Fault diagnosis using vibration signals is a major challenge for industrial manufacturing. Obtaining defect information is an important step to make decisions about the maintenance in prognostic and health management systems. Existing studies mostly considers vibration signals collected from elements such as rolling element bearings and hydraulic presses. In this paper, we use the vibration signals obtained from the mechanical transfer press during metal forming process and analyze them from the signal processing point of view. Experimental results reveals that spectral analysis is a good candidate for fault diagnosis and it provides important information about the localized faults embedded in the vibration signals.