In this thesis, identification of damage in a thick steel beam, made of ST-52, is studied based on the guided ultrasonic wave propagation method. At the beginning, various iecting methods are introduced. Regarding many advantages over other methods, guided wave based damage identification is selected for the purpose of structural health monitoring (SHM) of the structure in hand. Fundamentals, theoretical and experimental aspects of this method are then illustrated. Digital signal processing (DSP) of guided wave (GW) signals, as the key parameter in identification process, is presented, next. For this intent, various DSP algorithms are introduced. Considering the high capability for processing in time-frequency domain, Wavelet transform is chosen for processing guided wave signals. Then, a DSP algorithm for identification of damage is presented and its different aspects are examined in terms of three major steps: pre-processing, processing and post-processing. With the aim of eliminating unwanted and redundant information, a set of signal filtering techniques like sampling, windowing, averaging, DC offsetting, normalizing and de-noising are applied to raw acquired GW signals. Then, pre-processed signals are applied to a number of processing procedures in order to extract damage related features. For this intent, continuous wavelet transform (CWT) together with scaled averaged-wavelet power (SAP) method are used. A set of features are then extracted from processed signals in order to detect and characterize different damage parameters. Taking advantage of Time-of-flight (ToF) feature, damage which is in terms of a saw-cut generated crack, is localized. In order to determine the damage severity (crack depth), the peak amplitude of damage reflected wave, is used which is obtained from the energy distribution of SAP. As the next step, a novel feature extraction technique named Damage Characteristic Points (DCPs) is presented so that the damage could be anonymously and automatically identified within the use of artificial intelligence (AI) techniques. Using the extracted DCPs, a multilayer feedforward artificial neural network (ANN) is developed and trained. As far as developing an efficient damage parameter database using the experimental tests, is pretty expensive and time consuming, finite element method (FEM) through a parameterized modeling is employed for generating the damage parameter database, cost effectively. For this intent, commercial FEM software ABAQUS® together with a Python® code are used. For evaluating the accuracy of the proposed damage identification scheme, experimental tests are then conducted. Two piezo-ceramic transducers are employed, one for generating and the other one for sensing the GW signals. A program compiled in C language is used for remote and precise control of experimental procedures. Experimental, as well as numerical validations are implemented by identifying actual cracks in a steel (ST52 alloy) specimen by the active actuator/sensor path and the proposed online structural health monitoring system. Excellent quantitative diagnosis results for damage parameters (presence, location and severity) are achieved. It should be noted that the employed DSP code is developed and implemented in MATLAB ® . Keywords: Structural health monitoring, Online monitoring, Guided ultrasonic wave propagation method, Thick steel beam, Damage, Crack, Sensor, Actuator, Piezoelectric, Digital signal processing, Continuous wavelet transform, Scaled averaged-wavelet power, Damage characteristic points, Finite element simulation, Experimental test, Artificial neural network.