Dvood Rezaei Sabet - Alamut Metallurgy Powder Co., Qazvin, Iran
Amir Movafeghi - Nuclear Science and Technology Research Institute, Tehran, Iran
Effat Yahaghi - Imam Khomeini International University, Qazvin, Iran

Manufacturing of different industrial components by powder metallurgy (PM) is one of the repeatable mass production methods in net-shape production, particularly, for automotive industry. Defected PM parts can be removed from the manufacturing lines and production cycle by the inspection system and this can increase efficiency. Radiography is a nondestructive testing (NDT) technique that allows the inspection of the outlet parts to increase the system performance and eliminate the defective parts as much as possible. The detection of the different cracks and the other types of the defects on radiography images can be improved with the image processing techniques, such as de-blurring algorithms. The original radiography images are often blurred and defects are difficult to detect with the naked eye. In the digital radiography image, the defects are appeared as gradients. Therefore, the methods which are based on gradient (e.g. Gaussian diffusion method) can improve the defect detection probability. The method relies on a diffusion process formulated by parabolic partial differential equations as an isotropic diffusion equation. In this research, a set of PM automotive parts with different size and defects were studied at Alamut Powder Metallurgy Co. The initial radiography images of parts were provided by a computed radiography (CR) digital radiography system. Then, the Gaussian diffusion method was used to improve visibility of the region defects in the radiography images of the parts. The results show that after subtraction of smoothed image from the initial image, the reconstructed images show that the contrast of defect regions improved and internal defects were better specified.