Data Science

NDT aims to increase inspection accuracy by providing reliable indicators of the damage state of a structure or material. Signal theory-based models and feature extraction is a physical approach that consists in determining the descriptors of an acoustic event (AE event, vibrational response…) to predict an anomaly or a state of damage.

Deep learning approach consists in providing raw-acoustic signals and use a convolutional deep neural networks to allow the model extracting the features.

Deep Learning algorithms can also be used to build surrogate models in order to predict materials properties when the outcome can not be measured or easily computed.