تعیین موقعیت سنسور برای نگهداری پیشگویانه ماشین آلات دوار

The monitoring by measurement and analysis of vibration is largely used to detect the defects in revolving machines. The determination of the best sensor positions is one of the main research goals in the field of predictive maintenance. This paper proposes a numerical methodology based on a finite element model and a spectral analysis in order to find optimum sensor positions. The bearing is a key component for the vibration propagation from the moving parts to static ones. An analytical bearing model and its numerical implementation in a finite element code are presented. The tangent stiffness matrix of the bearing element is obtained by the Newton–Raphson method and then used for the modal and spectral analyses. Several techniques are used to find the most sensitive zones to common defects. The proposed numerical approach correlate well with the experimental results. The numerical modeling of a grinder shows the interests in industrial applications.

The follow-up of the damage of some parts in a rotating machine by vibration analysis is a widely used technique in the predictive maintenance. The purpose of this type of maintenance, advantageous to the curative and periodic maintenance, is to carry out an intervention on a part just before its mechanical failure (AFNOR, 1995). This requires the monitoring and analysis of the evolution of vibration spectrums at one or several points on the machine in order to detect the characteristic peaks of common defects (Max, 1987). For the vibration follow-up of the bearings, it is possible to calculate in advance the frequencies of ring or ball defects according to the bearing geometry and its rotating speed (Morel, 1992). In most cases, the ideal measurement points are situated near the parts to be followed up, but the size of some machines and the accessibility to certain areas makes it difficult, even impossible, to take the measurements in these places. Although the reliability of defect detecting in the predictive maintenance has made enormous improvements, mainly due to the computer treatment of vibratory signals, it is nevertheless greatly dependant on the quality of signal analysis and positioning of sensors. This study proposes a methodology based on a numerical approach in order to find an optimum sensor implementation on a revolving machine. The numerical modeling allows to determine the number and the location of measurement points. Particular consideration is given to a common component on revolving machines: the bearing which is the only material link between the moving part and the immobile part in the vibration transmission. Firstly an analytical model for the rolling ball bearing is presented. The relations between the displacements (rotations) and the forces (torques) are obtained by using the cinematic relations and the Hertz contact theory. The formulation of the bearing element and its implementation in a finite element software are described. The model and the spectral analyses are used to determine the most sensitive zones for given defects. This numerical methodology is validated with an experimental academic example. Moreover, the comparison between the numerical and experimental results of an industrial grinder shows the interest of this application in the sensor monitoring setting for the predictive maintenance.

In this paper, a numerical methodology is proposed to determine optimum sensor positions for predictive maintenance. An analytical bearing model is adopted to avoid complicated computations. An explicit tangent stiffness matrix is obtained with the Newton–Raphson method. A bearing element is implemented into the commercial code ABAQUS. The non linear FE modeling allows to determine the tangent stiffness matrix at the given pre-load lever, then this matrix is used for the dynamic analysis. The defect excitations are assumed to be periodic such as the defects on the inner or external rings. The analysis of the response vectors in the frequency domain allows to find the most sensitive zones to the defect excitations. An elementary unit shaft-ball bearing housing was modeled to study its spectral response to defect excitations. The comparison between the experimental and numerical results shows a good correlation. An industrial application was carried out and validated experimentally. This experiment shows that it is possible to choose, with a finite element modeling, an optimum measurement area on a revolving machine. This zone will be used to monitor chipping defects on a bearing of the grinder. This methodology gives more information to help engineers to design revolving machines with the point of view of predictive maintenance.