Convolutional neural network for ultrasonic weldment flaw classification in noisy conditions.

Ultrasonic flaw classification in weldment is an active area of research and many artificial intelligence approaches have been applied to automate this process. However, in the industrial applications, the ultrasonic flaw signals are not noise free and automatic intelligent defect classification algorithms show relatively low classification performance. In addition, most of the algorithms require some statistical or signal processing techniques to extract some features from signals in order to make classification easier. In this article, the convolutional neural network (CNN) is applied to noisy ultrasonic signatures to improve classification performance of weldment defects and applicability. The result shows that CNN is robust, does not require specific feature extraction methods and give considerable high defect classification accuracies even for noisy signals.

[1]  S. Sambath,et al.  Automatic Defect Classification in Ultrasonic NDT Using Artificial Intelligence , 2011 .

[2]  Xiaodong Cui,et al.  Data Augmentation for Deep Neural Network Acoustic Modeling , 2015, IEEE/ACM Transactions on Audio, Speech, and Language Processing.

[3]  Geoffrey E. Hinton,et al.  Deep Learning , 2015, Nature.

[4]  A. Masnata,et al.  Neural network classification of flaws detected by ultrasonic means , 1996 .

[5]  F C Cruz,et al.  Efficient feature selection for neural network based detection of flaws in steel welded joints using ultrasound testing , 2017, Ultrasonics.

[6]  Zhe Gan,et al.  Learning Deep Sigmoid Belief Networks with Data Augmentation , 2015, AISTATS.

[7]  P. C. M. A. Farias,et al.  Flexible decision support system for ultrasound evaluation of fiber–metal laminates implemented in a DSP , 2016 .

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

[9]  Sung-Jin Song,et al.  Investigation of deep neural network with drop out for ultrasonic flaw classification in weldments , 2018 .

[10]  David A. Bradley,et al.  The use of neural networks in ultrasonic flaw detection , 1999 .

[11]  Sung-Jin Song,et al.  Ultrasonic flaw classification in weldments using probabilistic neural networks , 1992 .

[12]  Gaoliang Peng,et al.  A deep convolutional neural network with new training methods for bearing fault diagnosis under noisy environment and different working load , 2018, Mechanical Systems and Signal Processing.

[13]  Jin H. Huang,et al.  Detection of cracks using neural networks and computational mechanics , 2002 .

[14]  P. Testoni,et al.  Artificial neural networks for non-destructive evaluation with ultrasonic waves in not accessible pipes , 2005, Fourtieth IAS Annual Meeting. Conference Record of the 2005 Industry Applications Conference, 2005..

[15]  Min Meng,et al.  Ultrasonic signal classification and imaging system for composite materials via deep convolutional neural networks , 2017, Neurocomputing.

[16]  Jürgen Schmidhuber,et al.  Deep learning in neural networks: An overview , 2014, Neural Networks.

[17]  Bixio Rimoldi,et al.  Principles of Digital Communication: A Top-Down Approach , 2016 .

[18]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[19]  Redouane Drai,et al.  Time frequency and wavelet transform applied to selected problems in ultrasonics NDE , 2002 .

[20]  Sung-Jin Song,et al.  Development of an intelligent system for ultrasonic flaw classification in weldments , 2002 .

[21]  R. Gallager.  Principles of Digital Communication , 2008 .

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

[23]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.