Weld Flaw Recognition By Neural Network Using Phase Array Images

Nowadays, modern steel productions are used widely inmany fields like aviation, petroleum, chemicals, electricity, railway and soon. Welding plays a very important role in joining parts of equipment of mentioned industries. The craft of welding can be improved in terms of welding tools ,welding technology and welding inspection. Quality control and welding inspection have been very important complicated problem, special trained people can operate and survey the result of NDT equipment such as phase array ut, MT,UT and RT .In this research according to electronic hardware development and inspection methods improvement from 1D to 3D , possibibillity confidence to new methods, replacement from old ways e.g. x-ray that it is harmful for health to phase array ,will be discussed. By applying image processing techniques on PA and x-ray images, feature extraction is done. The techniques includes image segmentation, edge detection, and image enhancement (thresholding, noise elimination). for flaw detection, firstly combination of x -ray and phase array images parameters are applied to neural network for training and testing then both train and test by phase array features will be done. a multi layer neural network with backpropagation algorithm selected as classifier.

Keywords:

Multi layer perceptron NN , Back propagation algorithm , Segmentation , Thresholding , Clustering , Phase array transducers