Neural network for automating diagnosis in aircraft inspection

As and when aircraft aged, fatigue problems appeared and the necessity for detecting cracks in the lap joint parts became obvious. Numerous systems were then developed to follow up cracks appearance and growth in rivet holes. If most of the inspection used to be manual, it is now partially automated since AEROSPATIALE developed the System of Inspection Assisted by Microprocessor (SIAM) in which an EC probe slides on the rivet lines [1]. At present, our research is focused on the diagnosis, last phase of the inspection. In order to process the inspection data, to speed up the control, and to improve the detection performances, a neural network has been developed