A method, device and medium for predicting remaining life of bearing

Abstract

This application discloses a method, device, and medium for predicting the remaining life of a bearing, relating to the field of bearing life prediction. The method comprises: extracting features from the time series of the vibration signals of the target bearing to obtain time-domain features; screening the time-domain features based on robustness, monotonicity, and predictability, assigning feature weights, and performing weighted calculations to obtain a feature vector to be used; performing a multivariate regression operation on the set of available time-domain features and the set of available feature vectors using the partial least squares method to achieve feature fusion and obtain a comprehensive degradation feature quantity; constructing a prediction model based on the comprehensive degradation feature quantities corresponding to each vibration signal time series using a preset exponential model, and then updating and estimating the prediction model parameters using Bayesian updating and expectation maximization algorithms; and correcting prediction errors using a strong tracking filtering algorithm to obtain a bearing remaining life prediction result. This application can improve the accuracy of bearing remaining life prediction.

Description

Technical Field

[0001] The present application relates to the field of bearing life prediction, and in particular to a method, device and medium for predicting the remaining life of a bearing. Background Art

[0002] Bearings, as a vital component, are widely used in industry, and their performance directly impacts production efficiency. However, bearings are subject to harsh operating environments, such as high loads and high speeds, making them prone to failure. Therefore, monitoring bearing data and predicting their remaining life are essential. This process can lead to different types of failures depending on the bearing's operating conditions, and bearings are susceptible to interference from external factors such as noise during operation. Traditional data feature extraction methods suffer from shortcomings such as insensitivity to key data and difficulty processing high-dimensional data. Summary of the Invention

[0003] The purpose of this application is to provide a bearing remaining life prediction method, device and medium, which can improve the accuracy of bearing remaining life prediction.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a method for predicting the remaining life of a bearing, comprising:

[0006] Obtain a set of vibration signal time series of the target bearing and perform feature extraction on each vibration signal time series to obtain time domain features;

[0007] Based on robustness, monotonicity and predictability, the time domain features corresponding to all the vibration signal time series are screened to obtain a set of ready-to-use time domain features;

[0008] Based on robustness, monotonicity and predictability, each of the standby time domain features in the standby time domain feature set is assigned a feature weight and weighted calculation is performed to obtain a standby feature vector; a plurality of the standby feature vectors constitute a standby feature vector set;

[0009] Using partial least squares method, a multivariate regression operation is performed on the set of unused time domain features and the set of unused feature vectors to achieve feature fusion and obtain a comprehensive degradation feature quantity;

[0010] Based on the comprehensive degradation characteristic quantities corresponding to the time series of each vibration signal, a preset exponential model is used to construct a prediction model. The Bayesian update and expectation maximization algorithms are used to update and estimate the parameters of the prediction model. The strong tracking filtering algorithm is used to correct the prediction error to obtain the remaining life prediction result of the bearing.

[0011] In a second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a bearing remaining life prediction method.

[0012] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements a method for predicting the remaining life of a bearing when executed by a processor.

[0013] According to the specific embodiments provided by the present application, the present application has the following technical effects: The present application provides a method, device and medium for predicting the remaining life of a bearing. First, based on robustness, monotonicity and predictability, the time domain features corresponding to the time series of the vibration signal of the target bearing are subjected to feature analysis and screening, and standby time domain features that are more in line with and more conducive to degradation research are extracted. Then, based on robustness, monotonicity and predictability, feature weights are assigned to the standby time domain features and weighted calculations are performed to obtain standby feature vectors, providing a more practical data basis for subsequent processing. The partial least squares method is used to perform a multivariate regression operation on the set of standby time domain features and the set of standby feature vectors to synthesize all standby degradation feature quantity features to obtain a comprehensive degradation feature quantity, thereby improving the sensitivity to key degradation data and the interpretability of features. Finally, a preset exponential model is used to construct the model, and the Bayesian update and expectation maximization algorithms are used to update and estimate the parameters of the prediction model. The strong tracking filtering algorithm is used to correct the prediction error to ensure the prediction accuracy and obtain a more accurate prediction result of the remaining life of the bearing. The strong tracking filtering technology used in this step enables the preset exponential model to dynamically adjust the parameters in time to adapt to the changes when the degradation state changes suddenly, thereby improving the generalization ability of the model, realizing dynamic tracking of the degradation state, and increasing the robustness of the prediction, so that it can face different degradation stages and mutations in the degradation stage, achieve effective processing, and thus obtain a more accurate prediction result of the remaining life of the bearing. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 This is a diagram of the application environment of the bearing remaining life prediction method in one embodiment of the present application.

[0016] Figure 2 A flow chart of a bearing remaining life prediction method provided in one embodiment of the present application.

[0017] Figure 3 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0019] In order to make the purpose, features and advantages of this application more obvious and easy to understand, this application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0020] The bearing remaining life prediction method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the vibration signal time series set of the target bearing to the server 104. After receiving it, the server 104 extracts the features of the vibration signal time series to obtain the time domain features. Based on robustness, monotonicity and predictability, the time domain features corresponding to all vibration signal time series are screened, the feature weights are assigned and weighted calculations are performed to obtain the feature vectors to be used. The partial least squares method is used to perform a multivariate regression operation on the feature vector set to achieve feature fusion and obtain a comprehensive degradation feature quantity. Finally, based on the comprehensive degradation feature quantity corresponding to each vibration signal time series, a preset exponential model is used to construct a model, and the Bayesian update and expectation maximization algorithms are used to update and estimate the parameters of the prediction model. The strong tracking filter algorithm is used to correct the prediction error to obtain the remaining life prediction result of the bearing. The server 104 may feed back the bearing remaining life prediction model or the corresponding remaining life to the terminal 102. In some embodiments, the bearing remaining life prediction method may also be implemented by the server 104 or the terminal 102 alone.

[0021] In an exemplary embodiment, Figure 2 As shown, a method for predicting the remaining life of a bearing is provided. The method is executed by a computer device, specifically a computer device such as a terminal or a server, or a terminal and a server. In the embodiment of the present application, the method is applied to Figure 1The server 104 in the example is used for explanation, including the following steps 201 to 205.

[0022] Step 201 : Acquire a set of vibration signal time series of a target bearing, and perform feature extraction on each vibration signal time series to obtain a time domain feature.

[0023] In practical applications, multiple sampling points are determined on the target bearing. At each sampling point, multiple vibration signals are sampled to generate a vibration signal time series. This vibration signal time series contains k data points, corresponding to k acquisition time points. Feature extraction of the vibration signal time series yields time-domain features that also include k observed variables. These time-domain features are the mean, RMS value, RMS amplitude, absolute mean, skewness, kurtosis, variance, maximum, minimum, and peak values.

[0024] Step 202, based on robustness, monotonicity and predictability, screen the time domain features corresponding to all the vibration signal time series to obtain a set of time domain features to be used. In an application example, step 202 includes the following steps (21)-(24).

[0025] (21) The following formula is used to evaluate the feature robustness:

[0026] .

[0027] Among them, Rob ( F i ) is the i Time domain features F i Robustness evaluation results of , f i (1) i The first observed variable in the time domain features, f i (2) i The second observed variable in the time domain characteristics, For the i Among the time domain features j observed variables, For the i Among the time domain features k observed variables, j=1,…,k; For The observed variable after smoothing.

[0028] Taking the maximum value as an example, if the maximum value is the first time domain feature, then Represents the matrix containing the corresponding k largest values.

[0029] (22) The following formula is used to evaluate the monotonicity of the feature:

[0030] .

[0031] Among them, Mon( F i ) is the i Time domain features F i Monotonicity evaluation results of ; is a unit step function, f i ( j -1) is the i Among the time domain features j -1 observed variable.

[0032] (23) The following formula is used to evaluate the predictive performance of the feature:

[0033] .

[0034] Among them, Pro( F i ) is the i Time domain features F i Predictive evaluation results; std() is the standard deviation function, and mean() is the mean function.

[0035] (24) From the time domain features corresponding to all the vibration signal time series, the time domain features whose robustness evaluation results, monotonicity evaluation results, and predictability evaluation results all meet the preset threshold range are screened out and marked as stand-by time domain features to obtain a set of stand-by time domain features.

[0036] In practical applications, the preset threshold range can be set to [0, 1]. After the three evaluations are completed, the time domain features that are not in the range of [0, 1] will be eliminated, and n time domain features that meet the conditions will be retained, thereby ensuring the interpretability and significance of each time domain feature.

[0037] In step 203, based on robustness, monotonicity, and predictability, each of the standby time domain features in the standby time domain feature set is assigned a feature weight and weighted calculation is performed to obtain a standby feature vector; a plurality of the standby feature vectors constitute a standby feature vector set. In one application example, step 203 includes the following steps:

[0038] Based on robustness, monotonicity and predictability, the following matrix to be used is determined according to the set of time-domain features to be used:

[0039] .

[0040] Among them, Mon( F 1) is the first standby time domain feature F 1 monotonicity evaluation results, Mon( F 2) is the second standby time domain feature F 2 monotonicity evaluation results, Mon( F n ) is the nth time domain feature to be used F n Monotonicity evaluation results of Rob( F 1) is the first standby time domain feature F 1 robustness evaluation results, Rob( F 2) is the second standby time domain feature F 2 robustness evaluation results, Rob( F n ) is the nth time domain feature to be used F n Robustness evaluation results of Pro( F 1) is the first time domain feature to be used F 1's predictive evaluation results, Pro( F 2) is the second standby time domain feature F 2 predictive evaluation results, Pro( F n ) is the nth time domain feature to be used F n predictive evaluation results.

[0041] Using K to refer to each evaluation result, it is expressed as follows:

[0042] .

[0043] in, K 11 Refers to Mon( F 1), K 1n Refers to Mon( F n ), K 21 Refers to Rob( F 1), K 2n Refers to Rob( F n ), K 31 Refers to Pro( F 1), K 3n Refers to Pro( F n ).

[0044] The feature weight of the to-be-used time domain feature is calculated using the following formula:

[0045] .

[0046] .

[0047] .

[0048] .

[0049] .

[0050] .

[0051] .

[0052] Where H is the number of types of evaluation results, h=1,…,H, corresponding to the above, H=3; z is the coefficient, is the ratio, K hn is the hth evaluation result of the nth time domain feature to be used, For K hn Sum by row, e hn 、 d n 、 are all intermediate parameters. For e hn Sum by row, is the information entropy, is the feature weight, is the initial value of the weight, and the initial value is 1.

[0053] The standby feature vectors are obtained by weighted calculation, and the corresponding set of standby feature vectors is expressed as .

[0054] Step 204, using partial least squares method, performs multiple regression operation on the set of unused time domain features and the set of unused feature vectors to achieve feature fusion and obtain comprehensive degradation feature quantity. In this step, the introduction of traditional partial least squares method can find the linear relationship between the independent variable and the dependent variable, and extract the component with the largest amount of information from these relationships, so as to achieve multiple inputs and single output. In an application example, step 204 includes the following steps (41)-(46) to construct a comprehensive n A new characteristic quantity of the degradation characteristic, namely the comprehensive degradation characteristic quantity, is used to simulate the bearing degradation process.

[0055] (41) Normalizing the set of unused time domain features and the set of unused feature vectors so that the values are within the range , determine independent variable data and dependent variable data; the independent variable data refers to the set of time-domain features to be used, and the dependent variable data refers to the comprehensive degradation feature quantity. Furthermore, the dependent variable data is user-defined data and is the sum of the independent variable data, Y = sum(X).

[0056] The formula for determining the independent variable data is:

[0057] .

[0058] in, represents the mean value of the time domain feature F to be used, represents the standard deviation of the time domain feature F to be used, X’ represents the independent variable data, F i is the i-th time domain feature to be used.

[0059] The formula for determining the dependent variable data is:

[0060] ; .

[0061] in, Y’ represents the dependent variable data, is the feature vector to be used, Represents the weighted sum Y The mean of Represents the weighted sum Y The standard deviation of n represents the number of time domain features to be used.

[0062] (42) Based on the independent variable data, determine the latent variable; the calculation formula of the latent variable is:

[0063] .

[0064] in, For the latent variables, For independent variable data The weight of .

[0065] (43) Performing regression analysis based on the latent variables and the dependent variable data to obtain regression coefficients; regression coefficients c k The calculation formula is:

[0066] .

[0067] (44) Based on the regression coefficient, the independent variable data and the dependent variable data are orthogonalized to obtain updated independent variable data and updated dependent variable data; the updated independent variable data is expressed as: ; The updated dependent variable data is expressed as: .

[0068] (45) Based on the updated independent variable data and the updated dependent variable data, regression is performed to obtain a comprehensive degradation feature quantity prediction model, which is: ;in, X (k+1) is the k+1 moment State quantity, Y (k+1) is the k+1 moment State quantity, Y’ is the dependent variable data, is the comprehensive degradation characteristic quantity, () T is the transpose of the matrix.

[0069] In step 205, a prediction model is constructed using a preset exponential model based on the comprehensive degradation characteristic quantities corresponding to the time series of each vibration signal. The prediction model is updated and estimated using the Bayesian update and expectation maximization algorithms. A strong tracking filter algorithm is used to correct the prediction error to obtain a prediction result for the remaining life of the bearing. In one application example, step 205 includes the following steps (51)-(54).

[0070] (51) Based on the comprehensive degradation characteristic quantity corresponding to each vibration signal time series, the bearing degradation path function is determined using a preset exponential model. In practical applications, the comprehensive degradation characteristic quantity is assigned to x k ,Right now , the preset index model used is:

[0071] .

[0072] in, Characterization at time points The degraded state, represents the initial degraded state, and is a random variable that satisfies , ; is a random error, satisfying , represents the diffusion parameter used to describe the uncertainty of the degradation process.

[0073] (52) Based on the bearing degradation path function, determine the degradation amount function and the degradation prediction value calculation function; specifically, define the time The degradation amount at , that is, the degradation function is:

[0074] .

[0075] in, , , , , . is the diffusion parameter to describe the uncertainty of degradation.

[0076] Defined in time Degradation prediction value at , that is, the degradation prediction value calculation function is:

[0077] .

[0078] (53) Based on the degradation amount function and the degradation prediction value calculation function, the Bayesian update and expectation maximization algorithm are used to update and estimate the model parameters of the bearing degradation path function; specifically, The parameters are randomly updated and unknown parameters are estimated using the Bayesian update and expectation maximization (EM) algorithms. The specific process is shown below. To simplify the expression of the formula, the subscript k is omitted in the following formula. The subscript k represents the time point of the corresponding data.

[0079] Among them, the update process of Bayesian update is:

[0080] .

[0081] .

[0082] .

[0083] .

[0084] .

[0085] in, yes The mean of the posterior estimate, yes The variance of the posterior estimate, is the correlation coefficient. The above data is the Bayesian update result, that is, the update result of the exponential model parameters.

[0086] The expectation maximization algorithm converts the state monitoring data into The unknown parameter set After update :

[0087] .

[0088] .

[0089] .

[0090] The expectation maximization algorithm mainly updates the parameter σ 2 , the other four parameters The same is also updated. In this step, the Bayesian update and expectation maximization algorithms have different focuses, that is, the updated parameters are different. In the next step, their values are replaced by the values after strong tracking filtering.

[0091] (54) A strong tracking filter algorithm is used to correct the prediction error to obtain the remaining life prediction result of the bearing. In this step, the strong tracking filter is used to correct the prediction error, so that the preset exponential model can dynamically adjust the parameters to adapt to the changes in time, so that the exponential degradation model is more consistent with the actual degradation path. Specifically, the following steps are included (541)-(545).

[0092] (541) Based on the degradation amount function and the degradation prediction value calculation function after preliminary parameter optimization, the system equation is established as follows:

[0093] .

[0094] .

[0095] The observation equation is:

[0096] .

[0097] (542) The fading factor is calculated using the following formula: :

[0098] .

[0099] .

[0100] .

[0101] . .

[0102] .

[0103] in, for t k The forecast residual at time t, For the forgetting factor, is the softening coefficient, To initially estimate the variance, the initial parameters need to be set . 、 It is the coefficient of the fading factor, which is used to adjust the tracking ability of the algorithm. The larger the value, the better the tracking performance. However, too large a value will affect the smoothness of the predicted trajectory.

[0104] (543) The following formula is used to update the status:

[0105] .

[0106] .

[0107] .

[0108] .

[0109] (544) passed , Update the model parameters.

[0110] (545) passed , Update the forecast variance.

[0111] The remaining life of the bearing is predicted in real time based on the updated parameters. In this process, the problem of estimating the remaining life of the bearing is transformed into the problem of the time required for the degradation process to reach the failure threshold ω for the first time, that is, t k Data collected within time W 1:k ,W(t+ t k )yes t + t k The predicted value at the moment obeys the Gaussian distribution. After obtaining the data at the new moment and the corresponding updated parameters θ′, β, Θ, t k The estimated remaining bearing life at time t is expressed as RUL k :

[0112] .

[0113] The corresponding probability distribution function is calculated as follows:

[0114] .

[0115] in, Calculates the probability distribution function.

[0116] In summary, the present application has the advantages of reducing the amount of calculation and strong real-time performance by analyzing and screening time domain features; it uses characteristic analysis and weighted calculation of features to screen features, and uses partial least squares method to fuse features, thereby improving the model's sensitivity to key degradation data and the interpretability of features. The use of a preset exponential model can make the present application scheme more consistent with the degradation process of bearings. The use of strong tracking filtering technology improves the exponential model's ability to track the state of different degradation stages and its generalization ability. In short, in terms of improving the accuracy of rolling bearing remaining service life prediction, the present application introduces characteristic analysis and weighted calculation of features and partial least squares method into the feature extraction of data, uses strong tracking filtering technology to correct prediction errors, enables the algorithm to adaptively adjust model parameters, ensures prediction accuracy, and obtains more accurate bearing remaining service life prediction results. The above settings can give full play to the advantages of each algorithm and form a complementary relationship, thereby forming a new rolling bearing remaining service life prediction method, which is of great significance and value to the practical application and research of bearings.

[0117] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a bearing remaining life prediction method is implemented.

[0118] Those skilled in the art will understand that Figure 3The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application and does not constitute a limitation on the computer device to which the solution of the present application is applied. A specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of the above-mentioned method embodiments when executing the computer program.

[0119] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0120] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0121] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0122] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0123] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0124] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0125] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for predicting the remaining life of a bearing, characterized in that: The bearing remaining life prediction method comprises: Obtain a set of vibration signal time series of the target bearing and perform feature extraction on each vibration signal time series to obtain time domain features; Based on robustness, monotonicity and predictability, the time domain features corresponding to all the vibration signal time series are screened to obtain a set of ready-to-use time domain features; Based on robustness, monotonicity and predictability, each of the standby time domain features in the standby time domain feature set is assigned a feature weight and weighted calculation is performed to obtain a standby feature vector; a plurality of the standby feature vectors constitute a standby feature vector set; Using partial least squares method, a multivariate regression operation is performed on the set of unused time domain features and the set of unused feature vectors to achieve feature fusion and obtain a comprehensive degradation feature quantity; Based on the comprehensive degradation characteristic quantities corresponding to the time series of each vibration signal, a preset exponential model is used to construct a prediction model. The Bayesian update and expectation maximization algorithms are used to update and estimate the parameters of the prediction model. The strong tracking filtering algorithm is used to correct the prediction error to obtain the remaining life prediction result of the bearing.

2. The method for predicting the remaining life of a bearing according to claim 1, characterized in that: The time domain features include a plurality of observed variables, and the time domain features are mean, root mean square value, root square amplitude, absolute mean value, skewness kurtosis, variance, maximum value, minimum value, and peak value.

3. The method for predicting the remaining life of a bearing according to claim 1, wherein: Based on robustness, monotonicity and predictability, the time domain features corresponding to all the vibration signal time series are screened to obtain a set of ready-to-use time domain features, including: The following formula is used to evaluate the feature robustness: ； Among them, Rob ( F i ) is the i Time domain features F i Robustness evaluation results of , f i (1) i The first observed variable in the time domain features, f i (2) i The second observed variable in the time domain characteristics, For the i Among the time domain features j observed variables, For the i Among the time domain features k observed variables, j=1,…,k; For Observed variables after smoothing; The following formula is used to evaluate the monotonicity of features: ； Among them, Mon( F i ) is the i Time domain features F i Monotonicity evaluation results of ; is a unit step function, f i ( j -1) is the i Among the time domain features j -1 observed variable; The following formula is used to evaluate the feature predictability: ； Among them, Pro( F i ) is the i Time domain features F i The predictive evaluation results; std() is the standard deviation function, mean() is the mean function; From the time domain features corresponding to all the vibration signal time series, the time domain features whose robustness evaluation results, monotonicity evaluation results and predictability evaluation results all meet the preset threshold range are screened out and marked as stand-by time domain features to obtain a set of stand-by time domain features.

4. The method for predicting the remaining life of a bearing according to claim 1, wherein: The step of assigning a feature weight to each of the standby time domain features in the standby time domain feature set based on robustness, monotonicity, and predictability includes: Based on robustness, monotonicity and predictability, the following matrix to be used is determined according to the set of time-domain features to be used: ； Among them, Mon( F 1) is the first standby time domain feature F 1 monotonicity evaluation results, Mon( F 2) is the second standby time domain feature F 2 monotonicity evaluation results, Mon( F n ) is the nth time domain feature to be used F n Monotonicity evaluation results of Rob( F 1) is the first standby time domain feature F 1 robustness evaluation results, Rob( F 2) is the second standby time domain feature F 2 robustness evaluation results, Rob( F n ) is the nth time domain feature to be used F n Robustness evaluation results of Pro( F 1) is the first standby time domain feature F 1's predictive evaluation results, Pro( F 2) is the second standby time domain feature F 2 predictive evaluation results, Pro( F n ) is the nth time domain feature to be used F n Predictive evaluation results; Using K to refer to each evaluation result, it is expressed as follows: ； in, K 11 Refers to Mon( F 1), K 1n Refers to Mon( F n ), K 21 Refers to Rob( F 1), K 2n Refers to Rob( F n ), K 31 Refers to Pro( F 1), K 3n Refers to Pro( F n ); The feature weight of the to-be-used time domain feature is calculated using the following formula: ； ； ； ； ； ； ； Where H is the number of types of evaluation results, h=1,…,H; z is the coefficient, is the ratio, K hn is the hth evaluation result of the nth time domain feature to be used, For K hn Sum by row, e hn 、 d n 、 are all intermediate parameters. For e hn Sum by row, is the information entropy, is the feature weight, is the initial value of the weight.

5. The method for predicting the remaining life of a bearing according to claim 1, wherein: A partial least squares method is used to perform a multivariate regression operation on the set of unused time domain features and the set of unused feature vectors to achieve feature fusion and obtain a comprehensive degradation feature quantity, including: Normalizing the set of time-domain features to be used and the set of feature vectors to be used to determine independent variable data and dependent variable data; wherein the independent variable data refers to the set of time-domain features to be used, and the dependent variable data refers to the comprehensive degradation feature quantity; determining latent variables based on the independent variable data; Performing regression analysis based on the latent variables and the dependent variable data to obtain regression coefficients; Based on the regression coefficient, orthogonalizing the independent variable data and the dependent variable data to obtain updated independent variable data and updated dependent variable data; Regression is performed based on the updated independent variable data and the updated dependent variable data to obtain a comprehensive degradation feature value.

6. The method for predicting the remaining life of a bearing according to claim 5, characterized in that: The formula for determining the independent variable data is: ； in, represents the mean value of the time domain feature F to be used, represents the standard deviation of the time domain feature F to be used, X’ represents the independent variable data, F i For the i Time domain features to be used; The formula for determining the dependent variable data is: ； ； in, Y’ represents the dependent variable data, is the feature vector to be used, Represents the weighted sum Y The mean of Represents the weighted sum Y The standard deviation of n represents the number of time domain features to be used; The calculation formula of the latent variable is: ； in, For the latent variables, For independent variable data The weight of The calculation formula of the regression coefficient is: ； in, c k is the regression coefficient, () T is the transpose of the matrix; The updated independent variable data is expressed as: ; The updated dependent variable data is expressed as: ; The comprehensive degradation feature quantity prediction model is: ; in, X (k+1) is the k+1 moment State quantity, Y (k+1) is the k+1 moment State quantity, is the comprehensive degradation characteristic.

7. The method for predicting the remaining life of a bearing according to claim 1, characterized in that: Based on the comprehensive degradation characteristics corresponding to the time series of each vibration signal, a preset exponential model is used to construct a prediction model. The Bayesian update and expectation maximization algorithms are used to update and estimate the parameters of the prediction model. The strong tracking filter algorithm is used to correct the prediction error to obtain the remaining life prediction results of the bearing, including: Based on the comprehensive degradation characteristics corresponding to the time series of each vibration signal, a preset exponential model is used to determine the bearing degradation path function; Determining a degradation amount function and a degradation prediction value calculation function based on the bearing degradation path function; Based on the degradation amount function and the degradation prediction value calculation function, the model parameters of the bearing degradation path function are updated and estimated using the Bayesian update and expectation maximization algorithms; A strong tracking filter algorithm is used to correct the prediction error to obtain the remaining life prediction result of the bearing.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for predicting the remaining life of a bearing according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting the remaining life of a bearing according to any one of claims 1 to 7 is implemented.

Images