Feature Fusion Deep Learning Model for Defects Prediction in Crystal Structures

Detection of defective crystal structures can help in refute such to decrease industrial defects. In our research, we are concerned with Silicon nitride crystals. There four types structure classes, namely no-defect structures, pristine random displacement and 25% vacancies structures. This paper proposes a deep learning model detect the high accuracy precision. The proposed consists both classification regression models new loss function definition. After training models, features extracted fused utilized as an input perceptron classifier identify A novel dense neural network (DNN) is multitasking tactic. developed multitask tactic validated using dataset 16,000 30% highly Crystal images captured under cobalt blue light. DNN achieves precision 97% 96% respectively. Also, average area curve (AUC) 0.96 on average, which outperforms existing detection methods for experiments depict computational time comparison single epoch versus state-of-the-art models. performed diffraction image database twelve batches. It be realized that prediction least 21 s.