A system capable of detecting any presence of mechanical defects on the lateral surface of magnetic rotors has been developed for a leading company in the sector.

Thanks to the vision algorithm, a high frame rate and high-resolution linear camera; the system develops a reconstruction image of the lateral surface of the object.

Starting from the reconstructed image with the use of an adequately trained neural network, an inference is performed which allows the recognition of cracks and other anomalies. And then it distinguishes between the acceptable and unacceptable rotors as defined by customer.

For instance, based on the size of a crack, it can be assessed whether it is to be considered defective or not.
Furthermore, it is possible to evaluate the possible presence of external dust particles and stains as non-defective.