Normalization of vibration signals generated under highly varying speed and load with application to signal separation

The paper presents a normalization dedicated to transform non-stationary vibration signals into signals characterized by purely stationary properties. For this purpose, a novel class of generalized periodic signals is defined followed by a proposition of a normalization technique, which takes advantage of available, instantaneous values of operational parameters. Within the paper, a well-known discrete-random separation (DRS) technique is recalled as an exemplary technique, which has been restricted to stationary signals so far. The authors present a step-by-step adoption of the DRS to non-stationary signals. The method is applied to simulated signal, test rig signal, and a vibration signal from industrial object. Additionally, for the purpose of synthesis of simulated signal, a new model of multi-component vibrations generated under varying regime is proposed. The presented method aims to expand existing solutions dealing with varying frequency to a more general solution dealing with independent, simultaneous varying frequency and amplitude of signal components.