A neural network predictor is designed for analyzing vibration parameters of the rotating system. The vibration parameters (amplitude, velocity, acceleration in vertical direction) are measured at the bearing points. The system's vibration and noise are analyzed with and without load. The designed neural predictor has three (input, hidden, Output) layers. In the hidden layer, 10 neurons are used for approximation. The results show that the network is useful as ail analyzer of such systems in experimental applications. The neural networks are validated for reduced test data with unknown faults.