Detecting corrosion in the tendons of post-tensioned (PT) concrete structures is technically challenging. The general inaccessibility of the tendons makes evaluation difficult, costly and often inconclusive. This study had two objectives. The first objective was to design, implement and validate the use of an innovative monitoring system based on embedded arrays of low profile piezoelectric transducers to generate and receive guided ultrasonic waves (GUW) over the length of a pre-stressing tendon. The second objective was to develop a new signal processing technique based on fractal theory, to enable the monitoring of the corrosion progression. To accomplish the objectives of the study, accelerated corrosion tests were carried out on two seven wire steel strands embedded in two concrete blocks. Signal generation and data acquisition were achieved with a National Instruments (NI), modular PXI unit. This unit included an arbitrary waveform generator card and one, multi-channel digitizers. In addition, a high voltage amplifier was used to amplify the excitation to the ultrasonic transmitters. Toneburst signals, consisting of 3.5 cycles, were excited by sweeping the generation frequency from 50 kHz to 400 kHz. LabVIEW software developed at the University at Buffalo (UB) was used to control the sensors, acquire and process the data. A box counting algorithm was used to calculate the fractal dimension (FD) of the GUW signals. The changes in FD as a function of time at the anchorages and inside the beam were analyzed. The following corrosion stages were observed at the anchorages: 1) an initial stage in which the FD shown no significant change in values; 2) a second stage characterized by sharp drops, indicated that a significant change in the signal shape occurred due to the initiation of the corrosion in the strand (as corrosion progress scattering, multiple reflections and mode conversion are expected). Inside the beam two corrosioninduced damage mechanisms were observed: 1) loss of bond between strand and grout that caused an increasing FD and 2) deterioration of the strand which resulted in a decreasing FD. In addition, an outlier detection algorithm based on the fractal dimension of GUWs was developed to automatically identify the initiation of the corrosion. Finally, the proposed approach may play an important role in decision-making processes by providing a relatively “early warning” of the corrosion process and allow the authorities for the planning and implementation of corrosion control strategies at a point where it is less expensive and invasive than when visible surface signs of corrosion have been observed. However, more theoretical investigations and formal tests need to be carried out to verify the robustness of the approach. In addition, large scale specimens should be examined.