Gears are machine parts widely implemented in industry, and several problems different equipment present are due to cog issues, making corrective maintenance potentially expensive. Techniques that allow for diagnosis while these dispositives are operational are therefore necessary, addressing the problem in the context of predictive maintenance, through a model as realistic as possible. This work studies vibration signals for a healthy signal and for a signal where different fracture levels are present. In this model, the fractures consist of a series of straight increments with variable angles, and we evaluate the signals through a number of different indices, namely: Mean Value, Variance, RMS, Skewness, Kurtosis, Sr, Sα and Form Factor. Firstly, it is observed that the stiffness of meshing tends to decrease with the increase of the fracture, and then the vibration signals, which have periodic peaks that protrude in proportion to the increase of the fracture, are analyzed. For the model adopted, both the pinion and the gear provide good references for analysis. In the evaluation of the indices, it is observed that the Mean and the RMS present the best performances in the early diagnosis of the fracture, while indices such as Skewness, Kurtosis, Sr, and Sα present higher values of responses, with, however, a slower diagnosis rate. Finally, the Variance and the Form Factor present medium performances. We concluded that the ideal scenario involves a combined analysis of the previous indices.
|
