Abstarct: In recent decades, damage detection and structural health monitoring with three main objectives of reducing maintenance costs, improving safety and reliability of the structure have been considered. Advances in the technology of sensors and processing the response of structure have made non-destructive methods as useful, easy, and cost-effective methods for identifying structural damage. Periodic inspections to prevent the development and aggravation of damage is an important factor in structure health monitoring. Therefore, developing methods that can detect structural damage only by using vibrational responses over a certain period of time makes them more prominent. In this research, an attempt is made to determine the damage sensitive features by using data processing tools in such a way that the existence of damage, its location, and severity can be determined only by using recorded acceleration vibration response. Initially, the empirical mode decomposition is applied as a method to remove noises from the data. Time series modeling is a powerful and efficient method that can be used to transform the vibrational response of a structure into a mathematical model and extract damage-sensitive features from the vibrational responses of the structure. These methods require statistical decision-making criterion to determine variations in observed data. In this research, using Autoregressive time series modeling, structural damage sensitive features including model parameters are extracted. The probability density function is a common tool to detect the deviations and changes in data. Therefore, by calculating the probability density function of the damage sensitive features, we can observe and examine the changes in the trend of data that occurred due to damage in the structure. However, using the probability density function of the features cannot conclude the existence of damage. Therefore, using the distance correlation method, which is a statistical decision-making method, it is possible to make good decisions about these changes. The main advantage of using distance correlation is that it shows the independence of two random variables by a numerical value. Hence, a new fast correlation distance method is used. To evaluate the proposed method, structural damage detection is used on two benchmark structures in structural health monitoring. The results show that the proposed method baxsed on the use of the probability density function of the damage sensitive features from the time series modeling has been able to identify the location of the damage in these two benchmark examples.