Closing the Sim-to-Real Gap in Guided Wave Damage Detection with Adversarial Training of Variational Auto-Encoders

Khurjekar, Ishan D.; Harley, Joel B.

doi:10.1109/icassp43922.2022.9746196

Cited by 8 publications

(6 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The rootmean-square error (RMSE) of the network output is used as a DI while the wavelet packet energy relative entropy is used for localization. Khurjekar and Harley 59 proposed using an ensemble of VAEs for guided wave damage detection subject to non-uniform temperature variations. Anaissi et al 60 used frequency domain data reconstruction probability of VAEs to detect damage in an anomaly detection setting.…”

Section: Deep Variational Encoder Architecturementioning

confidence: 99%

Uncertainty‐aware structural damage warning system using deep variational composite neural networks

Eltouny

Liang

2023

Earthq Engng Struct Dyn

View full text Add to dashboard Cite

Structural health monitoring (SHM) is, without a doubt, one of the most important assets for building resilient communities. The vast and rapidly advancing research in data science and machine learning has provided researchers in the civil engineering community with various tools that can facilitate the processing of significant amounts of gathered data. However, deep learning models are prone to mistakes, and with the catastrophic consequences that can happen due to damage misidentification, damage diagnosis models’ predictions should not be taken for granted. In this study, we present an uncertainty‐aware early‐warning system that can provide near real‐time SHM. The system utilizes a deep composite encoder‐decoder network that combines elements from convolutional neural networks, recurrent neural networks, and variational inference (VI) to provide damage index distributions. The framework can detect anomalies in the structural system during seismic events and provide a measure of uncertainty that can be used to question the model's predictions. To assess the system's validity and practicality, we apply our proposal to three real structures, two of which suffered damage during the 1994 Northridge earthquake. We found that the early warning system delivers an accurate, yet cautious, continuous monitoring that is capable of sending warning signals when damage occurs in the course of seismic events. Source code is available at: https://github.com/keltouny/VSCAN.

show abstract

Section: Deep Variational Encoder Architecturementioning

confidence: 99%

Uncertainty‐aware structural damage warning system using deep variational composite neural networks

Eltouny

Liang

2023

Earthq Engng Struct Dyn

View full text Add to dashboard Cite

show abstract

“…The design of this VAE architecture was guided by Khurjekar and Harley, 25 and refinements were made based on the nature of the input data employed in this study. The architecture used in the study by Khurjekar and Harley 25 was for only one input variable and consisted of only three different types of layers; the Conv1D layers for extracting the underlying structure in the input signals, the Dense layers to decrease and increase dimensionality, and to connect the encoder and the decoder through the latent space, and the Conv1DTranspose layers to apply a transposed 1D convolution operation. In this paper, we also include Embedding layers since the VAE needs to enhance the input baseline signals with damage signatures according to the user-specified positions.…”

Section: Vae Fusing Damage Signature With Virtual Experimental Datamentioning

confidence: 99%

“…In this paper, we also include Embedding layers since the VAE needs to enhance the input baseline signals with damage signatures according to the user-specified positions. In reference, 25 the Embedding layers were not used as the VAE architecture was used to reconstruct the input data without adding new features. By adding new features through embedding layers, the VAE can learn to capture the relationship between the original features in the baseline signals and the newly introduced damage features and learn the meaning and interpretation of how the features vary with respect to the distance and the side indicator inputs.…”

Section: Vae Fusing Damage Signature With Virtual Experimental Datamentioning

confidence: 99%

See 1 more Smart Citation

A data-driven hybrid approach to generate synthetic data for unavailable damage scenarios in welded rails for ultrasonic guided wave monitoring

Ramatlo,

Wilke,

Loveday

2023

Structural Health Monitoring

View full text Add to dashboard Cite

Developing reliable ultrasonic-guided wave monitoring systems requires a significant amount of inspection data for each application scenario. Experimental investigations are fundamental but require a long period and are costly, especially for real-life testing. This is exacerbated by a lack of experimental data that includes damage. In some guided wave applications, such as pipelines, it is possible to introduce artificial damage and perform lab experiments on the test structure. However, in rail track applications, laboratory experiments are either not possible or meaningful. The generation of synthetic data using modelling capabilities thus becomes increasingly important. This paper presents a variational autoencoder (VAE)-based deep learning approach for generating synthetic ultrasonic inspection data for welded railway tracks. The primary aim is to use a VAE model to generate synthetic data containing damage signatures at specified positions along the length of a rail track. The VAE is trained to encode an input damage-free baseline signal and decode to reconstruct an inspection signal with damage by adding a damage signature on either side of the transducer by specifying the distance to the damage signature as an additional variable in the latent space. The training data was produced from a physics-based model that computes virtual experimental response signals using the semi-analytical finite element and the traditional finite element procedures. The VAE reconstructed response signals containing damage signatures were almost identical to the original target signals simulated using the physics-based model. The VAE was able to capture the complex features in the signals resulting from the interaction of multiple propagating modes in a multi-discontinuous waveguide. The VAE model successfully generated synthetic inspection data by fusing reflections from welds with the reflection from a crack model at specified distances from the transducer on either the right or left side. In some cases, the VAE did not exactly reconstruct the peak amplitude of the reflections. This study demonstrated the potential and highlighted the benefit of using a VAE to generate synthetic data with damage signatures as opposed to using superposition to fuse the damage-free responses containing reflections from welds with a damage signature. The results show that it is possible to generate realistic inspection data for unavailable damage scenarios.

show abstract

“…Compared with traditional deep learning algorithms, a relatively small training data set was used and it was verified that the CAE had a high generalization capability. Deep learning [29,30] has great potential and advantages in the field of damage detection. Traditional deep learning networks only need to learn one code to reproduce the input, but the distribution of data in the hidden space is not uniform.…”

Section: Introductionmentioning

confidence: 99%

Deep learning-based corrosion-like defect localization technique for high-voltage cable aluminum sheaths using guided waves

Zhang

Luo

Hong

et al. 2023

Meas. Sci. Technol.

View full text Add to dashboard Cite

As an important part of the power system, high voltage cable accessories need to ensure the reliability of the electrical connection. However, its aluminum sheath is prone to corrosion with complex working conditions, which has a bad effect on the normal operation of the electric system. Ultrasonic guided wave detection is a promising non-destructive testing method suited to detecting corrosion of aluminum sheaths in complex structures of high-voltage cable accessories. However, current ultrasonic guided wave detection methods still require manual extraction of the signal features and have a high reliance on professional knowledge. This paper proposes a deep learning-based corrosion-like defect localization technique for high-voltage cable aluminum sheath using guided wave. Firstly, the original ultrasonic guided wave signals of corrosion defects at different locations are obtained using ultrasonic guided wave detection platform. Then, the original signals are input into the variable auto-encoder (VAE) network to obtain a low-dimensional representation for automatic feature extraction. Finally, the low-dimensional representation is input into a gate recurrent unit (GRU) based recurrent network for corrosion defects localization. In the feature extraction stage, the VAE can automatically extract the effective features and avoid the interference of noisy signals. In the defect localization stage, GRU method can accurately identify the location of corrosion defects. The maximum error of the training median is 0.072 m, and the maximum error of the test median is 0.0021 m. The experimental results indicate that the VAE-GRU method is capable to accurately identify corrosion defects based on the original signals.

show abstract

Closing the Sim-to-Real Gap in Guided Wave Damage Detection with Adversarial Training of Variational Auto-Encoders

Cited by 8 publications

References 17 publications

Uncertainty‐aware structural damage warning system using deep variational composite neural networks

Uncertainty‐aware structural damage warning system using deep variational composite neural networks

A data-driven hybrid approach to generate synthetic data for unavailable damage scenarios in welded rails for ultrasonic guided wave monitoring

Deep learning-based corrosion-like defect localization technique for high-voltage cable aluminum sheaths using guided waves

Contact Info

Product

Resources

About