Vibration sensor fault detection method

Abstract

The invention relates to a vibration sensor fault detection method, which comprises the following steps of: acquiring vibration data of vibration sensors arranged at different measuring points, and extracting the amplitude of the vibration data at a characteristic frequency point as a target characteristic amplitude; for any characteristic frequency point of any measuring point, calculating the amplitude ratio of the target characteristic amplitude between the current measuring point and the other measuring points to obtain a target amplitude ratio; and comparing the target amplitude ratio with a reference ratio parameter, and if the ratio of the target amplitude ratio and the reference ratio parameter is imbalance, determining that a sensor fault occurs. The fault of the vibration sensor can be effectively detected.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sensor detection, in particular to a vibration sensor fault detection method. BACKGROUND

[0002] Due to the fact that a certain type of transmission vibration sensor has more arranged measuring points, the vibration measuring points arranged by the vibration sensor include power input shaft, oil pump, power output shaft and other parts. When the channel data of a certain measuring point is misreported due to sensor failure, sensor loosening and poor channel data contact, the system is shut down, causing huge economic losses. Therefore, it is necessary to verify the effectiveness of the data of the sensor and the related channel. The current vibration sensor fault tolerance technology mainly aims at the scheme of multiple sensors for the same measuring point, and there is no better method for the case where the measuring points are arranged far apart. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a vibration sensor fault detection method which can effectively detect vibration sensor faults.

[0004] The technical solution adopted by the present application to solve the technical problem is: a vibration sensor fault detection method is provided, comprising the following steps:

[0005] Obtain the vibration data of the vibration sensors placed at different measuring points, and extract the amplitude of the vibration data at the characteristic frequency point as the target characteristic amplitude;

[0006] For any measuring point, calculate the amplitude ratio of the target characteristic amplitude between the current measuring point and the remaining measuring points to obtain the target amplitude ratio;

[0007] Compare the target amplitude ratio with the reference ratio parameter, and if the proportion is out of adjustment, it is determined that a sensor fault has occurred.

[0008] Further, the reference ratio parameter includes a reference amplitude ratio mode and a reference amplitude ratio mode mean square error.

[0009] Further, if the distance between the target amplitude ratio of the current measuring point within a set sampling period and the reference amplitude ratio mode is greater than the set multiple of the reference amplitude ratio mode mean square error, it is determined that the proportion between the target amplitude ratio and the corresponding reference ratio parameter is out of adjustment.

[0010] Further, if the proportion between all the target amplitude ratios of the current measuring point at all the characteristic frequency points and the corresponding reference ratio parameters is out of adjustment, it is determined that the current measuring point has a sensor fault.

[0011] Further, it further comprises the step of estimating the related data of the fault measuring point by using the target characteristic amplitude of the non-fault measuring point.

[0012] Further, the reference ratio parameter further comprises a magnitude ratio concentration.

[0013] Further, the magnitude of the fault measuring point at the target characteristic frequency point is a weighted sum of the target characteristic magnitude of the non-fault measuring point and the product of the mode of the magnitude ratio and the target characteristic magnitude of the fault measuring point, with the magnitude ratio concentration as the weight.

[0014] Further, the reference ratio parameter is obtained by the following method:

[0015] Reference vibration data of the vibration sensor placed at different measuring points when working normally is obtained, and the magnitude of the reference vibration data at the characteristic frequency point is extracted as a reference characteristic magnitude;

[0016] The reference ratio parameter of all the characteristic frequency points of all the measuring points is obtained according to the magnitude ratio of the reference characteristic magnitudes of the same characteristic frequency point between any two measuring points.

[0017] Further, before the step of obtaining the reference ratio parameter of all the characteristic frequency points of all the measuring points, the method further comprises:

[0018] The magnitude ratio of the same characteristic frequency point between any two measuring points of different sampling periods is obtained as magnitude ratio data;

[0019] The mode and mode mean square deviation of the magnitude ratio data are calculated;

[0020] It is judged whether the difference between any of the magnitude ratio data and the mode of the magnitude ratio data is greater than the set multiple of the mode mean square deviation of the magnitude ratio data, and if so, the current magnitude ratio data is deleted as an abnormal value and the previous step is returned.

[0021] Advantages

[0022] Compared with the prior art, the present application has the following advantages and positive effects: the present application obtains the magnitude ratio of the vibration sensor at the characteristic frequency point when working normally, calculates the reference ratio parameter according to the data, compares the target magnitude ratio calculated according to the target vibration data with the reference ratio parameter, and identifies the fault and corrects the data based on the result, which can effectively detect the sensor fault of the multi-measuring-point multi-vibration-sensor system. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 is a flowchart of the embodiment of the present application;

[0024] Figure 2 is a flowchart of the establishment of the magnitude ratio relationship network of the embodiment of the present application;

[0025] Figure 3is a flow chart of determining sensor fault of the embodiment of the present application;

[0026] Figure 4 is a flow chart of determining fault measurement point of the embodiment of the present application. DETAILED DESCRIPTION

[0027] The present application will be further described with reference to the following examples. It should be understood that these examples are intended to illustrate the present application and are not intended to limit the scope of the present application. Furthermore, it should be understood that various modifications and changes can be made to the present application by those skilled in the art upon reading the contents of the present application, and such equivalent forms are also within the scope of the appended claims.

[0028] The embodiment of the present application relates to a vibration sensor fault detection method, aiming at a plurality of vibration sensors arranged at different measurement points of a transmission, when a certain vibration sensor fails, using data of other vibration sensors to identify the fault and reconstruct data of the fault sensor.

[0029] It is found through a large amount of data that a vibration frequency point can be observed on data of multiple measurement points, such as input shaft gear meshing frequency and output shaft gear meshing frequency, which can be observed on vibration signals of the above-mentioned sensors. When one of the sensors fails, the characteristic proportional relationship established by the remaining sensors and the failed sensor can be compared; in this way, fault identification and data correction can be performed through multiple vibration sensors, so as to determine sensor failure and prevent false alarms from occurring, and then a plurality of one voting mode is used to determine sensor failure and normal sensor data is used to reproduce characteristic data of the fault sensor.

[0030] As shown in Figure 1 For the transmission of each device, collect normal data of all vibration sensors, calculate characteristic frequency points of all single vibration sources of each data packet and measurement points with larger vibration amplitudes. Calculate amplitude ratios between each pair of measurement points of the above-mentioned frequency points at the same time point. Statistically analyze all amplitude ratios, give mode and mode mean square deviation of each pair of amplitude ratios, set a distance between the amplitude ratio and the mode of the amplitude ratio exceeding 5σ (σ is the mode mean square deviation) as an abnormal value, calculate the mode and the mode mean square deviation after removing the abnormal value. Calculate the data concentration degree, and select the measurement point characteristic ratio with higher concentration degree for subsequent fault identification and data reconstruction.

[0031] For the subsequent collected vibration data, all frequency point data of all measuring points on the transmission are calculated, and the amplitude ratio between each other is calculated. The amplitude ratio between each pair of measuring points is checked. If the distance between the calculated amplitude ratio and the previous amplitude ratio mode (amplitude ratio mode in Tables 1 and 2) in the continuous 5 groups of data is greater than 5σ, it is determined that one of the two measuring points is a sensor failure. The failure measuring point is determined through the ratio of multiple groups of data, and then the failure measuring point is removed. The amplitude of the remaining measuring points is multiplied by the amplitude ratio to reconstruct the data of the failure measuring point.

[0032] The specific steps of the present embodiment will be further described below. The present embodiment can be divided into two parts: constructing an amplitude ratio relationship network and determining whether the amplitude ratio of the measuring point is out of adjustment using the amplitude ratio relationship network.

[0033] As shown in Figure 1 , the part of constructing the amplitude ratio relationship network mainly includes the following steps.

[0034] Step 1: Collecting and processing characteristic data

[0035] Collect time domain data of the transmission vibration sensor under the condition of no failure for a period of time (usually more than 10 hours of normal time). Perform short-time FFT calculation on all channel data (the sampling period of FFT is fixed, and a sampling period of 2 seconds can usually be used). Save the amplitude of each channel characteristic frequency point of the mechanism under the condition of rated speed in tabular form, and save the corresponding time information, that is, each row of data records the amplitude of each characteristic frequency point in the same sampling period. In the present embodiment, the characteristic frequency points include the input shaft gear meshing frequency and the output shaft gear meshing frequency, but are not limited thereto. If other vibrations are more obvious and the characteristics of the vibrations can be detected on all sensor measuring points, they can be extracted to increase reliability.

[0036] Step 2: Calculate the amplitude ratio between all channels of the main reducer at the characteristic frequency points, and determine the proportional relationship and data proportion between the channels.

[0037] Set the vibration channels to include 5 channel measuring points of the input measuring point, the oil pump measuring point, the left shell measuring point, the right shell measuring point, and the output measuring point. Calculate the amplitude ratio between the above-mentioned measuring points (the ratio of the amplitude at the input shaft gear meshing frequency and the output shaft gear meshing frequency after FFT calculation) two by two, that is, the amplitude ratio in each sampling period includes 2xC25=20 amplitude ratio data. Save all the above-mentioned amplitude ratio data.

[0038] Statistically analyze the mode, mode mean square deviation, and concentration degree of each group of amplitude ratio data as reference ratio parameters.

[0039] Set the amplitude ratio data of one group as u i , that is, the values are u i(i from 0 to N-1) data points, the mode, the mode of the mean square error formula is:

[0040] Mode: divide the interval, a group of u i Data is divided into M intervals according to its maximum and minimum value, M can be an integer not less than 50, to ensure that the length of each interval is less than 0.2; the number of data points falling in each interval is counted, and the midpoint value of the interval with the most data points is the mode of the amplitude ratio of the group, and the mode is u m .

[0041] Mode of mean square error:

[0042] Concentration calculation method: count the absolute value of the difference between each point u i and the mode u m Within 50% of the mode value u m , the proportion of the number of data that meets the above conditions to the total data is the concentration of the amplitude ratio of the group.

[0043] Abnormal value removal: check each point u i in a group of numbers and the mode u m Whether the difference is within 5 times the mode of the mean square error, if greater than, it is determined as an abnormal value, and the abnormal value is removed before the above steps; usually, the abnormal value of a normal group of amplitude ratios of the sensor is far less than 1%, if the abnormal value is large (more than 1%), the feature cannot be used as the amplitude ratio of the two measuring points for fault identification, and if the abnormal value of all features of the two measuring points is more than 1%, the two measuring points cannot be used for fault identification, and the mode is set to 0.

[0044] Step three: construct the measuring point relationship map

[0045] According to the input shaft gear meshing frequency and the output shaft gear meshing frequency, the relationship map between all measuring points is constructed, which contains all reference ratio parameters. The mode of the amplitude ratio is obtained by the above steps, the mode of the amplitude ratio is used as the reference amplitude ratio for measuring the ratio of the two measuring points under the feature condition (i.e. reference amplitude ratio), and the mode of the amplitude ratio is used as the variance of the amplitude ratio (i.e. reference amplitude ratio variance), to establish the relationship between each two sensors, and thus to establish the network between the measuring point sensors of the entire reduction and transmission mechanism. The amplitude ratio network relationship is shown in the following table:

[0046] Table 1 Amplitude ratio mode network (input shaft gear meshing frequency)

[0047] Amplitude ratio Input measurement point Lubricating oil pump measurement point Housing left measurement point Housing right measurement point Output measurement point Input measurement point 1 a1,2 a1,3 a1,4 a1,5 Lubricating oil pump measurement point a2,1 1 a2,3 a2,4 a2,5 Housing left measurement point a3,1 a3,2 1 a3,4 a3,5 Housing right measurement point a4,1 a4,2 a4,3 1 a4,5 Output measurement point a5,1 a5,2 a5,3 a5,4 1

[0048] Table 2 Amplitude ratio mode network (output shaft gear meshing frequency)

[0049] Amplitude ratio Input measurement point Lubricating oil pump measurement point Housing left measurement point Housing right measurement point Output measurement point Input measurement point 1 b1,2 b1,3 b1,4 b1,5 Lubricating oil pump measurement point b2,1 1 b2,3 b2,4 b2,5 Housing left measurement point b3,1 b3,2 1 b3,4 b3,5 Housing right measurement point b4,1 b4,2 b4,3 1 b4,5 Output measurement point b5,1 b5,2 b5,3 b5,4 1

[0050] Table 3 Amplitude ratio dispersion network (input / output shaft gear meshing frequency)

[0051] Amplitude ratio mean square error Input measurement point Lubricating oil pump measurement point Housing left measurement point Housing right measurement point Output measurement point Input measurement point σ1,1 σ1,2 σ1,3 σ1,4 σ1,5 Lubricating oil pump measurement point σ2,1 σ2,2 σ2,3 σ2,4 σ2,5 Housing left measurement point σ3,1 σ3,2 σ3,3 σ3,4 σ3,5 Housing right measurement point σ4,1 σ4,2 σ4,3 σ4,4 σ4,5 Output measurement point σ5,1 σ5,2 σ5,3 σ5,4 σ5,5

[0052] Table 4 Amplitude ratio concentration network (input / output shaft gear meshing frequency)

[0053]

[0054]

[0055] a in the above two tables i,j represents the mode of the amplitude ratio between the input shaft gear meshing frequency of the ith channel and the jth channel; b i,j represents the mode of the amplitude ratio between the output shaft gear meshing frequency of the ith channel and the jth channel; σ of Table 3 i,j represents the mode dispersion of the amplitude ratio between the input shaft gear meshing frequency or the output shaft gear meshing frequency of the ith channel and the jth channel; Ω of Table 4 i,j represents the concentration of the amplitude ratio between the input shaft gear meshing frequency or the output shaft gear meshing frequency of the ith channel and the jth channel. In this way, the network relationship of the amplitude ratio between each measuring point is established.

[0056] As Figure 2 shown, the part of determining whether the amplitude ratio of the measuring points is out of adjustment using the amplitude ratio relationship network mainly includes the following steps.

[0057] Step four: fault identification on subsequent data.

[0058] a) Form the above-mentioned amplitude ratio network through the processing of the early normal data. The method of fault identification on the later collected signals (online or offline) is as follows:

[0059] b) Perform FFT calculation on the data of one sampling period of all measuring points, and calculate the two characteristic frequency points of each measuring point: input shaft gear meshing frequency and output shaft gear meshing frequency;

[0060] c) Calculate the amplitude ratio of the characteristic frequency point of each channel and the characteristic frequency points of the remaining channels, and compare the amplitude ratio with the corresponding amplitude ratio mode in the above table. When it is found that the distance between the amplitude ratio calculated from the data of five consecutive sampling periods and the corresponding amplitude ratio in the table exceeds 5 times the mode dispersion, it is determined that the proportion of the two measuring points is out of adjustment;

[0061] d) When the proportion of two measuring points is found to be out of adjustment, check the proportion relationship of each measuring point and the remaining measuring points by repeating the step b); as Figure 3The sensor fault is determined by the voting method of minority obeying majority, that is, when a sensor measuring point is proportionally out of order with the other sensors at two characteristic frequencies, the sensor is determined to be faulty. When there are two frequency points that are proportionally out of order between only two measuring points, and the two measuring points are normal with the other measuring points, the two measuring points are determined to be fault-free.

[0062] Step five: data reconstruction is performed on the fault data.

[0063] When a measuring point is determined to be faulty, the characteristic of the measuring point can be reconstructed using the same characteristics of the other measuring points. Taking the output shaft gear meshing frequency as an example, if the output shaft measuring point is determined to be faulty, the amplitude value of the output shaft gear meshing frequency of the output shaft measuring point is represented by a linear combination of the amplitude values of the other measuring points at the frequency point with the concentration degree as the weight. For example, the number of the output shaft measuring point is recorded as 5, the amplitude values of the other measuring points at the frequency point are F i (i = 1, 2, 3, 4), and the mode of the amplitude value ratio of the measuring point to the other measuring points is b 5,i , and the concentration degree of the mode of the amplitude value ratio is Ω 5,i , and the data reconstruction formula is:

[0064]

Claims

1. A method of fault detection for a vibration sensor, the method comprising: The method comprises the following steps: Obtaining vibration data of vibration sensors placed at different measuring points, and extracting the amplitude of the vibration data at a characteristic frequency point as a target characteristic amplitude; For any measuring point and any characteristic frequency point, calculating the amplitude ratio of the target characteristic amplitude between the current measuring point and the remaining measuring points to obtain a target amplitude ratio; Comparing the target amplitude ratio with a reference ratio parameter, and determining that a sensor fault has occurred if the two are out of proportion.

2. The method according to claim 1, characterized in that The reference ratio parameter includes a reference amplitude ratio mode and a reference amplitude ratio mode mean square deviation.

3. The method of claim 2, wherein, If the distance between the target amplitude ratio of the current measuring point and the reference amplitude ratio mode mean square deviation is greater than a set multiple of the reference amplitude ratio mode mean square deviation within a set sampling period, it is determined that the target amplitude ratio and the corresponding reference ratio parameter are out of proportion.

4. The method of claim 3, wherein, If all the target amplitude ratios of the current measuring point at all the characteristic frequency points are out of proportion with the corresponding reference ratio parameters, it is determined that the current measuring point has a sensor fault.

5. The method of claim 4, wherein, The method further comprises the step of estimating the relevant data of the fault measuring point using the target characteristic amplitudes of the non-fault measuring points.

6. The method of claim 4, wherein, The reference ratio parameter further includes a reference amplitude ratio concentration.

7. The method of claim 6, wherein, The amplitude of the target characteristic frequency point of the fault measuring point is a weighted sum of the target characteristic amplitudes of the non-fault measuring points and the product of the amplitude ratio mode and the reference amplitude ratio concentration.

8. The method of claim 1, wherein, The reference ratio parameter is obtained by the following method: Obtaining reference vibration data of the vibration sensors placed at different measuring points when they are working normally, and extracting the amplitude of the reference vibration data at a characteristic frequency point as a reference characteristic amplitude; According to the amplitude ratio of the reference characteristic amplitudes of the same characteristic frequency point between any two measuring points, the reference ratio parameters of all the characteristic frequency points of all the measuring points are calculated and obtained.

9. The method of claim 8, wherein, The method further comprises the following steps before the step of calculating and obtaining the reference ratio parameters of all the characteristic frequency points of all the measuring points: Obtaining the amplitude ratio of the same characteristic frequency point between any two measuring points at different sampling periods as amplitude ratio data; Calculating the mode and mode mean square deviation of the amplitude ratio data; Determining whether the difference between any amplitude ratio data and the mode of the amplitude ratio data is greater than a set multiple of the mode mean square deviation of the amplitude ratio data, and if it is greater, deleting the current amplitude ratio data as an abnormal value and returning to the previous step.

Images