چكيده به لاتين
Bearings are commonly used in all rotary machines in a variety of industries, including railways, and more than half of industrial machinery failures are due to defective bearings and their failure is a very important factor in the disruption of the machines and the resulting damage. Therefore, technology in the industry that recognizes the primary defects of bearings to prevent the loss of widespread efficiency and performance is crucial, and extensive research has been done on tracking and troubleshooting the bearings.
In this thesis, first, the general defects of the bearings and the reasons for their creation, the generalities of the monitoring of the situation and the methods of troubleshooting are introduced. In the following, a brief introduction of the mathematics and basic concepts of Hilbert-Huang's method, which is simpler than other methods and can be used both in transient and static systems and in linear and nonlinear systems, is described and the algorithms of empirical mode decomposition and intrinsic mode functions are described. A brief description of the wavelet transform is described. Then, the conditions and how to perform the test are used to extract the vibration signals required for the analysis taken from the CWRU bearing data center.
To analyze the signals obtained from the test, a new method for improving troubleshooting is presented based on a combination of previous methods and a new method for estimating defect size.
In the troubleshooting section, firstly, different modes are examined and it is shown that the first mode more than other modes indicates defects characteristics. It is then shown by Hilbert transformation that in the time-frequency domain, if the defect is too small, the defect can be determined, but for a larger defect it is not a suitable method. In the following, the proposed method is verified and for all defects, the troubleshooting process is done with certainty.
. In the dimension estimation, the signal is investigated in the time domain, so that the signal is firstly decomposed with the aid of a discrete wavelet transform as described, and the size of the defect is estimated using the extracted relations. At the end of the measurement, the estimated defects are compared to the actual defects, and it is shown that the method error is very small and can be used to estimate small sizes
