The Research Of Neural Network Application In The Electromagnet Nondestructive Testing

With the industry rapid development, the quality control of metal work piece is always a difficult problem. Electromagnet nondestructive testing that is applied widely is a testing method nondestructively. The artificial intelligence technique is developing, which has been promoting the research of intelligent NDT device. So, I have been beginning to research NDT based on artificial intelligence neural network system. To settle problem, I study the neural network fuzzy ISODATA arithmetic and obtain some success. Because neural network has self-studying ability, self-organizing memory, non-linear and function inferring, it has the ability to learn the physical law hidden in the testing data and do great deal of information-dealing work that other means can't realize. According to the characteristic of neural network, establishing the HNNS network model under the VC++6.0 platform combining with the ISODATA arithmetic merits to dispose of the data being from the DSP system. According to the result of experimentation, this system can work out completely and correctly the hardness of bearing and the crack of bolt. At the same time, researching the communication technique, orthodox serial port cann't meet fast data transmission. USB is the best of interface communication, analyzing indicate the USB technique will optimize the instrument.