Rail corrugation is a common type of wear in all tracks. With the rising axle loads and train speeds for modern railway administrations, the demand of early detection has taken prominence. With new methodologies and algorithms arising, it remains a challenge for industries to decide which tool is the best fit. This project proposed a machine learning architecture for the detection of rail corrugation. The software developed in this project is based on the dataset obtained from an advanced FBG sensing system. A data pre-processing software was developed to extract signatures from the interaction of train wheels and rail. A J48 decision tree algorithm followed by advanced sampling techniques were utilized to ascertain its level of influence on the formation of corrugation. The problem was then modelled using various classifiers and their performance was evaluated by comparing their classification accuracies. The results demonstrate the effectiveness of the proposed approach with 97% accuracy. Hence, this method is fit for industrial application and its universal setting makes it applicable to various engineering disciplines. In short, the proposed approach offers robust performance advantages and, as fault detection being the preliminary stage of predictive maintenance, it also establishes a strong foundation for future development.