Structural health monitoring systems have a return of investment of over 13 times [1] and hence has attracted interest from different sectors. SHM systems look at damage detection as an anomaly detection problem. The techniques identify changes in the system with respect to a known healthy condition. If the change in the damage sensitive feautre is greater than a certain pre-determined threshold the structure is flagged as damaged. In real applications due to the changing ambient conditions, the comparison of the measured quantities with the known healthy condition is not easy. There need to be elaborate compensation techniques and signal processing to allow this. This is not always possible and hence there is a need for reference free damage detection. This paper makes use of the so-called self-referencing [2] configuration of the FBG sensors for guided waves (GW) sensing. The self-referencing configuration under some circumstances allows simultanaeous measurement of the baseline and the damaged signal. Some authors have used this technique for damage detection. In the present work the authors expand this work using machine learning algorithms for damage detection and localization. The method is a realized on an isotropic plate with 2 FBG sensors in self-referencing configuration. The results indicate that analytical techniques based on time of flight are not capable of damage detection and localization but through the use of data-driven tools the localization is indeed possible.