Ultrasonic guided waves represent a new development in the field of non-destructive testing. Longitudinal guided waves are mostly used to monitor the damage of steel bars, but the received signal is usually degraded and noisy owing to its dispersive propagation and multimodal behavior, making its implementation and location challenging. The torsional mode of T (0, 1) is not dispersive in the propagation of a steel bar and only produces circumferential displacement. It was chosen, in this study, to conduct guided wave-based damage monitoring on steel bars to reduce the signal processing complexity. The defects of steel bars, including circular surface defects, internal defects, and uniform damage defects, were thoroughly investigated, respectively, using numerical simulation. The waves were excited and received using the pitch-and-catch technique and the collected monitoring signals were processed using Hilbert transformation to highlight the amplitude and time-of-flight values of the wave signals, which were used for defect identification. In this paper, the reflectivity of guided waves is compared between torsional waves and longitudinal waves, in each case. The impact of defect size changes on damage monitoring is studied and the sensitivity of both the wave frequency and the wave mode (L and T) is also discussed. The results show that the monitoring method based on the torsional wave T (0, 1) is more sensitive to surface defects than the conventional method based on longitudinal waves. The reflectivity of the torsional wave T (0, 1) can be twice that of the longitudinal wave L (0, 1) when the depth of the defect in the circumferential grooves is less than 50% of the diameter of the steel bar. It is more sensitive to shallow surface defects within half of the bar’s radius, and it can also effectively identify defects under the conditions of the uniform damage defects of steel bars, even when the measurements are heavily noise-polluted. This proves the superiority of the torsional guided wave T (0, 1) in defect monitoring and provides a theoretical basis for the application of the torsional guided wave T (0, 1) in actual monitoring.