Rolling bearing residual life prediction method based on EDM and Self-Attention-BiGRU

Abstract

The invention discloses a rolling bearing residual life prediction method based on EDM and Self-Attention-BiGRU, and particularly relates to the technical field of bearing residual life prediction, and the method comprises the following steps: collecting an original vibration signal from an experiment platform of an accelerated degradation rolling bearing; performing multilayer decomposition on the original vibration signal by using empirical mode decomposition to obtain a plurality of mode function components and corresponding residual components; on the basis of the modal function component, obtaining an original feature set by calculating dimensionless statistical features; based on the original feature set, performing evaluation and screening by using a self-attention mechanism to obtain an optimized feature data set, and dividing the optimized feature data set into training set data and test set data; based on the training set data, constructing and training a model based on a bidirectional gating circulation unit and a self-attention mechanism to obtain a model for predicting the residual life of the rolling bearing; and based on the model for predicting the residual life of the rolling bearing, verifying the accuracy of the model for predicting the residual life of the rolling bearing by using test set data.

Description

Technical Field

[0001] The present invention belongs to the technical field of bearing remaining life prediction, and particularly relates to a rolling bearing remaining life prediction method based on EDM and Self-Attention-BiGRU. Background Art

[0002] With the development of industrial technology, modern mechanical equipment is developing towards the direction of large-scale, comprehensive and complex. As an important component in rotating machinery, rolling bearings are widely used in industrial equipment. During the long-term operation of industrial mechanical equipment, it may be affected by overload, impact, etc., and it is inevitable to have a series of fault problems such as wear and fatigue. More seriously, these faults will accelerate the degradation of the bearings and have a serious impact on the safe and reliable operation of mechanical equipment. Therefore, how to effectively monitor the operating conditions and health status of machine equipment and accurately predict the remaining service life (RUL) of its important components is the focus of current research. Among the current methods for predicting the remaining life of bearings, the method based on physical models needs to establish an accurate simulation degradation model according to the bearing mechanism and operating conditions. However, due to the complex and changeable operating environment of the equipment, there are many difficulties in predicting the remaining service life of bearings by establishing a general failure model. In view of this, a feature extraction method based on empirical mode decomposition (EMD) and a rolling bearing remaining service life prediction method based on three modules of bidirectional gated recurrent unit (BiGRU) and self-attention mechanism (Self-Attention) are proposed. This method can not only extract more recognizable features from the original signal, but also ensure that the local features on different time scales are efficiently utilized. At the same time, for shaft movement data, it can easily extract effective features, meet the requirements of small data dimensions, make the prediction model tend to be simple, and the prediction results are accurate. Summary of the Invention

[0003] The purpose of the present invention is to provide a rolling bearing remaining life prediction method based on EDM and Self-Attention-BiGRU, aiming to solve the problems that it is difficult to extract effective features from the current rolling bearing vibration data, the data dimension is small and difficult to meet the requirements, the prediction model tends to be complex and causes high calculation cost and low prediction accuracy.

[0004] To achieve the above purpose, a rolling bearing remaining life prediction method based on EDM and Self-Attention-BiGRU is provided, and the prediction method includes the following steps:

[0005] S1. Collect the original vibration signal from the experimental platform of the accelerated degradation rolling bearing;

[0006] S2. Use empirical mode decomposition to perform multi-layer decomposition on the original vibration signal to obtain a plurality of mode function components and corresponding residual components;

[0007] S3. Based on the mode function components and corresponding residual components, calculate dimensionless statistical features to obtain an original feature set;

[0008] S4. Based on the original feature set, use the self-attention mechanism for evaluation and screening to obtain an optimized feature data set, and divide the optimized feature data set into training set data and test set data;

[0009] S5. Based on the training set data, construct and train a model based on a bidirectional gated recurrent unit and a self-attention mechanism to obtain a model for predicting the remaining life of a rolling bearing;

[0010] S6. Based on the model for predicting the remaining life of the rolling bearing, use the test set data to verify the accuracy of the model for predicting the remaining life of the rolling bearing.

[0011] Further, the method for collecting the original vibration signal from the experimental platform of the accelerated degradation rolling bearing in step S1 specifically includes:

[0012] The original vibration signal is measured by an acceleration sensor, and the measurement interval time and sampling time are fixed;

[0013] Perform wavelet processing and noise reduction processing on the original vibration signal to complete preliminary data preprocessing.

[0014] Further, the method for using empirical mode decomposition to perform multi-layer decomposition on the original vibration signal to obtain a plurality of mode function components and corresponding residual components in step S2 specifically includes:

[0015] For a given original time series signal x(t), its empirical mode decomposition process is as follows:

[0016] (1) Find the local maximum and minimum points of x(t), and use the cubic spline interpolation method to fit the sequences of maximum and minimum points respectively to obtain the upper envelope and the lower envelope; take the average of the upper and lower envelopes to obtain m1(t);

[0017] (2) Calculate the difference p1(t) between the original time series data x(t) and the average envelope m1(t), that is:

[0018] p1(t) = x(t) - m1(t)

[0019] (3) If p1(t) satisfies the conditions of the modal function components, it is the first modal function component; otherwise, take p1(t) as the new original time series data and repeat steps (1) and (2) until the conditions of the modal function components are satisfied;

[0020] (4) After obtaining the first modal function component p1(t), decompose it from the original time series x(t) to obtain the residual component u1(t), that is:

[0021] u1(t) = x(t) - p1(t)

[0022] (5) Take the residual component u1(t) as the new data input, and re - execute steps (1) to (5) to obtain a new residual component u2(t) and the second modal function component p2(t); and so on, until the residual component u k (t) of the k - th modal function component p k (t) is a constant or a monotonic function and cannot be decomposed further, and the entire empirical mode decomposition process is completed; at this time, the original time series x(t) can be expressed as:

[0023]

[0024] In the formula, u k (t) can be regarded as the trend or mean of x(t); q1(t), q2(t), …, q k (t) are the modal function components of x(t), representing the high - frequency components to low - frequency components of the original time series data;

[0025] That is, perform multi - layer empirical mode decomposition on the input original signal to obtain the corresponding intrinsic mode function components and the corresponding residual components.

[0026] Further, the method of obtaining the original feature set by calculating dimensionless statistical features based on the modal function components and the corresponding residual components in step S3 specifically includes:

[0027] A variety of the dimensionless statistical features are used to characterize each modal function component obtained after empirical mode decomposition processing. These dimensionless statistical features include: root - mean - square value, energy, Shannon entropy, peak factor, skewness, variance, mean, standard deviation, and kurtosis;

[0028] Based on the statistical analysis of a variety of dimensionless statistical features, capture the key features of the fault data to obtain a multi - dimensional original feature set.

[0029] Further, the method of evaluating and screening based on the original feature set using the self-attention mechanism in step S4 to obtain an optimized feature data set and dividing the optimized feature data set into training set data and test set data specifically includes:

[0030] The calculation of the self-attention mechanism is divided into two steps:

[0031] The first step: Calculate the attention weights between any two vectors in the input sequence;

[0032] The second step: Calculate the weighted average of the input sequence according to these attention weights;

[0033] The specific operation method is as follows:

[0034] Q = XW q

[0035] K = XW k

[0036] V = XW ν

[0037]

[0038] In the formula: Q is the query matrix; K is the key matrix; V is the value matrix; dim is the dimension of Q, K, and V.

[0039] The three evaluation feature indicators used for screening the original feature set are: correlation (Corr(f,t)), monotonicity (Mon(f)), and variance (Var(f)) to screen features that can effectively reflect the degradation process;

[0040] Using the central moving method, the feature f is regarded as a random process and divided into a trend part, f T representing the average trend and a random part, f R representing the residual, as shown in the following formula:

[0041] f(t k ) = f T (t k ) + f R (t k )

[0042] Among them: f(t k ) is the degradation feature at time t k ;

[0043] The three evaluation indicators are as follows:

[0044]

[0045] Among them, k is the total number of observations; h(t) is the step function;

[0046] The screening equation is as follows:

[0047] Ce = 0.2Var(f) + 0.5Mon(f) + 0.3Corr(f,t)

[0048] The weight value before each feature evaluation represents the importance of this index, and the range of the weight value is within (0,1);

[0049] The feature dataset after screening is finally divided into: a training set and a test set.

[0050] Furthermore, the method for constructing and training a model based on a bidirectional gated recurrent unit and a self-attention mechanism to obtain a rolling bearing remaining life prediction model based on the training set data in step S5 specifically includes:

[0051] (1) Design a network structure of a 14-layer bidirectional gated recurrent unit and a self-attention mechanism;

[0052] (2) For the network structure of the bidirectional gated recurrent unit and the self-attention mechanism, a self-attention mechanism is added after the bidirectional gated recurrent unit layer;

[0053] (3) During the training of the model based on the bidirectional gated recurrent unit and the self-attention mechanism, the mean square error is used as the loss function, and the Adam optimizer optimization algorithm is selected for model training;

[0054] (4) High-scoring features are selected as the input of the rolling bearing remaining life prediction model through the feature screening equation to obtain the rolling bearing remaining life prediction model.

[0055] Furthermore, the method for verifying the accuracy rate of the rolling bearing remaining life prediction model using the test set data based on the rolling bearing remaining life prediction model in step S6 specifically includes:

[0056] For the verification of the test set data, a univariate experimental method is used to verify different parameters.

[0057] Compared with the prior art, the beneficial effects of the present invention are:

[0058] A method for predicting the remaining life of a rolling bearing based on EDM and Self-Attention-BiGRU of the present invention: 1. Solve the problem of model information overload; 2. Improve the accuracy rate and robustness; 3. The prediction model is simple and the cost is low; 4. Facilitate the extraction of effective features of bearing vibration signals.

[0059] Based on the implementation manners provided in the above aspects, the present application can be further combined to provide more implementation manners. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0061] Figure 1 It is a flowchart for predicting the remaining life of the bearing of the present invention;

[0062] Figure 2 It is a structural diagram of the self-attention mechanism of the present invention;

[0063] Figure 3 It is the Self-Attention-BiGRU network structure of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] In order to enable those skilled in the art to better understand the solutions of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. If not specifically specified, the technical means used in the embodiments are conventional means well known to those skilled in the art.

[0065] The embodiments of the present invention provide a method for predicting the remaining life of a rolling bearing based on EDM and Self-Attention-BiGRU. Refer to Figure 1 , Figure 1 It is a flowchart of a method for predicting the remaining life of a rolling bearing based on EDM and Self-Attention-BiGRU provided by an embodiment of the present invention. The prediction method includes the following steps:

[0066] S1. Collect the original vibration signal from the experimental platform of the accelerated degradation rolling bearing;

[0067] S2. Use empirical mode decomposition to perform multi-layer decomposition on the original vibration signal to obtain a plurality of modal function components and corresponding residual components;

[0068] S3. Based on the modal function components and the corresponding residual components, calculate dimensionless statistical features to obtain an original feature set;

[0069] S4. Based on the original feature set, use the self-attention mechanism for evaluation and screening to obtain an optimized feature data set, and divide the optimized feature data set into training set data and test set data;

[0070] S5. Based on the training set data, construct and train a model based on bidirectional gated recurrent unit and self-attention mechanism to obtain a model for predicting the remaining life of rolling bearings;

[0071] S6. Based on the model for predicting the remaining life of rolling bearings, use the test set data to verify the accuracy of the model for predicting the remaining life of rolling bearings.

[0072] Based on a method for predicting the remaining life of rolling bearings based on EDM and Self-Attention-BiGRU provided by an embodiment of the present invention, first, collect the original vibration signals from the experimental platform of the accelerated degradation rolling bearings, and use empirical mode decomposition (EMD) for multi-layer decomposition to separate multiple modal function components and residual components, thereby improving the effectiveness of feature extraction; then, calculate dimensionless statistical features based on these components to form an original feature set, and use the self-attention mechanism to evaluate and screen out the optimized feature data set, enhancing the accuracy of feature selection; subsequently, divide the optimized feature data set into a training set and a test set, construct and train a model based on bidirectional gated recurrent unit and self-attention mechanism based on the training set, and finally obtain a model for predicting the remaining life of rolling bearings; finally, verify the accuracy of the model through the test set data to ensure its stability and reliability in practical applications; this method not only improves the accuracy of feature extraction and selection, but also optimizes the model training process, thereby realizing more accurate and reliable prediction of the remaining life of rolling bearings.

[0073] In one embodiment, the method for collecting the original vibration signals from the experimental platform of the accelerated degradation rolling bearings in step S1 specifically includes:

[0074] The original vibration signals are measured by an acceleration sensor, and the measurement interval time and sampling time are fixed; the original vibration signals are subjected to wavelet processing and noise reduction processing to complete preliminary data preprocessing.

[0075] Specifically, the collected data is based on an experimental platform for the accelerated aging of rolling bearings. The vibration data of the entire life cycle of the bearings in the horizontal and vertical directions is collected using acceleration sensors. The data is collected once every 10 s, the sampling duration is 0.1 s, and the sampling frequency is 25.6 kHz. That is, the length of the data collected each time in the horizontal and vertical directions is 2,560 points. At the same time, through experimental verification, the collected bearing degradation data can be better used for RUL prediction when the equipment is in a light load condition, that is, the data collected for model establishment in this method is for the equipment under light load conditions; the horizontal vibration signal is used as the original data input for EMD; and wavelet processing and noise reduction processing are used to preliminarily process the collected data after collection.

[0076] In one embodiment, the method for using empirical mode decomposition to perform multi-layer decomposition on the original vibration signal to obtain a plurality of mode function components and corresponding residual components in step S2 specifically includes:

[0077] Finding the extreme points of the original vibration signal; fitting the extreme points with an envelope line, performing upper envelope fitting on the maximum values and lower envelope fitting on the minimum values, and calculating the mean value of the envelope line; calculating the intermediate signal, where the intermediate signal is the difference between the original signal and the mean value; determining whether the intermediate signal satisfies the mode function component condition, and if not, returning to the first step; if the intermediate signal that satisfies the mode function component condition is used as the original signal, repeat the above steps to continue the decomposition.

[0078] Specifically, empirical mode decomposition (EDM) is an adaptive data processing method; it has obvious advantages in processing non-stationary and non-linear data; the basic idea of EMD mainly lies in being able to decompose a complex original signal into a group of independent, near-periodic intrinsic mode functions (IMFs); among them, the IMF must meet the following conditions: the number of zeros and extreme points of the IMF must be equal, or the difference between the two does not exceed 1; the upper and lower envelope lines respectively composed of the local maximum and minimum values of the IMF have a mean value of 0 at any time.

[0079] For a given original time series signal x(t), its EMD process is as follows:

[0080] 1) Find the local maximum and minimum points of x(t), and use the cubic spline interpolation method to fit the sequences of maximum and minimum points respectively to obtain the upper and lower envelope lines; take the average of the upper and lower envelope lines to obtain m1(t).

[0081] 2) Calculate the difference p1(t) between the original time series data x(t) and the average envelope line m1(t), that is:

[0082] p1(t) = x(t) - m1(t)

[0083] 3) If p1(t) satisfies the conditions of the IMF component, it is the first IMF component; otherwise, take p1(t) as the new original time series data and repeat steps 1) and 2) until the conditions of the IMF component are met.

[0084] 4) After obtaining the first IMF component p1(t), decompose it from the original time series x(t) to obtain the residual component u1(t), that is:

[0085] u1(t) = x(t) - p1(t)

[0086] 5) Take the residual component u1(t) as the new data input and repeat steps 1) to 5) to obtain a new residual component u2(t) and the second IMF component p2(t); and so on until the residual component u k (t) of the kth IMF component p k (t) is a constant or a monotonic function and cannot be decomposed further, and the entire EMD process is completed; at this time, the original time series x(t) can be expressed as:

[0087]

[0088] Among them, u k (t) can be regarded as the trend or mean of x(t); q1(t), q2(t), …, q k (t) are the IMF components of x(t), representing the high-frequency components to low-frequency components of the original time series data.

[0089] In summary, the above method performs multi-layer EDM on the input original signal to obtain the corresponding intrinsic mode function (IMF) and the corresponding residual component.

[0090] In one embodiment, the method for obtaining the original feature set by calculating dimensionless statistical features based on the modal function component and the corresponding residual component in step S3 specifically includes:

[0091] A variety of the dimensionless statistical features are used to characterize each modal function component obtained after empirical mode decomposition processing. These dimensionless statistical features include but are not limited to: root mean square value, energy, Shannon entropy, peak factor, skewness, variance, mean, standard deviation, and kurtosis, etc.; the signals to be processed are the components obtained after EDM processing, and the features of statistical fault data are extracted from these components, so as to obtain a multi-dimensional original feature set; the main formulas for feature extraction are as follows:

[0092] Root mean square:

[0093]

[0094] Energy: used to measure the overall energy of a signal in the time domain

[0095]

[0096] where x i is each sample value of the signal, and N is the number of samples of the signal.

[0097] Shannon entropy: an index used in information theory to measure the uncertainty or amount of information

[0098]

[0099] where p(x i ) is the probability that the random variable X takes the value x i , and log2 is the logarithm operation with base 2; the unit of Shannon entropy is usually bit, and when the base is 2, the unit of Shannon entropy can be expressed as bit.

[0100] Trigonometric functions: related to the periodic and oscillatory characteristics of the signal;

[0101] Peak: the maximum value that the signal reaches within a period of time

[0102] Peak = max(x1, x2,..., x N )

[0103] Variance: measures the degree of dispersion of the signal data

[0104]

[0105] where μ is the mean of the signal, x i is each sample value of the signal, and N is the number of samples of the signal.

[0106] Mean:

[0107]

[0108] where x i is the sample value of the signal, and N is the number of samples.

[0109] Standard deviation: the square root of the variance, indicating the degree of dispersion of the signal data

[0110]

[0111] where μ is the mean of the signal, x i is each sample value of the signal, and N is the number of samples of the signal.

[0112] Skewness: describes the degree of skewness of the signal data distribution

[0113]

[0114] Among them, μ is the mean of the signal, σ is the standard deviation of the signal, x i is each sample value of the signal, and N is the number of samples of the signal.

[0115] Kurtosis: Describes the peakedness of the signal data distribution

[0116]

[0117] Among them, μ is the mean of the signal, σ is the standard deviation of the signal, x i is each sample value of the signal, and N is the number of samples of the signal.

[0118] In one embodiment, the method of using the self-attention mechanism to evaluate and screen based on the original feature set in step S4 to obtain an optimized feature data set and dividing the optimized feature data set into training set data and test set data specifically includes:

[0119] The calculation of the self-attention mechanism is divided into two steps:

[0120] The first step: Calculate the attention weights between any two vectors in the input sequence;

[0121] The second step: Calculate the weighted average of the input sequence according to these attention weights;

[0122] The three evaluation feature indicators used for screening the original feature set are: correlation (Corr(f,t)), monotonicity (Mon(f)), and variance (Var(f));

[0123] The selected screening equation is as follows:

[0124] Ce = 0.2Var(f) + 0.5Mon(f) + 0.3Corr(f,t).

[0125] Specifically, this method uses the self-attention mechanism (Self-Attention) to screen data features. This mechanism is improved on the basis of the attention mechanism (Attention). At the same time, this mechanism can not only quickly screen out key information, reduce the attention to other irrelevant information, but also reduce the dependence on external information; by introducing the self-attention mechanism, not only the problem of model information overload is solved, but also the accuracy and stability of the network are improved; among them, the self-attention mechanism structure is as Figure 2 shown.

[0126] Figure 2 Among them: x i is the input sequence, i = 1, 2, 3,...t; ν i(i = 1, 2, 3…t) is the value vector generated from the input sequence; x ti (i = 1, 2, 3…t) is the result after the input sequence is operated with its respective q and k vectors and passed through the SoftMax function; b i (i = 1, 2, 3…t) is the result after the attention mechanism operation between the i-th position information in the input sequence and all position information.

[0127] The calculation of Self-Attention is divided into two steps. The first step: calculate the attention weights between any two vectors in the input sequence; The second step: calculate the weighted average of the input sequence according to these attention weights.

[0128] The specific operation method is as follows:

[0129] Q = XW q

[0130] K = XW k

[0131] V = XW ν

[0132]

[0133] In the formula, Q is the query matrix; K is the key matrix; V is the value matrix; dim is the dimension of Q, K and V.

[0134] The three evaluation feature indicators used in the present invention: correlation (Corr(f, t)), monotonicity (Mon(f)) and variance (Var(f)) are used to screen the features that can effectively reflect the degradation process.

[0135] Adopt the central moving method to regard the feature f as a random process, and divide it into a trend part, f T represents the average trend and the random part, f R represents the residual, as shown in the following formula:

[0136] f(t k ) = f T (t k ) + f R (t k )

[0137] Among them, f(t k ) is the degradation feature at time t k ;

[0138] The three evaluation indicators are as follows

[0139]

[0140] where k is the total number of observations; h(t) is the step function;

[0141] The screening equation is as follows:

[0142] Ce = 0.2Var(f) + 0.5Mon(f) + 0.3Corr(f, t)

[0143] The weight value before each feature evaluation represents the importance of the index, and the range of the weight value is within (0, 1).

[0144] The screened feature dataset is finally divided into: a training set and a test set.

[0145] In one embodiment, the method for constructing and training a model based on a bidirectional gated recurrent unit and a self-attention mechanism to obtain a model for predicting the remaining life of a rolling bearing based on the training set data in step S5 specifically includes:

[0146] (1) Design a network structure of a 14-layer bidirectional gated recurrent unit and a self-attention mechanism as Figure 3 shown;

[0147] (2) For the network structure of the bidirectional gated recurrent unit and the self-attention mechanism, a self-attention mechanism is added after the bidirectional gated recurrent unit layer;

[0148] (3) During the training process of the model based on the bidirectional gated recurrent unit and the self-attention mechanism, the mean squared error is used as the loss function, and the Adam optimizer optimization algorithm is selected for model training;

[0149] (4) High-scoring features are selected as the input of the rolling bearing remaining life prediction model through the feature screening equation to obtain the rolling bearing remaining life prediction model.

[0150] Specifically for model establishment, the Self-Attention-BiGRU network structure is as Figure 3 shown, with a total of 14 layers; high-scoring features are selected as the input of the RUL prediction model through the feature screening equation for predicting the remaining service life of the bearing; in the Figure 3 network structure, a self-attention mechanism is added after the BiGRU layer, reducing the attention to irrelevant information, reducing the dependence on external information, solving the problem of model information overload, and helping to improve the accuracy of RUL prediction.

[0151] We can roughly divide the life cycle of a bearing into four stages: the normal working stage, in which the amplitude of the vibration signal of the bearing is relatively low; the early degradation stage, in which the amplitude of the vibration signal gradually increases over time, and in this stage, the RUL can be predicted; in the mid-degradation stage, the amplitude of the vibration signal will continue to increase; and in the late degradation stage, when the amplitude of the signal further increases significantly, the bearing is considered to have completely failed. Based on this, the data in the training set is input into the model for prediction, and the corresponding remaining useful life curve (predicted value curve) is plotted.

[0152] During the training process of the model, the mean squared error (MSE) is used as the loss function, and the Adam optimizer optimization algorithm is selected to train the model.

[0153] In one embodiment, the method for verifying the accuracy of the model for predicting the remaining life of the rolling bearing based on the data of the test set in step S6 specifically includes:

[0154] For the verification of the test set data, a univariate experimental method is used to verify different parameters.

[0155] Specifically, for the test set verification, in order to verify the accuracy of the method based on EDM and Self-Attention BiGRU, according to the RUL prediction model established in step 5, the remaining life of the bearing is verified through the data in the test set; according to the data in the test set, it is input into the prediction model, and the corresponding remaining useful life curve (true value curve) is plotted. Finally, by comparing the predicted value curve and the true value curve, it is checked whether they can coincide, so as to conduct the verification.

[0156] The beneficial effects of the present invention: When predicting the RUL of a bearing, different hyperparameter values need to be set for verification. Therefore, the optimizer, learning rate, sample number, and iteration parameters need to be considered. This method selects the Adam optimizer optimization algorithm to train the model because it has the characteristics of simple algorithm implementation, high computational efficiency, and fast convergence speed. The learning rate and sample parameters need to be selected according to the data requirements because too large or too small learning rate and sample number will affect the accuracy of RUL prediction. Therefore, for the three hyperparameters of learning rate, sample number, and iteration times, in order to determine the optimal parameter values, the present invention uses a univariate experimental method to verify different parameters.

[0157] The above has described a specific embodiment of the present invention in detail, but the content is only the preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the application of the present invention should still fall within the scope covered by the patent of the present invention.

Claims

1. A method for predicting the remaining useful life of rolling bearings based on EDM and Self-Attention-BiGRU, characterized in that, The prediction method includes the following steps: S1. Collect the original vibration signals from the experimental platform of the accelerated degradation rolling bearing; S2. Use empirical mode decomposition to perform multi-layer decomposition on the original vibration signals to obtain multiple modal function components and corresponding residual components; S3. Based on the modal function components and corresponding residual components, obtain the original feature set by calculating dimensionless statistical features; S4. Based on the original feature set, use the self-attention mechanism for evaluation and screening to obtain the optimized feature data set, and divide the optimized feature data set into training set data and test set data; S5. Based on the training set data, construct and train a model based on the bidirectional gated recurrent unit and the self-attention mechanism to obtain a model for predicting the remaining life of the rolling bearing; S6. Based on the model for predicting the remaining life of the rolling bearing, use the test set data to verify the accuracy of the model for predicting the remaining life of the rolling bearing.

2. The method for predicting the remaining life of a rolling bearing based on EDM and Self-Attention-BiGRU according to claim 1, wherein The method for collecting the original vibration signals from the experimental platform of the accelerated degradation rolling bearing in step S1 specifically includes: The original vibration signals are measured by an acceleration sensor, and the measurement interval time and sampling time are fixed; Perform wavelet processing and noise reduction processing on the original vibration signals to complete preliminary data preprocessing.

3. The method for predicting the remaining life of a rolling bearing based on EDM and Self-Attention-BiGRU according to claim 1, wherein The method for using empirical mode decomposition to perform multi-layer decomposition on the original vibration signals to obtain multiple modal function components and corresponding residual components in step S2 specifically includes: For the given original time series signal x(t), its empirical mode decomposition process is as follows: (1) Find the local maximum and minimum points of x(t), and use the cubic spline interpolation method to fit the sequences of maximum and minimum points respectively to obtain the upper envelope and the lower envelope; take the average of the upper and lower envelopes to obtain m1(t); (2) Calculate the difference p1(t) between the original time series data x(t) and the average envelope m1(t), that is: p1(t) = x(t) - m1(t) (3) If p1(t) meets the conditions of the modal function component, it is the first modal function component; otherwise, take p1(t) as the new original time series data and repeat steps (1) and (2) until the conditions of the modal function component are met; (4) After obtaining the first modal function component p1(t), decompose it from the original time series x(t) to obtain the residual component u1(t), that is: u1(t) = x(t) - p1(t) (5) Take the residual component u1(t) as the new data input, and repeat steps (1) to (5) to obtain a new residual component u2(t) and the second mode function component p2(t); and so on, until the residual component u of the k-th mode function component p k (t) is a constant or a monotonic function and cannot be decomposed further, and the entire empirical mode decomposition process is completed; at this time, the original time series x(t) can be expressed as: k (t) is a constant or a monotonic function and cannot be decomposed further, and the entire empirical mode decomposition process is completed; at this time, the original time series x(t) can be expressed as: where u k (t) can be regarded as the trend or mean value of x(t); q1(t), q2(t), …, q k (t) are the modal function components of x(t), representing the high-frequency components to low-frequency components of the original time series data; That is, perform multi-layer empirical mode decomposition on the input original signal to obtain the corresponding intrinsic modal function components and corresponding residual components.

4. The remaining life prediction method of the rolling bearing based on EDM and Self-Attention-BiGRU according to claim 1, wherein The method for obtaining the original feature set by calculating dimensionless statistical features based on the modal function components and corresponding residual components in step S3 specifically includes: A variety of the dimensionless statistical features are used to characterize each modal function component obtained after empirical mode decomposition processing. These dimensionless statistical features include: root mean square value, energy, Shannon entropy, peak factor, skewness, variance, mean, standard deviation, and kurtosis; Based on the statistical analysis of multiple dimensionless statistical features, the key features of the fault data are captured, and a multi-dimensional original feature set is obtained.

5. The method for predicting the remaining life of a rolling bearing based on EDM and Self-Attention-BiGRU according to claim 1, wherein The method of evaluating and screening using the self-attention mechanism based on the original feature set in step S4 to obtain an optimized feature data set and dividing the optimized feature data set into training set data and test set data specifically includes: The calculation of the self-attention mechanism is divided into two steps: The first step: Calculate the attention weights between any two vectors in the input sequence; The second step: Calculate the weighted average of the input sequence according to these attention weights; The specific operation method is as follows: Q = XW q K = XW k V = XW ν In the formula: Q is the query matrix; K is the key matrix; V is the value matrix; dim is the dimension of Q, K, and V. The three evaluation feature indicators used for screening the original feature set are: correlation (Corr(f,t)), monotonicity (Mon(f)), and variance (Var(f)) to screen the features that can effectively reflect the degradation process; The feature f is regarded as a random process using the central moving method and is divided into a trend part, f T representing the average trend and a random part, f R representing the residual, as shown in the following formula: f(t k ) = f T (t k ) + f R (t k ) where: f(t k ) is the degradation characteristic of time t k ; The three evaluation indicators are as follows: Where: k is the total number of observations; h(t) is the step function; (Is h(t) the same as h?) The screening equation is as follows: Ce = 0.2Var(f) + 0.5Mon(f) + 0.3Corr(f,t) The weight value before each feature evaluation represents the importance of the indicator, and the range of the weight value is within (0,1); The screened feature data set is finally divided into: training set and test set.

6. The method for predicting the remaining life of a rolling bearing based on EDM and Self-Attention-BiGRU according to claim 1, wherein The method of constructing and training a model based on a bidirectional gated recurrent unit and self-attention mechanism to obtain a model for predicting the remaining life of a rolling bearing based on the training set data in step S5 specifically includes: (1) Design a 14-layer network structure of a bidirectional gated recurrent unit and self-attention mechanism; (2) For the network structure of the bidirectional gated recurrent unit and self-attention mechanism, a self-attention mechanism is added after the bidirectional gated recurrent unit layer; (3) Use the mean square error as the loss function during the training of the model based on the bidirectional gated recurrent unit and self-attention mechanism, and select the Adam optimizer optimization algorithm for model training; (4) Screen out the features with high scores through the feature screening equation as the input of the model for predicting the remaining life of the rolling bearing to obtain a model for predicting the remaining life of the rolling bearing.

7. The method for predicting the remaining life of a rolling bearing based on EDM and Self-Attention-BiGRU according to claim 1, wherein The method of verifying the accuracy of the model for predicting the remaining life of the rolling bearing based on the model for predicting the remaining life of the rolling bearing using the test set data in step S6 is specifically to verify the test set data using a univariate experimental method for different parameters.

Images