At present, vibration signal analysis is a foundation of fault diagnosis. The most widely used method is to analyze the characteristic quantities extracted from an original vibration waveform of a particular measuring point at a particular moment, but it lacks abilities in fault classification. Based on the information fusion process, two kinds of information exergy indicators are proposed and defined to represent the change rule between different states in the same process and the rule of corresponding state changes in different processes. On this basis, a fault diagnosis method of information exergy of rotating machinery is proposed that is based on the space time feature spectrum in the frequency domain. This method has been used to analyze vibration signals and diagnose faults of a rotor test platform. The numeric results indicate that it is an effective fault diagnosis method.