers trained exclusively on synthetic data and evaluated on real signals. As a side effect of the signal reconstruction process, the network can also compress the signals by 98.44% while retaining the damage index information they carry.

show abstract