Research Group
Diagnosis and Maintenance of Guided Transport Systems

DIAGUS
Diagnosis of internal rail defect by ultrasonic signature analysis

The railway networks achieve inboard inspection of the internal state of the rail at reguar intervals with the help of several ultrasonic probres. The detection of some type of defect is still difficult. The improvement of such kind of measuring systems and signal processing capabilities allows the implementation of more complex efficient detection procedures, multisensor fusion and better classification modulus. For example, the B-SCAN data use (instead of A-SCAN) leads the way to recognize horizontal or vertical splits and to estimate the defect gravity.

With SNCF partner (departement of Infrastructure), the Diagnosis group works since 2001 on this theme. This project tackles the signal 2D processing (wavelet denoising, Radon transform, transfor parametrization, mathematical morphology...) and information fusion. A CIFRE PhD thesis has been conducted, in collaboration with UTC (P. Simard). The main contributions involve :

• Ultrasonic field 3D simulations by finite elements (CIVA™ -CEA software)

• New real time detection tool adapted to the angle-variable case : the Smoothed Radon Transform

• 2D wavelet denoising and defect metrology by neural networks

• improving the well detection rate (+15%) and decreasing the false alarm rate (-20%) without any modification of sensor arrangement.

Staff: Patrice Aknin, Hervé Cygan