CN-121981287-A - Bearing real-time abnormality detection method and related equipment based on self-adaptive network fuzzy inference system

CN121981287ACN 121981287 ACN121981287 ACN 121981287ACN-121981287-A

Abstract

The application discloses a real-time abnormality detection method and related equipment for a bearing based on a self-adaptive network fuzzy inference system, wherein the method comprises the steps of determining a first characteristic index group and a second characteristic index group according to a first sensor signal and an abnormality detection result in a bearing data set, designing a fuzzy inference rule and calculating an abnormality score of the first characteristic index group, taking the second characteristic index group as an input of the self-adaptive network fuzzy inference system, and training the system according to the abnormality score; and extracting a third characteristic index group of the bearing to be detected, inputting the third characteristic index group into a trained self-adaptive network fuzzy inference system for reasoning, and completing abnormality detection according to a negative abnormal characteristic value prediction result obtained by reasoning and a preset decision rule. The embodiment of the application can realize bearing abnormality detection by combining a fuzzy reasoning system and a self-adaptive network fuzzy reasoning system, and has higher prediction accuracy and operation efficiency. The method can be widely applied to the technical field of bearing abnormality detection.

Inventors

ZHANG QINGHUA
HUANG JUNHANG
CAI YEBIN
FAN YILE
LU BINGYU
ZHU GUANHUA
XU GUANGMING
LIU MEI
LU YUSHEN
ZHUO XIAOKUI

Assignees

广东石油化工学院

Dates

Publication Date: 20260505
Application Date: 20251210

Claims (10)

1. The real-time bearing abnormality detection method based on the self-adaptive network fuzzy inference system is characterized by comprising the following steps of: acquiring a first sensor signal and an abnormality detection result in a bearing data set; Determining a first characteristic index group and a second characteristic index group according to the first sensor signal and the abnormality detection result; Analyzing the bearing data set, designing a fuzzy inference rule according to an analysis result, and calculating an abnormal score of the first characteristic index set through the fuzzy inference rule; Constructing a self-adaptive network fuzzy inference system, taking the second characteristic index set as the input of the self-adaptive network fuzzy inference system, and training the self-adaptive network fuzzy inference system according to the output of the self-adaptive network fuzzy inference system and the anomaly score; acquiring a second sensor signal of a bearing to be detected, and determining a third characteristic index group according to the second sensor signal; Processing the third characteristic index group through the trained self-adaptive network fuzzy inference system to obtain a negative abnormal characteristic value prediction result; And carrying out abnormality detection on the bearing to be detected according to a preset decision rule, the second sensor signal and the negative abnormality characteristic value prediction result.
2. The method for real-time anomaly detection of a bearing based on an adaptive network fuzzy inference system of claim 1, wherein the first sensor signal comprises a first horizontal pressure sensor signal and a first vertical pressure sensor signal, the first set of characteristic indicators comprises a first high-low frequency root mean square ratio change rate and a sudden deviation mark accumulation value, and the determining the first set of characteristic indicators based on the first sensor signal and the anomaly detection result comprises: filtering the first sensor signal to obtain a high-frequency component and a low-frequency component of the first sensor signal; converting the high frequency component and the low frequency component into root mean square voltage signals; Determining the first high-low frequency root mean square ratio change rate according to the root mean square voltage signal; Modeling the distribution of the first high-low frequency root mean square ratio change rate through a histogram, and determining a standard deviation constant according to the histogram; And determining the burst deviation mark accumulated value according to the standard deviation constant and the first high-low frequency root mean square ratio change rate.
3. The method for detecting real-time anomalies of bearings based on the adaptive network fuzzy inference system according to claim 2, wherein the expression of the first high-low frequency root mean square ratio change rate is: Wherein, the For the first high and low frequency root mean square ratio change rate, For a first high-low frequency root mean square ratio of the first sensor signal in a kth time interval, For the root mean square voltage of the high frequency component in the kth time interval, For the root mean square voltage of the low frequency component in the kth time interval.
4. The method for detecting real-time anomalies of a bearing based on an adaptive network fuzzy inference system according to claim 2, wherein said determining the burst bias flag accumulated value from the standard deviation constant and the first high-low frequency root mean square ratio change rate comprises: judging whether a condition of a first mathematical formula is met or not according to the first high-low frequency root mean square ratio change rate and the standard deviation constant; when the condition of the first mathematical formula is met, adding one to the value of the burst deviation mark accumulated value; Wherein the first mathematical formula is: for the first high and low frequency root mean square ratio change rate, Is the standard deviation constant.
5. The method for detecting real-time anomalies of bearings based on the adaptive network fuzzy inference system according to claim 2, wherein the second characteristic index group includes the burst deviation mark accumulated value, a second high-low frequency root mean square ratio change rate, a current time interval sequence number, and a high-low frequency root mean square ratio deviation value, and the determining the second characteristic index group according to the first sensor signal and the anomalies detection result includes: Determining the second high-low frequency root mean square ratio change rate and the high-low frequency root mean square ratio deviation value according to the root mean square voltage signal; And determining the sequence number of the current time interval according to the first sensor signal.
6. The method for detecting real-time anomalies of bearings based on adaptive network fuzzy inference system according to claim 1, wherein the training of the adaptive network fuzzy inference system according to the output of the adaptive network fuzzy inference system and the anomaly score comprises: determining a loss value according to the output of the adaptive network fuzzy inference system and the anomaly score; And optimizing parameters and rules of the self-adaptive network fuzzy inference system through a least square method and a gradient descent algorithm according to the loss value.
7. The method for detecting the bearing real-time abnormality based on the adaptive network fuzzy inference system according to claim 1, wherein the second sensor signal includes a second horizontal pressure sensor signal and a second vertical pressure sensor signal, the abnormality detection of the bearing to be detected according to a preset decision rule, the second sensor signal and the negative abnormality characteristic value prediction result includes: determining a second high-low frequency root mean square ratio from the second horizontal pressure sensor signal; determining a third high-low frequency root mean square ratio from the second vertical pressure sensor signal; judging whether the third high-low frequency root mean square ratio is smaller than the second high-low frequency root mean square ratio; judging whether a negative abnormal characteristic value occurs or not and whether the occurrence times of the negative abnormal characteristic value in a preset time interval are larger than a preset first threshold value or not according to the negative abnormal characteristic value prediction result; activating an abnormality flag when the third high-low frequency root mean square ratio is smaller than the second high-low frequency root mean square ratio and the negative abnormality characteristic value does not appear; when the third high-low frequency root mean square ratio is not smaller than the second high-low frequency root mean square ratio, the negative abnormal characteristic value appears, the occurrence frequency of the negative abnormal characteristic value in a preset time interval is not larger than a preset first threshold value, and an abnormal sign is activated; Activating an anomaly warning when the third high-low frequency root mean square ratio is less than the second high-low frequency root mean square ratio and the negative anomaly characteristic value occurs; And when the third high-low frequency root mean square ratio is not smaller than the second high-low frequency root mean square ratio, and the occurrence frequency of the negative abnormal characteristic value in a preset time interval is larger than a preset first threshold value, activating an abnormal warning.
8. The utility model provides a real-time abnormal detection device of bearing based on self-adaptation network fuzzy inference system which characterized in that, the device includes: the training data acquisition module is used for acquiring a first sensor signal and an abnormality detection result in the bearing data set; The training data characteristic index construction module is used for determining a first characteristic index group and a second characteristic index group according to the first sensor signal and the abnormality detection result; the fuzzy reasoning module is used for analyzing the bearing data set, designing a fuzzy reasoning rule according to an analysis result, and calculating the abnormal score of the first characteristic index set through the fuzzy reasoning rule; The self-adaptive network fuzzy inference system training module is used for constructing a self-adaptive network fuzzy inference system, taking the second characteristic index set as the input of the self-adaptive network fuzzy inference system, and training the self-adaptive network fuzzy inference system according to the output of the self-adaptive network fuzzy inference system and the abnormal score; The to-be-processed data characteristic index construction module is used for acquiring a second sensor signal of the bearing to be detected and determining a third characteristic index group according to the second sensor signal; the negative abnormal characteristic value prediction module is used for processing the third characteristic index group through the trained self-adaptive network fuzzy inference system to obtain a negative abnormal characteristic value prediction result; and the abnormality detection module is used for carrying out abnormality detection on the bearing to be detected according to a preset decision rule, the second sensor signal and the negative abnormality characteristic value prediction result.
9. An electronic device comprising a memory storing a computer program and a processor implementing the method of any of claims 1 to 7 when the computer program is executed by the processor.
10. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the method of any one of claims 1 to 7.

Description

Bearing real-time abnormality detection method and related equipment based on self-adaptive network fuzzy inference system Technical Field The application relates to the technical field of bearing abnormality detection, in particular to a real-time bearing abnormality detection method and related equipment based on a self-adaptive network fuzzy inference system. Background Rolling machines play a vital role in industrial applications such as aircraft engines, gas turbines, wind turbines and various machine tools. For the rolling mechanical bearing, the efficient early abnormality detection can help enterprises to formulate more effective maintenance strategies, so that premature or delayed intervention is avoided, maintenance cost is finally reduced, and equipment availability and production efficiency are improved. The existing bearing fault prediction method mainly comprises frequency domain analysis, time-frequency analysis, traditional machine learning methods and deep learning methods which are emerging in recent years based on vibration signals. The method can predict the running state of the bearing to a certain extent, but has the defects that on one hand, the prediction accuracy is limited by the feature extraction quality depending on manual feature design, and on the other hand, part of the methods have high calculation cost and poor instantaneity, and are difficult to meet the requirement of industrial application on the running efficiency. In summary, the technical problems in the related art are to be improved. Disclosure of Invention The embodiment of the application mainly aims to provide a bearing real-time abnormality detection method and related equipment based on a self-adaptive network fuzzy inference system, which can be used for carrying out real-time prediction on the abnormal state of a bearing by combining the fuzzy inference system and the self-adaptive network fuzzy inference system, and has higher prediction accuracy and operation efficiency. In order to achieve the above object, an aspect of an embodiment of the present application provides a method for detecting real-time anomalies of a bearing based on an adaptive network fuzzy inference system, the method comprising: acquiring a first sensor signal and an abnormality detection result in a bearing data set; Determining a first characteristic index group and a second characteristic index group according to the first sensor signal and the abnormality detection result; Analyzing the bearing data set, designing a fuzzy inference rule according to an analysis result, and calculating an abnormal score of the first characteristic index set through the fuzzy inference rule; Constructing a self-adaptive network fuzzy inference system, taking the second characteristic index set as the input of the self-adaptive network fuzzy inference system, and training the self-adaptive network fuzzy inference system according to the output of the self-adaptive network fuzzy inference system and the anomaly score; acquiring a second sensor signal of a bearing to be detected, and determining a third characteristic index group according to the second sensor signal; Processing the third characteristic index group through the trained self-adaptive network fuzzy inference system to obtain a negative abnormal characteristic value prediction result; And carrying out abnormality detection on the bearing to be detected according to a preset decision rule, the second sensor signal and the negative abnormality characteristic value prediction result. In some embodiments, the first sensor signal includes a first horizontal pressure sensor signal and a first vertical pressure sensor signal, the first set of characteristic indicators includes a first high-low frequency root mean square ratio rate of change and a burst bias flag accumulation value, the determining the first set of characteristic indicators from the first sensor signal and the anomaly detection result includes: filtering the first sensor signal to obtain a high-frequency component and a low-frequency component of the first sensor signal; converting the high frequency component and the low frequency component into root mean square voltage signals; Determining the first high-low frequency root mean square ratio change rate according to the root mean square voltage signal; Modeling the distribution of the first high-low frequency root mean square ratio change rate through a histogram, and determining a standard deviation constant according to the histogram; And determining the burst deviation mark accumulated value according to the standard deviation constant and the first high-low frequency root mean square ratio change rate. In some embodiments, the first high-low frequency root mean square ratio rate of change is expressed as: Wherein, the For the first high and low frequency root mean square ratio change rate,For a first high-low frequency root mean square ratio of the first sensor signal in a kth