Abstarct: Structural health monitoring aims to provide a report on condition of materials, members, and connections and integrity of a structure, during its useful life cycle. Damage can change state of the structure in its static or dynamic behavior. And certain parameters of the structure change. Monitoring health of a structure provides useful information about the serviceability, integrity and safety of structures. Therefore, maintaining continuous performance depends on monitoring the occurrence and propagation of damage. Several monitoring and detection methods have been developed to provide practical tools for the early detection of structural damage. Also, a significant effort has been made to develop vibration-baxsed methods that use the vibration response of the monitored structure to assess its condition and identify structural damage. It has become more practical and is widely used in the calculation of vibration-baxsed structural damage detection. This thesis deals with a one-dimensional convolutional neural network structure that is able to detect small changes in stiffness and mass as damage in the structure, using experimental data. The proposed structure was examined. In the proposed network structure, a convolution laxyer is added after the input laxyer, which is different from previous studies. By this change in the structure in the case studies related to the two benchmark structures of wooden bridge and three-story laboratory frxame, accuracy of the model has increased and the network will be able to predict the structural states. The input data of this network is the acceleration data extracted from the sensors installed on the structure. Also, acceleration data is not pre-processed and directly used as training and validation data. After training, the discussed neural network identifies the locations of small local changes in the mass and stiffness of the structure, which indicates the high sensitivity of the proposed network to small structural changes.