Skip to Main Content
It was developed the special neural network classifier, which provides a flexible and stable base of knowledge about the possible defects of honeycomb panels, and which effectively operates with data vectors of large dimension. This classifier has ability to adapt the architecture of the generated network for new changes and an opportunity to get high reliability of control. As a result of the work, was developed system of standardless diagnosis and classification the technical state of products from composite materials, which allows to identify defective parts and objects under control, provides their classification by the degree of damage. Application of artificial neural network for processing of the experimental data shows that it is possible to automate this process and decision making on the results of NDT. The use of the system is feasible and enables to achieve high accuracy control - 97 - 98%. Conducted experiments and the results showed promising application of neural networks as the core of information-diagnostic system for nondestructive testing and classification of defects in products from composite materials. It was determined that for solving the problem of standardless diagnostic of cell panels Fuzzy-ART neural network with sensitivity coefficient of p = 0,92 is needed.
Date of Conference: 25-27 Aug. 2011