A Two-Stage Bearing Life Prediction Method and System

Abstract

The present application discloses a two-stage bearing life prediction method and system. The method includes: obtaining online monitoring data; processing and calculating the bearing fault impact signal data and bearing operation mileage data in the online monitoring data to obtain a bearing degradation characteristic value, where the bearing degradation characteristic value is a smoothed impact characteristic change rate value; determining the current degradation stage according to the bearing degradation characteristic value and an empirical threshold, where the degradation stage includes a stable operation stage and a rapid degradation stage; if the current degradation stage is the stable operation stage, using a pre-established stable operation stage prediction model to perform life prediction to obtain the remaining operation mileage of the current bearing; if the current degradation stage is the rapid degradation stage, using a pre-established rapid degradation stage prediction model to perform life prediction to obtain the remaining operation mileage of the current bearing. The present application can improve the accuracy of bearing life prediction.

Description

Technical Field

[0001] The present application relates to the technical field of bearing life prediction, and particularly to a two-stage bearing life prediction method and system. Background Art

[0002] As a key component in the transmission system of mechanical equipment, bearings are widely used in mechanical equipment, and their operating state has an important impact on the health of rotating mechanical equipment. Due to the operation of bearings under a certain load and the influence of complex environmental factors, they are prone to failure, which will ultimately lead to equipment failure. At the same time, the condition-based maintenance of mechanical equipment has gradually developed and attracted attention. When formulating a maintenance plan, the life prediction results of mechanical equipment will be fully considered. Therefore, the accuracy of life prediction directly affects the rationality of maintenance decisions. In view of this, in order to ensure the safety and reliability of equipment, avoid the occurrence of disaster accidents, save maintenance costs and reduce economic losses, the life prediction of bearings has attracted wide attention from scholars and engineering technicians.

[0003] Due to the complex and diverse failure forms of bearings, it is difficult to describe the true degradation physical process of the equipment, thus making it difficult to accurately predict the life. And currently, in the prediction of the remaining life of rotating components of equipment, there are problems that the monotonicity and trend of the constructed characteristic factors are not ideal enough, resulting in low accuracy of life prediction results.

[0004] Therefore, how to improve the accuracy of bearing life prediction is a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0005] To solve the above technical problems, the present application provides a two-stage bearing life prediction method, which can improve the accuracy of bearing life prediction. The present application also provides a two-stage bearing life prediction system, which has the same technical effect.

[0006] The first object of the present application is to provide a two-stage bearing life prediction method.

[0007] The above object one of the present application is achieved through the following technical solutions:

[0008] A two-stage bearing life prediction method includes:

[0009] Obtaining online monitoring data, where the online monitoring data includes bearing fault impact signal data and bearing operation mileage data;

[0010] Preprocessing the bearing fault impact signal data to obtain preprocessed bearing fault impact signal data;

[0011] Process the preprocessed bearing fault impact signal data and the bearing operation mileage data to obtain bearing degradation characteristic values, where the bearing degradation characteristic values are the shock characteristic change rate values after smoothing;

[0012] Determine the current degradation stage according to the bearing degradation characteristic values and the empirical threshold, where the degradation stage includes a stable operation stage and a rapid degradation stage;

[0013] If the current degradation stage is the stable operation stage, use the pre-established stable operation stage prediction model for life prediction to obtain the remaining operation mileage of the current bearing;

[0014] If the current degradation stage is the rapid degradation stage, use the pre-established rapid degradation stage prediction model for life prediction to obtain the remaining operation mileage of the current bearing.

[0015] Preferably, in the bearing two-stage life prediction method, the process of processing the preprocessed bearing fault impact signal data and the bearing operation mileage data to obtain bearing degradation characteristic values includes:

[0016] According to the preprocessed bearing fault impact signal data and the bearing operation data, use the moving average window method to obtain the root mean square value of the bearing fault impact signal within the interval and the average mileage value within the interval;

[0017] Perform smoothing processing on the root mean square value to obtain the smoothed root mean square value;

[0018] Perform differencing and comparison on the smoothed root mean square value and the average mileage value obtained for each window respectively to obtain the shock characteristic change rate value;

[0019] Perform smoothing processing on the shock characteristic change rate value to obtain the bearing degradation characteristic value.

[0020] Preferably, in the bearing two-stage life prediction method, the process of determining the current degradation stage according to the bearing degradation characteristic values and the empirical threshold includes:

[0021] Determine the first predicted value according to the empirical threshold;

[0022] Judge whether the bearing degradation characteristic value is greater than or equal to the first predicted value. If so, determine that the current degradation stage is the rapid degradation stage; if not, determine that the current degradation stage is the stable operation stage.

[0023] Preferably, in the bearing two-stage life prediction method, the process of using the pre-established stable operation stage prediction model for life prediction to obtain the remaining operation mileage of the current bearing includes:

[0024] Combined with the historical bearing degradation eigenvalue, parameter estimation is performed on the pre-established prediction model for the steady operation stage to obtain a first parameter estimation value, where the historical bearing degradation eigenvalue is the smoothed impact feature change rate value obtained based on historical monitoring data, and the historical monitoring data includes bearing historical fault impact signal data and bearing historical operation mileage data;

[0025] According to the first parameter estimation value and the prediction model for the steady operation stage, life prediction is performed to obtain the remaining operation mileage of the current bearing;

[0026] Correspondingly, the life prediction using the pre-established prediction model for the rapid degradation stage to obtain the remaining operation mileage of the current bearing includes:

[0027] Combined with the historical bearing degradation eigenvalue, parameter estimation is performed on the pre-established prediction model for the rapid degradation stage to obtain a second parameter estimation value, where the historical bearing degradation eigenvalue is the smoothed impact feature change rate value obtained based on historical monitoring data, and the historical monitoring data includes bearing historical fault impact signal data and bearing historical operation mileage data;

[0028] According to the second parameter estimation value and the prediction model for the rapid degradation stage, life prediction is performed to obtain the remaining operation mileage of the current bearing.

[0029] Preferably, in the two-stage bearing life prediction method, the pre-established prediction model for the steady operation stage is a Weibull distribution model:

[0030]

[0031] In the formula, η is the scaling factor, m is the shape parameter, and both η and m are statistical empirical values, and t is the mileage value.

[0032] Preferably, in the two-stage bearing life prediction method, the combining the historical bearing degradation eigenvalue and performing parameter estimation on the pre-established prediction model for the steady operation stage to obtain a first parameter estimation value includes:

[0033] Combined with the historical bearing degradation eigenvalue, the likelihood equation of the pre-established prediction model for the steady operation stage is iteratively solved using a numerical method to obtain a first parameter estimation value.

[0034] Preferably, in the two-stage bearing life prediction method, the performing life prediction according to the first parameter estimation value and the pre-established prediction model for the steady operation stage to obtain the remaining operation mileage of the current bearing includes:

[0035] Based on the first parameter estimation value and the steady operation stage prediction model, solve the confidence interval estimation result for the steady operation stage according to the confidence levels of the given upper and lower quantiles.

[0036] According to the confidence interval estimation result, obtain the remaining operating mileage prediction interval from the current mileage in the steady operation stage to the first prediction point.

[0037] Add the remaining operating mileage prediction interval from the current mileage in the steady operation stage to the first prediction point to the remaining life experience value after the first prediction point to obtain the remaining operating mileage of the current bearing.

[0038] Preferably, in the two-stage bearing life prediction method, the rapid degradation stage prediction model is an exponential distribution model with an error term:

[0039] X(t) = aexp(bt) + ε

[0040] where a and b represent the rapid degradation stage parameters, ε represents white noise, ε ∼ N(0, σ 2 ), and t is the mileage value.

[0041] Preferably, in the two-stage bearing life prediction method, the step of combining historical bearing degradation characteristic values to estimate the parameters of the pre-established rapid degradation stage prediction model to obtain the second parameter estimation value includes:

[0042] Combine historical bearing degradation characteristic values and use the maximum likelihood method to estimate the parameters of the pre-established rapid degradation stage prediction model to obtain the second parameter estimation value.

[0043] Preferably, in the two-stage bearing life prediction method, the step of performing life prediction according to the second parameter estimation value and the rapid degradation stage prediction model to obtain the remaining operating mileage of the current bearing includes:

[0044] Based on the second parameter estimation value and the rapid degradation stage prediction model, use Monte Carlo simulation to obtain the point estimate and interval estimate of the remaining operating mileage of the current bearing.

[0045] Preferably, in the two-stage bearing life prediction method, the step of preprocessing the bearing fault impact signal data to obtain the preprocessed bearing fault impact signal data includes:

[0046] Remove the outliers in the bearing fault impact signal data and use the truncated normal distribution to interpolate the missing values in the bearing fault impact signal data to obtain the preprocessed bearing fault impact signal data.

[0047] Preferably, in the bearing two-stage life prediction method, after obtaining the online monitoring data, the following steps are further included:

[0048] Correct the bearing operation mileage data to obtain the corrected bearing operation mileage data;

[0049] Correspondingly, processing the preprocessed bearing fault impact signal data and the bearing operation mileage data to obtain the bearing degradation characteristic value includes:

[0050] Process the preprocessed bearing fault impact signal data and the corrected bearing operation mileage data to obtain the bearing degradation characteristic value.

[0051] Preferably, in the bearing two-stage life prediction method, correcting the bearing operation mileage data to obtain the corrected bearing operation mileage data includes:

[0052] According to the principles of monotonicity and continuity, correct the abnormal data in the bearing operation mileage data to obtain the corrected bearing operation mileage data.

[0053] The second object of the present application is to provide a bearing two-stage life prediction system.

[0054] The above object two of the present application is achieved by the following technical solutions:

[0055] A bearing two-stage life prediction system includes:

[0056] A data acquisition module for acquiring online monitoring data, where the online monitoring data includes bearing fault impact signal data and bearing operation mileage data;

[0057] A data processing module for preprocessing the bearing fault impact signal data to obtain the preprocessed bearing fault impact signal data;

[0058] A feature extraction module for processing the preprocessed bearing fault impact signal data and the bearing operation mileage data to obtain the bearing degradation characteristic value, where the bearing degradation characteristic value is the impact characteristic change rate value after smoothing;

[0059] A stage identification module for determining the current degradation stage according to the bearing degradation characteristic value and the empirical threshold, where the degradation stage includes a stable operation stage and a rapid degradation stage;

[0060] A first prediction module for, when the current degradation stage is the stable operation stage, using the pre-established stable operation stage prediction model to perform life prediction to obtain the remaining operation mileage of the current bearing;

[0061] A second prediction module, configured to perform life prediction by using a pre-established rapid degradation stage prediction model to obtain the remaining operating mileage of the current bearing when the current degradation stage is the rapid degradation stage.

[0062] Based on the online monitoring data, the above technical solution can comprehensively consider the dynamics in the time dimension and the differences in failure modes during the operation of the bearing when predicting the life of the bearing; by preprocessing the bearing fault impact signal data in the online monitoring data, and then based on the preprocessed bearing fault impact signal data and the bearing operating mileage data, a bearing degradation characteristic value is obtained, that is, the impact characteristic change rate value after smoothing. The bearing degradation characteristic value can well reflect the physical failure characteristics of the bearing degradation trajectory and the bearing fault development law. At the same time, as the online monitoring data is updated, the dynamic extraction of the bearing physical degradation characteristics can be realized, which can solve the problems that the bearing failure mechanism is complex and the physical degradation characteristics are difficult to extract and random in the physical model life prediction method; by combining the multi-stage complex degradation behaviors of the bearing in different failure modes and using the pre-established two-stage remaining operating mileage prediction model, that is, the stable operation stage prediction model and the rapid degradation stage prediction model, the prediction of the remaining operating mileage of the bearing is realized, solving the problems such as the complex bearing degradation stage and the difficult determination of the failure threshold at present. And as the online monitoring data is continuously updated, the phased dynamic real-time remaining life prediction of the bearing can be realized. In addition, the above technical solution is applicable to the life prediction of all bearing types. In summary, the above technical solution can improve the accuracy of bearing life prediction. Description of the Drawings

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0064] Figure 1 It is a schematic flowchart of a two-stage bearing life prediction method provided in an embodiment of the present application;

[0065] Figure 2 It is a bearing health status degradation trend diagram provided in an embodiment of the present application;

[0066] Figure 3 It is a schematic flowchart of the bearing degradation characteristic value calculation process provided in an embodiment of the present application;

[0067] Figure 4 It is a schematic structural diagram of a two-stage bearing life prediction system provided in an embodiment of the present application. Detailed implementation manners

[0068] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of this application.

[0069] In the embodiments provided in this application, it should be understood that the disclosed methods and systems can be implemented in other ways. The system embodiments described below are only illustrative. For example, the division of units and modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or modules can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed with each other can be through some interfaces, and the indirect coupling or communication connection of devices or modules can be electrical, mechanical, or other forms.

[0070] In addition, each functional unit in the embodiments of this application can be all integrated in a processor, or each unit can be separately used as a device, or two or more units can be integrated in a device; each functional unit in the embodiments of this application can be implemented in the form of hardware, or in the form of a combination of hardware and software functional units.

[0071] Those of ordinary skill in the art can understand that all or part of the steps of implementing the following method embodiments can be completed through program instructions and related hardware. The foregoing program instructions can be stored in a computer-readable storage medium. When the program instructions are executed, the steps of the following method embodiments are executed; and the foregoing storage medium includes: various media that can store program codes such as removable storage devices, read-only memory (ROM), magnetic disks, or optical discs.

[0072] In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" can explicitly or implicitly include one or more of such features. In the description of this application, the meanings of "a plurality" and "several" are two or more, unless otherwise clearly and specifically defined.

[0073] The embodiments of this application are written in a progressive manner.

[0074] Such asFigure 1 As shown, the embodiment of the present application provides a two-stage bearing life prediction method, comprising:

[0075] S101. Obtaining online monitoring data, wherein the online monitoring data includes bearing fault impact signal data and bearing operating mileage data;

[0076] In S101, the online monitoring data can be obtained through an online monitoring system, or directly obtained through other methods. The specific acquisition method does not affect the implementation of this embodiment. The bearing fault impact signal data can be impact signal data generated by the bearing during operation due to component failure. Specifically, the bearing fault impact signal data can include bearing outer ring impact data, etc. It should be noted that when the actual operating time or number of revolutions of the bearing can be obtained, the bearing operating mileage data can also be replaced with the actual operating time or number of revolutions of the bearing. This application does not impose any restrictions on this.

[0077] S102. Preprocessing the bearing fault impact signal data to obtain preprocessed bearing fault impact signal data;

[0078] In S102, the preprocessed bearing fault impact signal data is used for subsequent bearing life prediction, and the preprocessing process can improve the accuracy of subsequent model predictions. Considering the existence of data anomalies and data missing during data collection, the preprocessing process may include outlier processing and missing value processing. Specifically, one implementation method of this step may include the following steps: removing outliers in the bearing fault impact signal data, and interpolating missing values in the bearing fault impact signal data using a truncated normal distribution to obtain preprocessed bearing fault impact signal data. For example, when the bearing fault impact signal data includes bearing outer ring impact data, outliers in the bearing outer ring impact data may be removed; wherein, the truncated normal distribution (Truncated_normal_distribution) is a distribution that defines the range of values of random variables in the normal distribution. The truncation range can be divided into four cases: no truncation; lower bound truncation; upper bound truncation; upper and lower bound truncation. The probability density function formula of the truncated normal distribution is as follows:

[0079]

[0080] Where φ[·] represents the probability density function of the normally distributed random variable, Φ[·] represents the distribution function of the normally distributed random variable, μ represents the mean of the standard normal distribution, σ represents the standard deviation of the standard normal distribution, and x1 and x2 represent the upper and lower bounds of the value of the truncated normally distributed random variable.

[0081] Based on the characteristics of data collection and the causes of equipment failures, the above implementation steps analyze the principle of data anomalies and use an imputation method for missing impact information based on statistical analysis to solve problems such as discontinuous and severely missing bearing monitoring data. Compared with methods that preprocess abnormal data from a pure data perspective, this method can make the processed data more conform to physical logic and closer to the true value.

[0082] S103. Process the preprocessed bearing fault impact signal data and the bearing operation mileage data to obtain a bearing degradation characteristic value, where the bearing degradation characteristic value is a smoothed impact characteristic change rate value.

[0083] In S103, by performing calculation processing on the preprocessed bearing fault impact signal data and the bearing operation mileage data, a smoothed impact characteristic change rate value can be obtained as the bearing degradation characteristic value. The bearing degradation characteristic value can well reflect the physical failure characteristics of the bearing degradation trajectory and the bearing fault development law. At the same time, with the update of on-line monitoring data, dynamic extraction of the bearing physical degradation characteristics can be realized, and the problem that it is difficult to extract the complex physical degradation characteristics of the bearing failure mechanism and has randomness in the physical model life prediction method can be solved.

[0084] S104. Determine the current degradation stage according to the bearing degradation characteristic value and the empirical threshold, where the degradation stage includes a stable operation stage and a rapid degradation stage. If the current degradation stage is the stable operation stage, then execute S105; if the current degradation stage is the rapid degradation stage, then execute S106.

[0085] In S104, the empirical threshold can be obtained based on historical test analysis. Considering the characteristics of the impact signal of mechanical components, a three-stage degradation process of the bearing is established, namely the running-in period, the stable operation stage, and the rapid degradation stage. At the same time, four key points are determined, namely the change point between the running-in period and the stable operation stage, the change point between the stable operation stage and the rapid degradation stage, the first prediction point, and the failure point. Since the bearing impact characteristics are unstable during the running-in period when the bearing is initially put into use and cannot show the degradation trend characteristics, this embodiment does not consider including the running-in period in the feature prediction process. At the same time, in this embodiment, since the bearing impact characteristics increase non-linearly during the rapid degradation stage, considering the need for original data accumulation in degradation modeling and life prediction, after the bearing enters the rapid degradation stage and experiences a certain amount of data accumulation, the remaining operation mileage prediction starts from the first prediction point and does not use the change point between the stable operation stage and the rapid degradation stage as a criterion. The bearing health status degradation trend diagram is as Figure 2As shown. One implementation of this step may include the following steps: Determine the first predicted value according to an empirical threshold; Determine whether the bearing degradation eigenvalue is greater than or equal to the first predicted value. If so, determine that the current degradation stage is the rapid degradation stage. If not, determine that the current degradation stage is the stable operation stage. Among them, the first predicted value is used to represent the position of the first prediction point.

[0086] S105. Use the pre-established stable operation stage prediction model to predict the remaining operating mileage of the current bearing.

[0087] In S105, when the current degradation stage is the stable operation stage, use the pre-established stable operation stage prediction model to predict the remaining operating mileage. The stable operation stage prediction model can preferably be a Weibull distribution model:

[0088]

[0089] In the formula, η is the scaling factor, m is the shape parameter, and both η and m are statistical empirical values, and t is the mileage value.

[0090] The Weibull distribution is a continuous distribution that can describe the distribution law of failure data of various types of mechanical components. It has been applied to a certain extent in life data analysis, reliability design, fatigue reliability analysis, maintenance decision-making, warranty strategy formulation, etc. Using the Weibull distribution model can better fit the degradation process in the stable operation stage. The stable operation stage prediction model can also be any existing model as long as it conforms to the degradation process in the stable operation stage. This embodiment does not limit this.

[0091] S106. Use the pre-established rapid degradation stage prediction model to predict the remaining operating mileage of the current bearing.

[0092] In S106, when the current degradation stage is the rapid degradation stage, use the pre-established rapid degradation stage prediction model to predict the remaining operating mileage. Specifically, the point estimate and interval estimate of the remaining operating mileage of the bearing at a certain confidence level can be predicted and calculated using the rapid degradation stage prediction model. The rapid degradation stage prediction model can preferably be an exponential distribution model with an error term:

[0093] X(t) = aexp(bt) + ε

[0094] In the formula, a and b represent the rapid degradation stage parameters, t is the mileage value, ε represents white noise, and ε ~ N(0, σ 2)。The exponential distribution model with an error term can better fit the degradation process in the rapid degradation stage. The rapid degradation stage prediction model can also be any existing model as long as it conforms to the degradation process in the rapid degradation stage, and this is not limited in this embodiment.

[0095] Since the failure modes of bearings are complex and diverse, it is difficult to characterize the true physical degradation process of the equipment, making it difficult to accurately predict the life. And currently, in the prediction of the remaining life of rotating components of equipment, there are problems such as the monotonicity and trend of the constructed characteristic factors not being ideal enough, resulting in low accuracy of the life prediction results.

[0096] In the above embodiment, based on the online monitoring data, the life of the bearing is predicted, which can comprehensively consider the dynamics in the time dimension and the differences in failure modes during the operation of the bearing; by preprocessing the bearing fault impact signal data in the online monitoring data, and then based on the preprocessed bearing fault impact signal data and the bearing operation mileage data, the bearing degradation characteristic value is obtained, that is, the impact characteristic change rate value after smoothing. The bearing degradation characteristic value can well reflect the physical failure characteristics of the bearing degradation trajectory and the bearing fault development law. At the same time, with the update of the online monitoring data, the dynamic extraction of the bearing physical degradation characteristics can be realized, which can solve the problems that the bearing failure mechanism is complex and the physical degradation characteristics are difficult to extract and have randomness in the physical model life prediction method; by combining the multi-stage complex degradation behaviors of the bearing under different failure modes, using the pre-established two-stage remaining operation mileage prediction model, that is, the stable operation stage prediction model and the rapid degradation stage prediction model, the prediction of the remaining operation mileage of the bearing is realized, solving the problems such as the complex degradation stage of the current bearing and the difficult determination of the failure threshold, and with the continuous update of the online monitoring data, the phased dynamic real-time remaining life prediction of the bearing can be realized. In addition, the above embodiment is applicable to the life prediction of all bearing types. In summary, the above embodiment can improve the accuracy of bearing life prediction.

[0097] As Figure 3 shown, on the basis of the above embodiment, the processing of the preprocessed bearing fault impact signal data and the bearing operation mileage data to obtain the bearing degradation characteristic value includes:

[0098] S201. According to the preprocessed bearing fault impact signal data and the bearing operation data, using the moving average window method, obtain the root mean square value of the bearing fault impact signal within the interval and the average mileage value within the interval;

[0099] In S201, specifically, an equal mileage sliding interval with a sliding interval length of N1 and a right shift of 0.5N1 km each time can be adopted to calculate the root mean square value y of the bearing fault impact signal within the interval rmsand the average mileage value t within this interval mean .

[0100] S202. Smooth the root mean square value to obtain the smoothed root mean square value;

[0101] In S202, specifically, a moving average filter can be used to smooth the root mean square value y of the bearing fault impact signal within the interval rms to obtain the smoothed root mean square value y rms-smooth , where the continuous sampling data is regarded as a sequence with a fixed length of N2. After each smoothing, the first data is removed, new data is inserted at its tail, and then arithmetic operations are performed on this new sequence. The specific calculation formula is as follows:

[0102]

[0103] In the formula, N2 represents the window length of the moving filter, and t j represents the position corresponding to each mileage within the interval.

[0104] S203. After taking the difference between the smoothed root mean square value obtained for each window and the average mileage value respectively and then comparing them, obtain the impact characteristic change rate value;

[0105] In S203, the calculation formula for the impact characteristic change rate value is as follows:

[0106]

[0107] In the formula, y rms-smooth-rate represents the impact characteristic change rate value.

[0108] S204. Smooth the impact characteristic change rate value to obtain the bearing degradation characteristic value.

[0109] In S204, in the same way as the formula in S202 above, smooth the impact characteristic change rate value y rms-smooth-rate with the moving filter window length of N4 to obtain the smoothed impact characteristic change rate value y rms-smooth-rate-smooth , that is, the bearing degradation characteristic value.

[0110] The above implementation steps, based on the moving average window method, can denoise the data, calculate the smoothed impact characteristic change rate, and can well reflect the trend change of the bearing degradation trajectory.

[0111] In other embodiments of the present application, after the step of obtaining the online monitoring data, the following steps may further be included:

[0112] S107. Modify the bearing operation mileage data to obtain the modified bearing operation mileage data;

[0113] Correspondingly, the processing of the preprocessed bearing fault impact signal data and the bearing operation mileage data to obtain the bearing degradation eigenvalue specifically includes: processing the preprocessed bearing fault impact signal data and the modified bearing operation mileage data to obtain the bearing degradation eigenvalue.

[0114] In S107, the modified bearing operation mileage data is used for subsequent bearing life prediction, and the modification process can improve the accuracy of subsequent model prediction. Considering that the development trend of the already-operated mileage data should be continuously and monotonically increasing, but when a new host is replaced and the mileage is reset to zero, there are errors in the host file record of the data recording device, or on-site mileage calibration is performed, the already-operated mileage data will show abnormal decreases or increases. One implementation method of this step may include the following steps: According to the principles of monotonicity and continuity, correct the abnormal data in the bearing operation mileage data to obtain the modified bearing operation mileage data. It should be noted that the execution order of S107 and S102 can be interchanged or they can be executed simultaneously, which does not affect the implementation of this embodiment.

[0115] In other embodiments of the present application, the use of the pre-established steady operation stage prediction model for life prediction to obtain the remaining operation mileage of the current bearing includes:

[0116] S301. Combine the historical bearing degradation eigenvalues to perform parameter estimation on the pre-established steady operation stage prediction model to obtain the first parameter estimation value, where the historical bearing degradation eigenvalue is the smoothed impact feature change rate value obtained based on historical monitoring data, and the historical monitoring data includes bearing historical fault impact signal data and bearing historical operation mileage data;

[0117] In S301, the historical bearing degradation eigenvalue is the rate of change of the impact feature after smoothing based on historical monitoring data, which can be obtained in advance before parameter estimation. Its specific acquisition method may include the following steps: obtaining historical monitoring data; preprocessing the bearing historical fault impact signal data to obtain the preprocessed bearing historical fault impact signal data; correcting the bearing historical operating mileage data to obtain the corrected bearing historical operating mileage data; processing the preprocessed bearing historical fault impact signal data and the corrected bearing historical operating mileage data to obtain the historical bearing degradation eigenvalue. Among them, the historical monitoring data can be obtained through an online monitoring system, and it can also be directly obtained through other means. Its specific acquisition method does not affect the implementation of this embodiment. The bearing historical fault impact signal data may be the impact signal data generated by the bearing during operation due to historical component failures. The time period for collecting the historical monitoring data can be determined according to actual needs. For the specific calculation process of the historical bearing degradation eigenvalue, reference can be made to Figure 3 S201 - S204 in the embodiment shown. One implementation manner of this step may include the following steps: Combining the historical bearing degradation eigenvalue, using numerical methods to iteratively solve the likelihood equation of the pre - established prediction model for the steady - running stage to obtain the first parameter estimate value.

[0118] S302. According to the first parameter estimate value and the prediction model for the steady - running stage, perform life prediction to obtain the remaining operating mileage of the current bearing;

[0119] In S302, specifically, taking the prediction model for the steady - running stage as the Weibull distribution model as an example, one implementation manner of this step may include the following steps:

[0120] S3021. Based on the first parameter estimate value and the prediction model for the steady - running stage, solve the confidence interval estimation result for the steady - running stage according to the confidence level of the given upper and lower quantiles;

[0121] In S3021, the calculation method for the upper and lower interval values of the confidence interval estimation result is as follows:

[0122] t L = η[-ln(a / 2)] 1 / m

[0123] t U = η[-ln(1 - α / 2)] 1 / m

[0124] In the formula, t L represents the upper interval value, t UDenote the lower interval value, α is the significance level, η is the scaling factor, m is the shape parameter, and both η and m are statistical empirical values.

[0125] S3022. Obtain the predicted interval of the remaining operating mileage from the current mileage at the steady operation stage to the first prediction point according to the confidence interval estimation result;

[0126] In S3022, specifically, according to the confidence interval estimation result, obtain the predicted interval of the remaining operating mileage from the current mileage at the steady operation stage t i The predicted interval of the remaining operating mileage from the current mileage at the steady operation stage to the first prediction point is [t L -t i , t U -t i .

[0127] S3023. Add the predicted interval of the remaining operating mileage from the current mileage at the steady operation stage to the first prediction point to the empirical value of the remaining life after the first prediction point to obtain the remaining operating mileage of the current bearing.

[0128] In S3023, specifically, add the predicted interval of the remaining operating mileage from the current mileage at the steady operation stage t i to the empirical value of the remaining life after the first prediction point to obtain the remaining operating mileage of the current bearing. The empirical value of the remaining life after the first prediction point can be obtained based on historical test analysis.

[0129] Correspondingly, the life prediction using the pre-established rapid degradation stage prediction model to obtain the remaining operating mileage of the current bearing includes:

[0130] S401. Combine the historical bearing degradation characteristic values to estimate the parameters of the pre-established rapid degradation stage prediction model to obtain the second parameter estimation value. Among them, the historical bearing degradation characteristic values are the smoothed impact characteristic change rate values obtained based on historical monitoring data, and the historical monitoring data includes bearing historical fault impact signal data and bearing historical operating mileage data;

[0131] In S401, the acquisition methods of the historical bearing degradation characteristic values and the historical monitoring data can refer to the above S301. One implementation method of this step may include the following steps: Combine the historical bearing degradation characteristic values and estimate the parameters of the pre-established rapid degradation stage prediction model by the maximum likelihood method to obtain the second parameter estimation value. Taking the rapid degradation stage prediction model as an exponential distribution model with an error term as an example, the maximum likelihood function equation of the exponential model with an error term in the rapid degradation stage can be solved. First, find the least squares solution of parameter b, and then substitute it back into other parameter equations to find the optimal closed solution of its parameters to obtain the second parameter estimation value.

[0132] S402. Based on the second parameter estimation value and the rapid degradation stage prediction model, perform life prediction to obtain the remaining operating mileage of the current bearing.

[0133] In S402, specifically, one implementation of this step may include the following steps: Based on the second parameter estimation value and the rapid degradation stage prediction model, use Monte Carlo simulation to obtain the point estimate and interval estimate of the remaining operating mileage of the current bearing. Taking the exponential distribution model with error terms of the rapid degradation stage prediction model as an example, the process is as follows:

[0134] Substitute the second parameter estimation value into the RUL (Remaining Useful Life) calculation formula:

[0135]

[0136] In the formula, a and b represent the rapid degradation stage parameters, w is the extracted bearing degradation eigenvalue y rms-smooth-rate-smooth , t k is the current operating mileage, and the point estimate result of the remaining operating mileage of the bearing can be calculated.

[0137] At mileage t k moment, randomly generate N k groups of parameters according to the parameter estimation mean and variance, where the parameters a and b respectively follow the normal distribution corresponding to their means and variances, that is:

[0138]

[0139]

[0140] In the formula, j = 1, 2, ···, N k .

[0141] Substitute each group of parameters (a k,j , b k,j ) into the RUL calculation formula to obtain N k estimation results r k =(r k,j |j = 1, 2, ···, N k ), fit r k with the normal distribution, and calculate the mean μ and variance σ 2 ;

[0142] Given the confidence level of the upper and lower quantiles, the values of the upper and lower quantiles corresponding to the normal distribution N(mean(r k ), var(r k )) are the interval estimation results of RUL, denoted as Among them,

[0143]

[0144]

[0145] in the formula, α is the significance level, F -1 is the inverse normal distribution function, μ is the mean, and σ is the standard deviation.

[0146] As Figure 4 shown, in another embodiment of the present application, a bearing two-stage life prediction system is further provided, including:

[0147] A data acquisition module 11, configured to acquire online monitoring data, where the online monitoring data includes bearing fault impact signal data and bearing operation mileage data;

[0148] A data processing module 12, configured to preprocess the bearing fault impact signal data to obtain preprocessed bearing fault impact signal data;

[0149] A feature extraction module 13, configured to process the preprocessed bearing fault impact signal data and the bearing operation mileage data to obtain bearing degradation characteristic values, where the bearing degradation characteristic values are smoothed impact characteristic change rate values;

[0150] A stage identification module 14, configured to determine the current degradation stage according to the bearing degradation characteristic values and empirical thresholds, where the degradation stage includes a stable operation stage and a rapid degradation stage;

[0151] A first prediction module 15, configured to use a pre-established stable operation stage prediction model for life prediction to obtain the remaining operation mileage of the current bearing when the current degradation stage is the stable operation stage;

[0152] A second prediction module 16, configured to use a pre-established rapid degradation stage prediction model for life prediction to obtain the remaining operation mileage of the current bearing when the current degradation stage is the rapid degradation stage.

[0153] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A two-stage bearing life prediction method, characterized in that Including: Obtain online monitoring data, where the online monitoring data includes bearing fault impact signal data and bearing operation mileage data; Preprocess the bearing fault impact signal data to obtain preprocessed bearing fault impact signal data; Process the preprocessed bearing fault impact signal data and the bearing operation mileage data to obtain bearing degradation characteristic values, where the bearing degradation characteristic values are impact characteristic change rate values after smoothing; Determine the current degradation stage according to the bearing degradation characteristic values and empirical thresholds, where the degradation stage includes a stable operation stage and a rapid degradation stage; If the current degradation stage is the stable operation stage, use a pre-established stable operation stage prediction model for life prediction to obtain the remaining operation mileage of the current bearing; If the current degradation stage is the rapid degradation stage, use a pre-established rapid degradation stage prediction model for life prediction to obtain the remaining operation mileage of the current bearing; Among them, the use of the pre-established stable operation stage prediction model for life prediction to obtain the remaining operation mileage of the current bearing includes: Combined with historical bearing degradation characteristic values, perform parameter estimation on the pre-established stable operation stage prediction model to obtain a first parameter estimation value, where the historical bearing degradation characteristic values are impact characteristic change rate values after smoothing based on historical monitoring data, and the historical monitoring data includes bearing historical fault impact signal data and bearing historical operation mileage data; According to the first parameter estimation value and the stable operation stage prediction model, perform life prediction to obtain the remaining operation mileage of the current bearing; Correspondingly, the use of the pre-established rapid degradation stage prediction model for life prediction to obtain the remaining operation mileage of the current bearing includes: Combined with historical bearing degradation characteristic values, perform parameter estimation on the pre-established rapid degradation stage prediction model to obtain a second parameter estimation value, where the historical bearing degradation characteristic values are impact characteristic change rate values after smoothing based on historical monitoring data, and the historical monitoring data includes bearing historical fault impact signal data and bearing historical operation mileage data; According to the second parameter estimation value and the rapid degradation stage prediction model, perform life prediction to obtain the remaining operation mileage of the current bearing; The pre-established stable operation stage prediction model is a Weibull distribution model: ； In the formula, is the scaling factor, is the shape parameter, and are both statistically empirical values, is the mileage value; The rapid degradation stage prediction model is an exponential distribution model with an error term: ； where a and b represent the parameters in the rapid degradation stage, ε represents white noise, , represents the variance of the normal distribution.

2. The method according to claim 1, wherein The processing of the preprocessed bearing fault impact signal data and the bearing operation mileage data to obtain bearing degradation characteristic values includes: According to the preprocessed bearing fault impact signal data and the bearing operation mileage data, use the moving average window method to obtain the root mean square value of the bearing fault impact signal within the interval and the average mileage value within the interval; Perform smoothing processing on the root mean square value to obtain a smoothed root mean square value; The smoothed root mean square value and the average mileage value obtained for each window are respectively differentiated and then compared to obtain the impact feature change rate value; The impact feature change rate value is smoothed to obtain the bearing degradation feature value.

3. The method according to claim 1 or 2, characterized in that, The determining the current degradation stage according to the bearing degradation feature value and the empirical threshold includes: Determining the first predicted value according to the empirical threshold; Judging whether the bearing degradation feature value is greater than or equal to the first predicted value. If so, determining that the current degradation stage is the rapid degradation stage; if not, determining that the current degradation stage is the stable operation stage.

4. The method according to claim 1, wherein The parameter estimation of the pre-established stable operation stage prediction model by combining the historical bearing degradation feature values to obtain the first parameter estimation value includes: Combining the historical bearing degradation feature values and using numerical methods to iteratively solve the likelihood equation of the pre-established stable operation stage prediction model to obtain the first parameter estimation value.

5. The method according to claim 1, characterized in that, The remaining operating mileage of the current bearing is obtained by performing life prediction according to the first parameter estimation value and the pre-established stable operation stage prediction model, including: Based on the first parameter estimation value and the stable operation stage prediction model, the confidence interval estimation result of the stable operation stage is solved according to the given confidence level of the upper and lower quantiles; According to the confidence interval estimation result, the remaining operating mileage prediction interval from the current mileage of the stable operation stage to the first predicted point is obtained; Adding the remaining operating mileage prediction interval from the current mileage of the stable operation stage to the first predicted point to the empirical value of the remaining life after the first predicted point to obtain the remaining operating mileage of the current bearing.

6. The method according to claim 1, characterized in that The parameter estimation of the pre-established rapid degradation stage prediction model by combining the historical bearing degradation feature values to obtain the second parameter estimation value includes: Combining the historical bearing degradation feature values and performing parameter estimation on the pre-established rapid degradation stage prediction model by the maximum likelihood method to obtain the second parameter estimation value.

7. The method according to claim 1, wherein The remaining operating mileage of the current bearing is obtained by performing life prediction according to the second parameter estimation value and the rapid degradation stage prediction model, including: Based on the second parameter estimation value and the rapid degradation stage prediction model, the point estimation and interval estimation of the remaining operating mileage of the current bearing are obtained by using Monte Carlo simulation.

8. The method according to claim 1, wherein The preprocessing of the bearing fault impact signal data to obtain the preprocessed bearing fault impact signal data includes: Removing the outliers in the bearing fault impact signal data and using the truncated normal distribution to interpolate the missing values in the bearing fault impact signal data to obtain the preprocessed bearing fault impact signal data.

9. The method according to claim 1, wherein After obtaining the online monitoring data, it further includes: Correcting the bearing operating mileage data to obtain the corrected bearing operating mileage data; Correspondingly, the processing of the preprocessed bearing fault impact signal data and the bearing operating mileage data to obtain the bearing degradation feature value includes: Processing the preprocessed bearing fault impact signal data and the corrected bearing operating mileage data to obtain the bearing degradation feature value.

10. The method according to claim 9, characterized in that, Correct the bearing operation mileage data to obtain the corrected bearing operation mileage data, including: According to the principles of monotonicity and continuity, correct the abnormal data in the bearing operation mileage data to obtain the corrected bearing operation mileage data.

11. A bearing two-stage life prediction system, characterized in that, Applied to the bearing two-stage life prediction method described in claim 1, including: A data acquisition module for acquiring online monitoring data, where the online monitoring data includes bearing fault impact signal data and bearing operation mileage data; A data processing module for preprocessing the bearing fault impact signal data to obtain the preprocessed bearing fault impact signal data; A feature extraction module for processing the preprocessed bearing fault impact signal data and the bearing operation mileage data to obtain bearing degradation characteristic values, where the bearing degradation characteristic values are the impact characteristic change rate values after smoothing; A stage identification module for determining the current degradation stage according to the bearing degradation characteristic values and empirical thresholds, where the degradation stage includes a steady operation stage and a rapid degradation stage; A first prediction module for, when the current degradation stage is the steady operation stage, using a pre-established steady operation stage prediction model to perform life prediction to obtain the remaining operation mileage of the current bearing; A second prediction module for, when the current degradation stage is the rapid degradation stage, using a pre-established rapid degradation stage prediction model to perform life prediction to obtain the remaining operation mileage of the current bearing.

Images