CONDITION MONITORING OF GEARS UNDER MEDIUM ROTATIONAL SPEED

This paper presents a pre-processing technique to select the optimal filtering frequency band identifying gear fault signature and use the filtered signal to calculate statistical features. The selected technique the cyclic spectral correlation, as proposed by Antoni [11]. It allows the visualization of excitations of the signal on a map. According to the literature, this technique appears to be more powerful than the time-frequency methods as it exploits the cyclostationary nature of signals, emitted by rotating machinery. This work analyses the benefits of filtering the signal before proceeding to feature extraction. The last step of this analysis is related to dimensionality reduction using the Principal Component Analysis method. In order to understand the influence of different gear fault conditions, healthy, chipped-tooth and missing tooth measurements have been performed in a Machine Fault Simulator from SpectraQuest under different speed regimes. The central goal of this paper is to obtain key features that allow the identification of the condition of the gear at different speeds.status: publishe