Monitoring the condition of mechanical systems, including gear transmissions, can be efficiently achieved through vibration analysis and signal processing techniques. Identifying potential faults accurately and reliably in these applications often depends on having comprehensive data that represents various health states, which usually requires a substantial number of experimental measurements. This study introduces an innovative CM approach employing machine learning models and deep learning models trained on data generated through vibration signatures recorded through an in-house developed experimental rig. An experimental test rig was developed to measure gear vibration, accommodating gear sets with varying shaft center distances and known lead and profile errors of gears. A magnetic brake applies low torque to ensure gear mesh integrity. Eddy current proximity probes, mounted on a rigid base plate, capture horizontal and vertical shaft displacements, recording the vibration response while the current probe records the current data. The novelty is in identifying the class of profile deviation of the gears across different gear sets by utilizing experimental data for supervised classification and predicting the value of deviation using regression. The classifier successfully identifies gear faults and profile deviations, demonstrating the potential of the proposed model-based framework for CM. Deviations in gear profiles affect vibration magnitudes. This highlights the necessity of high-precision manufacturing processes to minimize performance deficiencies. These observations reinforce the importance of implementing stringent quality control procedures during gear production.