Structural damage detection using deep learning of ultrasonic guided waves

Structural health monitoring using ultrasonic guided waves relies on accurate interpretation of guided wave propagation to distinguish damage state indicators. However, traditional physics based models do not provide an accurate representation, and classic data driven techniques, such as a support vector machine, are too simplistic to capture the complex nature of ultrasonic guide waves. To address this challenge, this paper uses a deep learning interpretation of ultrasonic guided waves to achieve fast, accurate, and automated structural damaged detection. To achieve this, full wavefield scans of thin metal plates are used, half from the undamaged state and half from the damaged state. This data is used to train our deep network to predict the damage state of a plate with 99.98% accuracy given signals from just 10 spatial locations on the plate, as compared to that of a support vector machine (SVM), which achieved a 62% accuracy.Structural health monitoring using ultrasonic guided waves relies on accurate interpretation of guided wave propagation to distinguish damage state indicators. However, traditional physics based models do not provide an accurate representation, and classic data driven techniques, such as a support vector machine, are too simplistic to capture the complex nature of ultrasonic guide waves. To address this challenge, this paper uses a deep learning interpretation of ultrasonic guided waves to achieve fast, accurate, and automated structural damaged detection. To achieve this, full wavefield scans of thin metal plates are used, half from the undamaged state and half from the damaged state. This data is used to train our deep network to predict the damage state of a plate with 99.98% accuracy given signals from just 10 spatial locations on the plate, as compared to that of a support vector machine (SVM), which achieved a 62% accuracy.

[1]  Geoffrey E. Hinton,et al.  Rectified Linear Units Improve Restricted Boltzmann Machines , 2010, ICML.

[2]  Christopher M. Bishop,et al.  Neural networks for pattern recognition , 1995 .

[3]  G. H. James,et al.  Damage detection in aircraft structures using dynamically measured static flexibility matrices , 1996 .

[4]  Joel B. Harley,et al.  Application of Mellin transform features for robust ultrasonic guided wave structural health monitoring , 2012 .

[5]  Davar Khalili,et al.  Daily Outflow Prediction by Multi Layer Perceptron with Logistic Sigmoid and Tangent Sigmoid Activation Functions , 2010 .

[6]  Charles R. Farrar,et al.  Damage identification and health monitoring of structural and mechanical systems from changes in their vibration characteristics: A literature review , 1996 .

[7]  Lawrence D. Jackel,et al.  Handwritten Digit Recognition with a Back-Propagation Network , 1989, NIPS.

[8]  Keith Worden,et al.  An introduction to structural health monitoring , 2007, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences.

[9]  Dirk P. Kroese,et al.  The Cross-Entropy Method: A Unified Approach to Combinatorial Optimization, Monte-Carlo Simulation and Machine Learning , 2004 .

[10]  Keith Worden,et al.  Damage identification using support vector machines , 2001 .

[11]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[12]  Joel B. Harley,et al.  Predictive Guided Wave Models Through Sparse Modal Representations , 2016, Proceedings of the IEEE.