Design of an Intelligent System Using RNNs to Detect Steel Plate Surface Defects

For the steel manufacturing sector, steel defect diagnostics is crucial since it has a direct impact on both production efficiency and product quality. Although product quality control is less automated and unreliable in identifying steel imperfections in the surface, it suffers from a real-time diagnostic capacity. This paper introduces a Recurrent Neural Network (RNN) approach for detecting defects in steel plate manufacturing. The steel manufacturing plants may encounter a variety of flaws, including scratches, holes, crazing, and dirt. In the proposal, the first step is to take a different number of defective and non-defective images and then extract the feature using the wavelet transform. Prepare a feature matrix with 13 features for each image. The completed data set is fed into an RNN to assess the suggested algorithm's effectiveness during testing and training. The proposed method is evaluated using both online and industry data. Also feeding different numbers of images to determine the accuracy of suggested algorithms. The proposed approach is implemented using MATLAB software. The proposed strategies have an accuracy of 98.25% based on empirical data.