Application of Convolutional Neural Networks for Recognizing Long Structural Elements of Rails in Eddy-Current Defectograms

To ensure traffic safety of railway transport, non-destructive test of rails is regularly carried out by using various approaches and methods, including eddy-current flaw detection methods. An automatic analysis of large data sets (defectograms) that come from the corresponding equipment is an actual problem. The analysis means a process of determining the presence of defective sections along with identifying structural elements of railway tracks in defectograms. This article is devoted to the problem of recognizing images of long structural elements of rails in eddy-current defectograms. Two classes of rail track structural elements are considered: 1) rolling stock axle counters, 2) rail crossings. Long marks that cannot be assigned to these two classes are conditionally considered as defects and are placed in a separate third class. For image recognition of structural elements in defectograms a convolutional neural network is applied. The neural network is implemented by using the open library TensorFlow. To this purpose each selected (picked out) area of a defectogram is converted into a graphic image in a grayscale with size of 30 x 140 points.

1. A. A. Markov and E. A. Kuznetsova, Rails flaw detection. Formation and analysis of signals. Book 1. Principles. St. Petersburg: KultInformPress, 2010.

2. A. A. Markov and E. A. Kuznetsova, Rails flaw detection. Formation and analysis ofsignals. Book 2. Data interpretation. St. Petersburg: Ultra Print, 2014.

3. V. F. Tarabrin, A. V. Zverev, O. E. Gorbunov, and E. V. Kuzmin, “About Data Filtration of the Defectogram Automatic Interpretation by Hardware and Software Complex ASTRA”, NDT World, vol. 64, no. 2, pp. 5-9, 2014.

4. E. V. Kuzmin, O. E. Gorbunov, P. O. Plotnikov, V. A. Tyukin, and V. A. Bashkin, “Application of Neural Networks for Recognizing Rail Structural Elements in Magnetic and Eddy Current Defectograms”, Automatic Control and Computer Sciences, vol. 53, no. 7, pp. 628-637, 2019.

5. E. V. Kuzmin, O. E. Gorbunov, P. O. Plotnikov, and V. A. Tyukin, “An Efficient Algorithm for Finding the Level of Useful Signals on Interpretation of Magnetic and Eddy Current Defectograms”, Automatic Control and Computer Sciences, vol. 52, no. 7, pp. 867-870, 2018.

6. E. V. Kuzmin, O. E. Gorbunov, P. O. Plotnikov, and V. A. Tyukin, “Finding the Level of Useful Signals on Interpretation of Magnetic and Eddy-Current Defectograms”, Automatic Control and ComputerSciences, vol. 52, no. 7, pp. 658-666, 2018.

7. I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. MIT Press, 2016.

8. F. Chollet, Deep Learning with Python. Manning Publications Co., 2018.

9. TensorFlow. [Online]. Available: https://www.tensorflow.org/.