Bearing health index construction method and system based on Gaussian coincidence index

Abstract

The present invention discloses a method and system for constructing a bearing health index based on a Gaussian coincidence index, the method comprising: obtaining a sample data set of the entire life cycle of the bearing; the sample data set comprises N data sample points, and each data sample point is a vibration signal of the bearing; based on the sample data set, a Gaussian mixture model is used to evaluate H standard data sample points to obtain a Gaussian reference distribution; the standard data sample point is a data sample point when the bearing is in a healthy state; a Gaussian mixture model is used to evaluate the remaining data sample points to obtain a Gaussian distribution of each remaining data sample point; the remaining data sample points are data sample points in the sample data set other than the H standard data sample points; the Gaussian coincidence between the Gaussian reference distribution and each Gaussian distribution is calculated, and then a health index that can measure the degree of bearing degradation is obtained. The present invention can determine the health index of the bearing without human interference and at low cost.

Description

Technical Field

[0001] The present invention relates to the technical field of rolling bearing health status assessment, and in particular to a method and system for constructing a bearing health index based on a Gaussian coincidence index. Background Art

[0002] As one of the most commonly used parts of mechanical equipment, the health status of rolling bearings directly affects whether the mechanical equipment can operate stably. In recent years, with the rapid development of industry, related fields have put forward high load and high speed requirements for mechanical equipment. The complex and harsh working environment can easily cause serious faults such as wear and pitting of rolling bearings, which will cause problems such as shutdown and damage of mechanical equipment, and in severe cases, cause casualties. Therefore, accurately obtaining the health status of rolling bearings can ensure the safe and stable operation of mechanical equipment and guarantee the normal operation of production and life.

[0003] The health status of rolling bearings is determined by the health indicators of rolling bearings. The current health indicators of rolling bearings are mainly constructed using supervised learning methods, that is, the degradation trend of rolling bearings is constrained by labels, which brings about human interference to a certain extent, and the production of full-life rolling bearing labels will increase additional costs. In addition, the current health indicator construction method is also difficult to uniformly define the failure threshold. Therefore, it is particularly important to study the unsupervised health indicator construction method with a unified threshold. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for constructing a bearing health index based on a Gaussian coincidence index, which can determine the bearing health index without human interference and at low cost.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A method for constructing a bearing health index based on a Gaussian coincidence index, comprising:

[0007] Acquire a sample data set of the entire life cycle of the bearing; the sample data set includes N data sample points, and each of the data sample points is a vibration signal of the bearing;

[0008] Based on the sample data set, a Gaussian mixture model is used to evaluate H standard data sample points to obtain a Gaussian reference distribution; the standard data sample points are data sample points when the bearing is in a healthy state;

[0009] The Gaussian mixture model is used to evaluate the remaining data sample points to obtain the Gaussian distribution of each of the remaining data sample points; the remaining data sample points are data sample points other than the H standard data sample points in the sample data set;

[0010] The Gaussian overlap between the Gaussian reference distribution and each of the Gaussian distributions is calculated, thereby obtaining a health indicator that can measure the degree of bearing degradation.

[0011] Optionally, the obtaining of a sample data set of the entire life cycle of a bearing specifically includes:

[0012] Obtain vibration signals throughout the entire life cycle of the bearing;

[0013] The wavelet denoising algorithm and the first-order difference algorithm are used to remove singular values ​​from the vibration signal to obtain a sample data set for the entire life cycle of the bearing.

[0014] Optionally, the calculating the Gaussian overlap between the Gaussian reference distribution and each of the Gaussian distributions specifically includes:

[0015] According to the formula GCRM(N1||N2)=1-S 2 =S 3 +S 4 (ieS 1 ) Calculate the Gaussian coincidence degree of Gaussian distribution N1 and Gaussian distribution N2;

[0016] Among them, GCRM(N1||N2) represents the Gaussian coincidence value of Gaussian distribution N1 and Gaussian distribution N2; area S 1 represents the non-overlapping area of ​​Gaussian distribution N2, area S 2 represents the overlapping area of ​​Gaussian distribution N1 and Gaussian distribution N2, area S 3 and area S 4 They respectively represent two non-overlapping areas of the Gaussian distribution N1; the Gaussian distribution N1 represents the Gaussian reference distribution, and the Gaussian distribution N2 represents the Gaussian distribution corresponding to any remaining data sample points.

[0017] Optionally, also include:

[0018] The health index is evaluated using a monotonicity index, a correlation index, and a robustness index.

[0019] Optionally, also include:

[0020] A failure threshold is determined based on the health indicator.

[0021] A bearing health index construction system based on Gaussian coincidence index, comprising:

[0022] A data acquisition module is used to acquire a sample data set of the entire life cycle of the bearing; the sample data set includes N data sample points, and each of the data sample points is a vibration signal of the bearing;

[0023] A Gaussian reference distribution determination module is used to evaluate H standard data sample points based on the sample data set using a Gaussian mixture model to obtain a Gaussian reference distribution; the standard data sample points are data sample points when the bearing is in a healthy state;

[0024] A Gaussian distribution determination module, configured to use the Gaussian mixture model to evaluate the remaining data sample points to obtain the Gaussian distribution of each of the remaining data sample points; the remaining data sample points are data sample points other than the H standard data sample points in the sample data set;

[0025] The health index determination module is used to calculate the Gaussian overlap between the Gaussian reference distribution and each Gaussian distribution, and then obtain a health index that can measure the degree of bearing degradation.

[0026] Optionally, the data acquisition module specifically includes:

[0027] A vibration signal acquisition unit, used to acquire the vibration signal of the bearing throughout its life cycle;

[0028] The sample data set construction unit is used to remove singular values ​​from the vibration signal using a wavelet noise reduction algorithm and a first-order difference algorithm to obtain a sample data set for the entire life cycle of the bearing.

[0029] Optionally, in the method for calculating the Gaussian overlap between the Gaussian reference distribution and each Gaussian distribution, the health indicator determination module is used to:

[0030] According to the formula GCRM(N1||N2)=1-S 2 =S 3 +S 4 (ieS 1 ) Calculate the Gaussian coincidence degree of Gaussian distribution N1 and Gaussian distribution N2;

[0031] Among them, GCRM(N1||N2) represents the Gaussian coincidence value of Gaussian distribution N1 and Gaussian distribution N2; area S 1 represents the non-overlapping area of ​​Gaussian distribution N2, area S 2 represents the overlapping area of ​​Gaussian distribution N1 and Gaussian distribution N2, area S 3 and area S 4 They respectively represent two non-overlapping areas of the Gaussian distribution N1; the Gaussian distribution N1 represents the Gaussian reference distribution, and the Gaussian distribution N2 represents the Gaussian distribution corresponding to any remaining data sample points.

[0032] Optionally, also include:

[0033] The evaluation module is used to evaluate the health index by using a monotonicity index, a correlation index, and a robustness index.

[0034] Optionally, also include:

[0035] A failure threshold determination module is used to determine a failure threshold according to the health indicator.

[0036] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0037] The present invention provides a method and system for constructing a bearing health index based on a Gaussian coincidence index, which determines the health index of a bearing through Gaussian distribution and Gaussian coincidence, thereby avoiding human interference caused by constructing the health index using a supervised learning method and adding additional costs when producing full-life bearing labels. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0039] Figure 1 Schematic diagram of probability density function of two Gaussian distributions of the present invention;

[0040] Figure 2 It is the position distribution diagram of three Gaussian distributions of the present invention;

[0041] Figure 3 A schematic diagram of the Gaussian distribution and GCRM ​​value of the simulation set of the present invention;

[0042] Figure 4 It is a flow chart of a method for constructing a bearing health index based on a Gaussian coincidence index according to the present invention;

[0043] Figure 5 This is a physical picture of the French PRONOSTIA test bench of the present invention;

[0044] Figure 6 A health index graph of two data sets determined using the technical solution of the present invention;

[0045] Figure 7 A health index graph of multiple data sets determined using the technical solution of the present invention;

[0046] Figure 8 It is a structural schematic diagram of a bearing health index construction system based on Gaussian coincidence index of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0049] Gaussian distribution distance indicator:

[0050] For any two Gaussian distributions N1 and N2, their probability density functions are as follows: Figure 1 As shown. Figure 1 It can be seen that the two Gaussian distributions can be divided into four parts by two density function curves, among which area S 1 represents the non-overlapping area of ​​Gaussian distribution N2, S 2 Represents the overlap area of ​​two probability density functions, S 3 and S 4 They respectively represent two non-overlapping areas of Gaussian distribution N1.

[0051] For Gaussian distribution, the larger the overlapping area is, the more similar the two distributions are. Therefore, the area of ​​the overlapping area can be used to evaluate the distance between the two Gaussian distributions. Therefore, the present invention calculates the distance between the two Gaussian distributions by using the Gaussian contact ratio metric (GCRM), and the calculation formula is as follows:

[0052] GCRM(N1||N2)=1-S 2 =S 3 +S 4 (ieS 1 ) (1);

[0053] The meaning of ie is: that is.

[0054] according to Figure 1 And formula (1), we can draw the following conclusions:

[0055] 1) The GCRM ​​indicator has a probabilistic characteristic; 2) As the GCRM ​​indicator becomes smaller, the overlap between the two Gaussian distributions becomes larger, and when the GCRM ​​indicator is 0, the two Gaussian distributions are exactly the same; 3) According to the definition of the GCRM ​​indicator, it can be concluded that GCRM(N1||N1)≥0, which satisfies the non-negativity; 4) Because GCRM(N1||N2)=GCRM(N2||N1), the GCRM ​​indicator also satisfies the symmetry.

[0056] For a distribution distance metric, in addition to satisfying non-negativity and symmetry, it should also satisfy triangle inequality.

[0057] First, assume that three Gaussian distributions are defined as N 1 =(μ 1 ,σ 1 ),N 2 =(μ 2 ,σ 2 ) and N 3 =(μ 3 ,σ 3 ), where σ 1 >σ 2 >σ 3 According to the symmetry of Gaussian distribution, three cases need to be considered separately, such as Figure 2 shown.

[0058] Assumptions Represents two Gaussian distributions N i and Gaussian distribution N j The kth region, GCRM ij (Right now ) is represented by two Gaussian distributions N i and Gaussian distribution N j The distribution distance of . Based on the previous assumptions, the proof of triangle inequality is as follows:

[0059] Proof: For the first case, from Figure 2 (a) and formula (1) show that:

[0060]

[0061] according to Figure 2 (a) Easy to obtain:

[0062]

[0063] Due to GCRM 13 In three indicators GCRM 11 , GCRM 12 and GCRM 13 is the largest in the equation (3).

[0064] For the second case, according to Figure 2 (b) and formula (1) show that:

[0065]

[0066] Similar ones include:

[0067]

[0068] At the same time based on Figure 2 (b) There are:

[0069]

[0070] Formulas (4)-(6) show that this index also satisfies the triangle inequality.

[0071] Situation 3 Figure 2 As shown in (c), according to formula (1), we can get:

[0072]

[0073] because yes The largest one among them, then according to formula 1:

[0074]

[0075] In summary, the Gaussian coincidence index proposed in the present invention meets the requirements of measuring distribution distance, namely, non-negativity, symmetry and triangle inequality.

[0076] In order to further demonstrate the evaluation ability of the Gaussian coincidence index, the present invention uses a simulation set of 10 Gaussian distributions, whose distribution conditions and corresponding GCRM ​​values ​​are as follows: Figure 3 As shown. The first two Gaussian distributions have the same mean and variance, and the mean and variance of the remaining Gaussian distributions gradually increase. Taking the first Gaussian distribution as a comparison benchmark, the GCRM ​​value can be calculated according to formula (1), as follows: Figure 3 (b) is shown. Figure 3 It can be seen that the first two GCRM ​​values ​​are zero, indicating that the distance between the two Gaussian distributions is small, and the GCRM ​​index continues to rise with the mean and variance. Therefore, the GCRM ​​value can effectively evaluate the distance between Gaussian distributions. In addition, this characteristic of the GCRM ​​index can be used to obtain the health index of the bearing based on the distribution of the bearing vibration signal.

[0077] Currently, supervised learning is mainly used to construct health indicators. However, this method not only introduces human interference, but also the constructed indicators have no accurate physical meaning and cannot uniformly define the failure threshold. Since the vibration signal of rotating machinery can well reflect its operating status and the Gaussian mixture model can accurately extract the distribution parameters of vibration signal data, the proposed GCRM ​​indicator can be combined to construct a health indicator with physical meaning and a uniform definition of the failure threshold. First, the Gaussian distribution p of the healthy sample is 0 When the bearing is healthy, the Gaussian distribution p 0There is no significant difference from the Gaussian distribution of the collected samples. On the contrary, obvious errors appear when the bearing fails, and the errors generally increase with the increase of the fault severity.

[0078] Based on the above, this embodiment provides a method for constructing a bearing health index based on a Gaussian coincidence index, such as Figure 4 As shown, including:

[0079] Step 401: Acquire a sample data set of the entire life cycle of a bearing; the sample data set includes N data sample points, and each of the data sample points is a vibration signal of the bearing.

[0080] Step 401 is specifically as follows:

[0081] The vibration signal of the bearing throughout its life cycle is obtained; then a wavelet denoising algorithm and a first-order difference algorithm are used to remove singular values ​​from the vibration signal to obtain a sample data set of the bearing throughout its life cycle.

[0082] The data set of the entire life cycle of the bearing is represented by X = (x 1 ,x 2 ,…,x N ), where x i =(x 1 ,x 2 ,…,x n ), N represents the number of data sample points, and n represents the length of the data sample points.

[0083] Step 402: Based on the sample data set, a Gaussian mixture model is used to evaluate H standard data sample points to obtain a Gaussian reference distribution; the standard data sample points are data sample points when the bearing is in a healthy state.

[0084] This step is as follows: Take the first H bearing health data sample points x 1 ,…,x H , and then the Gaussian mixture model is used to evaluate its data distribution, that is, the Gaussian benchmark distribution p 0 Among them, the Gaussian component of the Gaussian mixture model is set to 4.

[0085] Step 403: Use the Gaussian mixture model to evaluate the remaining data sample points to obtain the Gaussian distribution of each of the remaining data sample points; the remaining data sample points are the data sample points in the sample data set except the H standard data sample points.

[0086] This step is specifically as follows: a Gaussian mixture model with 4 Gaussian components is used to calculate the remaining vibration data x H+1 ,…,x N The distribution parameters are denoted as p H+1 ,…,p N.

[0087] Step 404: Calculate the Gaussian overlap between the Gaussian reference distribution and each of the Gaussian distributions, and then obtain a health indicator that can measure the degree of bearing degradation.

[0088] According to the formula GCRM(N1||N2)=1-S 2 =S 3 +S 4 (ieS 1 ) Calculate the Gaussian reference distribution p 0 With Gaussian distribution p = [p H+1 ,…,p N ], that is, according to the formula GCRM(N1||N2)=1-S 2 =S 3 +S 4 (ieS 1 ) calculates the Gaussian coincidence of Gaussian distribution N1 and Gaussian distribution N2; wherein GCRM(N1||N2) represents the Gaussian coincidence value of Gaussian distribution N1 and Gaussian distribution N2; region S 1 represents the non-overlapping area of ​​Gaussian distribution N2, area S 2 represents the overlapping area of ​​Gaussian distribution N1 and Gaussian distribution N2, area S 3 and area S 4 They respectively represent two non-overlapping areas of the Gaussian distribution N1; the Gaussian distribution N1 represents the Gaussian reference distribution, and the Gaussian distribution N2 represents the Gaussian distribution corresponding to any remaining data sample points.

[0089] Then, all the calculated Gaussian coincidences are summarized to obtain the health index GCRHI, which can measure the degree of bearing degradation.

[0090] Furthermore, the method provided in this embodiment also includes:

[0091] The health index is evaluated using a monotonicity index, a correlation index, and a robustness index.

[0092] A failure threshold is determined based on the health indicator.

[0093] The technical solution provided by the present invention is verified by specific experiments below.

[0094] Experiment 1:

[0095] The full life cycle data of the experimental bearing is collected from the French PRONOSTIA platform. Figure 5As shown. The experimental platform is mainly composed of a motor, a reducer, a coupling, a bearing to be tested, a loading device, several sensors and an acquisition card. In order to accelerate the degradation of the bearing, the loading device applies radial force to the bearing. The vibration signal is collected by two accelerometers. The sampling frequency is 25.6kHz, the recording interval is 10s, and the sampling length is 0.1s. In this experiment, the bearing1_1 dataset and the bearing1_3 dataset will be used to verify the technical solution proposed in this invention. For these two datasets, the rotation speed is 1800rpm and the radial force is 4kN. The bearing1_1 dataset and the bearing1_3 dataset have 2803 and 2375 samples, respectively.

[0096] For the bearing1_1 dataset and the bearing1_3 dataset, the first 200 samples are taken as healthy samples and used to calculate p 0 Then, the GCRHI values ​​of bearing1_1 and bearing1_3 are calculated according to the Gaussian coincidence health index construction steps (i.e., the bearing health index GCRHI is constructed using GCRM), as shown in Figure 6 As shown in the figure, it can be clearly seen that the constructed health index can clearly reflect the healthy stage and degradation process of the bearing. At the same time, the value of GCRHI is strictly constrained between 0 and 1, which indicates the distance between the standard distribution and the data sample. That is, when the GCRHI value is close to 0, the bearing is in a healthy state, and when the GCRHI value is closer to 1, the bearing is more likely to fail. Therefore, the proposed GCRHI can well reflect the degradation trajectory of the bearing and can uniformly define the failure threshold.

[0097] To demonstrate the superiority of the proposed GCRHI indicator, it is compared with RMS, VAE HI, JRD HI and KLD HI. Monotonicity, correlation, robustness and comprehensive indicators are used for quantitative comparison. To calculate these evaluation indicators, HI (health indicator) is decomposed into the average trend part and the random part:

[0098] X(t k )=X T (t k )+X R (t k ) (9);

[0099] In this formula, X(t k ) indicates that at t k Health indicators at all times, X T (t k ) represents its average trend, X R (t k ) represents the random fluctuation part. The formulas of the three indicators are as follows:

[0100] 1) Monotonicity:

[0101]

[0102] 2) Relevance:

[0103]

[0104] 3) Robustness:

[0105]

[0106] 4) Comprehensive evaluation indicators:

[0107] CI=ω 1 Mon(X)+ω 2 Corr(X,T)+ω 3 Rob(X) (13);

[0108] Where ω>0 and Select ω 1 =0.2,ω 2 =0.5 and ω 3 =0.3, the bearing health index evaluation can be obtained, and the results of GCRHI, RMS, VAE HI, JRD HI and KLD HI are shown in Table 1. The above evaluation indicators reflect the performance of HI from multiple aspects. Obviously, the health indicators with better degradation performance can better reflect the degradation condition of the bearing. As can be seen from Table 1, GCRM ​​has the best monotonicity, correlation and comprehensive performance for these two data sets. Taking Bearing1_1 as an example, the comprehensive index of GCRHI is 12%, 4.9%, 6.7% and 13.8% higher than that of RMS, VAE HI, JRD HI and KLD HI, respectively. Although RMS has the highest robustness, other performance indicators are not as good as GCRHI. Therefore, the GCRHI proposed in the present invention has a better ability to evaluate bearing degradation.

[0109] Table 1 Evaluation index table of different bearing health indicators

[0110]

[0111] Experiment 2:

[0112] The experimental data is the XJTU-SY bearing dataset. The test bench consists of an AC motor, a motor speed controller, two support bearings, a support shaft, a bearing under test, a loading system, and a measurement system. The adjustable working conditions include load and speed. The loading system generates radial force to accelerate bearing degradation. The speed is adjusted by a speed controller. Vibration samples are collected by two accelerometers. The sampling frequency, sampling length, and recording interval are set to 25.6kHz, 1.28s, and 1min, respectively. In the experiment, six bearing datasets are used to verify the proposed indicator construction method, as shown in Table 2.

[0113] Table 2 Test bearings under different working conditions

[0114]

[0115]

[0116] According to the proposed HI construction method, the GCRHIs of the six bearings are calculated, as shown in Figure 7 As shown. Figure 7 It can be seen that each GCRHI has a different degradation trajectory, and its value also varies in the range of [0,1]. All GCRHIs have a unified failure threshold (the failure point is determined by the experiment, and the failure threshold is determined by the health indicator value corresponding to the failure point), which is set to 0.95, that is, the 95% probability failure threshold. Compared with the average failure threshold and weighted failure threshold, setting a 95% probability failure threshold in GCRM ​​can greatly reduce the number and cost of experiments.

[0117] The above two sets of experiments have proved that the proposed GCRHI has the ability to accurately describe the bearing degradation stage and degradation trajectory, and the performance of GCRHI is quantitatively evaluated through multiple evaluation indicators such as monotonicity and correlation. This indicator not only has better degradation performance but also can uniformly set the failure threshold.

[0118] Embodiment 2

[0119] like Figure 8 As shown, this embodiment provides a bearing health index construction system based on Gaussian coincidence index, including:

[0120] The data acquisition module 801 is used to acquire a sample data set of the entire life cycle of the bearing; the sample data set includes N data sample points, and each of the data sample points is a vibration signal of the bearing.

[0121] The Gaussian reference distribution determination module 802 is used to evaluate H standard data sample points based on the sample data set using a Gaussian mixture model to obtain a Gaussian reference distribution; the standard data sample points are data sample points when the bearing is in a healthy state.

[0122] The Gaussian distribution determination module 803 is used to evaluate the remaining data sample points using the Gaussian mixture model to obtain the Gaussian distribution of each of the remaining data sample points; the remaining data sample points are the data sample points in the sample data set except the H standard data sample points.

[0123] The health index determination module 804 is used to calculate the Gaussian overlap between the Gaussian reference distribution and each of the Gaussian distributions, and then obtain a health index that can measure the degree of bearing degradation.

[0124] The data acquisition module 801 specifically includes:

[0125] The vibration signal acquisition unit is used to acquire the vibration signal of the bearing throughout its life cycle.

[0126] The sample data set construction unit is used to remove singular values ​​from the vibration signal using a wavelet noise reduction algorithm and a first-order difference algorithm to obtain a sample data set for the entire life cycle of the bearing.

[0127] In the method for calculating the Gaussian coincidence between the Gaussian reference distribution and each Gaussian distribution, the health index determination module 804 is used to:

[0128] According to the formula GCRM(N1||N2)=1-S 2 =S 3 +S 4 (ieS 1 ) Calculate the Gaussian coincidence degree of Gaussian distribution N1 and Gaussian distribution N2;

[0129] Among them, GCRM(N1||N2) represents the Gaussian coincidence value of Gaussian distribution N1 and Gaussian distribution N2; area S 1 represents the non-overlapping area of ​​Gaussian distribution N2, area S 2 represents the overlapping area of ​​Gaussian distribution N1 and Gaussian distribution N2, area S 3 and area S 4 They respectively represent two non-overlapping areas of the Gaussian distribution N1; the Gaussian distribution N1 represents the Gaussian reference distribution, and the Gaussian distribution N2 represents the Gaussian distribution corresponding to any remaining data sample points.

[0130] Furthermore, the system provided in this embodiment also includes:

[0131] The evaluation module is used to evaluate the health index by using a monotonicity index, a correlation index, and a robustness index.

[0132] A failure threshold determination module is used to determine a failure threshold according to the health indicator.

[0133] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0134] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for constructing a bearing health index based on a Gaussian coincidence index, characterized in that: include: Obtain a sample data set of the entire life cycle of a bearing; The sample data set includes N data sample points, and each of the data sample points is a vibration signal of a bearing; The obtaining of a sample data set for the entire life cycle of a bearing specifically includes: Obtain vibration signals of the bearing throughout its life cycle; A wavelet denoising algorithm and a first-order difference algorithm are used to remove singular values ​​from the vibration signal to obtain a sample data set for the entire life cycle of the bearing; The data set of the entire life cycle of a bearing is represented by X = (x1, x2, L, x N ), where x i =(x 1 ,x 2 ,L,x n ), N represents the number of data sample points, and n represents the length of the data sample points; Based on the sample data set, a Gaussian mixture model is used to evaluate H standard data sample points to obtain a Gaussian benchmark distribution; the standard data sample points are data sample points when the bearing is in a healthy state; specifically, the H bearing healthy data sample points x1, L, x H , and then the Gaussian mixture model is used to evaluate its data distribution, that is, the Gaussian benchmark distribution p0; wherein the Gaussian component of the Gaussian mixture model is set to 4; The Gaussian mixture model is used to evaluate the remaining data sample points to obtain the Gaussian distribution of each of the remaining data sample points; the remaining data sample points are data sample points other than H standard data sample points in the sample data set; specifically: the Gaussian mixture model with 4 Gaussian components is used to calculate the remaining vibration data x H+1 ,L,x N The distribution parameters are denoted as p H+1 ,L,p N ; Calculating the Gaussian overlap between the Gaussian reference distribution and each Gaussian distribution, and then obtaining a health indicator that can measure the degree of bearing degradation; Then, all the calculated Gaussian coincidences are summarized to obtain the health index GCRHI that can measure the degree of bearing degradation; The calculating of the Gaussian overlap between the Gaussian reference distribution and each of the Gaussian distributions specifically includes: The Gaussian coincidence of Gaussian distribution N1 and Gaussian distribution N2 is calculated according to the formula GCRM ​​(N1||N2) = 1-S2 = S3+S4 (ieS1); Among them, GCRM(N1||N2) represents the Gaussian overlap value between Gaussian distribution N1 and Gaussian distribution N2; area S1 represents the non-overlapping area of ​​Gaussian distribution N2, area S2 represents the overlapping area of ​​Gaussian distribution N1 and Gaussian distribution N2, area S3 and area S4 respectively represent two non-overlapping areas of Gaussian distribution N1; Gaussian distribution N1 represents the Gaussian reference distribution, and Gaussian distribution N2 represents the Gaussian distribution corresponding to any remaining data sample points.

2. A method for constructing a bearing health index based on a Gaussian coincidence index according to claim 1, characterized in that: Also includes: The health index is evaluated using a monotonicity index, a correlation index, and a robustness index.

3. A method for constructing a bearing health index based on a Gaussian coincidence index according to claim 1, characterized in that: Also includes: A failure threshold is determined based on the health indicator.

4. A bearing health index construction system based on Gaussian coincidence index, characterized in that: include: Data acquisition module, used to obtain sample data sets of the entire life cycle of the bearing; The sample data set includes N data sample points, and each of the data sample points is a vibration signal of a bearing; The obtaining of a sample data set for the entire life cycle of a bearing specifically includes: Obtain vibration signals of the bearing throughout its life cycle; A wavelet denoising algorithm and a first-order difference algorithm are used to remove singular values ​​from the vibration signal to obtain a sample data set for the entire life cycle of the bearing; The data set of the entire life cycle of a bearing is represented by X = (x1, x2, L, x N ), where x i =(x 1 ,x 2 ,L,x n ), N represents the number of data sample points, and n represents the length of the data sample points; A Gaussian reference distribution determination module is used to evaluate H standard data sample points based on the sample data set using a Gaussian mixture model to obtain a Gaussian reference distribution; the standard data sample points are data sample points when the bearing is in a healthy state; specifically, the H bearing healthy data sample points x1, L, x H , and then the Gaussian mixture model is used to evaluate its data distribution, that is, the Gaussian benchmark distribution p0; wherein the Gaussian component of the Gaussian mixture model is set to 4; A Gaussian distribution determination module is used to evaluate the remaining data sample points using the Gaussian mixture model to obtain the Gaussian distribution of each of the remaining data sample points; the remaining data sample points are data sample points other than H standard data sample points in the sample data set; specifically: a Gaussian mixture model with 4 Gaussian components is used to calculate the remaining vibration data x H+1 ,L,x N The distribution parameters are denoted as p H+1 ,L,p N ; A health index determination module, used to calculate the Gaussian overlap between the Gaussian reference distribution and each Gaussian distribution, and then obtain a health index that can measure the degree of bearing degradation; Then, all the calculated Gaussian coincidences are summarized to obtain the health index GCRHI that can measure the degree of bearing degradation; In the method for calculating the Gaussian coincidence between the Gaussian reference distribution and each Gaussian distribution, the health indicator determination module is used to: The Gaussian coincidence of Gaussian distribution N1 and Gaussian distribution N2 is calculated according to the formula GCRM ​​(N1||N2) = 1-S2 = S3+S4 (ieS1); Among them, GCRM(N1||N2) represents the Gaussian overlap value between Gaussian distribution N1 and Gaussian distribution N2; area S1 represents the non-overlapping area of ​​Gaussian distribution N2, area S2 represents the overlapping area of ​​Gaussian distribution N1 and Gaussian distribution N2, area S3 and area S4 respectively represent two non-overlapping areas of Gaussian distribution N1; Gaussian distribution N1 represents the Gaussian reference distribution, and Gaussian distribution N2 represents the Gaussian distribution corresponding to any remaining data sample points.

5. A system for constructing a bearing health index based on a Gaussian coincidence index according to claim 4, characterized in that: Also includes: The evaluation module is used to evaluate the health index by using a monotonicity index, a correlation index, and a robustness index.

6. A system for constructing a bearing health index based on a Gaussian coincidence index according to claim 4, characterized in that: Also includes: A failure threshold determination module is used to determine a failure threshold according to the health indicator.

Images