DETECTION CORROSION BY USING CONVOLUTIONAL NEURAL NETWORKS TECHNIQUE

There are traditional methods for detecting corrosion, such as using cameras and physical examinations, but they are not useful for many types of corrosion. Artificial intelligence AI predict corrosion into metals these are smart technologies will contribute to enhancing efficiency, productivity and sustainability across the various stages of the metals process. Knowing the initial corrosion helps to identifying corrosion area of the surface and taking the necessary measures to prevent it. In this research, we propose an automatic method for examining images of metal surfaces and classifying the pitting corrosion strength through Convolutional Neural Networks (CNNs). It use open source dataset of steel bridge corrosion was created along with the annotations that matched them. This database contains 2530 images are separated into two sections: 980 with no corrosion and 1550 with corrosion. Initially, image processing is done to make the image specifications suitable for the CNN model. After that, a number of CNN models are compared, with pre-trained models being used with transfer learning technology. The best evaluation accuracy results are 98% in the training set. As a result, the outcomes show how well the suggested algorithm works as an assistance model for corrosion detection.