Fault diagnosis is a complex problem that concerns effective decision-making. Air brake system is a crucial safety unit and failure of which leads to the loss of vehicle yaw stability and increase in stopping distance of the vehicle. Fault detection and isolation in brake system is critical for continuity of the performance and the safe running of autonomous vehicles. Carrying out timely system diagnosis whenever a fault occur is important to prevent component degradation and vehicle breakdown. This work use the Rough Set Theory to develop fault diagnostic scheme for classifying the "fault" and "Nofault" conditions of air brake system with the knowledge of wheel speed sensor data. The rough set reduction principle is applied to find all reducts, and then a set of generalized classification rules for predicting the faults of Air brake system was extracted. The results show that the presented method can effectively enrich the vehicle condition monitoring system, able to give a feedback to the driver regarding the working condition of air brake system, detect and identify the location of faults, minimize the resources, such as time, cost of maintenance, etc.