The usage of big data and data analytics methods has significantly increased during the Industry 4.0 era. An example is the application of Machine Learning (ML) in Predictive Maintenance (PdM). PdM tries to avoid the premature and costly repair of a system while ensuring a timely repair before failure. This paper conducts literature studies from several journals to examine the PdM concept, the ML methods used, the level of accuracy, and the correlation with SDGs goals. As a result, the Neural Network is the most commonly used method and the most accurate method with an accuracy rate of 95.69%. It is followed by Logistic Regression at 92.40%, Gradient Boosting Tree at 86.20%, and Random Forest at 84.40%. PdM is also aligned with SDG goals in the 7th goal in energy efficiency, the 9th goal regarding increasing equipment reliability, and the 12th goal in terms of optimizing resource utilization and reducing waste. The challenges are the possibilities of integrating with real-time telemetric monitoring and developing the digital twin as a future prospect. For future research, it is recommended to apply these methods in real life and estimate the impact on operation and maintenance costs.