Vibration bearing diagnostic system machine learning

Abstract

The article shows the possibility of using a mathematical model of a bearing with various types of defects for machine learning of the vibration diagnostic system. A plane geometric model of a radial bearing with two degree of freedom is considered in detail, a system of differential equations describing the nature of the motion of moving masses in the mechanical bearing system is presented. A formula is obtained for calculating the total kinematic disturbance, the calculation of which is associated with significant computational costs. The article proposes an alternative calculation method, namely, modeling the profile of kinematic disturbances in the time domain, which will minimize computational costs. The adequacy of the proposed method is confirmed by calculations in Matlab and comparison with experimental data. The development of equipment maintenance systems for the actual technical condition is relevant as in the case of operation of facilities of increased danger Gerike, and in the case of the operation of expensive equipment Pisarev and Vaganov, where the diagnosis with disassembling the machine leads to large financial losses. Diagnostics of equipment by vibration parameters allows fixing defects in the early stages during normal operation without the need to disassemble the units and take them out of service.