A method for bearing life prediction based on health indicators

Abstract

The application discloses a kind of based on new health index bearing life prediction method, specifically related to bearing residual life prediction technical field, comprising the following steps: S1, the vibration signal of rolling bearing is collected using vibration sensor, and the vibration signal is pretreated;S2, the health index of bearing is constructed from the multi-domain feature extracted from the vibration signal after pretreatment;S3, establish health index degradation model, set failure threshold, then according to the probability distribution of the time when the first failure threshold is reached by equipment degradation, determine the prediction of bearing residual life;The present application reduces the fluctuation caused by noise through completely self-adaptive noise ensemble empirical mode decomposition and kernel principal component analysis;Compared with the current residual service life prediction method, multi-domain features are extracted, so that the bearing health index constructed contains rich degradation information, and the accuracy of the prediction result is improved.

Description

Technical Field

[0001] This invention belongs to the field of bearing remaining life prediction technology, and particularly relates to a method for predicting bearing life based on health indicators. Background Technology

[0002] With the advancement of science and technology, mechanical equipment has become more complex and precise. Predictive and health management have become effective solutions for improving the reliability, safety, and maintainability of mechanical equipment. Therefore, research on motor bearing failure prediction is of great significance.

[0003] Previous research on bearing failure prediction mainly relied on a single health index to predict remaining service life. Current research has systematically summarized recent studies on rolling bearing failure prediction using a single health index. However, this approach has several drawbacks. First, it neglects fluctuations caused by noise. Second, the constructed single health index for bearings lacks rich degradation information.

[0004] For the reasons mentioned above, it is necessary to propose a new health index to predict bearing failure by extracting multi-domain features. Summary of the Invention

[0005] The purpose of this invention is to provide a method for predicting bearing life based on health indicators, which aims to solve the problems in the existing technology that relies on a single health indicator to predict the remaining service life of rolling bearings, ignoring the fluctuations caused by noise and the lack of rich degradation information in the single health indicator of the constructed bearing.

[0006] To achieve the above objectives, a method for predicting bearing life based on health indicators is provided, the prediction method comprising the following steps:

[0007] S1. Use a vibration sensor to collect the vibration signal of the rolling bearing, and preprocess the vibration signal;

[0008] S2. Extract multi-domain features from the preprocessed vibration signal to construct the bearing health index;

[0009] S3. A health index degradation model is established based on the nonlinear Wiener process, and a high-precision prediction of the remaining life of the bearing is achieved by combining the failure threshold.

[0010] Furthermore, in step 1, the preprocessing of the vibration signal is performed using continuous wavelet transform, expressed as follows:

[0011]

[0012] In the formula: Represents a one-dimensional vibration signal. This represents the complex conjugate operation. Scale factor The translation factor is... The mother wavelet is t, where t is time. The time-frequency representation used is TFR.

[0013] Furthermore, in step S2, the process of extracting multi-domain features from the preprocessed vibration signal to construct the bearing health index includes:

[0014] S2-1: Obtain the nonlinear health index HI of the bearing using fully adaptive noise ensemble empirical mode decomposition and kernel principal component analysis;

[0015] S2-2: Constructing multiple health indicators (HIs) in the dual-channel transformer network-convolutional block attention module (DCTN-CBAM) architecture based on the multi-domain characteristics of bearings;

[0016] S2-3: The constructed bearing multi-health index (HIs) will be used for first prediction time (FPT) detection.

[0017] Furthermore, the fully adaptive noise set empirical mode decomposition process described in step S2-1 includes:

[0018] (1) Construct an adaptive noise sequence and add it to the original signal to generate multiple sets of random test signals;

[0019] (2) Perform fully adaptive noise set empirical mode decomposition (CEEMD) on each set of random test signals to obtain a set of intrinsic mode function (IMF) functions;

[0020] (3) Combine and weight each set of intrinsic mode IMF functions to generate a total intrinsic mode IMF function;

[0021] (4) Perform fully adaptive noise set empirical mode decomposition (CEEMD) on the total intrinsic mode IMF function to obtain a new set of intrinsic mode IMF functions;

[0022] (5) Check whether the newly generated intrinsic mode IMF function meets the convergence condition. If it does not meet the convergence condition, return to step (1) to adjust the adaptive noise.

[0023] (6) If the convergence condition is met, repeat steps (2) to (5) for iteration until the number of IMF functions no longer increases.

[0024] Further, in step 2-1, the kernel principal component analysis process:

[0025] (1) Calculate the kernel matrix K:

[0026] ;

[0027] In the formula, and All of these are sample points from the dataset;

[0028] (2) Centering of the kernel matrix:

[0029]

[0030] In the formula, It is an all-one matrix used to eliminate the influence of the mean;

[0031] (3) Calculate the eigenvalues ​​and eigenvectors of the kernel matrix:

[0032] ;

[0033] In the formula, These are the eigenvalues ​​of the kernel matrix. It is the corresponding feature vector;

[0034] (4) Select the eigenvectors corresponding to the K largest eigenvalues ​​as the new reduction space.

[0035] Furthermore, the dual-channel transformer network-convolutional block attention module DCTN-CBAM architecture described in step S2-2 consists of a dual-channel transformer network and a convolutional block attention module. The dual-channel transformer network has one channel for processing time-domain and frequency-domain features and another channel for extracting degradation information from the reduced time-frequency representation (TFR). The convolutional block attention module includes a channel attention module and a spatial attention module.

[0036] Furthermore, the channel attention module is represented as follows:

[0037]

[0038] In the formula, , For the weight matrix, , The features are those obtained after average pooling and max pooling. As input features, For the feature mapping after the channel attention module, This refers to channel attention module operations on input features.

[0039] Furthermore, the spatial attention module is represented as follows:

[0040]

[0041] In the formula, This refers to performing spatial attention module operations on the input features. This represents a convolution operation with a kernel size of n. This represents the feature map of the stack.

[0042] Furthermore, the process of detecting the first prediction time (FPT) in steps S2-3 includes: determining the first prediction time (FPT) using multiple health indicators (HIs) of the bearing, and using 3σ to construct the monitoring interval for the health status. The first prediction time (FPT) is defined as follows:

[0043]

[0044] In the formula, Let m be the (t+i)th value of the health indicator HI. These represent the mean and standard deviation of the health status monitoring interval, respectively.

[0045] Further, in step S3, the process of establishing a health indicator degradation model, setting a failure threshold, and then determining the predicted remaining life of the bearing based on the probability distribution of the time when the equipment degradation first reaches the failure threshold includes:

[0046] (1) To address the nonlinear characteristics of the rolling bearing health index HI, a nonlinear Wiener process is used to characterize the degradation characteristics of HI; let {X(t), t > 0} be the degradation process of the health index HI, and the nonlinear Wiener process is defined as follows:

[0047]

[0048] In the formula, For drift parameters, Let be the diffusion parameter, and b be the unknown parameter in the nonlinear function of the drift coefficient. This represents the standard Brownian motion process that reflects the stochastic dynamics of the degradation process and the time-varying uncertainty inherent in the performance degradation process.

[0049] (2) A failure threshold is predefined. The life of the rolling bearing ends when its health index HI exceeds the failure threshold, and the remaining service life probability is predicted. The process is as follows:

[0050] First, define the failure threshold C, which is defined as follows:

[0051]

[0052] In the formula, The remaining useful life is defined as the time when the health indicator first exceeds the failure threshold C;

[0053] Then, the probability distribution of the time when the equipment degradation first reaches the failure threshold (i.e., the remaining lifetime) is calculated, and its expression is:

[0054]

[0055] In the formula, Here, b is the diffusion parameter, b is the unknown parameter in the nonlinear function of the drift coefficient, and C is the failure threshold. For drift parameters standard deviation For drift parameters The mean;

[0056] The remaining service life (RUL) of the rolling bearing at the kth observation time is represented by L. k Represented as:

[0057] ;

[0058] In the formula, Let k be the time point of the observation. The time of the kth observation The remaining service life at that time.

[0059] Compared with the prior art, the beneficial effects of the present invention are:

[0060] 1. This invention reduces fluctuations caused by noise by using fully adaptive noise set empirical mode decomposition and kernel principal component analysis.

[0061] 2. Compared with current methods for predicting remaining useful life, this method extracts multi-domain features, enabling the constructed bearing health index to contain rich degradation information and improving the accuracy of the prediction results.

[0062] Based on the implementation methods provided in the above aspects, this application can be further combined to provide more implementation methods. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 This is a flowchart of the bearing remaining life prediction process of the present invention;

[0065] Figure 2 Flowchart for constructing health indicators according to the present invention;

[0066] Figure 3 This is a diagram illustrating the processing steps of the convolutional attention module.

[0067] Figure 4This is a comparison chart of the prediction results for five rolling bearings of the same model under different operating conditions. Detailed Implementation

[0068] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention. Unless otherwise specified, the technical means used in the embodiments are conventional means well known to those skilled in the art.

[0069] This invention provides a method for predicting bearing life based on health indicators. See also... Figure 1 , Figure 1 This is a flowchart illustrating a method for predicting bearing life based on health indicators, provided in an embodiment of the present invention. First, vibration signals from the rolling bearing during operation are collected using vibration sensors. These signals are then processed using continuous wavelet transform. Multi-domain features are extracted from the processed vibration signals to construct the bearing's health indicators. Firstly, the nonlinear health indicator HI of the bearing is obtained using fully adaptive noise ensemble empirical mode decomposition and kernel principal component analysis. Subsequently, HIs are constructed within a DCTN-CBAM architecture based on the bearing's multi-domain features. Next, the constructed rolling bearing HIs are used for FPT detection. Finally, degradation modeling and remaining service life probability prediction are performed to obtain the final prediction result. The specific implementation steps are as follows:

[0070] Step 1: Collect vibration signals from the rolling bearing using a vibration sensor and preprocess the vibration signals.

[0071] The vibration signals collected by the vibration sensor during operation are processed using continuous wavelet transform, and the expression is as follows:

[0072]

[0073] In the formula: Represents a one-dimensional vibration signal. This represents the complex conjugate operation. Scale factor The translation factor is... The mother wavelet is t, where t is time. The TFR used.

[0074] Step 2: Extract multi-domain features from the preprocessed vibration signal to construct the bearing health index.

[0075] Since a single health indicator does not contain rich degradation information, a multi-domain feature extraction approach is adopted to calculate the time-domain, frequency-domain, and time-frequency-domain features of the vibration signal, forming a multi-dimensional feature vector. This constructs a novel health indicator to comprehensively characterize the bearing's condition.

[0076] The process of constructing health indicators: such as Figure 2 The specific process of constructing health indicators is illustrated below: First, fully adaptive noise ensemble empirical mode decomposition is used to adaptively decompose the preprocessed vibration signal to suppress noise interference. Combined with kernel principal component analysis, high-dimensional mapping and dimensionality reduction of nonlinear degradation features are performed to construct a nonlinear health indicator HI. Subsequently, using the DCTN-CBAM architecture, the time-domain statistical features and frequency-domain spectral features of the vibration signal are extracted through parallel convolutional channels, and composite health indicators HIs are constructed using channel attention modules and spatial attention modules. Finally, FPT detection is used to capture the critical point where the health indicator exceeds the preset failure threshold, and 3σ is used to construct the monitoring interval of the health status.

[0077] Specifically, the fully adaptive noise set empirical mode decomposition process includes:

[0078] (1) Construct an adaptive noise sequence and add it to the original signal to generate multiple sets of random test signals;

[0079] (2) Perform fully adaptive noise set empirical mode decomposition (CEEMD) on each set of random test signals to obtain a set of intrinsic mode function (IMF).

[0080] (3) Combine and weight each group of intrinsic mode IMF functions to generate a total intrinsic mode IMF function.

[0081] (4) Perform fully adaptive noise set empirical mode decomposition (CEEMD) on the total intrinsic mode IMF function to obtain a new set of intrinsic mode IMF functions.

[0082] (5) Check whether the newly generated intrinsic mode IMF function meets the convergence condition. If it does not meet the convergence condition, return to step (1) to adjust the adaptive noise.

[0083] (6) If the convergence condition is met, repeat steps (2) to (5) for iteration until the number of IMF functions no longer increases.

[0084] Specifically, the process of kernel principal component analysis:

[0085] (1) Calculate the kernel matrix K:

[0086]

[0087] In the formula, and All of these are sample points from the dataset.

[0088] (2) Centering of the kernel matrix:

[0089]

[0090] In the formula, It is an all-one matrix used to eliminate the influence of the mean.

[0091] (3) Calculate the eigenvalues ​​and eigenvectors of the kernel matrix:

[0092] .

[0093] In the formula, These are the eigenvalues ​​of the kernel matrix. It is the corresponding feature vector.

[0094] (4) Select the eigenvectors corresponding to the K largest eigenvalues ​​as the new reduction space.

[0095] Specifically, the dual-channel transformer network-convolutional block attention module (DCTN-CBAM) architecture consists of a dual-channel transformer network and a convolutional block attention module. The dual-channel transformer network has one channel for processing time-domain and frequency-domain features, and another channel for extracting degradation information from the reduced time-frequency representation (TFR). The convolutional block attention module includes a channel attention module and a spatial attention module. The processing procedure of the convolutional attention module is as follows: Figure 3 As shown.

[0096] Specifically, the channel attention module is represented as follows:

[0097]

[0098] In the formula, , For the weight matrix, , The features are those obtained after average pooling and max pooling. As input features, Feature mapping after the channel attention module.

[0099] Specifically, the spatial attention module is represented as follows:

[0100]

[0101] In the formula, This refers to performing spatial attention module operations on the input features. This represents a convolution operation with a kernel size of n. This represents the feature map of the stack.

[0102] The FPT testing process includes: determining the FPT based on the bearing HIs, and using 3σ to construct the health status monitoring interval. The FPT is defined as follows:

[0103]

[0104] In the formula, Let m be the (t+i)th value of the health indicator HI. These represent the mean and standard deviation of the health status monitoring interval, respectively.

[0105] Step 3: Establish a health index degradation model based on the nonlinear Wiener process, and combine it with the failure threshold to achieve high-precision prediction of the remaining life of the bearing.

[0106] (1) To address the nonlinear characteristics of the rolling bearing health index HI, a nonlinear Wiener process is used to characterize the degradation characteristics of HI; let {X(t), t > 0} be the degradation process of the health index HI, and the nonlinear Wiener process is defined as follows:

[0107]

[0108] In the formula, For drift parameters, Let be the diffusion parameter, and b be the unknown parameter in the nonlinear function of the drift coefficient. This represents the standard Brownian motion process that reflects the stochastic dynamics of the degradation process and the time-varying uncertainty inherent in the performance degradation process.

[0109] (2) A failure threshold is predefined. The life of the rolling bearing ends when its health index HI exceeds the failure threshold, and the probabilistic remaining service life is predicted. The process is as follows:

[0110] First, define the failure threshold C, which is defined as follows:

[0111]

[0112] In the formula, The remaining service life is defined as the time when the health indicator first exceeds the failure threshold C;

[0113] Then, the probability distribution of the time when the equipment degradation first reaches the failure threshold (i.e., the remaining lifetime) is calculated, and its expression is:

[0114]

[0115] In the formula, Here, b is the diffusion parameter, b is the unknown parameter in the nonlinear function of the drift coefficient, and C is the failure threshold. For drift parameters standard deviation For drift parameters The mean.

[0116] The RUL of the rolling bearing at the kth observation time is L. k Represented as:

[0117] ;

[0118] In the formula, Let k be the time point of the observation. The time of the kth observation The remaining service life at that time.

[0119] To accurately evaluate the predictive accuracy of the proposed novel health indicator, life expectancy prediction was conducted using both the novel health indicator and a single health indicator in the experiment, and the mean absolute error was selected as the indicator of the accuracy of the prediction model.

[0120] Since horizontal vibration signals contain richer degradation information, vibration sensors are installed in the horizontal direction to sample the vibration signals. Five rolling bearings of the same model are run under different operating conditions, and two types of health indicators are used to predict their performance. The mean absolute error is calculated to measure the difference between the predicted and actual values.

[0121] The five operating conditions are as follows:

[0122] Operating conditions Voltage (V) Load (N) Operating Condition 1 100 0 Operating Condition 2 100 25 Operating Condition 3 100 50 Operating Condition 4 100 75 Operating Condition 5 100 100

[0123] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for predicting bearing life based on health indicators, characterized in that, The prediction method includes the following steps: S1. Use a vibration sensor to collect the vibration signal of the rolling bearing, and preprocess the vibration signal; S2. Extract multi-domain features from the preprocessed vibration signal to construct the health index of the bearing; S3. Establish a health indicator degradation model, set a failure threshold, and then determine the prediction of the remaining life of the bearing based on the probability distribution of the time when the equipment degradation first reaches the failure threshold. Step S2, the process of extracting multi-domain features from the preprocessed vibration signal to construct the bearing health index, includes: S2-1: Obtain the nonlinear health index HI of the bearing using fully adaptive noise ensemble empirical mode decomposition and kernel principal component analysis; S2-2: Constructing multiple health indicators (HIs) in the dual-channel transformer network-convolutional block attention module (DCTN-CBAM) architecture based on the multi-domain characteristics of bearings; S2-3: The constructed bearing multi-health index (HIs) will be used for first prediction time (FPT) detection.

2. The method for predicting bearing life based on health indicators according to claim 1, characterized in that, In step 1, the vibration signal is preprocessed using continuous wavelet transform, as expressed by: ； In the formula: Represents a one-dimensional vibration signal. This represents the complex conjugate operation. Scale factor The translation factor is... The mother wavelet is t, where t is time. The time-frequency representation used is TFR.

3. The method for predicting bearing life based on health indicators according to claim 1, characterized in that, The fully adaptive noise set empirical mode decomposition process described in step S2-1 includes: (1) Construct an adaptive noise sequence and add it to the original signal to generate multiple sets of random test signals; (2) Perform fully adaptive noise set empirical mode decomposition (CEEMD) on each set of random test signals to obtain a set of intrinsic mode function (IMF) functions; (3) Combine and weight each set of intrinsic mode IMF functions to generate a total intrinsic mode IMF function; (4) Perform fully adaptive noise set empirical mode decomposition (CEEMD) on the total intrinsic mode IMF function to obtain a new set of intrinsic mode IMF functions; (5) Check whether the newly generated intrinsic mode IMF function meets the convergence condition. If it does not meet the convergence condition, return to step (1) to adjust the adaptive noise. (6) If the convergence condition is met, repeat steps (2) to (5) for iteration until the number of IMF functions no longer increases.

4. The method for predicting bearing life based on health indicators according to claim 1, characterized in that, The kernel principal component analysis process described in step S2-1: (1) Calculate the kernel matrix K: ； In the formula, and All of these are sample points from the dataset; (2) Centering of the kernel matrix: ； In the formula, It is an all-one matrix used to eliminate the influence of the mean; (3) Calculate the eigenvalues ​​and eigenvectors of the kernel matrix: ； In the formula, These are the eigenvalues ​​of the kernel matrix. It is the corresponding feature vector; (4) Select the eigenvectors corresponding to the K largest eigenvalues ​​as the new reduction space.

5. The method for predicting bearing life based on health indicators according to claim 1, characterized in that, The dual-channel transformer network-convolutional block attention module DCTN-CBAM architecture described in step S2-2 consists of a dual-channel transformer network and a convolutional block attention module. The dual-channel transformer network has one channel for processing time-domain and frequency-domain features and another channel for extracting degradation information from the reduced time-frequency representation (TFR). The convolutional block attention module includes a channel attention module and a spatial attention module.

6. The method for predicting bearing life based on health indicators according to claim 5, characterized in that, The channel attention module is represented as follows: ； In the formula, , For the weight matrix, , The features are those obtained after average pooling and max pooling. As input features, For the feature mapping after the channel attention module, This refers to channel attention module operations on input features.

7. The method for predicting bearing life based on health indicators according to claim 5, characterized in that, The spatial attention module is represented as follows: ； In the formula, This refers to performing spatial attention module operations on the input features. This represents a convolution operation with a kernel size of n. This represents the feature map of the stack.

8. The method for predicting bearing life based on health indicators according to claim 1, characterized in that, The process of detecting the first prediction time (FPT) in steps S2-3 includes: determining the first prediction time (FPT) using multiple health indicators (HIs) of the bearing, and constructing a monitoring interval for the health status using 3σ. The first prediction time (FPT) is defined as follows: ； In the formula, Let m be the (t+i)th value of the health indicator HI. These represent the mean and standard deviation of the health status monitoring interval, respectively.

9. The method for predicting bearing life based on health indicators according to claim 1, characterized in that, In step S3, the process of establishing a health indicator degradation model, setting a failure threshold, and then determining the predicted remaining life of the bearing based on the probability distribution of the time when the equipment degradation first reaches the failure threshold includes: (1) To address the nonlinear characteristics of the rolling bearing health index HI, a nonlinear Wiener process is used to characterize the degradation characteristics of HI; let {X(t), t > 0} be the degradation process of the health index HI, and the nonlinear Wiener process is defined as follows: ； In the formula, For drift parameters, Let be the diffusion parameter, and b be the unknown parameter in the nonlinear function of the drift coefficient. This represents the standard Brownian motion process that reflects the stochastic dynamics of the degradation process and the time-varying uncertainty inherent in the performance degradation process. (2) A failure threshold is predefined. The life of the rolling bearing ends when its health index HI exceeds the failure threshold, and the remaining service life probability is predicted. The process is as follows: First, define the failure threshold C, which is defined as follows: ； In the formula, The remaining service life is defined as the time when the health indicator first exceeds the failure threshold C; Then, the probability distribution of the time when the equipment degradation first reaches the failure threshold is calculated, and its expression is: ； In the formula, Here, b is the diffusion parameter, b is the unknown parameter in the nonlinear function of the drift coefficient, and C is the failure threshold. For drift parameters standard deviation For drift parameters The mean; The remaining service life (RUL) of the rolling bearing at the kth observation time is represented by L. k Represented as: ； In the formula, Let k be the time point of the observation. The time of the kth observation The remaining service life at that time.

Images