Bearing life prediction method and device, electronic equipment and storage medium

Abstract

The embodiment of the application provides a bearing life prediction method and device, electronic equipment and storage medium, including at least one target factor given according to a bearing prediction model, performing feature extraction on the detection signal of the to-be-tested bearing, obtaining the signal feature corresponding to each target factor of the detection signal, using the bearing prediction model, performing prediction according to the signal feature corresponding to each target factor of the detection signal, and determining the bearing life of the to-be-tested bearing. Therefore, the application can accurately predict the bearing life, improve the operation reliability of the mechanical system, and reduce the maintenance cost.

Description

Technical Field

[0001] This application relates to the field of signal analysis technology, and in particular to a bearing life prediction method, device, electronic device, and computer storage medium. Background Technology

[0002] Rolling bearings are widely used supporting components in rotating machinery and are also among the most vulnerable parts. Their health directly affects the reliable and stable operation of the equipment. According to spectrum analyzer test results, rolling bearings account for 60% of all mechanical failures. Therefore, predicting the remaining life of rolling bearings is of great significance for improving the overall operational reliability of mechanical systems and reducing maintenance costs.

[0003] Current bearing life prediction schemes mainly suffer from the following problems: there is a lack of bearing life data, and key parameters are difficult to measure, making it difficult to predict bearing life; bearing trend prediction results are inaccurate; and qualitative analysis results are inconsistent with model quantitative analysis results.

[0004] In view of this, there is an urgent need for a bearing life prediction scheme to improve the various problems existing in the current technology. Summary of the Invention

[0005] To address the aforementioned problems, embodiments of this application provide a bearing life prediction method, apparatus, electronic device, and computer storage medium, which at least partially solve the problems described above.

[0006] According to a first aspect of the present application, a bearing life prediction method is provided, comprising: performing feature extraction on a detection signal of a bearing under test based on at least one target factor given by a bearing prediction model, and obtaining signal features of the detection signal corresponding to each target factor; and using the bearing prediction model to perform prediction based on the signal features of the detection signal corresponding to each target factor, and determining the bearing life of the bearing under test.

[0007] Optionally, the bearing prediction model is trained by: performing feature extraction on the sample signal based on multiple candidate factors related to the peak vibration energy to obtain the signal features of the sample signal corresponding to each candidate factor; determining the true peak vibration energy of the sample signal based on the vibration acceleration data of the sample signal; and training the bearing prediction model based on the true peak vibration energy and the signal features of at least one target factor determined from each candidate factor to obtain a trained bearing prediction model.

[0008] Optionally, the candidate factors include at least one time-domain factor and at least one frequency-domain factor; wherein, the signal characteristics of the at least one time-domain factor include at least one of the following: waveform factor characteristics, root mean square characteristics, kurtosis characteristics, kurtosis index characteristics, and margin index characteristics; and the signal characteristics of the at least one frequency-domain factor include at least one of the following: bearing characteristic frequency characteristics and bearing sideband energy ratio characteristics.

[0009] Optionally, training the bearing prediction model based on the actual vibration peak energy and the signal characteristics of at least one target factor determined from each candidate factor to obtain a trained bearing prediction model includes: determining an original sequence based on the vibration acceleration data in the sample signal, and constructing a factor sequence for each candidate factor based on the signal characteristics of the sample signal corresponding to each candidate factor; a target factor determination step, performing correlation calculation based on the original sequence and the factor sequence of each candidate factor to determine at least one target factor from each candidate factor; using the bearing prediction model to perform prediction based on the signal characteristics of each target factor to obtain the predicted vibration peak energy of the sample signal; obtaining the residual result of the bearing prediction model based on the actual vibration peak energy and the predicted vibration peak energy; if the residual result is not satisfied with the given training termination condition, updating the factor sequence of each candidate factor, and returning to execute the target factor determination step until the residual result satisfies the training termination condition.

[0010] Optionally, the step of determining the original sequence based on the vibration acceleration data in the sample signal and constructing the factor sequence of each candidate factor based on the signal characteristics of the sample signal corresponding to each candidate factor includes: obtaining the original sequence of the sample signal based on the vibration acceleration data of the sample signal corresponding to each sampling point; and performing an accumulation calculation based on the signal characteristics of each sampling point corresponding to each candidate factor to obtain the factor sequence of each candidate factor.

[0011] Optionally, the target factor determination step includes: determining a candidate factor as the current factor; performing a correlation calculation on the current factor based on the original sequence and the factor sequence of the current factor to determine the correlation value of the current factor; determining the current factor whose correlation value is greater than a given correlation threshold as the target factor; returning to the step of determining a candidate factor as the current factor until all candidate factors are determined as the current factor;

[0012] Optionally, the correlation threshold is 0.55.

[0013] Optionally, the step of performing correlation calculation on the current factor based on the original sequence and the factor sequence of the current factor to determine the correlation value of the current factor includes: using a correlation coefficient conversion formula to calculate the correlation coefficient value of the current factor based on the original sequence and the factor sequence of the current factor; and using a correlation conversion formula to calculate the correlation value of the current factor based on the correlation coefficient value of the current factor.

[0014] The correlation coefficient conversion formula is expressed as follows:

[0015]

[0016] Wherein, the ξ i (k) represents the correlation coefficient value of the feature signal at the k-th sampling point of the i-th candidate factor; the X (0) (k) represents the original sequence of vibration acceleration data corresponding to k sampling points of the sample signal. A factor sequence representing the signal characteristics of k sampling points of the i-th candidate factor; where ρ is the weight value;

[0017] The correlation conversion formula is expressed as follows:

[0018]

[0019] Wherein, the r i This represents the correlation value of the i-th candidate factor, where N is the total number of sampling points k.

[0020] Optionally, the step of using the bearing prediction model to perform prediction based on the signal characteristics of each target factor to obtain the predicted vibration peak energy of the sample signal includes: constructing an ordinary differential equation based on the original sequence and the neighbor mean and other complete sequences determined based on the factor sequence of each target factor, and solving the differential equation parameters in the ordinary differential equation; substituting the solved differential equation parameters into the prediction formula obtained by transforming the ordinary differential equation to obtain the predicted vibration peak energy of the current factor.

[0021] Optionally, the ordinary differential equation is expressed as:

[0022]

[0023] Wherein, X (0) (k) represents the original sequence, where k represents the k-th sampling point, and Z (1) (k) represents the neighbor mean equal whole sequence, and X i (1)(k) represents the factor sequence of signal features of k sampling points of the i-th target factor, where M is the total number of target factors, and A and b i The parameters of the differential equation to be solved;

[0024] The prediction formula obtained from the transformation of the ordinary differential equation is expressed as follows:

[0025]

[0026] Among them, the This indicates the predicted peak vibration energy corresponding to the (k+1)th sampling point of the sample signal.

[0027] Optionally, the method includes: acquiring vibration acceleration data corresponding to each sampling point of the sample signal, and determining the true peak vibration energy corresponding to each sampling point of the sample signal.

[0028] Optionally, obtaining the residual result of the bearing prediction model based on the actual vibration peak energy and the predicted vibration peak energy includes: obtaining the residual value of the bearing prediction model corresponding to each sampling point based on the difference between the actual vibration peak energy and the predicted vibration peak energy corresponding to the same sampling point of the sample signal; and determining the residual ratio of the bearing prediction model based on the residual value of the bearing prediction model corresponding to the same sampling point and the predicted vibration peak energy.

[0029] Optionally, if the residual result is not satisfied with the given training termination condition, updating the factor sequence of each candidate factor and returning to execute the target factor determination step until the residual result satisfies the training termination condition includes: if the residual ratio is not greater than a given residual ratio threshold, reconstructing the factor sequence of each candidate factor according to the given weakened neighborhood mean weight, and returning to execute the target factor determination step until the residual ratio is greater than the residual ratio threshold; wherein the residual ratio threshold is 0.9.

[0030] Optionally, the method further includes: performing a posterior error test on the bearing prediction model based on the residual value of the bearing prediction model, the residual mean determined based on the residual value, and the residual standard deviation, to obtain a small probability error value of the bearing prediction model; if the small probability error value is not less than the posterior error threshold, reconstructing the factor sequence of each candidate factor according to the weakened neighborhood mean weight, and returning to execute the target factor determination step until the small probability error value is less than the posterior error threshold; wherein, the posterior error threshold is set to 0.05.

[0031] Optionally, the method further includes: determining the bearing peak energy warning value based on the training results of training the bearing prediction model using multiple sample signals.

[0032] Optionally, the step of using the bearing prediction model to perform prediction based on the signal characteristics of the detection signal corresponding to each target factor to determine the bearing life of the bearing under test includes: using the bearing prediction model to perform prediction based on the signal characteristics of the current sampling point of the detection signal corresponding to each target factor to obtain the predicted vibration peak energy of the detection signal corresponding to the predicted sampling point, wherein the predicted sampling point is a sampling point following the current sampling point; and determining the remaining bearing life of the bearing under test at the predicted sampling point based on the predicted vibration peak energy of the predicted sampling point and the bearing peak energy warning value.

[0033] Optionally, the method further includes: acquiring the vibration acceleration data of the bearing under test in real time to obtain the detection signal of the bearing under test.

[0034] According to a second aspect of the embodiments of this application, a bearing life prediction device is provided, comprising: a feature extraction module, configured to perform feature extraction on a detection signal of a bearing under test based on at least one target factor given by a bearing prediction model, and obtain signal features of the detection signal corresponding to each target factor; and a bearing prediction model, configured to perform prediction based on the signal features of the detection signal corresponding to each target factor, and determine the bearing life of the bearing under test.

[0035] According to a third aspect of the present application, an electronic device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; the memory is used to store at least one executable instruction, wherein the executable instruction causes the processor to perform the operation corresponding to the bearing life prediction method described in the first aspect.

[0036] According to a fourth aspect of the embodiments of this application, a computer storage medium is provided, on which a computer program is stored, which, when executed by a processor, can implement the bearing life prediction method described in the first aspect above.

[0037] The bearing life prediction schemes provided in the embodiments of this application perform feature extraction on the detection signals of the bearing under test based on the target factors given in the bearing prediction model, so as to obtain various signal features that are highly correlated with the bearing life, and perform prediction. This can obtain bearing life prediction results with high accuracy and realize real-time prediction analysis of bearing life. In addition, it can be quickly and automatically applied to different application fields to meet the life prediction needs of different types of mechanical bearings.

[0038] The bearing life prediction schemes provided in the various embodiments of this application use vibration peak energy of different cycles to train the bearing prediction model. Since vibration peak energy can not only characterize the characteristics of signal changes of typical rotating equipment with the cycle, but also is very sensitive to bearing damage such as wear, fatigue spalling and other anomalies, it can effectively track the wear of the bearing throughout its entire life cycle to the greatest extent, thereby improving the accuracy of the bearing life prediction results of the bearing prediction model.

[0039] The bearing life prediction schemes provided in the embodiments of this application combine time-domain and frequency-domain features in the sample signal to perform bearing life prediction based on features in multiple dimensions, thereby improving the accuracy and objectivity of the prediction results.

[0040] The bearing life prediction schemes provided in the embodiments of this application construct a factor sequence of candidate factors through an accumulation algorithm to perform target factor screening and bearing prediction model training. The factor sequence generated by the accumulation algorithm can weaken the volatility and randomness of random sequences and support modeling analysis with small samples, solving the limitation of insufficient sample size. It can meet the model prediction analysis requirements under small sample training set conditions. At the same time, it can also obtain the changing patterns between features and generate a strongly regular data sequence to make up for the lack of features in the small sample training set and improve the model prediction accuracy.

[0041] The bearing life prediction schemes provided in the various embodiments of this application can ensure the effectiveness between the target factors and the peak energy of bearing vibration by performing grey relational analysis, so that the bearing prediction model can accurately predict the peak energy of the bearing.

[0042] The bearing life prediction schemes provided in the embodiments of this application utilize the complete sequence, such as the neighbor mean, obtained based on the sequence of each factor corresponding to each target factor to construct the ordinary differential equation of the bearing prediction model. This can accurately predict the peak energy of vibration, which not only helps to improve the accuracy of bearing life prediction results, but also allows the prediction model constructed by this method to be quickly and automatically applied to different application fields to meet the life prediction needs of different types of mechanical bearings.

[0043] The bearing life prediction schemes provided in the embodiments of this application obtain the residual values ​​of the bearing prediction model corresponding to each sampling point based on the difference between the actual peak vibration energy and the predicted peak vibration energy of the sample signal at each sampling point. By using this residual verification method to train the bearing prediction model, a bearing prediction model with better prediction performance can be obtained, and the robustness of the model prediction results can be improved.

[0044] The bearing life prediction schemes provided in the embodiments of this application can further improve the accuracy and robustness of the model prediction results by performing a post-test on the test results of the bearing prediction model.

[0045] The bearing life prediction schemes provided in the various embodiments of this application calculate the bearing peak energy warning value based on the training dataset, and accurately predict the remaining bearing life of the bearing under test, so as to improve the overall operational reliability of the mechanical system and reduce maintenance costs. Attached Figure Description

[0046] The accompanying drawings are intended only to illustrate and explain this application and do not limit the scope of this application.

[0047] Figure 1 This is a flowchart illustrating the bearing life prediction method as an exemplary embodiment of this application.

[0048] Figure 2 This is a flowchart illustrating a bearing life prediction method as another exemplary embodiment of this application.

[0049] Figure 3 This is a flowchart illustrating a bearing life prediction method as another exemplary embodiment of this application.

[0050] Figure 4 This is a flowchart illustrating a bearing life prediction method as another exemplary embodiment of this application.

[0051] Figure 5 This is a flowchart illustrating a bearing life prediction method as another exemplary embodiment of this application.

[0052] Figure 6 This is a flowchart illustrating a bearing life prediction method as another exemplary embodiment of this application.

[0053] Figure 7 This is a flowchart illustrating a bearing life prediction method as another exemplary embodiment of this application.

[0054] Figure 8 This is a structural block diagram of a bearing life prediction device, which is an exemplary embodiment of this application.

[0055] Figure 9 This is a structural block diagram of an electronic device that is an exemplary embodiment of this application.

[0056] Explanation of reference numerals in the attached figures:

[0057] 800, Bearing life prediction device; 802, Feature extraction module; 804, Bearing prediction model; 900, Electronic equipment; 902, Processor; 904, Communication interface; 906, Memory; 908, Communication bus; 910, Computer program. Detailed Implementation

[0058] To provide a clearer understanding of the technical features, objectives, and effects of the embodiments of this application, the specific implementation methods of the embodiments of this application will now be described with reference to the accompanying drawings.

[0059] Rolling bearings are widely used support components in rotating machinery and are also among the most vulnerable parts. Therefore, predicting the remaining life of rolling bearings is of great significance for improving the overall operational reliability of rotating machinery systems and reducing maintenance costs.

[0060] Current bearing life prediction methods largely rely on bearing mechanisms, while the basic approach to monitoring mechanical vibration is to use condition monitoring systems (such as Siemens' CMS system) or offline analysis using experienced vibration analyzers. These methods have several limitations: First, current bearing life prediction schemes are not real-time, making timely early warning and online analysis difficult. Second, different types of machines have different mechanical characteristics; traditional condition monitoring systems are too general for typical machine vibration analysis, while vibration analyzers are too expensive, preventing current prediction methods from meeting the needs of different types of mechanical bearings. Furthermore, current bearing prediction schemes are mostly based on theoretical concepts and mechanisms, making it difficult to quickly and automatically adapt to different application areas.

[0061] In view of this, this application provides a bearing life prediction scheme that can solve the various problems existing in the prior art.

[0062] Figure 1 The processing flow of the bearing life prediction method according to an exemplary embodiment of this application is illustrated. As shown in the figure, this embodiment mainly includes the following processing steps:

[0063] Step S102: Based on at least one target factor given by the bearing prediction model, perform feature extraction on the detection signal of the bearing under test to obtain the signal features of the detection signal corresponding to each target factor.

[0064] Optionally, the vibration acceleration data of the bearing under test can be collected in real time to obtain the detection signal of the bearing under test.

[0065] Specifically, the vibration acceleration data of the bearing under test corresponding to each sampling point can be collected in real time according to the preset interval time of the sampling points in order to obtain the detection signal of the bearing under test.

[0066] Optionally, the bearing prediction model may include a grayscale prediction model.

[0067] Optionally, the target factors are the various factors in the detection signal of the bearing under test that are related to the peak vibration energy.

[0068] Optionally, the target factor may include at least one time-domain factor and at least one frequency-domain factor.

[0069] Optionally, the signal characteristics of time-domain factors may include, but are not limited to, at least one of the following: waveform factor characteristics, root mean square (RMS) characteristics, kurtosis characteristics, kurtosis index characteristics, and margin index characteristics.

[0070] Optionally, the signal characteristics of the frequency domain factors may include, but are not limited to, at least one of the following: bearing characteristic frequency characteristics and bearing sideband energy ratio characteristics.

[0071] Step S104: Using the bearing prediction model, perform prediction based on the signal characteristics of the detection signal corresponding to each target factor to determine the bearing life of the bearing under test.

[0072] Optionally, the bearing prediction model can perform prediction based on the signal characteristics of the current sampling point corresponding to each target factor, obtain the predicted vibration peak energy of the detection signal corresponding to the predicted sampling point, and determine the remaining bearing life of the bearing under test at the predicted sampling point based on the predicted vibration peak energy of the detection signal corresponding to the predicted sampling point and the preset bearing peak energy warning value.

[0073] In this embodiment, the predicted sampling point is a sampling point that follows the current sampling point.

[0074] In this embodiment, the bearing peak energy warning value can be determined based on the training dataset of the bearing prediction model.

[0075] In summary, the bearing life prediction method of this embodiment can obtain signal features with high correlation to bearing life from the detection signals of the bearing under test according to the target factors given in the bearing prediction model, and perform bearing life prediction operation accordingly, so as to obtain bearing life prediction results with high accuracy.

[0076] Furthermore, the bearing life prediction method in this embodiment can achieve real-time prediction and analysis of bearing life by acquiring the detection signal of the bearing under test in real time.

[0077] Figure 2 This is a flowchart illustrating a bearing life prediction method according to another embodiment of this application. This embodiment mainly shows the training scheme for the bearing prediction model. As shown in the figure, this embodiment mainly includes the following steps:

[0078] Step S202: Based on multiple candidate factors related to the peak vibration energy, perform feature extraction on the sample signal to obtain the signal features of the sample signal corresponding to each candidate factor.

[0079] Optionally, the sample signal may include vibration acceleration data of the sample bearing corresponding to each sampling point.

[0080] Optionally, candidate factors related to the peak vibration energy may include at least one time-domain factor and at least one frequency-domain factor.

[0081] Optionally, the signal characteristics of time-domain factors may include, but are not limited to, at least one of the following: waveform factor characteristics, root mean square (RMS) characteristics, kurtosis characteristics, kurtosis index characteristics, and margin index characteristics.

[0082] Optionally, the signal characteristics of the frequency domain factors may include, but are not limited to, at least one of the following: bearing characteristic frequency characteristics and bearing sideband energy ratio characteristics.

[0083] Step S204: Determine the true peak vibration energy of the sample signal based on the vibration acceleration data of the sample signal.

[0084] Optionally, vibration acceleration data corresponding to each sampling point in the sample signal can be obtained to determine the true peak vibration energy of the sample signal corresponding to each sampling point.

[0085] Optionally, a constant hardening process can be performed on the vibration acceleration data of the sample signal to better determine the true peak vibration energy of the sample signal.

[0086] Optionally, the true peak vibration energy corresponding to each sampling point in the sample signal can be obtained based on the peak vibration energy conversion formula and the vibration acceleration data corresponding to each sampling point in the sample signal.

[0087] The formula for converting peak vibration energy is as follows: Formula 1:

[0088]

[0089] In Formula 1 above, x(n) is the vibration acceleration data of the nth sampling point, and N is the total number of sampling points of the sample signal.

[0090] In this embodiment, the range of N values ​​can be determined based on the actual bearing speed.

[0091] For example, the value of N can range from 8192 to 23438.

[0092] Step S206: Based on the actual vibration peak energy and the signal characteristics of at least one target factor determined from each candidate factor, train the bearing prediction model to obtain the trained bearing prediction model.

[0093] Optionally, a bearing prediction model can be used to perform prediction based on the signal characteristics of at least one target factor determined from each candidate factor, obtain the predicted vibration peak energy of each sampling point corresponding to the sample signal, calculate the difference between the actual vibration peak energy and the predicted vibration peak energy corresponding to the same sampling point, and train the bearing prediction model based on the difference result.

[0094] In summary, the bearing life prediction method provided in this embodiment trains the bearing prediction model based on the peak vibration energy corresponding to different periods of the sample signal. It utilizes the fact that peak vibration energy can not only characterize the signal changes of typical rotating equipment with the period, but also that peak vibration energy is very sensitive to bearing damage, such as wear and fatigue spalling, thereby achieving effective tracking of bearing wear and improving the accuracy of bearing life prediction results of the bearing prediction model.

[0095] Furthermore, the bearing life prediction method provided in this embodiment extracts time-domain and frequency-domain features from the sample signal as signal features to be predicted, so that the bearing prediction model can perform bearing life prediction based on multi-dimensional features, thereby improving the accuracy and objectivity of the prediction results.

[0096] Figure 3 This is a flowchart illustrating a bearing life prediction method according to another exemplary embodiment of this application, showing a specific implementation of step S206 described above. As shown in the figure, this embodiment mainly includes the following steps:

[0097] Step S302: Determine the original sequence based on the vibration acceleration data in the sample signal, and construct the factor sequence for each candidate factor based on the signal characteristics of the sample signal corresponding to each candidate factor.

[0098] Optionally, the original sequence (or vibration peak energy sequence) can be obtained from the vibration acceleration data corresponding to each sampling point of the sample signal.

[0099] In this embodiment, the original sequence can be represented as: X (0) (k), that is, the original sequence consisting of k vibration acceleration data from k sampling points of the sample signal.

[0100] Optionally, based on the signal characteristics of each sampling point corresponding to each candidate factor, an accumulation calculation can be performed to obtain the factor sequence (AGO sequence) of each candidate factor.

[0101] In this embodiment, the factor sequence of each candidate factor can be represented as: That is, the i-th candidate factor corresponds to the factor sequence of k signal features at k sampling points.

[0102] In this embodiment, the conversion formula for the factor sequence of each candidate factor is as follows: Formula 2:

[0103]

[0104] In Formula 2 above, This indicates that the i-th candidate factor corresponds to the signal feature of the k-th sampling point. A factor sequence representing the signal characteristics of k sampling points of the i-th candidate factor.

[0105] Step S304: Perform the target factor determination step. Based on the original sequence and the factor sequence of each candidate factor, perform correlation calculation to determine at least one target factor among the candidate factors.

[0106] Optionally, grey relational degree calculation can be performed based on the original sequence of the sample signal and the factor sequence of each candidate factor to determine the relational degree value of each candidate factor, and the target factor among each candidate factor can be determined based on the relational degree value of each candidate factor and a given relational degree threshold.

[0107] In this embodiment, the correlation threshold can be set to 0.55.

[0108] Step S306: Using the bearing prediction model, perform prediction based on the signal characteristics of each target factor to obtain the predicted vibration peak energy of the sample signal.

[0109] Optionally, an ordinary differential equation can be constructed based on the original sequence, neighbor mean, and other complete sequences, and the differential equation parameters in the ordinary differential equation can be solved. Then, the vibration peak energy of the sample signal can be predicted using the solved differential equation parameters.

[0110] In this embodiment, the entire sequence, such as the neighbor mean, can be determined based on the factor sequence of each target factor.

[0111] Step S308: Based on the actual peak vibration energy and the predicted peak vibration energy, obtain the residual results of the bearing prediction model, and train the bearing prediction model based on the residual results.

[0112] In this embodiment, the residual results of the bearing prediction model can be obtained based on the actual peak vibration energy and the predicted peak vibration energy. If the residual results are not satisfied with the given training termination condition, the factor sequence of each candidate factor is updated, and the execution of the target factor determination step (i.e., step S304) is returned until the residual results meet the training termination condition.

[0113] In summary, the bearing life prediction method provided in this embodiment constructs the original sequence of sample signals and the corresponding factor sequences of each candidate factor to perform target factor screening and bearing prediction model training. The factor sequences generated by the accumulation algorithm can weaken the volatility and randomness of random sequences and support modeling analysis of small samples. It can meet the model prediction analysis requirements under small sample training set conditions. At the same time, it can also obtain the changing patterns between features and generate strong regular data sequences to make up for the lack of features in small sample data and improve the model prediction accuracy.

[0114] Specifically, since the fission of bearings is almost exponential, which satisfies the assumptions of grayscale theory, in the case of small samples, data can be generated by the algorithm of accumulation or summation, and then grayscale prediction can be performed. This solves the problems of difficulty in model construction and inaccurate model prediction results caused by insufficient sample size.

[0115] Furthermore, the prediction model constructed and trained using the solution in this embodiment can be quickly and automatically applied to different application fields to meet the life prediction needs of different types of mechanical bearings.

[0116] Figure 4 This is a flowchart illustrating a bearing life prediction method according to another exemplary embodiment of this application. This embodiment shows a specific implementation of step S304 (target factor determination step). As shown in the figure, this embodiment mainly includes the following processing steps:

[0117] Step S402: Identify a candidate factor as the current factor.

[0118] Specifically, one can obtain a candidate factor in turn and determine it as the current factor.

[0119] Step S404: Based on the original sequence and the factor sequence of the current factor, perform the correlation calculation of the current factor to determine the correlation value of the current factor.

[0120] Optionally, the correlation coefficient conversion formula can be used to calculate the correlation coefficient value of the current factor based on the original sequence and the factor sequence of the current factor, and the correlation degree conversion formula can be used to calculate the correlation degree value of the current factor based on the correlation coefficient value of the current factor.

[0121] In this embodiment, the correlation coefficient conversion formula is as follows: Formula 3:

[0122]

[0123] In formula 3 above, ξ i (k) represents the correlation coefficient value of the characteristic signal at the k-th sampling point of the i-th candidate factor; X (0)(k) represents the original sequence of vibration acceleration data corresponding to k sampling points of the sample signal. The factor sequence represents the signal characteristics of the k sampling points of the i-th candidate factor; ρ is the weight value.

[0124] In this embodiment, ρ can be set to 0.5.

[0125] In this embodiment, the correlation degree conversion formula is as follows: Formula 4:

[0126]

[0127] In formula 4 above, r i represents the correlation value of the i-th candidate factor, and N is the total number of sampling points k.

[0128] In this embodiment, the range of values ​​for k in Formula 4 should be the same as the range of values ​​for n in Formula 1 above, both being N.

[0129] Step S406: Identify the current factors whose correlation values ​​are greater than the given correlation threshold as target factors.

[0130] In this embodiment, the correlation threshold can be set to 0.55.

[0131] If the correlation value of the current factor is greater than 0.55, then the current factor is determined as the target factor.

[0132] Step S408: Determine whether all candidate factors have been identified as the current factor. If not, return to step S402; if yes, proceed to step S306.

[0133] Specifically, if it is determined that there are candidate factors that have not been identified as the current factor, return to step S402 to obtain the next candidate factor as the current factor and continue to perform the judgment of the target factor. If it is determined that each of the candidate factors has been identified as the current factor, it means that the confirmation operation of the target factor has been completed, and then step S306 is executed to perform prediction using the bearing prediction model.

[0134] In summary, the bearing prediction method provided in this embodiment performs grey relational degree calculation based on the original sequence of the sample signal and the sequence of each candidate factor, which can ensure the effectiveness between the determined target factor and the peak energy of bearing vibration, so as to enable the bearing prediction model to accurately predict the peak energy of the bearing.

[0135] Figure 5 This is a flowchart illustrating a bearing life prediction method as another exemplary embodiment of this application. This embodiment shows a specific implementation of step S306 described above. As shown in the figure, this embodiment mainly includes the following processing steps:

[0136] Step S502: Based on the original sequence and the neighbor mean obtained from the factor sequence based on each target factor, construct an ordinary differential equation and solve for the differential equation parameters in the ordinary differential equation.

[0137] Optionally, the neighbor mean and other full sequence conversion formulas can be used to obtain the neighbor mean and other full sequences based on the factor sequence of each target factor.

[0138] In this embodiment, the formula for converting the entire sequence, such as the neighbor mean, is as follows: Formula 5:

[0139] Z (1) (k)=α*X i (1) (k)+0.5*X2 (1) (k)...0.5*X i (1) (k) (Formula 5)

[0140] In Formula 5 above, Z (1) (k) represents the complete sequence of signal characteristics of k sampling points of the sample signal, such as the neighbor mean; i represents the i-th target factor.

[0141] Optionally, ordinary differential equations can be constructed based on the original sequence, the neighbor mean and other full sequences, and the factor sequence of each target factor.

[0142] In this embodiment, the constructed ordinary differential equation is as shown in Formula 6 below:

[0143]

[0144] In Formula 6 above, X (0) (k) represents the original sequence, k represents the k-th sampling point, and Z (1) (k) represents the entire sequence with equal neighbor mean, X i (1) (k) represents the factor sequence of signal features of k sampling points of the i-th target factor, where M is the total number of target factors, and A and b are... i represents the parameters of the differential equation to be solved.

[0145] Where A is the development coefficient, b i Let be the grey effect of the i-th target factor.

[0146] Alternatively, the least squares method can be used to solve for the differential equation parameters A and b in the ordinary differential equation (Equation 6). i .

[0147] Step S504: Substitute the parameters of the solved differential equation into the prediction formula obtained by the transformation of the ordinary differential equation to obtain the predicted vibration peak energy of the current factor.

[0148] It can perform transposition and simplification transformations on ordinary differential equations to obtain the prediction formula.

[0149] In this embodiment, the prediction formula obtained from the transformation of the ordinary differential equation is as follows: Formula 7:

[0150]

[0151] In formula 7 above, The sample signal represents the predicted peak vibration energy corresponding to the (k+1)th sampling point, M is the total number of target factors, and i represents the i-th target factor.

[0152] In summary, the bearing life prediction method provided in this embodiment constructs the ordinary differential equation of a gray-scale prediction model (bearing prediction model) based on the original sequence and the neighbor mean obtained from the factor sequence based on the target factor. By establishing correlations with hidden feature variables, the accuracy of the model prediction results can be improved, and efficient model training can be achieved on small sample datasets.

[0153] Figure 6 This is a flowchart illustrating a bearing life prediction method as another exemplary embodiment of this application. This embodiment shows a specific implementation of step S308 described above. (Referring to...) Figure 3 and Figure 6 This embodiment mainly includes the following steps:

[0154] Step S602: Based on the difference between the actual peak vibration energy and the predicted peak vibration energy corresponding to the same sampling point of the sample signal, obtain the residual values ​​of the bearing prediction model corresponding to each sampling point.

[0155] In this embodiment, the difference between the actual vibration peak energy and the predicted vibration peak energy corresponding to the same sampling point (each sampling point after the second sampling point) of the sample signal can be calculated to obtain the residual values ​​of the bearing prediction model corresponding to each sampling point.

[0156] In this embodiment, the residual value of the bearing prediction model corresponding to each sampling point is expressed as:

[0157]

[0158] Where E(k) represents the residual value of the bearing prediction model corresponding to the k-th sampling point; X (0) (k) represents the true peak vibration energy at the k-th sampling point; Let N represent the predicted peak vibration energy at the k-th sampling point, where k = 2, 3, ..., N.

[0159] Step S604: Determine the residual ratio of the bearing prediction model based on the residual value of the bearing prediction model corresponding to the same sampling point and the predicted peak vibration energy.

[0160] In this embodiment, the residual ratio of the bearing prediction model can be calculated using the following formula 8:

[0161]

[0162] In Formula 8 above, Q(k) represents the residual ratio of the k-th sampling point, where k = 2, 3, ..., N.

[0163] Step S606: Determine whether the residual ratio is greater than the given residual ratio threshold. If not, proceed to step S608; if yes, proceed to step S610.

[0164] In this embodiment, the residual ratio threshold can be set to 0.9.

[0165] If the residual ratio of the bearing prediction model is not greater than (less than or equal to) the residual ratio threshold, it means that the bearing prediction model has not passed the residual verification, and step S608 is performed; otherwise, it means that the bearing prediction model has passed the residual verification, and step 610 is performed.

[0166] Step S608: Based on the given weakened neighborhood mean weight, reconstruct the factor sequence of each candidate factor, and return to execute step S304.

[0167] In this embodiment, the neighborhood mean weight is weakened, and it can be set arbitrarily according to actual testing needs. This application does not limit it in this regard.

[0168] For example, when the weight of the weakened neighborhood mean is set to 2, the summation and averaging of any two consecutive values ​​in the original sequence are performed to generate a reconstructed sequence. Similarly, when the weight of the weakened neighborhood mean is set to 3, the summation and averaging of any three consecutive values ​​in the original sequence are performed to generate a reconstructed sequence.

[0169] In this embodiment, the factor sequence of each candidate factor can be reconstructed based on the weakened neighborhood mean weight, and the execution step S304 can be returned to re-execute the target factor determination step based on the updated factor sequence of each candidate factor.

[0170] Step S610: Based on the residual values ​​of the bearing prediction model, the residual average value determined based on the residual values, and the residual standard deviation, perform a posterior error test on the bearing prediction model to obtain the small probability error value of the bearing prediction model.

[0171] In this embodiment, the small probability error value of the bearing prediction model can be calculated using the following formula 9:

[0172]

[0173] In Formula 9 above, P represents the small probability error value of the bearing prediction model; E(k) represents the residual value of the bearing prediction model corresponding to the kth sampling point. S1 represents the residual mean; S2 represents the residual standard deviation.

[0174] In this embodiment, the residual standard deviation S1 can be obtained using the following formula 10:

[0175]

[0176] In Formula 10 above, X (0) (k) represents the true peak vibration energy at the k-th sampling point; This represents the predicted peak vibration energy at the k-th sampling point.

[0177] Step S612: Determine whether the low probability error value is less than the posterior difference threshold. If yes, proceed to step S614; otherwise, proceed to step S608.

[0178] Optionally, the posterior difference threshold can be set to 0.05.

[0179] In this embodiment, if the small probability error value P of the bearing prediction model is less than 0.05, it indicates that the bearing prediction model training is complete; otherwise, return to step S608 to reconstruct the factor sequence of each candidate factor according to the weakened neighborhood mean weight, and re-execute the target factor determination step (i.e., step S304).

[0180] Step S614: Obtain the trained bearing prediction model.

[0181] It should be noted that the posterior difference test steps in steps S610 and S612 of this embodiment are optional steps. In other embodiments, if the judgment result of step S606 is yes, the process can directly jump to step S614 to continue execution.

[0182] In summary, the bearing life prediction method provided in this embodiment, based on the difference between the actual peak vibration energy and the predicted peak vibration energy corresponding to each sampling point of the sample signal, uses the residual test method to verify the training effect of the bearing prediction model, and can obtain a bearing prediction model with better prediction performance.

[0183] Furthermore, the bearing life prediction method provided in this embodiment can further perform a post-hoc error test on the bearing prediction model that has passed the residual test, so as to further improve the accuracy and robustness of the bearing prediction model prediction results.

[0184] Figure 7This is a flowchart illustrating a bearing life prediction method according to another exemplary embodiment of this application. This embodiment shows a specific implementation of step S104 described above. As shown in the figure, this embodiment mainly includes the following steps:

[0185] Step S702: Using the bearing prediction model, prediction is performed based on the signal characteristics of the current sampling point corresponding to each target factor of the detection signal to obtain the predicted vibration peak energy of the detection signal corresponding to the predicted sampling point.

[0186] In this embodiment, the predicted sampling point is a sampling point that follows the current sampling point.

[0187] Step S704: Based on the predicted vibration peak energy corresponding to the predicted sampling point of the detection signal and the bearing peak energy warning value, determine the remaining bearing life of the bearing under test at the predicted sampling point.

[0188] Optionally, the bearing peak energy warning value (bearing scrap line) can be determined based on the training results of the bearing prediction model trained from multiple sample signals.

[0189] Optionally, if the difference between the predicted peak vibration energy and the bearing peak energy warning value at the predicted sampling point corresponding to the detection signal is greater than 3 times the standard deviation of the training data (i.e., 3 sigma), the prediction result of the bearing life of the bearing under test is obtained; otherwise, the remaining bearing life of the bearing under test at the predicted sampling point is calculated based on the predicted peak vibration energy and the bearing peak energy warning value.

[0190] In this embodiment, the remaining bearing life of the bearing under test can be calculated using the following formula 11:

[0191]

[0192] In summary, the bearing life prediction method provided in this embodiment can accurately predict the remaining service life of the bearing under test based on the bearing peak energy warning value, thereby improving the overall operational reliability of the mechanical system and reducing maintenance costs.

[0193] Figure 8 A structural block diagram of a bearing life prediction device according to an exemplary embodiment of this application is shown. As shown in the figure, the bearing life prediction device 800 of this embodiment mainly includes: a feature extraction module 802 and a bearing prediction model 804.

[0194] The feature extraction module 802 is used to perform feature extraction on the detection signal of the bearing under test based on at least one target factor given by the bearing prediction model, and to obtain the signal features of the detection signal corresponding to each target factor.

[0195] The bearing prediction model 804 is used to perform prediction based on the signal characteristics of the detection signal corresponding to each target factor, and to determine the bearing life of the bearing under test.

[0196] Optionally, the bearing life prediction device 800 includes a training module for training the bearing prediction model 804, which includes: performing feature extraction on a sample signal based on multiple candidate factors related to vibration peak energy to obtain signal features of the sample signal corresponding to each candidate factor; determining the true vibration peak energy of the sample signal based on the vibration acceleration data of the sample signal; and training the bearing prediction model based on the true vibration peak energy and the signal features of at least one target factor determined from each candidate factor to obtain a trained bearing prediction model.

[0197] Optionally, the candidate factors include at least one time-domain factor and at least one frequency-domain factor; wherein, the signal characteristics of the at least one time-domain factor include at least one of the following: waveform factor characteristics, root mean square characteristics, kurtosis characteristics, kurtosis index characteristics, and margin index characteristics; and the signal characteristics of the at least one frequency-domain factor include at least one of the following: bearing characteristic frequency characteristics and bearing sideband energy ratio characteristics.

[0198] Optionally, the training module is further configured to: determine an original sequence based on the vibration acceleration data in the sample signal, and construct a factor sequence for each candidate factor based on the signal characteristics of the sample signal corresponding to each candidate factor; a target factor determination step, performing correlation calculation based on the original sequence and the factor sequence of each candidate factor to determine at least one target factor from each candidate factor; using the bearing prediction model, performing prediction based on the signal characteristics of each target factor to obtain the predicted vibration peak energy of the sample signal; obtaining the residual result of the bearing prediction model based on the actual vibration peak energy and the predicted vibration peak energy; if the residual result is not satisfied with the given training termination condition, updating the factor sequence of each candidate factor, and returning to execute the target factor determination step until the residual result satisfies the training termination condition.

[0199] Optionally, the training module is further configured to: obtain the original sequence of the sample signal based on the vibration acceleration data corresponding to each sampling point of the sample signal; and perform cumulative calculation based on the signal characteristics of each sampling point corresponding to each candidate factor to obtain the factor sequence of each candidate factor.

[0200] Optionally, the training module is further configured to: identify a candidate factor as the current factor; perform a correlation calculation on the current factor based on the original sequence and the factor sequence of the current factor to determine the correlation value of the current factor; identify the current factors whose correlation value is greater than a given correlation threshold as target factors; and return to the step of identifying a candidate factor as the current factor until all candidate factors are identified as current factors.

[0201] Optionally, the correlation threshold is 0.55.

[0202] Optionally, the training module is further configured to: calculate the correlation coefficient value of the current factor based on the original sequence and the factor sequence of the current factor using a correlation coefficient conversion formula; and calculate the correlation value of the current factor based on the correlation coefficient value of the current factor using a correlation degree conversion formula.

[0203] The correlation coefficient conversion formula is expressed as follows:

[0204]

[0205] Wherein, the ξ i (k) represents the correlation coefficient value of the feature signal at the k-th sampling point of the i-th candidate factor; the X (0) (k) represents the original sequence of vibration acceleration data corresponding to k sampling points of the sample signal. A factor sequence representing the signal characteristics of k sampling points of the i-th candidate factor; where ρ is the weight value;

[0206] The correlation conversion formula is expressed as follows:

[0207]

[0208] Wherein, the r i This represents the correlation value of the i-th candidate factor, where N is the total number of sampling points k.

[0209] Optionally, the training module is further configured to: construct an ordinary differential equation based on the original sequence and the neighbor mean determined by the factor sequence for each target factor, and solve for the differential equation parameters in the ordinary differential equation; substitute the solved differential equation parameters into the prediction formula obtained by transforming the ordinary differential equation to obtain the predicted vibration peak energy of the current factor.

[0210] Optionally, the ordinary differential equation is expressed as:

[0211]

[0212] Wherein, X(0) (k) represents the original sequence, where k represents the k-th sampling point, and Z (1) (k) represents the neighbor mean equal whole sequence, and X i (1) (k) represents the factor sequence of signal features from k sampling points of the i-th target factor, where M is the total number of target factors, and A and b i The parameters of the differential equation to be solved;

[0213] The prediction formula obtained from the transformation of the ordinary differential equation is expressed as follows:

[0214]

[0215] Among them, the This indicates the predicted peak vibration energy corresponding to the (k+1)th sampling point of the sample signal.

[0216] Optionally, the feature extraction module 802 is further configured to: acquire vibration acceleration data corresponding to each sampling point of the sample signal, and determine the true peak vibration energy corresponding to each sampling point of the sample signal.

[0217] Optionally, the training module is further configured to: obtain the residual values ​​of the bearing prediction model corresponding to each sampling point based on the difference between the actual vibration peak energy and the predicted vibration peak energy corresponding to the same sampling point of the sample signal; and determine the residual ratio of the bearing prediction model based on the residual values ​​of the bearing prediction model corresponding to the same sampling point and the predicted vibration peak energy.

[0218] Optionally, the training module is further configured to: update the factor sequence of each candidate factor if the residual result is not satisfied with the given training termination condition, and return to execute the target factor determination step until the residual result satisfies the training termination condition, including: when the residual ratio is not greater than a given residual ratio threshold, reconstruct the factor sequence of each candidate factor according to the given weakened neighborhood mean weight, and return to execute the target factor determination step until the residual ratio is greater than the residual ratio threshold; wherein the residual ratio threshold is 0.9.

[0219] Optionally, the training module is further configured to: perform a posterior error test on the bearing prediction model based on the residual value of the bearing prediction model, the residual mean determined based on the residual value, and the residual standard deviation, to obtain a small probability error value of the bearing prediction model; if the small probability error value is not less than the posterior error threshold, reconstruct the factor sequence of each candidate factor according to the weakened neighborhood mean weight, and return to execute the target factor determination step until the small probability error value is less than the posterior error threshold; wherein, the posterior error threshold is set to 0.05.

[0220] Optionally, the feature extraction module is also used to: determine the bearing peak energy warning value based on the training results of training the bearing prediction model using multiple sample signals.

[0221] Optionally, the training module is further configured to: utilize the bearing prediction model to perform prediction based on the signal characteristics of the current sampling point corresponding to each target factor of the detection signal, and obtain the predicted vibration peak energy of the detection signal corresponding to the predicted sampling point, wherein the predicted sampling point is a sampling point following the current sampling point; and determine the remaining bearing life of the bearing under test at the predicted sampling point based on the predicted vibration peak energy of the predicted sampling point and the bearing peak energy warning value.

[0222] Optionally, the feature extraction module is further configured to: acquire the vibration acceleration data of the bearing under test in real time, and obtain the detection signal of the bearing under test.

[0223] The bearing life prediction device provided in this embodiment corresponds to the bearing life prediction method provided in various embodiments of this invention. Other descriptions can be referred to the descriptions of the bearing life prediction methods provided in various embodiments of this invention, and will not be repeated here.

[0224] Another embodiment of the present invention provides an electronic device, including: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus.

[0225] Figure 9 The following is a structural block diagram of an electronic device according to an exemplary embodiment of the present invention. As shown in the figure, the electronic device 900 of this embodiment may include a processor 902, a communication interface 904, and a memory 906.

[0226] The processor 902, communication interface 904, and memory 906 can communicate with each other via communication bus 908.

[0227] The communication interface 904 is used to communicate with other electronic devices such as terminal devices or servers.

[0228] The processor 902 is used to execute the computer program 910, specifically to execute the relevant steps in the above-described method embodiments, that is, to execute the steps in the bearing life prediction method as described in the above-described embodiments.

[0229] Specifically, computer program 910 may include program code that includes computer operation instructions.

[0230] Processor 902 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The electronic device may include one or more processors of the same type, such as one or more CPUs; or it may include processors of different types, such as one or more CPUs and one or more ASICs.

[0231] Memory 906 is used to store computer program 910. Memory 906 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0232] Another embodiment of the present invention provides a computer storage medium storing a computer program thereon, which, when executed by a processor, can implement the bearing life prediction method described in the above embodiments.

[0233] It should be noted that, depending on the implementation needs, the various components / steps described in the embodiments of the present invention can be broken down into more components / steps, or two or more components / steps or parts of the operation of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of the present invention.

[0234] The methods described above according to embodiments of the present invention can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CD-ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code originally stored on a remote recording medium or a non-transitory machine-readable medium and subsequently stored on a local recording medium, downloaded via a network. Thus, the methods described herein can be stored as software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components (e.g., RAM, ROM, flash memory, etc.) capable of storing or receiving software or computer code that, when accessed and executed by the computer, processor, or hardware, implements the bearing life prediction method described herein. Furthermore, when a general-purpose computer accesses the code used to implement the bearing life prediction method shown herein, the execution of the code transforms the general-purpose computer into a dedicated computer for executing the bearing life prediction method shown herein.

[0235] Those skilled in the art will recognize that the units and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the embodiments of the present invention.

[0236] The above embodiments are only used to illustrate the embodiments of the present invention, and are not intended to limit the embodiments of the present invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the embodiments of the present invention. Therefore, all equivalent technical solutions also fall within the scope of the embodiments of the present invention, and the patent protection scope of the embodiments of the present invention should be defined by the claims.

Claims

1. A method for predicting bearing life, characterized in that, include: Based on at least one target factor given by the bearing prediction model, feature extraction is performed on the detection signal of the bearing under test to obtain the signal features of the detection signal corresponding to each target factor. Using the aforementioned bearing prediction model, predictions are performed based on the signal characteristics of the detected signal corresponding to each target factor to determine the bearing life of the bearing under test. The bearing prediction model is trained in the following way: Based on multiple candidate factors related to the peak vibration energy, feature extraction is performed on the sample signal to obtain the signal features of the sample signal corresponding to each candidate factor. Based on the vibration acceleration data of the sample signal, determine the true peak vibration energy of the sample signal; Based on the actual vibration peak energy and the signal characteristics of at least one target factor determined from the candidate factors, the bearing prediction model is trained to obtain a trained bearing prediction model. The step of training the bearing prediction model based on the actual vibration peak energy and the signal characteristics of at least one target factor determined from each candidate factor to obtain the trained bearing prediction model includes: Based on the vibration acceleration data in the sample signal, the original sequence is determined, and based on the signal characteristics of the sample signal corresponding to each candidate factor, the factor sequence of each candidate factor is constructed. The target factor determination step involves performing correlation calculations based on the original sequence and the factor sequence of each candidate factor to determine at least one target factor from the candidate factors. Using the bearing prediction model, prediction is performed based on the signal characteristics of each target factor to obtain the predicted peak vibration energy of the sample signal; Based on the actual peak vibration energy and the predicted peak vibration energy, the residual results of the bearing prediction model are obtained. If the residual results are not satisfied with the given training termination condition, the factor sequence of each candidate factor is updated, and the target factor determination step is returned to be executed until the residual results satisfy the training termination condition.

2. The method according to claim 1, characterized in that, The candidate factors include at least one time-domain factor and at least one frequency-domain factor; wherein... The signal features of the at least one time-domain factor include at least one of the following: waveform factor features, root mean square features, kurtosis features, kurtosis index features, and margin index features; The signal characteristics of the at least one frequency domain factor include at least one of the following: bearing characteristic frequency characteristics and bearing sideband energy ratio characteristics.

3. The method according to claim 1, characterized in that, The step of determining the original sequence based on the vibration acceleration data in the sample signal, and constructing the factor sequence for each candidate factor based on the signal characteristics of the sample signal corresponding to each candidate factor, includes: The original sequence of the sample signal is obtained based on the vibration acceleration data corresponding to each sampling point of the sample signal; Based on the signal characteristics of each sampling point corresponding to each candidate factor, an accumulation calculation is performed to obtain the factor sequence of each candidate factor.

4. The method according to claim 1, characterized in that, The steps for determining the target factors include: Select a candidate factor as the current factor; Based on the original sequence and the factor sequence of the current factor, perform a correlation calculation on the current factor to determine the correlation value of the current factor; The current factors whose correlation values ​​are greater than a given correlation threshold are identified as target factors; Return to the step of identifying a candidate factor as the current factor, until all candidate factors have been identified as the current factor; The correlation threshold is 0.

55.

5. The method according to claim 4, characterized in that, The step of performing correlation calculation on the current factor based on the original sequence and the factor sequence of the current factor to determine the correlation value of the current factor includes: Using the correlation coefficient conversion formula, the correlation coefficient value of the current factor is calculated based on the original sequence and the factor sequence of the current factor; Using the correlation conversion formula, the correlation value of the current factor is calculated based on the correlation coefficient value of the current factor; The correlation coefficient conversion formula is expressed as follows: Among them, the Indicates the first The first candidate factor The correlation coefficient values ​​of the characteristic signals at each sampling point; This indicates that the sample signal corresponds to The original sequence of vibration acceleration data from each sampling point, Indicates the first One candidate factor The factor sequence of signal characteristics at each sampling point; These are weight values; The correlation conversion formula is expressed as follows: Among them, the Indicates the first The correlation value of each candidate factor, the Sampling points The total number.

6. The method according to claim 1, characterized in that, The step of using the bearing prediction model to perform prediction based on the signal characteristics of each target factor to obtain the predicted vibration peak energy of the sample signal includes: Based on the original sequence and the complete sequence including the neighbor mean determined by the factor sequence for each target factor, construct an ordinary differential equation and solve for the differential equation parameters in the ordinary differential equation. The parameters of the solved differential equation are substituted into the prediction formula obtained by transforming the ordinary differential equation to obtain the predicted vibration peak energy of each sampling point corresponding to the current factor.

7. The method according to claim 6, characterized in that, The ordinary differential equation is expressed as: Among them, the Represents the original sequence, the Indicates the first Each sampling point, the Represents the entire sequence with neighbor mean, the Indicates the first Each target factor The sequence of factors of signal characteristics at each sampling point, The total number of target factors, the and stated The parameters of the differential equation to be solved; The prediction formula obtained from the transformation of the ordinary differential equation is expressed as follows: Among them, the This indicates that the sample signal corresponds to the first... Predicted peak vibration energy at each sampling point.

8. The method according to any one of claims 1 to 7, characterized in that, The method includes: Obtain the vibration acceleration data corresponding to each sampling point of the sample signal, and determine the true peak vibration energy corresponding to each sampling point of the sample signal; The step of obtaining the residual result of the bearing prediction model based on the actual peak vibration energy and the predicted peak vibration energy includes: Based on the difference between the actual vibration peak energy and the predicted vibration peak energy corresponding to the same sampling point of the sample signal, the residual values ​​of the bearing prediction model corresponding to each sampling point are obtained. Based on the residual value of the bearing prediction model corresponding to the same sampling point and the predicted peak vibration energy, the residual ratio of the bearing prediction model is determined.

9. The method according to claim 8, characterized in that, If the residual result is not satisfied with the given training termination condition, the factor sequence of each candidate factor is updated, and the target factor determination step is returned to be executed until the residual result satisfies the training termination condition, including: When the residual ratio is not greater than a given residual ratio threshold, Based on the given weakened neighborhood mean weights, reconstruct the factor sequence for each candidate factor, and return to execute the target factor determination step until the residual ratio is greater than the residual ratio threshold. The residual ratio threshold is 0.

9.

10. The method according to claim 9, characterized in that, If the residual ratio is greater than the residual ratio threshold, the method further includes: Based on the residual values ​​of the bearing prediction model, the residual mean value determined based on the residual values, and the residual standard deviation, a posterior error test is performed on the bearing prediction model to obtain the small probability error value of the bearing prediction model. If the low probability error value is not less than the posterior difference threshold, the factor sequence of each candidate factor is reconstructed according to the weakened neighborhood mean weight, and the target factor determination step is returned to be executed until the low probability error value is less than the posterior difference threshold. The posterior difference threshold is set to 0.

05.

11. The method according to claim 2, characterized in that, The method further includes: Based on the training results of the bearing prediction model trained from multiple sample signals, the bearing peak energy warning value is determined. Furthermore, the step of using the bearing prediction model to perform prediction based on the signal characteristics of the detection signal corresponding to each target factor, and determining the bearing life of the bearing under test, includes: Using the bearing prediction model, prediction is performed based on the signal characteristics of the current sampling point corresponding to each target factor of the detection signal to obtain the predicted vibration peak energy of the detection signal corresponding to the predicted sampling point, wherein the predicted sampling point is a sampling point following the current sampling point; Based on the predicted peak vibration energy at the predicted sampling point and the bearing peak energy warning value, the remaining bearing life of the bearing under test at the predicted sampling point is determined.

12. The method according to claim 1 or 11, characterized in that, The method further includes: The vibration acceleration data of the bearing under test is collected in real time to obtain the detection signal of the bearing under test.

13. A bearing life prediction device, characterized in that, include: The feature extraction module is used to perform feature extraction on the detection signal of the bearing under test based on at least one target factor given by the bearing prediction model, and to obtain the signal features of the detection signal corresponding to each target factor. A bearing prediction model is used to perform predictions based on the signal characteristics of the detected signal corresponding to each target factor, thereby determining the bearing life of the bearing under test. The bearing prediction model is trained in the following way: Based on multiple candidate factors related to the peak vibration energy, feature extraction is performed on the sample signal to obtain the signal features of the sample signal corresponding to each candidate factor. Based on the vibration acceleration data of the sample signal, determine the true peak vibration energy of the sample signal; Based on the actual vibration peak energy and the signal characteristics of at least one target factor determined from the candidate factors, the bearing prediction model is trained to obtain a trained bearing prediction model. The step of training the bearing prediction model based on the actual vibration peak energy and the signal characteristics of at least one target factor determined from each candidate factor to obtain the trained bearing prediction model includes: Based on the vibration acceleration data in the sample signal, the original sequence is determined, and based on the signal characteristics of the sample signal corresponding to each candidate factor, the factor sequence of each candidate factor is constructed. The target factor determination step involves performing correlation calculations based on the original sequence and the factor sequence of each candidate factor to determine at least one target factor from the candidate factors. Using the bearing prediction model, prediction is performed based on the signal characteristics of each target factor to obtain the predicted peak vibration energy of the sample signal; Based on the actual peak vibration energy and the predicted peak vibration energy, the residual results of the bearing prediction model are obtained. If the residual results are not satisfied with the given training termination condition, the factor sequence of each candidate factor is updated, and the target factor determination step is returned to be executed until the residual results satisfy the training termination condition.

14. An electronic device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation corresponding to the bearing life prediction method as described in any one of claims 1 to 12.

15. A computer storage medium, characterized in that, It stores a computer program that, when executed by a processor, can implement the bearing life prediction method as described in any one of claims 1 to 12.

Images