Fault detecting technology based on neural network algorithm

The paper describes an automatic on_line detecting system which can detect inner parts of complex electromechanical products. In the system,digital image processing technology is used to preprocess X_ray images of the products, and neural network algorithm is applied to diagnose faults. The fault recognition model adopts an improved back_propagating neural network, which is trained by a series of standard X_ray images of correctly assembled products.During the process of detection, two images of objects in different directions are capable of acquiring the status of the key parts. After comparing the features of the two preprocessed images and standard images,the network can estimate different types of faults of the key parts. The detecting system combines digital radiography technology with digital image processing, and applies the back_propagating neural network algorithm in the fault recognition process. The system improves the speed and reliability of fault detection and has practicability in the field of industrial nondestructive detection.