Damage diagnosis systems for mechanical structures have a great importance and the development of such systems and methods of damage detection attracts many attentions to itself. It is expected that these methods provide enough information about structural health conditions and inform the operator of introduction, severity, and location of damage. Ultrasonic wave propagation with piezoelectric transducers is one of the most efficient and popular methods for detecting damage in structural health monitoring systems. Damage detection by using ultrasonic guided waves is facing issues such as complex wave propagation mechanism, the existence of multiple modes of guided waves in the same time and dispersive nature of ultrasonic waves. In addition, the biggest challenge to this goal is that the method is prone to contamination from the temperature variations and operational conditions. Damage-sensitive features are also sensitive to temperature variations. Therefore, conventional methods of damage detection lose their functionality in case of temperature changes effect on ultrasonic wave data and falsely alert the introduction of damage. The aim of this project is to remove undesired temperature effect from wave data in order to detect damage in the presence of temperature variability by using ultrasonic guided waves. To understand wave propagation mechanism, excitation conditions in pipes, finding the best sensor location and determination of the optimal number of actuators, three-dimensional and transient FEM simulations are performed. Temperature effect on guided waves in steel pipe and plate is studied experimentally. To detect damage in the presence of temperature variation and to remove undesired temperature effect on wave data, cointegration and fractal signal processing are used. The cointegration is a method that developed originally from the field of econometrics and relies on the analysis of stationary behavior whereas the fractal signal processing is based on wavelet transform for self-similarity analysis of time series in different scales. The idea behind using these methods is to find self-similarity pattern in cointegration residuals that are purged of the temperature effect. The influence of damage introduction on self-similarity pattern is studied, and severity of damage is Keywords: structural health monitoring, guided wave propagation, cointegration, wavelet transform, fractal signal processing, self-similarity, temperature variation, FEM simulation, experimental setup