The Air Production Unit (APU) is responsible for managing the load in the metro train by making sure that the load is evenly distributed on the wheels despite passenger congestion. The failure of APU results in a complete halt of operations. Therefore, it is pivotal to timely maintain the APU system. In this paper, a predictive maintenance (PdM) approach has been suggested using multiple state-of-the-art machine learning algorithms to predict failure in the APU. Sensor parameters that lead to failures in APUs of the electric trains were identified from the MetroPT train dataset by employing a verity of baseline models including linear regression (LR), decision tree regressor (DTR), random forest regressor (RFR), gradient boosting regressor (GBR) and XGBoost Regressor (XGBR) models. This will significantly reduce both the operational cost as well as the downtime of the trains. It will also help in the identification of faulty parts at a faster rate and reduce the failure rate of urban transportation systems.