Method and system for predicting and early warning health state of rolling bearing driven by digital-analog fusion under variable working conditions
Through the digital-to-analog fusion drive method, the characteristic data of rolling bearings is extracted and processed, the degradation model library is established, and the time-varying particle filtering algorithm is used to predict, which solves the accuracy of rolling bearing life prediction under variable working conditions and realizes the healthy state of rolling bearings Accurate prediction and early warning.
Patent Information
- Application Number
- CN202411073806.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-11-22
AI Technical Summary
The existing rolling bearing life prediction methods are difficult to accurately predict under variable working conditions. The traditional method is mainly based on constant working conditions and cannot be applied to the degradation model of bearings under variable working conditions, resulting in low prediction accuracy and poor interpretability.
By using digital-to-analog fusion-driven method, by obtaining the original operating data of rolling bearings, extracting the time-domain and time-frequency domain features, performing cumulative transformation and baseline conversion, selecting features related to fault development for dimensionality reduction fusion, establishing a degradation model library, and Use time-varying particle filtering algorithm to predict and alert health status.
Accurate prediction and early warning of rolling bearing health status under variable working conditions is realized, the accuracy and interpretability of prediction are improved, and the degradation model can be adaptively selected according to the dynamic deterioration of bearings, providing real-time health status information.
Smart Images

Figure CN119004260A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of rolling bearing state prediction, and in particular to a rolling bearing health state prediction and early warning method and system driven by digital-analog fusion under variable working conditions. Background Art
[0002] Rolling bearings are called "the joints of industry" and are one of the most vulnerable parts. Their operating status determines the safety performance of the entire machine. Among large rotating machinery failures, about 40% of major failures are caused by bearing defects. Accurate prediction of the remaining life of bearings can timely grasp the operating status and remaining working time of the equipment, and prevent major economic losses and casualties caused by unexpected failures. In actual industrial sites, rolling bearings often do not operate under constant conditions, which affects the degradation process of the bearings, making the mapping between characteristics and the health status of rolling bearings complicated, making it difficult to accurately predict the remaining life.
[0003] In the research on the remaining life prediction of rolling bearings, most of them are based on constant working conditions. The remaining life prediction methods are mainly divided into model-driven remaining life prediction and data-driven remaining life prediction. For model-driven, whether the model can accurately reflect the system plays a decisive role in the prediction effect. However, this method requires a full understanding of the degradation mechanism of the equipment and needs to consider the coupling of multiple external factors, which makes it difficult to determine the failure physical model of many equipment in practical applications. For data-driven remaining life prediction, it does not require model control, directly extracts relevant features from the historical operation data of the equipment and learns the system degradation behavior, and then predicts the remaining life. However, the accuracy of this method often depends on a large number of historical failure samples, which are difficult to obtain in practice, and the prediction process is poorly interpretable.
[0004] Because the suitable degradation model for rolling bearings under variable working conditions is different, and the constructed single model cannot be applicable to the entire life cycle, the traditional life prediction method for rolling bearings cannot accurately predict the remaining service life of rolling bearings under variable working conditions. Summary of the invention
[0005] The purpose of this application is to provide a rolling bearing health status prediction and early warning method and system driven by digital-analog fusion under variable working conditions, which can realize accurate prediction of the health status of rolling bearings under variable working conditions.
[0006] To achieve the above objectives, this application provides the following solutions:
[0007] In a first aspect, the present application provides a method for predicting and warning the health status of a rolling bearing driven by digital-analog fusion under variable working conditions, comprising the following steps:
[0008] The original operating data of the rolling bearing under the operating condition is obtained, and the sample points of the original operating data are screened based on the principle of small difference in initial conditions to obtain the screened original operating data.
[0009] For the filtered original operating data, several time domain features and time-frequency domain features are extracted respectively, and the slope of each feature is calculated and cumulatively transformed to form an original feature set.
[0010] The slope of each feature in the original feature set is baseline transformed, and the slope of each feature after baseline transformation is smoothed without changing the degradation trend to obtain a candidate feature set.
[0011] The preferred features that are highly correlated with fault development are selected from the candidate feature set, and the preferred features are fused by dimensionality reduction to obtain the fused health index.
[0012] According to the empirical model, multiple degradation models are selected to establish a rolling bearing degradation model library, and the first degradation time of the rolling bearing is determined based on the rolling bearing degradation model library and the fusion health index.
[0013] Based on the first degradation time, a time-varying particle filter algorithm and a rolling bearing degradation model library are used to predict the health status of the rolling bearing according to the measured operation data of the rolling bearing, and obtain the rolling bearing health status prediction result; the time-varying particle filter algorithm is used to select the optimal degradation model from the rolling bearing degradation model library for prediction based on the measured operation data of the rolling bearing within a sliding window of a preset length.
[0014] According to the rolling bearing health status prediction results and preset warning thresholds, rolling bearing health status warning is carried out.
[0015] Optionally, the original operating data includes operating data of the rolling bearing under operating conditions and acceleration detection data of each detection position on the rolling bearing under operating conditions; the operating data of the rolling bearing under operating conditions includes: temperature, rotation speed and load.
[0016] Based on the principle of small difference in initial conditions, sample points of the original operating data are screened to obtain screened original operating data. Specifically, by analyzing the initial conditions of the rolling bearing, sample points with small difference in initial conditions are screened to obtain screened original operating data.
[0017] Optionally, the time domain features include: absolute average amplitude, root mean square, maximum value, peak value, variance, root square amplitude, skewness, kurtosis coefficient, waveform index, peak index, margin index and pulse index; the time-frequency domain features are several wavelet packet coefficients obtained by decomposing the filtered original operating data through wavelet packet transform.
[0018] The slope of each feature is calculated and cumulatively transformed to form the original feature set, which specifically includes the following steps:
[0019] For several time domain features and time-frequency domain features, the difference method is used to calculate the differences between adjacent sample points to estimate the slope of each feature.
[0020] By cumulatively transforming the slope of each feature to form the original feature set, the overall information of the data can be mined.
[0021] Optionally, the cumulative transformation comprises the following steps:
[0022] Select a characteristic sequence of the rolling bearing under normal conditions, and calculate the average value of the characteristic sequence as the standard value.
[0023] Based on the standard value and the slope of each feature, the cumulative index of each feature is calculated.
[0024] Optionally, the mean value of the feature sequence is calculated according to the following formula:
[0025]
[0026] Among them, f nor is the average value of the feature sequence, m is the starting index of the feature sequence, n is the total number of sample points in the feature sequence, and f(i) is the i-th eigenvalue of the feature sequence.
[0027] The cumulative index of a feature is calculated according to the following formula:
[0028]
[0029] Among them, cf n Represents the cumulative index of the nth feature.
[0030] Optionally, the slope of the feature is baseline transformed according to the following formula:
[0031]
[0032] in, t k The baseline observation value at time, y k t k The slope of the characteristic at time s k t k Speed at the moment, s B is the baseline velocity, a and b are the baseline conversion parameters.
[0033] The baseline conversion parameters are determined according to the following formula:
[0034]
[0035] Among them, y i t i The slope of the characteristic at time s i t i The speed of the moment, is the baseline speed at t i Interpolation of time.
[0036] Optionally, a preferred feature highly correlated with the fault development is selected from the candidate feature set, specifically: a linear combination of time correlation and monotonicity is used as a comprehensive evaluation index, and a preferred feature highly correlated with the fault development is selected from the candidate feature set.
[0037] The comprehensive evaluation index of the feature is calculated according to the following formula:
[0038] W = Mon(X) + Corr(X, T).
[0039] Among them, W is the comprehensive evaluation index of the feature, Mon() is the monotonicity function, Corr() is the time correlation function, X is the feature sequence, and T is time.
[0040] Optionally, the selected features are subjected to dimensionality reduction fusion to obtain a fusion health index, which specifically includes the following steps:
[0041] The z-score standardization method is used to process the selected features and obtain the time series set.
[0042] Through the kernel principal component analysis method, the time series set is fused by dimensionality reduction to obtain the fused health index.
[0043] Optionally, the degradation models in the rolling bearing degradation model library include: a power function degradation model, an exponential function degradation model and a logarithmic function degradation model.
[0044] Based on the first degradation time, the time-varying particle filter algorithm and the rolling bearing degradation model library are used to predict the health status of the rolling bearing according to the measured operation data of the rolling bearing, which specifically includes the following steps:
[0045] When the rolling bearing begins to degrade based on the first degradation time, a sliding window with a preset length is set using a time-varying particle filter algorithm to obtain segment data from the measured operation data of the rolling bearing.
[0046] The power function degradation model, exponential function degradation model and logarithmic function degradation model are respectively used to fit the data under the window and determine the optimal degradation model; the optimal degradation model is used to predict the health status of the rolling bearing within a preset time period based on the segment data.
[0047] In the second aspect, the present application provides a rolling bearing health status prediction and early warning system driven by digital-analog fusion under variable working conditions, which is used to implement the rolling bearing health status prediction and early warning method driven by digital-analog fusion under variable working conditions as described above, and the rolling bearing health status prediction and early warning system driven by digital-analog fusion under variable working conditions includes the following modules:
[0048] The original operation data acquisition module is used to acquire the original operation data of the rolling bearing under the operation condition.
[0049] The original operation data screening module is used to screen the sample points of the original operation data based on the principle of small difference in initial conditions to obtain screened original operation data.
[0050] The feature extraction and cumulative transformation module is used to extract a number of time domain features and time-frequency domain features from the filtered original operation data, and calculate the slope of each feature for cumulative transformation to form an original feature set.
[0051] The baseline conversion and smoothing processing module is used to perform baseline conversion on the slope of each feature in the original feature set, and smooth the slope of each feature after baseline conversion without changing the degradation trend, so as to obtain a candidate feature set.
[0052] The feature optimization and dimensionality reduction fusion module is used to select the preferred features that are highly correlated with the fault development from the candidate feature set, and perform dimensionality reduction fusion on the preferred features to obtain a fusion health index.
[0053] The first degradation time determination module is used to select multiple degradation models according to the empirical model to establish a rolling bearing degradation model library, and determine the first degradation time of the rolling bearing based on the rolling bearing degradation model library and the fusion health index.
[0054] The rolling bearing health status prediction module is used to predict the health status of the rolling bearing based on the first degradation time, using a time-varying particle filter algorithm and the rolling bearing degradation model library, and according to the actual measured operation data of the rolling bearing to obtain a rolling bearing health status prediction result; the time-varying particle filter algorithm is used to select the optimal degradation model from the rolling bearing degradation model library for prediction based on the actual measured operation data of the rolling bearing within a sliding window of a preset length.
[0055] The rolling bearing health status warning module is used to provide rolling bearing health status warning based on the rolling bearing health status prediction results and preset warning thresholds.
[0056] According to the specific embodiments provided in this application, this application discloses the following technical effects:
[0057] The present application provides a method and system for predicting and warning the health status of rolling bearings driven by digital-analog fusion under variable working conditions. In the method, after obtaining the original operating data of the rolling bearing under the operating conditions, the sample points of the original operating data are screened based on the principle of small difference in initial conditions to avoid data differences; then, a number of time domain features and time-frequency domain features are extracted respectively, and the slope of each feature is calculated for cumulative transformation to form an original feature set; without changing the degradation trend, the slope of each feature in the original feature set is first baseline converted, and then smoothed to obtain a candidate feature set; selected preferred features that are highly correlated with fault development are selected from the candidate feature set, and the selected features are dimensionality reduced and fused to obtain a fused health index; then, multiple degradation models are selected according to the empirical model to establish a rolling bearing degradation model library, and based on the rolling bearing degradation model library, a rolling bearing degradation model library is established. The model library and the integrated health indicators are used to determine the first degradation time of the rolling bearing; that is, when judging the degradation risk of the rolling bearing based on the first degradation time, the time-varying particle filter algorithm and the rolling bearing degradation model library are used to predict and warn the health status of the rolling bearing according to the measured operation data of the rolling bearing and the preset warning threshold; on the one hand, the above-mentioned scheme of the present application takes into account the accumulation of bearing damage, realizes the purpose of mining the overall information of the data by accumulating the original features, and eliminates the excessive fluctuation of the characteristic observation values caused by the change of working conditions based on the baseline conversion and smoothing processing; on the other hand, a rolling bearing degradation model library based on the empirical model is established, and the time-varying particle filter algorithm is used to adaptively select the degradation model according to the dynamic degradation of the bearing, so as to realize the accurate prediction and warning of the health status of the rolling bearing under the changing working conditions. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0059] Figure 1 This is a flow chart of a method for predicting and warning the health status of a rolling bearing driven by digital-analog fusion under variable working conditions in one embodiment of the present application.
[0060] Figure 2 A schematic diagram of a horizontal vibration signal of a bearing used in a rolling bearing health status prediction and early warning method driven by digital-analog fusion under variable working conditions provided in another embodiment of the present application.
[0061] Figure 3 A schematic diagram of degradation period characteristics after baseline conversion and smoothing in a rolling bearing health status prediction and early warning method driven by digital-analog fusion under variable working conditions provided in another embodiment of the present application.
[0062] Figure 4 A bar chart of various characteristic indicators in a rolling bearing health status prediction and early warning method driven by digital-analog fusion under variable working conditions is provided in another embodiment of the present application.
[0063] Figure 5 A schematic diagram of a fused health index obtained by fusion of feature dimensionality reduction in a method for predicting and warning the health status of a rolling bearing driven by digital-analog fusion under variable working conditions provided in another embodiment of the present application.
[0064] Figure 6 A schematic diagram of prediction results of different algorithms in a rolling bearing health status prediction and early warning method driven by digital-analog fusion under variable working conditions provided in another embodiment of the present application.
[0065] Figure 7 A schematic diagram of the functional modules of a rolling bearing health status prediction and early warning system driven by digital-analog fusion under variable working conditions provided in one embodiment of the present application.
[0066] Figure 8 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION
[0067] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0068] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.
[0069] In an exemplary embodiment, Figure 1 As shown, a method for predicting and warning the health status of a rolling bearing driven by digital-analog fusion under variable working conditions is provided, comprising the following steps S1 to S6:
[0070] S1. Obtaining the original operating data of the rolling bearing under the operating condition, and filtering the sample points of the original operating data based on the principle of small difference in initial conditions, to obtain the filtered original operating data. In this embodiment, the original operating data includes the operating condition data of the rolling bearing under the operating condition and the acceleration detection data of each detection position on the rolling bearing under the operating condition; the operating condition data of the rolling bearing under the operating condition includes: temperature, speed and load.
[0071] In step S1, based on the principle of small difference in initial conditions, sample points of the original operation data are screened to obtain screened original operation data. Specifically, in this embodiment, the initial conditions of the rolling bearing are analyzed, sample points with small difference in initial conditions are screened to obtain screened original operation data. The initial conditions of the rolling bearing are the newness and oldness of the rolling bearing, environmental conditions, etc., and the conditions that are as similar as possible are selected for testing.
[0072] S2. For the original operation data after screening, several time domain features and time-frequency domain features are extracted respectively, and the slope of each feature is calculated for cumulative transformation to form the original feature set. The time domain features include: absolute average amplitude, root mean square, maximum value, peak value, variance, root amplitude, skewness, kurtosis coefficient, waveform index, peak index, margin index and pulse index; the time-frequency domain features are several wavelet packet coefficients obtained by decomposing the original operation data after screening through wavelet packet transform.
[0073] In this embodiment, the slope of each feature is calculated in step S2 and then accumulated and transformed to form an original feature set, which specifically includes the following steps:
[0074] For several time domain features and time-frequency domain features, the difference method is used to calculate the differences between adjacent sample points to estimate the slope of each feature.
[0075] By cumulatively transforming the slope of each feature to form the original feature set, the overall information of the data can be mined.
[0076] The cumulative transformation taken in step S2 specifically includes the following steps:
[0077] Select a characteristic sequence of the rolling bearing in normal state, and calculate the average value of the characteristic sequence as the standard value. Calculate the average value of the characteristic sequence according to the following formula:
[0078]
[0079] Among them, f nor is the average value of the feature sequence, m is the starting index of the feature sequence, n is the total number of sample points in the feature sequence, and f(i) is the i-th eigenvalue of the feature sequence.
[0080] According to the standard value and the slope of each feature, the cumulative index of each feature is calculated. Specifically, the cumulative index of the feature is calculated according to the following formula:
[0081]
[0082] Among them, cf n Represents the cumulative index of the nth feature.
[0083] S3, performing baseline conversion on the slope of each feature in the original feature set, and smoothing the slope of each feature after baseline conversion without changing the degradation trend, to obtain a candidate feature set. Specifically in this embodiment, the slope of the feature is baseline converted according to the following formula:
[0084]
[0085] in, t k The baseline observation value at time, y k t k The slope of the characteristic at time s k t k Speed at the moment, s B is the baseline velocity, a and b are the baseline conversion parameters.
[0086] The baseline conversion parameters are determined according to the following formula:
[0087]
[0088] Among them, y i t i The slope of the characteristic at time s i t i The speed of the moment, is the baseline speed at t i Interpolation of time.
[0089] S4. Select the preferred features that are highly correlated with the development of the fault from the candidate feature set, and perform dimensionality reduction fusion on the preferred features to obtain the fusion health index. In step S4, the preferred features that are highly correlated with the development of the fault are selected from the candidate feature set, specifically: the linear combination of time correlation and monotonicity is used as a comprehensive evaluation index, and the preferred features that are highly correlated with the development of the fault are selected from the candidate feature set.
[0090] In this embodiment, the comprehensive evaluation index of the feature is calculated according to the following formula:
[0091] W = Mon(X) + Corr(X, T).
[0092] Among them, W is the comprehensive evaluation index of the feature, Mon() is the monotonicity function, Corr() is the time correlation function, X is the feature sequence, and T is time.
[0093] After selecting the preferred features that are highly correlated with the fault development, the preferred features are fused by dimensionality reduction to obtain the fusion health index, which specifically includes the following steps:
[0094] The z-score standardization method is used to process the selected features and obtain the time series set.
[0095] Through the kernel principal component analysis method, the time series set is fused by dimensionality reduction to obtain the fused health index.
[0096] S5. Select multiple degradation models according to the empirical model to establish a rolling bearing degradation model library, and determine the first degradation time of the rolling bearing based on the rolling bearing degradation model library and the fusion health index. The degradation models in the rolling bearing degradation model library include: power function degradation model, exponential function degradation model and logarithmic function degradation model.
[0097] S6. Based on the first degradation time, a time-varying particle filter algorithm and a rolling bearing degradation model library are used to predict the health status of the rolling bearing according to the measured operation data of the rolling bearing. The time-varying particle filter algorithm is used to select the optimal degradation model from the rolling bearing degradation model library for prediction based on the measured operation data of the rolling bearing within a sliding window of a preset length. In this embodiment, step S6 specifically includes the following steps:
[0098] When the rolling bearing begins to degrade based on the first degradation time, a sliding window with a preset length is set using a time-varying particle filter algorithm to obtain segment data from the measured operation data of the rolling bearing.
[0099] The power function degradation model, exponential function degradation model and logarithmic function degradation model are respectively used to fit the data under the window and determine the optimal degradation model; the optimal degradation model is used to predict the health status of the rolling bearing within a preset time period based on the segment data.
[0100] S7. Perform a rolling bearing health status warning based on the rolling bearing health status prediction result and the preset warning threshold.
[0101] The following is a specific example to demonstrate the effect of the method provided by the present application. This example conducts an accelerated degradation experiment of the rolling bearing under variable speed. It is designed in combination with the background of the crane and the actual situation of the test bench (such as cost and other factors), simulating the common action of the crane increasing speed, maintaining a constant speed, and reducing speed. This scheme uses 6007 deep groove ball bearings, and the experimental conditions are shown in Table 1. In order to accurately collect data, the sampling frequency is set to 25.6kHz in the variable speed experiment, the single sampling time is 8s and continuous sampling is performed. Obvious abnormal noise occurs when the sample is collected for the 2130th time. In order to avoid safety problems caused by bearing damage, the experiment is stopped and it is determined that the rolling bearing fails at this time.
[0102] Table 1 Experimental conditions of accelerated degradation of rolling bearings over their entire life under variable speed
[0103]
[0104] Process the collected data as follows:
[0105] Step 1: Obtain the horizontal, vertical and axial acceleration detection data of the rolling bearing. The working condition data include temperature, speed, load, etc. By analyzing the initial condition of the bearing, select samples with small initial condition differences. The horizontal vibration signal of the bearing used in the algorithm verification of this embodiment is as follows: Figure 2 As shown, from Figure 2 It can be seen that the vibration signal changes little and fluctuates steadily in the early stage of bearing operation. As time goes by, its wear increases, and the vibration amplitude increases rapidly at the end of operation. The vibration amplitude reaches the failure threshold in a very short time.
[0106] Step 2: Extract 20 time domain and time-frequency domain features. The time domain features are absolute average amplitude, root mean square, maximum value, peak value, variance, root amplitude, skewness, kurtosis coefficient, waveform index, peak index, margin index, and pulse index. In the process of extracting time-frequency domain features, the signal is decomposed into 3 layers using wavelet packet transform to obtain 8 wavelet packet coefficients, and the energy of each wavelet packet coefficient is used as a feature. The slopes of the 20 features are calculated and cumulatively transformed to form the original feature set.
[0107] Specifically, in the above steps, several time domain and time-frequency domain features are extracted from the data after rolling bearing screening, and the difference method is used to calculate the difference between adjacent sample points to estimate the slope. The slope of the cumulative transformation of the original feature set is calculated. From the perspective of bearing damage accumulation, the purpose of mining the overall information of the data is achieved by cumulatively transforming the original feature slope.
[0108] The steps of the cumulative transformation algorithm are as follows:
[0109] (1) For a single feature sequence, select a section of the feature sequence under the normal state of the equipment and calculate the average value of the section as the standard value.
[0110]
[0111] Where m and n are the number of accumulated features; f(i) is the i-th value of the sequence; m,n∈N + .
[0112] (2) When calculating the nth cumulative feature, the eigenvalues from the first to the nth eigenvalues of the original feature sequence are successively subtracted from the standard value and squared, and then the n square values are summed and finally scaled.
[0113]
[0114] Among them, cf n Represents the cumulative index of the nth original feature.
[0115] Step 3: Perform baseline conversion on the features. To reduce the impact of randomness of vibration features, smooth the features without changing the degradation trend of the curve. Taking the root mean square in the time domain features as an example, the degradation period features after baseline conversion and smoothing are as follows: Figure 3 As shown in Figure 2, it can be seen that the speed change caused a large fluctuation in the original observation value, and the converted observation value obtained by the baseline conversion algorithm eliminated the fluctuation caused by the speed change.
[0116] In the above step 3, the characteristic value slope is converted to the baseline speed level. The baseline conversion is to combine the characteristic value and the speed to reduce the large fluctuation of the characteristic value. The health state observation value is (y1, y2, ..., y k )', the speed is (s1,s2,…,s k )'. The calculation formula of the baseline conversion algorithm is:
[0117]
[0118] in, t k The baseline observation value at time, y k t k The slope of the characteristic at time s k t k Speed at the moment, s B is the baseline speed, and a and b are the baseline transformation parameters. The baseline speed SB is specified as the minimum rotation speed. For observations at other speeds, their interpolated values at the baseline speed can be used to approximate their baseline observations. The model parameters a and b are estimated by minimizing the sum of the squared errors between the interpolated and transformed observations.
[0119]
[0120] Among them, y i t i The slope of the characteristic at time s i t i The speed of the moment, is the baseline speed at t i The interpolation parameters a and b are initialized to a0=1 and b0=0.
[0121] Step 4: Use the comprehensive evaluation index W to evaluate the extracted 20 candidate features. The calculated index of each feature is shown in Figure 4 In step 4, the linear combination of time correlation and monotonicity is first used as a comprehensive evaluation index to select the preferred features that are highly correlated with fault development in the candidate feature set. The calculation formula for the comprehensive evaluation index is as follows:
[0122] W = Mon(X) + Corr(X, T).
[0123] In the formula, W is the comprehensive evaluation index of the feature, X is the feature sequence, Mon() is the monotonicity function, and Corr() is the time correlation function. It can be seen from the formula that W is positively correlated with the two evaluation indicators. Therefore, the larger W is, the better it can reflect the degradation trend of the bearing and the more conducive to the prediction of the remaining life. In this embodiment, 6 features with a comprehensive score of more than 0.5 are selected to form a preferred feature set, as shown in Table 2.
[0124] Table 2 Six features with comprehensive index higher than 0.5
[0125] feature Monotonicity Relevance comprehensive WPE5 0.1729 0.5833 0.7562 RMS 0.1176 0.5137 0.6313 WPE7 0.0991 0.5046 0.6037 Margin Factor 0.1268 0.3907 0.5175 WPE3 0.0622 0.4410 0.5032 Crest Factor 0.1268 0.3753 0.5021
[0126] Then the z-score standardization method is used to process the optimized eigenvalues:
[0127]
[0128] Where, X i =(x i (t), t=1,2,…,n) is the i-th dimension condition monitoring time series composed of n data, μ i and σ i are the mean and standard deviation of the time series data respectively; corresponding to the time series, t is the discrete time scale; is the standardized i-th dimension time series.
[0129] Finally, the kernel principal component analysis (KPCA) method is used to reduce the dimension of the selected features and fuse them to obtain the fused health index. In this embodiment, KPCA is used to reduce the dimension of the above selected feature set to fully utilize the advantages of each typical feature. Before KPCA is performed, the monitoring data is processed by the z-score standardization method; the contribution rate of the first principal component obtained after KPCA dimensionality reduction and fusion is 95.6%, which contains most of the effective information and can be directly used as a health indicator, such as Figure 5 Although the fused first principal component is a dimensionless index and has no physical meaning, it can more completely represent the characteristics of the entire degradation process of the rolling bearing.
[0130] Step 5: Consider the speed, load and other working condition parameters in the model, select the power function, exponential function and logarithmic function that are consistent with the actual working conditions to establish the rolling bearing physical model library for the degradation model. Use the 3σ detection method to determine the first degradation time (FDT) of the health index, and determine that the degradation period begins from the 1960th sample. Figure 5It can be seen that the amplitude of the health index increased significantly when the 1960th sample of the rolling bearing was collected. The bearing began to degrade rapidly here, and the bearing failed completely at the 2130th sample. Before this, the bearing was in the health monitoring stage, and the health index fluctuated slightly, so this paper chose to study the rapid degradation stage, that is, the 1960th sample to the 2130th sample.
[0131] In the step 5, a degradation model library is established based on the empirical model, a rolling bearing physical model consistent with the actual working conditions is established, and the first degradation time is determined by the health index; for any dynamic system, its dynamic equation can be expressed as:
[0132] x k =f k (x k-1 )+v k-1 .
[0133] z k =h k (x k )+w k .
[0134] Where x k Indicates that the system is at t k The state variable at the moment z k Represents system t k The measured value at the moment; w k Represents system t k The measurement noise at the moment v k Represents system t k The state noise at the moment; f k (·) is the state transition function of the system; h k (·) is the measurement function of the system.
[0135] Select the empirical model of rolling bearings and establish the state equation with power function, exponential function and logarithmic function as degradation models respectively:
[0136] x k =ct d k-1 +v k-1 .
[0137]
[0138] x k =pln(qt k-1 )+v k-1 .
[0139] Among them, c and d are the parameters of the power function degradation model; f and g are the parameters of the exponential degradation model; p and q are the parameters of the logarithmic degradation model.
[0140] A physical model library for rolling bearings is established. According to the statistical characteristics of health indicators throughout the life cycle, the 3σ detection method is used to determine the first degradation time of health indicators, and the entire life cycle is divided into the healthy stage and the degradation stage.
[0141] Step 6: When the 1960th sample is collected, the TVPF algorithm is used to predict the remaining life of the rolling bearing to obtain the prediction result of the health status of the rolling bearing; and the health status warning of the rolling bearing is performed according to the prediction result of the health status of the rolling bearing and the preset warning threshold. In this embodiment, step 6 uses the time-varying particle filter TVPF to adaptively select the optimal degradation model, and dynamically updates the model parameters using the input real-time health index value to achieve accurate prediction of the remaining life of the rolling bearing. When the value of the prediction result reaches 70% of the preset warning threshold, the first level warning is activated; when the value of the prediction result reaches 90% of the preset warning threshold, a secondary alarm is performed.
[0142] Specifically in this embodiment, when it is monitored that the rolling bearing begins to degrade, the TVPF algorithm sets a sliding window of length L in the obtained health index and intercepts the latest data. The power function degradation model, exponential function degradation model and logarithmic function degradation model are respectively used to fit the data under the window and calculate their MMSE values, where the optimal model is used as the state equation of the particle filter PF:
[0143] best model = min(MMSE(model i )).
[0144] The prior probability density distribution of the current state:
[0145]
[0146] Among them, δ is the pulse function; is the weight of the i-th particle at time k, which can be obtained by importance sampling method
[0147] The optimal state x of the system at the current moment k The estimated value of is:
[0148]
[0149] Move the sliding window forward, which includes the new set of observations and L-1 previous observations; repeat the steps to calculate new degradation state values, and the calculation process will stop when the degradation state value exceeds the failure threshold. These estimates can accurately track the actual degradation state of the bearing.
[0150] The mean absolute percentage error (MAPE) and root mean square error (RMSE) were used to verify the effectiveness of the prediction model:
[0151]
[0152]
[0153] Where u is the number of samples, y i and is the i-th true value and predicted value.
[0154] In order to verify the effectiveness of the proposed algorithm, it is compared with three particle filter methods based on a single model. In order to ensure the fairness of the comparison, all prediction and warning methods use the same health indicators and set the same first degradation time and failure threshold. The prediction results are shown in Figure 2. Figure 6 shown.
[0155] from Figure 6 From the comparison of the life prediction results of the four methods, it can be seen that in the early and middle stages of prediction, the prediction results of the exponential particle filter, the power function particle filter and the logarithmic particle filter are much higher than the actual life, resulting in a large prediction error; the exponential particle filter can better fit the real life in the late stage of prediction, and the prediction accuracy in the late stage of prediction is higher; compared with the three filtering methods based on a single model, the proposed algorithm has the smallest volatility in the life prediction results during the entire degradation process, and the convergence speed is better than other methods, and it can converge to the real life the earliest, indicating that compared with other methods, the proposed algorithm is more stable during the entire degradation process.
[0156] In order to quantitatively measure the prediction performance of several methods, the values of MAPE and RMSE are calculated respectively, as shown in Table 3. As can be seen from the table, the root mean square error and mean absolute percentage error of the proposed algorithm are 16.32 and 0.1500 respectively. The proposed algorithm has the smallest prediction error in the entire prediction process, and the prediction accuracy is improved by more than 35% compared with the traditional sub-filter prediction method.
[0157] Table 3 Prediction performance evaluation criteria of different methods
[0158] Evaluation indicators Power function filtering Exponential Filtering Logarithmic filtering The algorithm proposed in this application MAPE 0.5965 0.2532 0.4187 0.1859 RMSE 47.47 20.74 26.45 12.55
[0159] Based on the same inventive concept, the embodiment of the present application also provides a rolling bearing health status prediction and warning system driven by digital-analog fusion under variable working conditions for realizing the above-mentioned rolling bearing health status prediction and warning method driven by digital-analog fusion under variable working conditions. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in the embodiments of the rolling bearing health status prediction and warning system driven by digital-analog fusion under variable working conditions provided below can be referred to the limitations of the rolling bearing health status prediction and warning method driven by digital-analog fusion under variable working conditions above, and will not be repeated here.
[0160] In an exemplary embodiment, Figure 7 As shown, a rolling bearing health status prediction and early warning system driven by digital-analog fusion under variable working conditions is provided, including:
[0161] The original operation data acquisition module is used to acquire the original operation data of the rolling bearing under the operation condition;
[0162] The original operation data screening module is used to screen the sample points of the original operation data based on the principle of small difference in initial conditions to obtain the screened original operation data;
[0163] A feature extraction and cumulative transformation module is used to extract a number of time domain features and time-frequency domain features from the filtered original operation data, and calculate the slope of each feature for cumulative transformation to form an original feature set;
[0164] A baseline conversion and smoothing processing module, used to perform baseline conversion on the slope of each feature in the original feature set, and smooth the slope of each feature after baseline conversion without changing the degradation trend, so as to obtain a candidate feature set;
[0165] A feature optimization and dimensionality reduction fusion module is used to select preferred features that are highly correlated with fault development from the candidate feature set, and perform dimensionality reduction fusion on the preferred features to obtain a fusion health index;
[0166] A first degradation time determination module, used to select multiple degradation models according to the empirical model to establish a rolling bearing degradation model library, and determine the first degradation time of the rolling bearing based on the rolling bearing degradation model library and the fusion health index;
[0167] The rolling bearing health status prediction module is used to predict the health status of the rolling bearing based on the first degradation time, using a time-varying particle filter algorithm and the rolling bearing degradation model library, and according to the actual measured operation data of the rolling bearing to obtain a rolling bearing health status prediction result; the time-varying particle filter algorithm is used to select the optimal degradation model from the rolling bearing degradation model library for prediction based on the actual measured operation data of the rolling bearing within a sliding window of a preset length.
[0168] The rolling bearing health status warning module is used to provide rolling bearing health status warning based on the rolling bearing health status prediction results and preset warning thresholds.
[0169] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 8As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a rolling bearing health status prediction and early warning method driven by digital-analog fusion under variable working conditions is implemented.
[0170] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0171] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.
[0172] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.
[0173] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.
[0174] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0175] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).
[0176] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.
[0177] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0178] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.
Claims
1. A rolling bearing health status prediction and early warning method driven by digital-analog fusion under variable working conditions, characterized in that: include: Acquire the original operating data of the rolling bearing under the operating condition, and filter the sample points of the original operating data based on the principle of small difference in initial conditions to obtain filtered original operating data; Extracting a plurality of time domain features and time-frequency domain features from the filtered original operation data, and calculating the slope of each feature for cumulative transformation to form an original feature set; Performing baseline conversion on the slope of each feature in the original feature set, and smoothing the slope of each feature after baseline conversion without changing the degradation trend, to obtain a candidate feature set; Selecting preferred features that are highly correlated with fault development from the candidate feature set, and performing dimensionality reduction fusion on the preferred features to obtain a fusion health index; Selecting multiple degradation models according to the empirical model to establish a rolling bearing degradation model library, and determining the first degradation time of the rolling bearing based on the rolling bearing degradation model library and the fusion health index; Based on the first degradation time, a time-varying particle filter algorithm and the rolling bearing degradation model library are used to predict the health state of the rolling bearing according to the measured operation data of the rolling bearing, so as to obtain a prediction result of the health state of the rolling bearing; the time-varying particle filter algorithm is used to select the optimal degradation model from the rolling bearing degradation model library for prediction according to the measured operation data of the rolling bearing within a sliding window of a preset length; According to the rolling bearing health status prediction result and the preset warning threshold, a rolling bearing health status warning is performed.
2. The method for predicting and warning the health status of rolling bearings driven by digital-analog fusion under variable working conditions according to claim 1 is characterized in that: The original operating data includes operating condition data of the rolling bearing under operating conditions and acceleration detection data of each detection position on the rolling bearing under operating conditions; The operating data of rolling bearings under operating conditions include: temperature, speed and load; Based on the principle of small difference in initial conditions, the sample points of the original operating data are screened to obtain the screened original operating data. Specifically, by analyzing the initial conditions of the rolling bearing, sample points with small difference in initial conditions are screened to obtain the screened original operating data.
3. The method for predicting and warning the health status of rolling bearings driven by digital-analog fusion under variable working conditions according to claim 1 is characterized in that: The time domain features include: absolute average amplitude, root mean square, maximum value, peak value, variance, root square amplitude, skewness, kurtosis coefficient, waveform index, peak index, margin index and pulse index; the time-frequency domain features are several wavelet packet coefficients obtained by decomposing the filtered original operation data through wavelet packet transform; The slope of each feature is calculated and cumulatively transformed to form the original feature set, including: For a plurality of the time domain features and the time-frequency domain features, using a difference method to calculate the difference between adjacent sample points to estimate the slope of each feature; By cumulatively transforming the slope of each feature to form the original feature set, the overall information of the data can be mined.
4. The method for predicting and warning the health status of rolling bearings driven by digital-analog fusion under variable working conditions according to claim 3 is characterized in that: The cumulative transformation comprises the following steps: Select a characteristic sequence of the rolling bearing in a normal state, and calculate the average value of the characteristic sequence as the standard value; The cumulative index of each feature is calculated based on the standard value and the slope of each feature.
5. The method for predicting and warning the health status of rolling bearings driven by digital-analog fusion under variable working conditions according to claim 4 is characterized in that: The average value of the characteristic sequence is calculated according to the following formula: Among them, f nor is the average value of the feature sequence, m is the starting index of the feature sequence, n is the total number of sample points in the feature sequence, and f(i) is the i-th eigenvalue of the feature sequence; The cumulative index of a feature is calculated according to the following formula: Among them, cf n Represents the cumulative index of the nth feature.
6. The method for predicting and warning the health status of rolling bearings driven by digital-analog fusion under variable working conditions according to claim 1 is characterized in that: The slope of the feature is baseline-shifted according to the following formula: in, t k The baseline observation value at time, y k t k The slope of the characteristic at time s k t k Speed at the moment, s B is the baseline speed, a and b are baseline conversion parameters; The baseline conversion parameters are determined according to the following formula: Among them, y i t i The slope of the characteristic at time s i t i The speed of the moment, is the baseline speed at t i Interpolation of time.
7. The method for predicting and warning the health status of rolling bearings driven by digital-analog fusion under variable working conditions according to claim 1 is characterized in that: Selecting preferred features that are highly correlated with fault development from the candidate feature set, specifically: taking a linear combination of time correlation and monotonicity as a comprehensive evaluation index, and selecting preferred features that are highly correlated with fault development from the candidate feature set; The comprehensive evaluation index of the feature is calculated according to the following formula: W = Mon(X) + Corr(X, T); Among them, W is the comprehensive evaluation index of the feature, Mon() is the monotonicity function, Corr() is the time correlation function, X is the feature sequence, and T is time.
8. The method for predicting and warning the health status of rolling bearings driven by digital-analog fusion under variable working conditions according to claim 7 is characterized in that: The selected features are subjected to dimensionality reduction fusion to obtain fusion health indicators, which specifically include: The preferred features are processed using a z-score standardization method to obtain a time series set; The time series set is fused by dimensionality reduction through kernel principal component analysis to obtain the fused health index.
9. The method for predicting and warning the health status of rolling bearings driven by digital-analog fusion under variable working conditions according to claim 1 is characterized in that: The degradation models in the rolling bearing degradation model library include: a power function degradation model, an exponential function degradation model and a logarithmic function degradation model; Based on the first degradation time, a time-varying particle filter algorithm and the rolling bearing degradation model library are used to predict the health status of the rolling bearing according to the measured operation data of the rolling bearing, specifically including: When the rolling bearing begins to degrade based on the first degradation time, a sliding window of a preset length is set by using a time-varying particle filter algorithm to intercept and obtain segment data from the measured operation data of the rolling bearing; A power function degradation model, an exponential function degradation model and a logarithmic function degradation model are respectively used to fit the data under the window and determine the optimal degradation model; the optimal degradation model is used to predict the health status of the rolling bearing within a preset time period based on the segment data.
10. A rolling bearing health status prediction and early warning system driven by digital-analog fusion under variable working conditions, characterized in that: The method for predicting and warning the health status of a rolling bearing driven by digital-analog fusion under variable working conditions according to any one of claims 1 to 9 is used to implement the method, wherein the system for predicting and warning the health status of a rolling bearing driven by digital-analog fusion under variable working conditions comprises: The original operation data acquisition module is used to acquire the original operation data of the rolling bearing under the operation condition; The original operation data screening module is used to screen the sample points of the original operation data based on the principle of small difference in initial conditions to obtain the screened original operation data; The feature extraction and accumulation transformation module is used to extract a number of time domain features and time-frequency domain features from the filtered original operation data, and calculate the slope of each feature for accumulation transformation to form an original feature set; A baseline conversion and smoothing processing module, used to perform baseline conversion on the slope of each feature in the original feature set, and smooth the slope of each feature after baseline conversion without changing the degradation trend, to obtain a candidate feature set; A feature optimization and dimensionality reduction fusion module is used to select preferred features that are highly correlated with fault development from the candidate feature set, and perform dimensionality reduction fusion on the preferred features to obtain a fusion health index; A first degradation time determination module, used to select multiple degradation models according to the empirical model to establish a rolling bearing degradation model library, and determine the first degradation time of the rolling bearing based on the rolling bearing degradation model library and the fusion health index; A rolling bearing health state prediction module, which is used to predict the health state of the rolling bearing based on the first degradation time, using a time-varying particle filter algorithm and the rolling bearing degradation model library, and according to the measured operation data of the rolling bearing, to obtain a rolling bearing health state prediction result; the time-varying particle filter algorithm is used to select the optimal degradation model from the rolling bearing degradation model library for prediction based on the measured operation data of the rolling bearing within a sliding window of a preset length; The rolling bearing health status warning module is used to provide a rolling bearing health status warning based on the rolling bearing health status prediction result and a preset warning threshold.
Citation Information
Cited By
Bearing quality detection method and system
CN121301821A
Health management method and device for crushing shaft group of crusher for underground crossheading of coal mine
CN122413382A
Health management method and device for coal mine underground crossheading crusher breaking shaft group
CN122413382B
