Vibration based condition monitoring plays an important role in fault detection of rotating machinery components. Conventional methods like time -based, frequency-based and time–frequency-based methods are often used to diagnose faults in bearings, shafts, rotating parts etc. However, these methods have limitations because they are not adaptive, and the signals obtained from faulty machinery are nonlinear and nonstationary. Therefore, mode decomposition techniques have become a promising alternative. This work presents a novel hybrid approach which includes the Nonlinear Mode Decomposition (NMD), Statistical Parameters, and Envelope Spectrum. First, data is acquired from the rotating machinery, and NMD is applied to the data, which decomposes the signal into different modes. These modes are then sorted using different statistical parameters such as Root Mean Square (RMS), Variance, Standard Deviation (SD), Kurtosis × RMS, RMS × SD, and RMS × Variance. The envelope spectrum is then formed for these selected modes, which show the fault signatures. Several experiments are performed on the experimental setup with multiple faults such as inner race fault, outer race fault, misalignment, etc. The proposed method successfully detected these faults, and the results, evaluated at different speeds, are promising.