A multi-weight index fusion rolling bearing health state degradation failure evaluation method

Abstract

The application discloses a kind of multi-weight index fusion's rolling bearing health state degradation failure evaluation method, comprising: collecting the vibration signal of rolling bearing full life stage, constructs the vibration signal full life data set of rolling bearing;Calculate the multiple characteristic indexes of each vibration signal in data set;Respectively calculate the spearman rank correlation coefficient of each characteristic index;Based on monotonicity, correlation and robustness, the weight coefficient of normalized feature index is calculated;The multiplication and addition of each screened feature index in the vibration signal full life data set of rolling bearing and its corresponding weight coefficient are carried out, to obtain the health factor of each vibration signal, and draw the rolling bearing health factor curve;According to rolling bearing health factor curve, determine the degradation state starting point and degradation state of rolling bearing and failure state starting point and failure state.The present application can accurately characterize the performance of rolling bearing slow degradation state.

Description

Technical Field

[0001] This invention relates to the field of rolling bearing testing technology, and in particular to a method for assessing the health status degradation and failure of rolling bearings by fusing multiple weighted indicators. Background Technology

[0002] As one of the most commonly used rotating machinery components, the health of rolling bearings plays a crucial role in the normal operation of mechanical equipment. If a rolling bearing fails, it can lead to equipment downtime and economic losses, or even accidents threatening personnel lives. Accurate health assessments throughout the entire service life of rolling bearings can determine their remaining service time in real time, allowing for timely maintenance and preventing major failures in the entire mechanical system.

[0003] To monitor and quantify the health degradation of rolling bearings during their service life, a health index (HI) is typically constructed to quantify the bearing's operating status. Plotting a performance degradation trend curve based on the HI allows for prediction of the bearing's operating condition, enabling prediction of remaining service life and early warning systems. However, the commonly used HI cannot accurately identify the initial failure of rolling bearings, especially during the healthy operating phase. It is easily affected by background noise, often misjudging the condition as a failure, severely impacting the accuracy and stability of subsequent predictions. In actual rolling bearing operation, the entire lifespan can be divided into healthy, degraded, and failed states. The traditional HI method can only determine the onset of rolling bearing failure, but cannot accurately and meticulously assess the overall health status of the rolling bearing throughout its lifespan.

[0004] Currently, there are two common methods for assessing the health status of rolling bearings: model-based methods and data-driven methods. Model-based methods construct specific models for rolling bearings to predict their lifespan, relying on expert knowledge and the physical degradation patterns of rolling bearings. These methods depend on the selection of initial parameters and are not flexible enough; they cannot promptly handle changes in data patterns, leading to significant biases in health status predictions. Data-driven methods, on the other hand, analyze monitoring data to explore the relationship between data changes and equipment operating conditions, enabling fault diagnosis and lifespan prediction. They can more flexibly detect data trends and make timely adjustments. However, current data-driven methods often employ deep learning, requiring a large amount of training data. Insufficient sample size can prevent the model from effectively learning the patterns of data change. However, rolling bearing failure conditions are complex and diverse, and the amount of full-lifecycle rolling bearing operating condition data used for training data-driven methods is limited, making it impossible to achieve good prediction results for all rolling bearings and resulting in insufficient generalization. Summary of the Invention

[0005] To address the above problems, this invention provides a method for assessing the health status degradation and failure of rolling bearings by fusing multiple weighted indices.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for assessing the health status degradation and failure of rolling bearings by fusing multiple weighted indices includes:

[0008] Vibration signals of rolling bearings throughout their entire lifespan are collected to construct a rolling bearing vibration signal lifespan dataset.

[0009] Calculate multiple characteristic indices for each vibration signal in the dataset;

[0010] Calculate the Spearman rank correlation coefficient for each feature index to filter out feature indices that are greater than a preset threshold, and then normalize the filtered feature indices.

[0011] The weight coefficients of the normalized feature indicators are calculated based on monotonicity, correlation, and robustness.

[0012] The health factor of each vibration signal is obtained by multiplying and adding the selected feature index and its corresponding weight coefficient in the whole life dataset of rolling bearing vibration signals. The health factor curve of rolling bearing is plotted with the sample sequence of each vibration signal in the whole life dataset of rolling bearing vibration signals as the x-axis and the health factor as the y-axis.

[0013] The starting point and state of degradation, as well as the starting point and state of failure, of rolling bearings are determined based on the health factor curve of the rolling bearing.

[0014] Furthermore: the rolling bearing vibration signal life-cycle dataset includes:

[0015] The whole-life dataset of rolling bearing vibration signals is represented as S={s1,s2,…s n The dataset includes n vibration time signal samples, where s i This represents the vibration time signal sample obtained from the i-th acquisition.

[0016] Furthermore, the various characteristic indicators of each vibration signal in the calculated dataset include:

[0017] The various characteristic indicators of each vibration signal in the computational dataset include:

[0018] For the whole-life dataset S of rolling bearing vibration signals, calculate the mean, standard deviation, skewness, kurtosis, waveform factor, shape factor, marginal factor, peak-to-peak value, root mean square, peak value, and rectified average value characteristic index F for each vibration time signal sample. i .

[0019] Further: the step of calculating the Spearman rank correlation coefficient for each feature index to filter out feature indices greater than a preset threshold, and then normalizing the filtered feature indices, includes:

[0020] The characteristic index Fi sequence corresponding to all vibration time signal samples throughout the entire lifespan of a rolling bearing is represented as follows: The Spearman rank correlation coefficient (SP) was calculated as follows:

[0021]

[0022] In the formula, For the characteristic index F i The sequence, where T is the number of vibration time signal samples, r k Let R be an element of vector R, and R be represented as follows:

[0023]

[0024] In the formula, for The sorting;

[0025] Set an SP threshold, calculate the SP for each feature indicator, filter out feature indicators with a SP greater than the SP threshold, and normalize the filtered feature indicators. The matrix constructed from all normalized feature indicators is represented as B. n×c (c≤11), represented as:

[0026]

[0027] In the formula, c represents the number of feature indicators after screening, and c≤11.

[0028] Furthermore: the weighting coefficients for the normalized feature indicators calculated based on monotonicity, correlation, and robustness include:

[0029] For matrix B n×c Represented as:

[0030]

[0031] In the formula, Q 1 Q 2 Q c They are represented as follows:

[0032]

[0033] In the formula, Q 1 Q 2 Q c These are vectors constructed from the vibration time signal sample sequences of the rolling bearing throughout its entire lifespan, representing the various selected feature indicators.

[0034] Calculate Q 1 Q 2 Q c The monotonicity index Mon() is as follows:

[0035]

[0036] In the formula, δ(k) is a unit step function, specifically expressed as follows:

[0037]

[0038] Get Q 1 Q 2 Q c The monotonicity index results are as follows: M1, M2, ..., M c ;

[0039] Calculate Q 1 Q 2 Q c The correlation metric Corr() is as follows:

[0040]

[0041] Get Q 1 Q 2 Q c The correlation index results are as follows: C1, C2, ..., C c ;

[0042] Calculate Q 1 Q 2 Q c The robustness index Rob() is as follows:

[0043]

[0044] In the formula, The result is the feature index sequence after Savitzky-Golay filtering.

[0045] Get Q 1 Q 2 Q c The correlation indices are: R1, R2, ..., R c ;

[0046] They are represented as follows:

[0047]

[0048] The above sequence was normalized, and the result is as follows:

[0049]

[0050] For the monotonicity, correlation, and robustness results of each category of feature indicators after normalization in the above formula, the results are summed according to the feature indicator category and averaged to obtain the weight coefficients for each category of feature indicators, denoted as w1, w2, ..., w c Wherein, the weight coefficient w of the i-th type of feature index i The calculation is as follows:

[0051]

[0052] In the formula, M i C i R i These represent the monotonicity, correlation, and robustness indices corresponding to the i-th type of feature index; M max C max R max The maximum values ​​of the monotonicity, correlation, and robustness results of all selected feature indicators are used to process the weight coefficients of each feature indicator as follows:

[0053]

[0054] In the formula, The sum of the weight coefficients of all feature indicators after filtering is used to determine the final weight coefficients for each category of feature indicators after filtering. These coefficients are represented as: W1, W2, ..., W... c .

[0055] Furthermore: the health factors obtained for each vibration signal include:

[0056] For the i-th vibration time signal sample s in the whole life dataset S of rolling bearing vibration signals i Feature indicators after screening And the weight coefficients W1, W2, ..., W corresponding to each feature index. c The corresponding feature indices are multiplied and added together to obtain the vibration time signal sample s. i Corresponding health factor HI i :

[0057]

[0058] The health factor HI of each vibration time signal sample in the whole life dataset S of rolling bearing vibration signal is calculated. The health factor HI curve of the whole life stage of rolling bearing is plotted with the sample sequence of each vibration time signal sample in the whole life dataset S as the abscissa and the health factor HI as the ordinate.

[0059] Furthermore, determining the starting point and degradation state of the rolling bearing based on the rolling bearing health factor curve includes:

[0060] Select the first 30% of the data from the health factor HI curve of rolling bearings throughout their entire lifespan, calculate their mean μ and standard deviation σ, and determine the 3σ interval [μ-3σ, μ+3σ].

[0061] Based on the order of the sample sequence, the health factor HI of each vibration time signal sample in the whole life dataset S of rolling bearing vibration signals is compared with the 3σ interval. If three consecutive health factors HI exceed the 3σ interval, the rolling bearing is considered to be in a degraded state. The coordinate point of the first health factor HI exceeding the 3σ interval is taken as the starting point of the rolling bearing's degraded state, denoted as HI. fdt .

[0062] Furthermore, determining the starting point and failure state of the rolling bearing based on the rolling bearing health factor curve includes:

[0063] Select the last 70% of the data from the health factor HI curve of rolling bearings throughout their entire lifespan, calculate their mean μ and standard deviation σ, and determine the 3σ interval [μ-3σ, μ+3σ].

[0064] Based on the order of the sample sequence, the health factor HI of each vibration time signal sample in the whole life dataset S of rolling bearing vibration signals is compared with the 3σ interval. If three consecutive health factors HI exceed the 3σ interval, the rolling bearing is considered to be in a failure state. The coordinate point of the first health factor HI exceeding the 3σ interval is taken as the starting point of the rolling bearing failure state, denoted as HI. fft ;

[0065] If there are no three consecutive health factors HI that exceed the 3σ interval, then the coordinate point of the first health factor HI where two consecutive health factors HI first exceed the 3σ interval is taken as the starting point of the rolling bearing failure state, denoted as HI. fft ;

[0066] If there are no two consecutive health factors HI that both exceed the 3σ interval, then the coordinate point of the first health factor HI that exceeds the 3σ interval is taken as the starting point of the rolling bearing failure state, denoted as HI. fft .

[0067] The technological advancements achieved by this invention compared to existing technologies are as follows:

[0068] This invention can accurately characterize both the gradual and rapid degradation states of rolling bearings, avoiding the adverse effects of complex operating conditions and signal fluctuations on the performance status assessment of rolling bearings. Furthermore, this invention proposes an adaptive identification method for the degradation and failure stages of rolling bearings, enabling a comprehensive assessment of the health status of rolling bearings throughout their entire lifespan. The resulting prediction method can also be applied to other rotating machinery components, providing feasible technical support for monitoring the performance degradation of mechanical parts. Attached Figure Description

[0069] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0070] In the attached diagram:

[0071] Figure 1 This is a flowchart of the present invention;

[0072] Figure 2 This is the time-domain waveform diagram of the entire lifetime of Bearing1-1 in this invention;

[0073] Figure 3 This is a time-domain waveform diagram of the entire lifetime of Bearing2-1 of the present invention;

[0074] Figure 4 This is the time-domain waveform diagram of the entire lifetime of Bearing3-3 in this invention;

[0075] Figure 5 This is a graph showing the health status degradation and failure assessment results of the Bearing1-1 rolling bearing of the present invention;

[0076] Figure 6 This is a graph showing the health status degradation and failure assessment results of the Bearing2-1 rolling bearing of the present invention;

[0077] Figure 7 This is a graph showing the results of the health status degradation and failure assessment of the Bearing3-3 rolling bearing of the present invention;

[0078] Figure 8 This is a diagram showing the health status assessment results of the Bearing1-1 rolling bearing of the present invention;

[0079] Figure 9 This is a diagram showing the health status assessment results of the Bearing2-1 rolling bearing of the present invention;

[0080] Figure 10 This is a diagram showing the health status assessment results of the Bearing3-3 rolling bearing of the present invention. Detailed Implementation

[0081] The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will be described below with reference to the accompanying drawings.

[0082] like Figure 1 As shown, this invention discloses a method for assessing the health status degradation and failure of rolling bearings by fusing multiple weighted indices, including:

[0083] Vibration signals of rolling bearings throughout their entire lifespan are collected to construct a rolling bearing vibration signal lifespan dataset.

[0084] Calculate multiple characteristic indices for each vibration signal in the dataset;

[0085] Calculate the Spearman rank correlation coefficient for each feature index to filter out feature indices that are greater than a preset threshold, and then normalize the filtered feature indices.

[0086] The weight coefficients of the normalized feature indicators are calculated based on monotonicity, correlation, and robustness.

[0087] The health factor of each vibration signal is obtained by multiplying and adding the selected feature index and its corresponding weight coefficient in the whole life dataset of rolling bearing vibration signals. The health factor curve of rolling bearing is plotted with the sample sequence of each vibration signal in the whole life dataset of rolling bearing vibration signals as the x-axis and the health factor as the y-axis.

[0088] The starting point and state of degradation, as well as the starting point and state of failure, of rolling bearings are determined based on the health factor curve of the rolling bearing.

[0089] 1. Collect vibration signals throughout the entire lifespan of the rolling bearing and construct a rolling bearing vibration signal lifespan dataset:

[0090] Vibration signals of rolling bearings throughout their entire lifespan, from normal operation to severe damage, are collected using accelerometers, thereby constructing a full-lifespan dataset of rolling bearing vibration signals. This dataset is represented as S = {s1, s2, ... s}. n The dataset contains n vibration time signal samples, where s i This represents the vibration time signal sample obtained from the i-th acquisition.

[0091] 2. Calculate multiple characteristic indices for each vibration signal in the dataset:

[0092] For the entire lifespan dataset S of rolling bearing vibration signals, 11 characteristic indices are calculated for each vibration time signal sample, including mean, standard deviation, skewness, kurtosis, waveform factor, shape factor, marginal factor, peak-to-peak value, root mean square, peak value, and rectified average. Vibration time signal sample s i The 11 calculated feature indices form a vector P, represented as: Calculate feature indices for all vibration time signal samples in the whole life dataset S of rolling bearing vibration signals, and construct feature index matrix A. n×11 , is represented as:

[0093]

[0094] The calculation methods for the above 11 features are shown in Table 1.

[0095] Table 1. Calculation Method of Feature Indicators

[0096]

[0097]

[0098] 3. Calculate the Spearman rank correlation coefficient for each feature indicator to filter out feature indicators that are greater than a preset threshold, and then normalize the filtered feature indicators:

[0099] Throughout the entire lifespan of a rolling bearing, the failure and degradation process exhibits a monotonic change over time. Therefore, for the 11 extracted feature indicators, it is necessary to select those with better monotonicity over time to overcome the problems of feature redundancy and feature representation conflicts associated with multiple feature indicators. For each of the 11 collected feature indicators, the Spearman rank correlation coefficient is calculated. Regarding feature indicator F... i The characteristic index F corresponding to all vibration signals throughout the entire lifespan of a rolling bearing i The sequence is represented as: The Spearman rank correlation coefficient (SP) was calculated as follows:

[0100]

[0101] In the formula, For the characteristic index F i The sequence, where T is the number of vibration time signal samples, r k For each element in vector R, R is represented as follows:

[0102]

[0103] In the formula, for The sorting.

[0104] For example Indicator F i The corresponding first 5 vibration time signal samples, assuming they are set as [50, 54, 52, 58, 56], then The sample sequence is then represented as rank(t) {i:T} Given R = [1,2,3,4,5] and R = [0,-1,1,-1,1], we can calculate... The Spearman rank correlation coefficient is 0.8.

[0105] A higher Spearman rank correlation coefficient (SP) indicates better monotonicity of the feature over the entire lifespan of the rolling bearing; a lower value indicates poorer monotonicity. Based on the actual operating conditions of rolling bearings, the initial stage is considered normal and relatively stable. Therefore, Spearman rank correlation coefficient analysis is performed on the characteristic indicators of vibration time signal samples from the latter 70% of the rolling bearing's lifespan. For example, if 100 signals are collected throughout the lifespan of a rolling bearing, the last 70 signals are selected, and 11 characteristic indicators are calculated for each signal. The Spearman rank correlation coefficient (SP) between each signal and the time series of the last 70 vibration time signal samples is then calculated. A threshold of 0.9 is set for SP, and the SP values ​​for each of the 11 characteristic indicators are calculated. All characteristic indicators greater than the pre-set threshold of 0.9 are selected. Then, each selected characteristic indicator is normalized to a value between 0 and 1, eliminating differences in the magnitude of the individual indicator values. The matrix constructed from all normalized characteristic indicators is represented as B. n×c (c≤11), represented as:

[0106]

[0107] In the formula, c represents the number of feature indicators after screening, and c≤11.

[0108] 4. Calculate the weight coefficients of the normalized feature indicators based on monotonicity, correlation, and robustness:

[0109] For the selected features, the rolling bearing degradation information contained in different features is also different, and it is necessary to further calculate the weight of the selected features. In this embodiment, three index methods are used to calculate the weight coefficient of each type of feature index: monotonicity Mon(), correlation Corr(), and robustness Rob(). Among them, the monotonicity index Mon() describes the consistency between the feature index sequence and the service performance degradation trend of the rolling bearing. The performance degradation of the rolling bearing during operation is irreversible, so the feature index should have a monotonic trend. The correlation index Corr() is the Pearson correlation coefficient between the feature index sequence and its time sample sequence, describing the degree of correlation between the feature sequence and the service performance change trend of the rolling bearing. The robustness index Rob() refers to the tolerance of the feature index sequence to abnormal state interference. The values ​​of the above three indices are all between [0, 1], and the larger the value, the better the performance of the corresponding index.

[0110] Matrix B constructed based on the selected feature indicators n×c Further expressed as:

[0111]

[0112] In the formula, Q 1 Q 2 Q c They are represented as follows:

[0113]

[0114] In the formula, Q 1 Q 2 Q c These are vectors constructed from the vibration time signal sample sequences of the rolling bearing throughout its entire lifespan, representing the various selected feature indicators.

[0115] Calculate Q 1 Q 2 Q c The monotonicity index Mon() is as follows:

[0116]

[0117] In the formula, δ(k) is a unit step function, specifically expressed as follows:

[0118]

[0119] Get Q 1 Q 2 Q c The monotonicity index results are as follows: M1, M2, ..., M c .

[0120] Calculate Q1 Q 2 Q c The correlation metric Corr() is as follows:

[0121]

[0122] Get Q 1 Q 2 Q c The correlation index results are as follows: C1, C2, ..., C c .

[0123] Calculate Q 1 Q 2 Q c The robustness index Rob() is as follows:

[0124]

[0125] In the formula, This is the result of the feature index sequence after Savitzky-Golay filtering.

[0126] Get Q 1 Q 2 Q c The correlation indices are: R1, R2, ..., R c .

[0127] The monotonicity, correlation, and robustness results of each type of feature indicator after screening, obtained through the above calculations, are shown below:

[0128]

[0129] The above sequence was normalized, and the result is as follows:

[0130]

[0131] For the monotonicity, correlation, and robustness results of each category of feature indicators after normalization in the above formula, the results are summed according to the feature indicator category and averaged to obtain the weight coefficients for each category of feature indicators, denoted as w1, w2, ..., w c Wherein, the weight coefficient w of the i-th type of feature index i The calculation is as follows:

[0132]

[0133] In the formula, M i C i R i These represent the monotonicity, correlation, and robustness indices corresponding to the i-th type of feature index; Mmax C max R max The maximum values ​​of the monotonicity, correlation, and robustness results of all feature indicators after screening are taken respectively. The weight coefficients of each type of feature indicator are then further processed as follows:

[0134]

[0135] In the formula, This is the sum of the weight coefficients of all feature indicators after filtering. Finally, the weight coefficients corresponding to each category of feature indicators after filtering are represented as: W1, W2, ..., W... c .

[0136] 5. Health index (Hi) curve plotting:

[0137] For the i-th vibration time signal sample s in the whole life dataset S of rolling bearing vibration signals i Feature indicators after screening And the weight coefficients W1, W2, ..., W corresponding to each type of feature index obtained above. c The corresponding feature indices are multiplied and summed to obtain the sample signal s. i Corresponding health factor HI i :

[0138]

[0139] Based on the above HI calculation process, the health factor HI of each vibration time signal sample in the whole life data of rolling bearing vibration signal is calculated. The sample sequence of each sample in the whole life data of rolling bearing vibration signal is used as the abscissa and the health factor HI is used as the ordinate to draw the rolling bearing health factor HI curve, thus obtaining the HI curve reflecting the whole life health status of rolling bearing.

[0140] 6. Determination of rolling bearing degradation status:

[0141] Select the health status HI curve data of the early operation stage of rolling bearing (the first 30% of the HI curve of the rolling bearing in the whole life stage), calculate its mean μ and standard deviation σ, and determine the 3σ interval [μ-3σ, μ+3σ].

[0142] Based on the order of the sample sequence, the health factor HI of each vibration time signal sample in the whole life dataset S of rolling bearing vibration signals is compared with the 3σ interval in the order of the sample sequence. If a certain health factor HI exceeds the 3σ interval, the coordinate point of the health factor HI is denoted as HI. fdt .

[0143] To avoid interference from noise, it is necessary to determine HI.fdt Do the two subsequent health factors (HI) also exceed the 3σ range?

[0144] 1) If HI fdt If the HI points for both subsequent health factors exceed the 3σ interval, then HI will be... fdt The starting point of the degradation state of a rolling bearing is considered to be the point where the rolling bearing enters a degradation state. After this point, the rolling bearing transitions from a healthy state to a degradation state, i.e., degradation behavior occurs.

[0145] 2) If HI fdt If neither of the two subsequent health factor HI points exceeds the 3σ interval, then HI will not be included. fdt The starting point of the deterioration state of the rolling bearing is taken as the coordinate point of the first health factor HI that exceeds the 3σ interval.

[0146] 7. Rolling bearing failure condition identification:

[0147] Select the health status HI curve data of the rolling bearing in the middle and late stage of operation (the last 70% of the rolling bearing's full life stage HI curve), calculate its mean μ and standard deviation σ, and determine the 3σ interval [μ-3σ, μ+3σ].

[0148] Based on the order of the sample sequence, the health factor HI of each vibration time signal sample in the whole life dataset S of rolling bearing vibration signals is compared with the 3σ interval in the order of the sample sequence. If a certain health factor HI exceeds the 3σ interval, the coordinate point of the health factor HI is denoted as HI. fft .

[0149] To avoid interference from noise, it is necessary to determine HI. fft Do the two subsequent health factors (HI) also exceed the 3σ range?

[0150] 1) If HI fft If the HI points for both subsequent health factors exceed the 3σ interval, then HI will be... fft If the rolling bearing is considered to be in a failure state as the starting point of the failure state, then the rolling bearing will move from the degraded state to the failure state, that is, failure behavior will occur.

[0151] 2) As the rolling bearing enters the final failure stage, the rolling bearing vibrates abnormally violently and the amplitude changes greatly. If three consecutive health factor HI points cannot be found to meet the conditions, the coordinate point of the first health factor HI where two consecutive health factors HI exceed the 3σ interval for the first time will be taken as the starting point of the rolling bearing failure state.

[0152] 3) If two consecutive health factor HI points cannot be found to satisfy the condition, the coordinate point of the first health factor HI point that exceeds the 3σ interval shall be taken as the starting point of the rolling bearing failure state.

[0153] 8. Experimental verification:

[0154] This invention uses the full-life data of the PHM2012 rolling bearing for evaluation. This experimental data comes from the PRONOSTIA test bench. The test bench consists of three parts: transmission, loading, and monitoring. The transmission part provides power and torque; the loading part uses a cylinder to apply radial pressure to the rolling bearing; and the monitoring part includes an accelerometer, thermocouple, pressure sensor, and speed sensor. The data sampling frequency is 25.6 kHz, with 2560 data points collected each time, and a sampling interval of 10 seconds. The experimental data includes three operating conditions:

[0155] Operating condition 1: Attached to 4000N, speed 1800r / min;

[0156] Operating condition 2: Attached at 4200N, speed 1650r / min;

[0157] Operating condition 3: Attached at 5000N, speed 1500r / min.

[0158] Rolling bearing life data (Bearing1-1, Bearing2-1, and Bearing3-3) were selected from three different operating conditions, and rolling bearing health condition degradation failure assessment analysis was performed according to the method of this invention. The full-life time-domain waveforms of Bearing1-1, Bearing2-1, and Bearing3-3 are shown below. Figure 2-4 As shown, the rolling bearing health factor curves and different health states obtained from the above three rolling bearing life data according to the method of the present invention are as follows. Figure 5-7 As shown in the figure, the health, degradation, and failure states of the rolling bearing throughout its entire lifespan can be clearly observed. For comparison, the traditional kurtosis index method was used to plot the kurtosis index trend charts for rolling bearings. The kurtosis index trend charts for Bearing1-1, Bearing2-1, and Bearing3-3 are shown below. Figure 8-10 As shown in the figure, it can be clearly seen that when the kurtosis index is used as a health factor, the peak of the curve fluctuates abnormally violently, and the fault initiation point determined according to the 3σ criterion is relatively late, which cannot accurately reflect the health status of the rolling bearing.

[0159] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for assessing the health status degradation and failure of rolling bearings by fusing multiple weighted indicators, characterized in that, include: Vibration signals of rolling bearings throughout their entire lifespan are collected to construct a rolling bearing vibration signal lifespan dataset. Calculate multiple characteristic indices for each vibration signal in the dataset; Calculate the Spearman rank correlation coefficient for each feature index to filter out feature indices that are greater than a preset threshold, and then normalize the filtered feature indices. The weight coefficients of the normalized feature indicators are calculated based on monotonicity, correlation, and robustness. The health factor of each vibration signal is obtained by multiplying and adding the selected feature index and its corresponding weight coefficient in the whole life dataset of rolling bearing vibration signals. The health factor curve of rolling bearing is plotted with the sample sequence of each vibration signal in the whole life dataset of rolling bearing vibration signals as the x-axis and the health factor as the y-axis. The starting point and state of degradation, as well as the starting point and state of failure, of the rolling bearing are determined based on the health factor curve of the rolling bearing. The rolling bearing vibration signal lifetime dataset includes: The whole life dataset of rolling bearing vibration signals is represented as follows The dataset includes n vibration time signal samples, where This represents the vibration time signal sample obtained from the i-th acquisition. The various characteristic indicators of each vibration signal in the computational dataset include: For the whole-life dataset S of rolling bearing vibration signals, calculate the mean, standard deviation, skewness, kurtosis, waveform factor, shape factor, marginal factor, peak-to-peak value, root mean square, peak value, and rectified average value characteristic index F for each vibration time signal sample. i ; For the monotonicity, correlation, and robustness results of each category of feature indicators after normalization, the results are summed according to the feature indicator category and averaged to obtain the weight coefficients for each category of feature indicators, denoted as w1, w2, ..., w c Wherein, the weight coefficient w of the i-th type of feature index i The calculation is as follows: In the formula, These are the results of monotonicity, correlation, and robustness indices corresponding to the i-th type of feature index, respectively. The maximum values ​​of the monotonicity, correlation, and robustness results of all selected feature indicators are used to process the weight coefficients of each feature indicator as follows: In the formula, The sum of the weight coefficients of all feature indicators after filtering is used to determine the final weight coefficient for each category of feature indicators after filtering. ; The health factors for obtaining each vibration signal include: For the i-th vibration time signal sample in the whole life dataset S of rolling bearing vibration signals Feature indicators after screening and the weighting coefficients corresponding to each feature index. The corresponding feature indices are multiplied and added together to obtain the vibration time signal sample. Corresponding health factor HI i : The health factor HI of each vibration time signal sample in the whole life dataset S of rolling bearing vibration signal is calculated. The health factor HI curve of the whole life stage of rolling bearing is plotted with the sample sequence of each vibration time signal sample in the whole life dataset S as the abscissa and the health factor HI as the ordinate.

2. The method for assessing the health status degradation and failure of rolling bearings by fusing multiple weighted indicators according to claim 1, characterized in that, The process of determining the starting point and degradation state of a rolling bearing based on its health factor curve includes: Data from the top 30% of the health factor HI curves across the entire lifespan of rolling bearings were selected, and their average value was calculated. and standard deviation , confirm 3 interval ; Based on the order of the sample sequence, the health factor HI of each vibration time signal sample in the whole life dataset S of rolling bearing vibration signals is compared with 3. For interval comparison, if the HI of three consecutive health factors all exceed 3... If the range is exceeded, the rolling bearing is considered to be in a degraded state, and the first value exceeding 3 is considered to be within the range. The coordinates of the interval health factor HI are taken as the starting point of the rolling bearing degradation state, denoted as HI. fdt .

3. The method for assessing the health status degradation and failure of rolling bearings by fusing multiple weighted indicators according to claim 2, characterized in that, The process of determining the starting point and failure state of rolling bearing based on the rolling bearing health factor curve includes: The average value was calculated by selecting the last 70% of the health factor HI curve data for the entire lifespan of rolling bearings. and standard deviation , confirm 3 interval ; Based on the order of the sample sequence, the health factor HI of each vibration time signal sample in the whole life dataset S of rolling bearing vibration signals is compared with 3. For interval comparison, if the HI of three consecutive health factors all exceed 3... If the range exceeds 3, the rolling bearing is considered to be in a failure state, and the first interval will be considered as such. The coordinates of the interval health factor HI are taken as the starting point of the rolling bearing failure state, denoted as HI. fft ; If there are no three consecutive health factors whose HI values ​​all exceed 3 The interval will be the first time that two consecutive health factors (HI) exceed 3. The coordinates of the first health factor HI in the interval are taken as the starting point of the rolling bearing failure state, denoted as HI. fft ; If there are no two consecutive health factors whose HI values ​​both exceed 3 The interval will be the first one exceeding 3. The coordinates of the health factor HI in the interval are taken as the starting point of the rolling bearing failure state, denoted as HI. fft .

Images