The great enhancement in the transmission media in computer networks has brought light to fiber optics because of the high data transfer rates and low signal attenuation. Nevertheless, the continual problem of reliability and efficiency of optical fiber networks persists to this day, especially in regard to fault detection and rectification. This paper provides a detailed overview of the fault detection techniques in optical fiber network with a background examining the types of faults as perceived by local monitoring centers known as Network Operations Centers. It emphasizes the need for the fault detection and fault classification and the consequences on the network’s performance. The paper briefly discusses historical approaches to fault detection and location, including OTDR, before identifying problems with their precision in determining fault location. For these challenges, the work considers the utilization of Machine Learning (ML) approaches for fault detection and prediction in optical networks. The work explores several fields in ML such as fault localization, predictive maintenance, anomaly detection, and signal quality optimization. Recent research activities and their impact on fault detection as well as fault classification in optical networks are also assessed. In conclusion, the paper points to the possibility of achieving a significant boost in fault identification accuracy when using ML-based methods, increase in network availability and reduction of outage times in optical fiber networks.