Structural Health Monitoring (SHM) has become a critical area of research due to its potential to enhance safety, and reliability, and reduce the life-cycle costs of structures. Traditional SHM techniques often rely on a healthy baseline for comparison, which can be impractical under varying operational conditions. Hence there is a need for reference free technique for SHM. In this paper, a special arrangement of FBG sensors called self-referencing is explored for achieving reference free detection [1,2]. The self-referencing configuration allows the simultaneous acquisition of signals from both healthy and damaged sections of a structure, removing the need for baseline comparisons. This paper explores the application of different supervised and unsupervised machine learning algorithm, including k-means clustering, decision trees, and different forest based algorithms for damage detection and localization. The comparative study is carried out on numerical and experimental results. The result identifies the relative merits of these techniques and determines their suitability for more generalized application of the self-referencing configuration.