The need for semi-autonomous or autonomous operations, communication delay, short contact periods as well as the need for survival in harsh environments poses unique challenges to Automotive Mechanical Transmission Systems (AMTS). Predictive health monitoring (PHM) systems are currently gaining in popularity due to their effectiveness in providing robust information about the system condition and reducing maintenance costs. This paper presents a PHM system for monitoring different gear faults using vibration analysis and Support Vector Machine (SVM) algorithms. Experiments were conducted on a multi-stage gearbox (Automotive Mechanical Transmission Systems) under three conditions, normal, external vibrational excitation and oiling system high temperature. Multi-class SVM based on developing a model for normal and faulty states; the model used for monitoring the upcoming sensory data, and classifies them as normal or faulty ones. The model is verified through additional experimental observations. The classifier algorithm was coded in Matlab and showed a good potential in classifying different failure mechanisms.