A method for diagnosing bearing faults in transmission components of mine hoists
The signal is collected through the vibration acceleration sensor and the information entropy and K-L divergence are extracted using the collective empirical modal decomposition method, which solves the problem of difficult-to-diagnose early bearing failures in the mine hoist, and improves the reliability and safety of the system.
Patent Information
- Application Number
- CN202110908516.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-08-09
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2041-08-09
AI Technical Summary
The prior art is difficult to effectively diagnose early failures of bearings in mine hoists, resulting in safety and reliability problems.
The signal is collected by the vibration acceleration sensor, the information entropy is extracted using the collective empirical modal decomposition method, and the information entropy comparison and K-L divergence calculation are used to realize the diagnosis of bearing failure.
Improve the reliability and safety of mine hoist transmission components, simplify calculation steps, and reduce equipment and cost requirements.
Smart Images

Figure CN113591248B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of bearing fault diagnosis, and relates to a bearing fault diagnosis method in a mine hoist transmission component. Background Art
[0002] Mine hoist refers to a transport machine that connects underground work with the surface in mining engineering. In a mine, a mine hoist is an automated machine for underground work and surface work. Mine hoist mainly consists of a three-phase motor, a reducer, a drum (or sliding friction wheel), a braking system, a depth warning sign software management system, a mobile speed measurement speed limit system and an operating software system.
[0003] At present, relevant algorithms dominated by machine learning have been successfully applied to detect bearing anomalies in mining machine hoists. However, this method tends to ignore some early failures that may occur in rolling bearings, thus causing safety and reliability problems in the entire mining machine hoisting system. Summary of the invention
[0004] The purpose of the present invention is to provide a bearing fault diagnosis method in a mine hoist transmission component that can diagnose early bearing faults, in view of the problem that early bearing faults are difficult to diagnose in the prior art.
[0005] To this end, the present invention adopts the following technical solutions:
[0006] A method for diagnosing bearing faults in a mine hoist transmission component comprises the following steps:
[0007] (1) Collecting vibration analog signals of the bearing in normal state and fault state through a vibration acceleration sensor, and converting the collected analog signals into digital signals;
[0008] (2) The digital signal is extracted by using the sample entropy feature extraction method of the ensemble empirical mode decomposition to obtain the information entropy H (X i ) and the information entropy H under fault conditions f1 (X i );
[0009] (3) Compare the information entropy of the bearing in normal state and fault state: H(X i )-H f1 (X i ).
[0010] Furthermore, in the step (1), the collected vibration analog signal is converted into a digital signal through signal conditioning and A / D conversion, and then the digital signal is processed in real time.
[0011] Furthermore, in step (2), the digital signal after real-time processing is subjected to ensemble empirical mode decomposition to extract data, and the data residual r between the actual output of the sensor and the expected output is obtained. i , through the data residual r i Calculate the information entropy H (X) of the vibration acceleration sensor in the normal state and fault state of the bearing i ) and H f1 (X i ), the calculation formula is:
[0012] (1)
[0013] Where, is the covariance matrix of the multivariate normal distribution, and n is the data dimension.
[0014] Furthermore, the formula (1) introduces H(X i )-H f1 (X i ),get
[0015] (2)
[0016] Where: is the probability density function of bearing failure.
[0017] Furthermore, the formula (2) conforms to the expression of KL divergence, namely
[0018] (3)
[0019] Furthermore, the calculation formula (1) is improved by introducing Mahalanobis transformation to obtain the following formula:
[0020] (4)
[0021] Where: is the covariance matrix of the multivariate normal distribution, and n is the data dimension.
[0022] Introducing formula (4) into formula (3) yields:
[0023] (5)
[0024] Where: is the covariance matrix of the bearing data when no fault occurs, and is the covariance matrix of the bearing when a fault occurs.
[0025] The beneficial effects of the present invention are:
[0026] (1) The relevant data of bearing operation is collected by vibration acceleration sensor and converted into digital signals. The relevant digital signals are extracted using the set empirical mode decomposition method to obtain the corresponding information entropy. The fault diagnosis of rolling bearing can be realized by comparing the information entropy, thereby improving the reliability and safety of the entire transmission components of the mine hoist.
[0027] (2) Converting information entropy into KL divergence further saves calculation steps and makes the calculation more convenient and faster.
[0028] (3) Vibration acceleration sensors are installed on the bearings to collect corresponding vibration signals and convert them into digital signals for subsequent processing. Fewer installation equipment are required and the cost is low. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 It is the time domain diagram of the original signal of the bearing;
[0030] Figure 2 is the time domain diagram of the IMF component;
[0031] Figure 3 It is the power spectrum density diagram of different IMF components;
[0032] Figure 4 KL divergence value diagram of different fault types with different sensitive components under the same fault diameter.
[0033] Figure 5 This is the KL divergence value diagram of inner ring fault under different diameters under the same fault type. DETAILED DESCRIPTION
[0034] The present invention is described in detail below with reference to the accompanying drawings and embodiments:
[0035] A method for diagnosing bearing faults in a mine hoist transmission component comprises the following steps:
[0036] (1) Install a vibration acceleration sensor on the above bearing and collect vibration signals in normal state, inner ring fault, rolling element (ball) fault and outer ring fault state at a sampling frequency of 12k (the collected original signal time domain is as follows: Figure 1 As shown in the figure, the collected analog signal is conditioned by the MAX3082 high-speed transceiver and converted into a standard signal, and then the standard signal is A / D converted to become a digital signal, and then the converted digital signal is communicated to the TMS320DM642DSP processor for real-time processing of the digital signal.
[0037] (2) The digital signal after real-time processing is extracted using the ensemble empirical mode decomposition method to obtain the data residual ri between the actual output of the sensor and the expected output. The information entropy H (X i ) and H f1 (X i ), where the calculation formula is:
[0038] (1)
[0039] In order to save computing resources, the traditional information entropy formula needs to be redesigned. By introducing the Mahalanobis transformation, the improved information entropy formula can be obtained as follows:
[0040] (2)
[0041] Where: is the covariance matrix of the multivariate normal distribution, and n is the data dimension.
[0042] The entropy difference between a bearing in normal state and when a fault occurs can be expressed as:
[0043] (3)
[0044] Where: is the probability density function when the sensor drifts.
[0045] By observing the results, it can be found that it just meets the expression of KL divergence, and the expression of KL can be expressed as:
[0046] (4)
[0047] KL divergence is introduced for fault diagnosis. However, due to the high computational complexity caused by the difficulty in estimating the residual probability density function in KL divergence, the new KL divergence expression is obtained by substituting formula (2) into formula (4):
[0048] (5)
[0049] Where: is the covariance matrix of the sensor when no fault occurs, is the sensor X when a fault occurs i The covariance matrix of .
[0050] In order to determine whether the above results are accurate, the present invention takes the rolling bearing fault diagnosis platform of Case Western Reserve University as an example, and the specific experimental parameters are as follows:
[0051] When the drive end bearing fault diameter is 0.1778 mm at a sampling frequency of 12K, the original signals of different fault types when the motor load is 3 horsepower are decomposed by EEMD, and the KL divergence (standard entropy) and correlation coefficient between the IMF component after EEMD decomposition and the original fault signal are shown in Table 1;
[0052] IMF1 IMF2 IMF3 IMF4 IMF5 IMF6 IMF7 IMF8 KL divergence -1.2757 0.2053 0.2040 0.7232 1.4958 1.8410 2.1212 2.6916 Correlation coefficient 0.2951 0.0153 0.0153 0.0054 0.0012 5.80438 3.3140 1.0591
[0053] According to Table 1, it can be seen that in the method of distinguishing sensitive components based on correlation coefficients, (refer to relevant literature research: Xiang Chuan, Ren Zejun, Zhao Jing, Zhou Jiahui. Research on rolling bearing fault diagnosis based on ISSAE and XGBoost [J]. Mechanical and Electrical Engineering, 2021, 38(06): 704-711) according to the literature, the threshold of the sensitive component is set to 0.1 times the maximum correlation coefficient value. Therefore, the threshold value can be obtained as 0.5804. Therefore, according to the correlation coefficient value, the sensitive IMF components can be obtained as IMF1, IMF2, IMF3, IMF4 and IMF5 respectively. According to the KL divergence to distinguish sensitive components, it can be seen that there are obvious differences in sensitive components between IMF4 and IMF5. Therefore, through the set threshold d=1, it can be seen that IMF1, IMF2, IMF3 and IMF4 are all less than the set threshold, so these four components can be determined as sensitive components.
[0054] like Figure 3 As shown in the figure, due to the discrepancy between the sensitive components identified by the correlation coefficient and the KL divergence method, the sensitive components are further judged by comparing the power spectrum density diagram of the determined sensitive components. From the figure, it can be seen that the power spectrum density diagram of the IMF component shows that the power spectrum density diagrams of the IMF4 component and the IMF5 component are similar, so IMF5 can be judged as a false sensitive component. Therefore, it can be seen that sensitive component extraction based on KL divergence is an effective method. Therefore, the final sensitive components are IMF1, IMF2, IMF3 and IMF4.
[0055] From this we get Figure 4 As shown in the figure, the KL divergence values of different fault types of different sensitive components under the same fault diameter, as well as Figure 5 As shown in the figure, when the motor load is 3 horsepower, the KL divergence values of inner ring faults with different fault diameters at different sensitive components.
[0056] From above Figure 4 It can be seen that when the fault diameter is 0.1778 mm, the KL divergence values of different sensitive components for different fault types are significantly different. Therefore, this method can be used to extract features of different fault types. Figure 5It can be seen that under the same fault type (inner race fault), the KL divergence values of different fault diameters with different sensitive components are significantly different.
[0057] pass Figure 4 It can be concluded that if H (X i )-H f1 (X i ) < 0.7, it is judged that the bearing rolling element is faulty; if 0.8 <H(X i )-H f1 (X i ) <1.4, it is judged that the inner ring of the bearing is faulty; if 1.4 <H(X i )-H f1 (X i ) < 2, it is judged that the outer ring of the bearing is faulty. Therefore, this method can effectively extract the features of different fault diameters.
[0058] The method can effectively extract the features of the early faults of the rolling bearings which are most prone to failure in the transmission components of the mine hoist, realize the fault diagnosis of the rolling bearings, and thus improve the reliability and safety of the entire transmission components of the mine hoist.
Claims
1. A method for diagnosing bearing faults in a mine hoist transmission component, characterized in that: The following steps are involved: (1) The vibration acceleration sensor is used to collect vibration analog signals when the bearing is in normal state and fault state, and the collected analog signals are converted into digital signals through signal conditioning and A / D, and then the digital signals are processed in real time; (2) The digital signal is extracted using the sample entropy feature extraction method of the ensemble empirical mode decomposition to extract data and obtain the data residual r between the actual output of the sensor and the expected output i , through the data residual r i Calculate the information entropy H (X) of the vibration acceleration sensor in the normal state and fault state of the bearing i ) and H f1 (X i ), the calculation formula is: Then introduce H(X) into formula (1) i )-H f1 (X i ),get Where: is the probability density function of the bearing failure, and formula (2) conforms to the expression of KL divergence, that is, Then, the Mahalanobis transformation is introduced into the calculation formula (1) to obtain the improved formula: Where: is the covariance matrix of the multivariate normal distribution, n is the data dimension, and then formula (4) is introduced into formula (3) to obtain: Where: is the covariance matrix of the bearing data when no fault occurs, Bearing failure The covariance matrix of (3) Compare the information entropy of the bearing in normal state and fault state, and determine the fault type through information entropy comparison.
