Early fault real-time detection method for rolling bearing

Abstract

The invention belongs to the field of fault detection and diagnosis of rotating machinery, and discloses a real-time detection method for an early fault of a rolling bearing. The method comprises the following steps: determining priori knowledge; acquiring N groups of periodic synchronous single-point sampling data; calculating to obtain fault features; accumulating weak fault information contained in the fault features to obtain a bearing fault index; judging whether an early failure occurs or not; and realizing adaptive data acquisition. According to the method, weak fault information contained in fault features is accumulated by using accumulation and CUSUM technologies, the reliability and sensitivity of early fault detection are improved, the data acquisition amount is effectively reduced under the condition that the fault detection reliability is ensured, and then the data transmission, processing and storage costs are reduced.

Description

Technical Field

[0001] The invention belongs to the field of rotating machinery fault detection and diagnosis, and in particular relates to a real-time detection method for early-stage rolling bearing faults. Background Art

[0002] Rotating machinery and equipment have been widely used in industry and life. Common rotating machinery and equipment include engines, motors, and gearboxes. Rotating machinery and equipment will inevitably degrade and fail during long-term operation, among which the most common failure of rotating machinery and equipment is caused by bearing failure. In order to timely understand the health status of bearings and avoid the impact of bearing failures on the working performance of the entire rotating machinery and equipment, it is necessary to detect early bearing failures in real time.

[0003] At present, existing real-time detection technologies for early bearing faults usually use a fixed sampling frequency to continuously collect bearing health status data. Continuous data collection with a fixed sampling frequency assumes that the bearing degradation rate is constant during its service. In fact, the degradation rate of the bearing changes during its service. The probability of a bearing failure during its service also changes. Therefore, the early bearing fault detection method based on a fixed sampling frequency will collect a large amount of data, thereby increasing the cost of data transmission, processing, storage, etc. Although the above problems can be solved by reducing the sampling frequency, a lower sampling frequency may reduce the reliability and sensitivity of early fault detection.

[0004] In addition, rotating machinery may work at different speeds, and changes in the operating speed will cause changes in the fault frequency. It may not be possible to efficiently extract fault features from data collected at a single sampling frequency.

[0005] In order to overcome the shortcomings of existing fault detection methods, the present invention discloses a real-time detection method for early faults of rolling bearings. The method disclosed in the present invention takes into account the degradation law of the bearing and adaptively determines the number of synchronously sampled data points per rotation period of the shaft, that is, when the bearing is slightly degraded, a larger sampling interval is used, thereby obtaining fewer sampled data points per rotation period of the shaft; and when the bearing is severely degraded, a smaller sampling interval is used, thereby obtaining more sampled data points per rotation period of the shaft. Taking into account the degradation law of the bearing, the present invention patent designs a fault detection method for adaptively determining the number of sampled data points per rotation period of the shaft. In addition, in order to reliably and sensitively identify early faults of motor bearings, the present invention patent introduces accumulation and CUSUM technology, and identifies early faults by accumulating and adaptively obtaining weak fault information in the data. The present invention patent reduces the amount of data collection while ensuring the reliability of early fault detection of bearings, effectively reduces the cost of data transmission, storage and processing, and improves the sensitivity and reliability of early fault detection. Summary of the invention

[0006] Purpose of the invention: In response to the problems pointed out in the background technology, the present invention discloses a real-time detection method for early-stage rolling bearing faults, which adaptively determines the number of sampling data points in each rotation period of the shaft, effectively reducing the cost of data collection, transmission and processing; in addition, the CUSUM technology is used to accumulate weak fault information, which can improve the sensitivity and reliability of early bearing fault detection.

[0007] Technical solution: The present invention discloses a real-time detection method for early failure of rolling bearings, which specifically includes the following steps: Step 1: Determine prior knowledge; Step 2: Obtain N groups of periodic synchronous single-point sampling data; Step 3: Calculate the fault characteristics; Step 4: Accumulate the weak fault information contained in the fault characteristics to obtain the bearing fault index; Step 5: Determine whether an early fault occurs and implement adaptive data acquisition.

[0008] Furthermore, in step 1, the prior knowledge includes: the installation positions of the vibration sensor and the speed sensor, the initial number N0 of synchronously sampled data points per rotation period of the shaft, the minimum number N min , the maximum number N max , shaft speed L, fault feature mean M, threshold T.

[0009] Furthermore, in step 1, a vibration sensor is installed at the bearing to be tested to collect bearing vibration data; a speed sensor is installed near the shaft to measure the shaft speed data.

[0010] Furthermore, according to the method for real-time detection of early failures of rolling bearings according to claim 1, it is characterized in that in step 1, the initial number N0 of synchronously sampled data points per rotation period of the shaft is determined according to the health status of the detected bearing; when the detected bearing is a brand new bearing, the initial number N0 of synchronously sampled data points per rotation period of the shaft is set to a minimum number N min ; When the health status of the detected bearing is unknown, the initial number of synchronous sampling data points N0 per rotation period of the shaft is set to Among them, round() is a rounding function, and the result is an integer; the shaft speed L is the shaft speed within a detection time window, which can also be regarded as the length of the time window; the fault feature mean M is the average value of the fault feature when the bearing is fault-free; the threshold T is the preset critical value of the bearing fault indicator.

[0011] Furthermore, in step 2, N groups of periodic synchronous single-point sampling data X=[X1, X2, X3, ..., Xi , …, X N ]; Since the number of rotations of the shaft for data acquisition is L, the length of each group of periodic synchronous single-point sampling data in data X is L; for the i-th group of periodic synchronous single-point sampling data X i , which can be expressed as X i =[x(i),x(i+N),x(i+2N),…,x(i+(L-1)N)].

[0012] Furthermore, in step 2, the number N of groups of periodic synchronous single-point sampling data X is an initial number N0 in the first detection time window, and is adaptively determined according to the bearing fault index in subsequent detection time windows.

[0013] Furthermore, in step 3, the standard deviation Z of N groups of periodic synchronous single-point sampling data is calculated respectively, and regarded as a fault feature. Z=[z(1),z(2),z(3),…,z(i),…,z(N)] T , in,

[0014] Furthermore, in step 4, the cumulative sum CUSUM technique is used to accumulate the weak fault information contained in the fault feature Z to obtain the cumulative sum C, and the final value c(N) of the cumulative sum C is recorded as the bearing fault index W; in the calculation process of the CUSUM method, the fault feature mean M is set as the expected value, and for each data in the data Z, the difference between it and the mean M is calculated to obtain the data Q, Q=[q(1),q(2),q(3),…,q(i),…q(N)] T , Where q(i) = z(i) - M; calculate the cumulative sum C of data Q, C=[c(1),c(2),c(3),…,c(i),…,c(N)],

[0015] Among them, c(i)=max(c(i-1)+q(i),0), c(0)=0, and the final value c(N) of the cumulative sum C is recorded as the bearing fault index W.

[0016] Furthermore, in step 5, the bearing fault index W is compared with the threshold value T. If the bearing fault index W is less than the threshold value T, it is determined that there is no fault, and the number of synchronous sampling data points N per rotation period of the shaft is adaptively adjusted to And return to step 2 to continue early fault detection; if the bearing fault index W is greater than or equal to the threshold T, it is considered that an early fault has occurred.

[0017] Beneficial effects:

[0018] 1. The method of the present invention discloses a real-time detection method for early rolling bearing faults, which overcomes the shortcomings of existing real-time detection methods for bearing faults and can adaptively determine the number of synchronously sampled data points per rotation period of the shaft. While ensuring the reliability of fault detection, the amount of data collection is effectively reduced, thereby reducing the cost of data transmission, processing and storage.

[0019] 2. The method of the present invention utilizes CUSUM technology to accumulate weak fault information in adaptively acquired data, which can improve the reliability and sensitivity of early fault detection.

[0020] 3. The computational complexity of the method of the present invention is low, and a control unit with a lower processing speed can be used, thereby reducing the cost of the control unit.

[0021] 4. The method of the present invention is not affected by the rotation speed of the rotating machinery. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a flow chart of the method of the present invention;

[0023] Figure 2 This is the test result when the bearing is fault-free;

[0024] Figure 3 It is the number of synchronously sampled data points in each rotation period of the shaft when the bearing is fault-free;

[0025] Figure 4 This is the detection result of the early bearing failure;

[0026] Figure 5 It is the number of synchronously sampled data points in each rotation period of the shaft when the bearing fails in the early stage. DETAILED DESCRIPTION

[0027] The implementation process of the method of the present invention is described below with reference to the accompanying drawings.

[0028] The present invention discloses a real-time detection method for early failure of rolling bearings, the process of which is as follows: Figure 1 As shown, the following steps are included:

[0029] Step 1: Determine the installation positions of the vibration sensor and the speed sensor, and determine the initial number of synchronous sampling data points N0 = 25 and the minimum number N per rotation period of the shaft. min =20, maximum number N max =30, shaft speed L=20, fault feature mean M=0.332, threshold T=1.

[0030] The vibration sensor needs to be installed at the pre-test bearing to collect the bearing vibration signal. The speed sensor is installed near the shaft to measure the shaft speed. Since the health status of the measured bearing is unknown, the initial number of synchronous sampling data points N0 per shaft rotation period is set to Among them, round() is a rounding function, and the result is an integer.

[0031] Step 2: Obtain N groups of periodically synchronized single-point sampling data X.

[0032] In the initial stage of early fault detection, N=N0=25. According to the number of synchronous sampling data points N in each rotation period of the shaft, N=25 groups of periodic synchronous single-point sampling data X are obtained.

[0033]

[0034] It should be noted that in some literature, the periodic synchronous single-point sampling data is also regarded as once-per-revolution sampling data, and the periodic synchronous multi-point sampling data is also regarded as multiple sampling per revolution data. N groups of periodic synchronous single-point sampling data X can be regarded as a group of periodic synchronous multi-point sampling data.

[0035] Step 3: Calculate the standard deviation Z of N groups of periodic synchronous single-point sampling data X respectively, and regard Z as the fault feature.

[0036] Z=[z(1)z(2)z(3)…z(i)…z(25)] T

[0037] in,

[0038] Step 4: Use the CUSUM technology to accumulate the weak fault information in the fault feature Z, obtain the cumulative sum data C, and obtain the fault index.

[0039] The fault feature mean M = 0.332 is set as the expected value in the CUSUM method. For each data point in the data Z, the difference between it and the mean M is calculated to obtain the data Q.

[0040] Q=[q(1),q(2),q(3),…,q(i),…q(25)] T

[0041] Where q(i) = z(i) - 0.332. Calculate the cumulative sum of data Q to get data C,

[0042] C=[c(1),c(2),c(3),…,c(i),…,c(25)],

[0043] Where c(i)=max(c(i-1)+q(i),0), c(0)=0; the final value c(N) of the cumulative sum C is recorded as the fault index W.

[0044] Step 5: Determine whether an early fault occurs and implement adaptive data acquisition.

[0045] The fault indicator W is compared with the threshold value T = 1. If the fault indicator W is less than the threshold value T, it is determined that there is no fault, and the number of synchronous sampling data points N in each rotation period of the shaft is adaptively adjusted to Among them, round() is a rounding function, the result is an integer, and the process returns to step 2 to continue early fault detection; if the fault indicator W is greater than or equal to the threshold T, it is considered that the bearing has an early fault.

[0046] Figure 2 is the detection result when the bearing is fault-free. As can be seen from the figure, the fault index W is less than the threshold T, which means that the bearing is in a fault-free state. The number of synchronous sampling data points N per rotation period of the shaft is adaptively adjusted to And return to step 2 to continue troubleshooting.

[0047] Figure 3 The number of synchronously sampled data points N per rotation period of the shaft when the bearing is fault-free is shown. In each fault detection, the number of synchronously sampled data points per rotation period of the shaft changes with the fault index. Since the bearing is fault-free, the number of synchronously sampled data points N per rotation period of the shaft is less than or equal to the maximum value of 30 during each detection process, which reduces the amount of data collection and thus reduces the cost of data transmission, processing and storage.

[0048] Figure 4 It is the detection result when the bearing fails. As can be seen from the figure, when the fault index W is greater than the threshold T in the 8th detection, it is considered that the bearing has an early fault.

[0049] Figure 5 is the number of synchronously sampled data points N in each rotation period of the shaft when the bearing has an early fault. In the eighth fault detection, the bearing has been detected in its early fault state, and the early fault detection work is terminated, and the number of synchronously sampled data points N in each rotation period of the shaft is no longer adaptively calculated for the next detection.

[0050] The method of the present invention accumulates weak fault information in the data through CUSUM technology, and adaptively adjusts the number of synchronous sampling data points N in each rotation period of the shaft according to the bearing degradation condition, increases or decreases the data collection amount in real time, optimizes the sampling data amount, effectively reduces the data transmission, storage and processing costs, and improves the sensitivity and reliability of early bearing fault detection. In addition, compared with traditional early fault detection methods, the method of the present invention has a lower computational complexity.

[0051] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A real-time detection method for early failure of rolling bearings, characterized in that: The following steps are involved: Step 1: Determine prior knowledge; Step 2: Obtain N groups of periodic synchronous single-point sampling data; Step 3: Calculate the fault characteristics; Step 4: Accumulate the weak fault information contained in the fault characteristics to obtain the bearing fault index; Step 5: Determine whether an early fault occurs and implement adaptive data acquisition.

2. A real-time detection method for early failure of rolling bearing according to claim 1, characterized in that: In step 1, the prior knowledge includes: the installation position of the vibration sensor and the speed sensor, the initial number N0 of synchronous sampling data points per rotation period of the shaft, the minimum number N min , the maximum number N max , shaft speed L, fault feature mean M, threshold T.

3. A real-time detection method for early failure of rolling bearing according to claim 1, characterized in that: In step 1, a vibration sensor is installed at a bearing to be tested to collect bearing vibration data; a speed sensor is installed near a rotating shaft to measure the rotating shaft speed data.

4. A real-time detection method for early failure of rolling bearing according to claim 1, characterized in that: In step 1, the initial number N0 of synchronously sampled data points per rotation period of the shaft is determined according to the health status of the detected bearing; when the detected bearing is a brand new bearing, the initial number N0 of synchronously sampled data points per rotation period of the shaft is set to the minimum number N min ; When the health status of the detected bearing is unknown, the initial number N0 of synchronous sampling data points per rotation period of the shaft is set to Among them, round() is a rounding function, and the result is an integer; the shaft rotation number L is the shaft rotation number within a detection time window, which can also be regarded as the time window length; The fault feature mean M is the average value of the fault feature when the bearing is fault-free; the threshold T is the preset critical value of the bearing fault indicator.

5. A real-time detection method for early failure of rolling bearing according to claim 4, characterized in that: In step 2, N groups of periodic synchronous single-point sampling data X=[X1, X2, X3, ..., X i , …, X N ]; Since the number of rotations of the shaft for data acquisition is L, the length of each group of periodic synchronous single-point sampling data in data X is L; for the i-th group of periodic synchronous single-point sampling data X i , which can be expressed as X i =[x(i),x(i+N),x(i+2N),…,x(i+(L-1)N)].

6. A real-time detection method for early failure of rolling bearing according to claim 5, characterized in that: In step 2, the number N of groups of periodic synchronous single-point sampling data X is an initial number N0 in the first detection time window, and is adaptively determined according to the bearing fault index in subsequent detection time windows.

7. A real-time detection method for early failure of rolling bearing according to claim 6, characterized in that: In step 3, the standard deviation Z of N groups of periodic synchronous single-point sampling data is calculated respectively, and regarded as a fault feature. Z=[z(1),z(2),z(3),…,z(i),…,z(N)] T , in, 8. A real-time detection method for early failure of rolling bearing according to claim 7, characterized in that: In step 4, the cumulative sum CUSUM technique is used to accumulate the weak fault information contained in the fault feature Z to obtain the cumulative sum C, and the final value c(N) of the cumulative sum C is recorded as the bearing fault index W; in the calculation process of the CUSUM method, the fault feature mean M is set as the expected value, and for each data in the data Z, the difference between it and the mean M is calculated to obtain the data Q, Q=[q(1),q(2),q(3),…,q(i),…q(N)] T , Where q(i) = z(i) - M; calculate the cumulative sum C of data Q, C=[c(1),c(2),c(3),…,c(i),…,c(N)], Among them, c(i)=max(c(i-1)+q(i),0), c(0)=0, and the final value c(N) of the cumulative sum C is recorded as the bearing fault index W.

9. A real-time detection method for early failure of rolling bearing according to claim 8, characterized in that: In step 5, the bearing fault index W is compared with the threshold value T. If the bearing fault index W is less than the threshold value T, it is determined that there is no fault, and the number of synchronous sampling data points N in each rotation period of the shaft is adaptively adjusted to And return to step 2 to continue early fault detection; if the bearing fault index W is greater than or equal to the threshold T, it is considered that an early fault has occurred.

Images