Abstarct: Abstract Structures, as vital components of infrastructure, are subjected to various factors such as fatigue and corrosion, which can threaten their safety and longevity. Therefore, developing methods capable of accurately detecting structural damages is essential. The primary objective of this study is to develop an advanced model utilizing convolutional neural networks (CNNs) for the precise identification of localized damage in flexural plates. This model is designed to analyze structural vibration data and detect subtle changes in structural properties, such as stiffness. The employed dataset consists of simulated accelerations under random loading and environmental noise, generated through theoretical modeling of flexural plates. To ensure the accuracy of the data generation process, a verification example has been designed and validated. During the research process, the proposed model was trained using the generated dataset and subsequently evaluated on benchmark samples and several numerical examples to assess its accuracy. The results indicate that the convolutional neural network is capable of accurately predicting damage locations. One of the prominent features of this model is the reduction in data preparation complexity, eliminating the need for preprocessing. The findings of this study demonstrate that integrating modern deep learning methods with simulated data can significantly enhance structural health monitoring systems. This achievement can contribute to reducing maintenance costs, increasing safety, and extending the lifespan of critical structures such as bridges and essential buildings.