Abstract: A method is provided for characterizing a part. The method includes a) carrying out non-destructive measurements using a sensor, the sensor being placed on the part or facing the part; b) using the measurements as input data of a neural network; and c) depending on the value of each node of the output layer of the neural network, characterizing the part. The method further includes prior to steps b) and c): constructing a first database, on a first model part; employing a first neural network parametrized by a first training operation, using the first database; constructing a second database, containing experimental measurements of the physical quantity performed on the part to be characterized; a second training operation, using the second database, so as to parametrize a second neural network, using the parametrization of the first neural network. In step c), the neural network used is the second neural network.

Abstract: A method is provided for characterizing a part includes: a) carrying out measurements using a sensor, the sensor being placed on the part or facing the part; b) forming at least one measurement matrix using the measurements performed in step a); c) using the matrix as input datum of a convolutional neural network including an extracting block, configured to extract features from each input datum; a classifying block, configured to classify the features extracted, the classifying block outputting to at least one node; and d) depending on each node, detecting the presence of a defect in the part. The neural network employed in step c) is established using the extracting block of another previously parametrized neural network.

Abstract: This method comprises: generating, only using the device, a low-frequency signal that makes the structure vibrate, generating a high-frequency signal in the structure, measuring a vibratory signal caused by the generated low-frequency and high-frequency signals at the same time then adaptively re-sampling these measurements to obtain a re-sampled vibratory signal the power spectrum of which comprises: a first frequency range [uBFmin; uBFmax] of width larger than 5 Hz that contains 95% of the power of the low-frequency signal, a second frequency range [uHFmin; uHFmax] of width systematically smaller than uBFmin that contains 95% of the power of the low-frequency signal, signaling a defect in the structure if an additional power lobe is detected outside of the ranges [uBFmin; uBFmax] and [uHFmin; uHFmax].

Abstract: Methods and devices are provided for analyzing a tubular structure including at least two electromagnetic-acoustic transducers (EMAT) and, called sensors, attachable or attached in, on or in the vicinity of the tubular structure; and computation and/or memory resources, that are accessed locally and/or remotely and that are configured to determine, for the pair of sensors, a function representing the impulse response of the tubular structure on the basis of the diffuse acousto-elastic noise present in the structure. Developments describe the use of rings supporting the sensors; translation and/or rotation movements; permanent or temporary installations; hinged rings; various computation modes, e.g., intercorrelation, a passive inverse filter, or correlation of the coda of the correlation; the use of artificial noise sources, imaging (e.g., tomography) for determining the existence of one or more defects in the structure. Software aspects are described.

Abstract: A method and system for inspecting a rail by guided waves, the rail being instrumented by sensors. The method comprises the steps of receiving elastic wave measurements from one or more sensors, as a train passes, releasing energy as guided waves into the rail; and of determining a function representative of the impulse response of the rail and the sensors. Developments describe how to determine the existence, position and characterisation of a defect in the rail (e.g. fracture, incipient fracture, etc.), the use of inter-correlation analyses, correlation of the coda of correlations, Passive Inverse Filter, imaging techniques. Other aspects are described for exploring rail defects: sensor position and movement, acquisition time, sampling frequency, frequency filters, amplifications, techniques for learning during successive train passes, signal injection by transducers. Software aspects are described.

Abstract: A device for imaging defects in a structure includes N transmitters and P receivers to be distributed over at least one surface of the structure and a central unit controlling the transmitters and receivers to sequentially record Q?N×P signals (S) obtained from electrical signals provided by the receivers of Q different transmitter/receiver pairs, after reception of mechanical waves transmitted by the transmitters of these Q pairs. It further stores Q first and Q second corresponding reference signals (SREF1, SREF2), representative of the structure without defects and differing by random noise. A central processing unit is programmed to: correlate each signal obtained with the corresponding first reference signal, in such a way as to construct an image of probabilities of defects; correlate each first reference signal with the corresponding second reference signal, in such a way as to construct a reference noisy image; and subtract the reference noisy image from the image of probabilities of defects.

Abstract: A device for imaging defects in a structure includes N transmitters and P receivers to be distributed over at least one surface of the structure and a central unit controlling the transmitters and receivers to sequentially record Q?N×P signals (S) obtained from electrical signals provided by the receivers of Q different transmitter/receiver pairs, after reception of mechanical waves transmitted by the transmitters of these Q pairs. It further stores Q first and Q second corresponding reference signals (SREF1, SREF2), representative of the structure without defects and differing by random noise. A central processing unit is programmed to: correlate each signal obtained with the corresponding first reference signal, in such a way as to construct an image of probabilities of defects; correlate each first reference signal with the corresponding second reference signal, in such a way as to construct a reference noisy image; and subtract the reference noisy image from the image of probabilities of defects.