Desulfurization pump main shaft bearing dynamic performance evaluation method based on statistical robust theory

Through a method based on statistical robust theory, the operation status of the spindle bearing of the high-power desulfurization pump is evaluated in real time, which solves the problems of its life uncertainty and high maintenance costs, and improves the reliability and use efficiency of the equipment.

CN119003969APending Publication Date: 2024-11-22HUADIAN NINGXIA LINGWU POWER GENERATION CO LTD +1
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410943815.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-15
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The uncertainty of the life of the spindle bearing of the high-power desulfurization pump leads to huge discreteness in the operation of the equipment, high maintenance costs and major accident risks, and it is difficult for the existing technology to effectively evaluate its dynamic performance and operating conditions.

Method used

Using a method based on statistical robust theory, the bearing vibration data is detected online, and the rotating bearing vibration data is classified using the fusion method of robust theory and non-parametric hypothesis testing, its intrinsic characteristics are obtained, and the operating status of the spindle bearing is evaluated in real time.

Benefits of technology

It improves the efficiency and reliability of the desulfurization pump, reduces maintenance costs, and effectively prevents equipment failures and accidents.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119003969A_ABST
    Figure CN119003969A_ABST
Patent Text Reader

Abstract

A dynamic performance evaluation method for a main shaft bearing of a desulfurization pump based on a statistical robust theory relates to the technical field of bearings, and comprises the following steps: acquiring vibration data of a rolling bearing of a special sensor based on the statistical robust theory, and dividing time stages by applying a robust theory and nonparametric hypothesis testing fusion method; analyzing the vibration data at different time stages by using a median estimation and Huber M estimation fusion method to obtain an intrinsic interval, a median, an average value and a variation characteristic coefficient of the vibration data of the desulfurization pump main shaft bearing; finally, the dynamic performance and the service condition of the desulfurization pump main shaft bearing are judged according to the variation characteristic coefficient, the robustness of a distribution time sequence can be unknown through a median and Huber M estimation fusion method involved in the method, and the boundary value and the confidence level of robust data are obtained; the non-parametric Smiunov hypothesis testing and robust theory fusion can effectively reduce the influence of unrobust data and reduce the risk caused by signal level selection.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of bearings, and in particular to a method for evaluating the dynamic performance of a desulfurization pump main shaft bearing based on statistical robust theory. Background Art

[0002] The main shaft bearing of the desulfurization pump bears the axial load and radial load from the uncertain performance of the arc blades. It is the reaction of the pump body vibration and the main shaft rotation stability. It is the core component of the high-power desulfurization pump and plays an important role in the safety and reliability of the pump operation. However, during the operation of the equipment, the blades of the high-power desulfurization pump are eroded by lime water, corroded by alkaline liquid acid gas, temperature, and solid impact generated during the reaction process, resulting in axial and radial loads of uncertain size and direction. This uncertain load is ultimately carried on the rolling bearing, making the bearing life uncertain.

[0003] There are three aspects of uncertainty in the life of the main shaft bearing of a high-power desulfurization pump. On the one hand, there is the uncertainty of the change in load size; on the other hand, there is the uncertainty of the change in load direction; and the third aspect is the coupling of the change in load size and direction. The combined effect of these three aspects results in a huge discreteness in the service life of the main shaft bearing of a high-power desulfurization pump, ranging from three months to three years. That is, under the same operating conditions, using the same materials, processes, equipment, and the same people for assembly, the life span varies from a few months to a few years, which is about ten times different. As a result, some main shaft bearings are still operating normally even though they have far exceeded their design life during operation, while some bearings have failed before reaching their expected life.

[0004] In order to improve the reliability of desulfurization pump operation, thermal power plants and industrial enterprises use regular maintenance, condition-based maintenance and post-event maintenance to maintain desulfurization pumps. Regular maintenance often sacrifices the service life of the equipment in exchange for higher reliability; condition-based maintenance is accompanied by great uncertainty and even serious consequences; post-event maintenance has already caused consequences or greatly reduced production efficiency. This method results in high maintenance costs, equipment utilization efficiency and the risk of major accidents, which ultimately results in only 40% equipment utilization efficiency, high equipment maintenance costs and some operating risks.

[0005] The main shaft bearing of the desulfurization pump is one of the core components that support the rotation of the equipment. It is also the most fragile and easily damaged component among the components. 65% of its failures are related to the main shaft bearing, which plays a key role in the reliability and safety of the equipment operation. At present, the main shaft bearing of the desulfurization pump is mainly maintained by regular replacement. This method will treat the main shaft bearing with normal operating performance as a failed component, causing great waste; it will also cause the main shaft bearing that does not reach the design life to fail, causing serious failure of the desulfurization pump, resulting in waste of resources and major accidents. Therefore, studying the dynamic performance of the main shaft bearing of the desulfurization pump and predicting its operating status online can not only prevent the desulfurization pump from decreasing its working accuracy, but also significantly improve the use efficiency of the bearing, protect the environment, and prevent major accidents. It has important research significance and engineering application value.

[0006] The vibration of the desulfurization pump main shaft bearing is affected by many factors, and its performance presents diverse, complex, and chaotic characteristics. The performance data presents unknown distribution and unknown confidence level, which causes great trouble in effectively evaluating the service characteristics of the desulfurization pump main shaft bearing. How to evaluate the desulfurization pump main shaft bearing is one of the main bottlenecks in the pump industry. Summary of the invention

[0007] In order to overcome the shortcomings of the desulfurization pump main shaft service performance evaluation in the background technology, the present invention provides a desulfurization pump main shaft bearing dynamic performance evaluation method based on statistical robust theory in view of the unknown distribution and confidence level of the main shaft bearing performance data. The present invention detects the vibration data of the desulfurization pump bearing online, adopts the robust theory and non-parametric hypothesis test fusion method, classifies the rolling bearing vibration data, obtains the intrinsic characteristics of the rolling bearing, analyzes the variation of its performance, and evaluates the operating status of the main shaft bearing in real time, thereby effectively improving the use efficiency and reliability of the desulfurization pump and reducing maintenance costs.

[0008] In order to achieve the above-mentioned purpose of the invention, the present invention adopts the following technical solution:

[0009] A dynamic performance evaluation method for a desulfurization pump main shaft bearing based on statistical robust theory is proposed. The evaluation method constructs a desulfurization pump main shaft bearing performance evaluation model with three coefficients: confidence level, median, and median variance of vibration data at different time stages to evaluate the operating condition of the desulfurization pump main shaft bearing. The specific method is as follows:

[0010] The first step is to measure the vibration of the main shaft bearing of the desulfurization pump through the sensor and collect data. The method is used to divide different time stages. The data of each time stage constitutes the statistic X i ,

[0011] X i ={x i (1),x i(2),…,x i (k),…,x i (n)}, i=1, ……, q; n=1, ……, N (1)

[0012] In the formula, x i (k) is the kth vibration data of the i-th time stage; i is the time stage, there are q time stages; k is the kth moment, there are N moments in total;

[0013] The second step is to calculate the absolute value of the vibration data, and obtain the robust data of the desulfurization pump main shaft bearing vibration according to the fusion method of median estimation and Huber M estimation;

[0014] Step 3: According to the statistical robustness theory, the absolute difference between the median and the mean is the smallest and the most robust to obtain the confidence level ɑ i ;

[0015] Step 4: Rolling bearing performance range (a i ,b i ):

[0016] a i 、b i are the critical values ​​of the statistical interval, where a i Slightly smaller than y min1 ; b i Slightly larger than y min2 , thus obtaining the rolling bearing performance interval (a i ,b i );y i (1) is the minimum value of the order statistics of the initial series and time series of bearing performance, y i (2N) is the maximum value of the order statistics of the initial series and time series of bearing performance;

[0017] According to the number N of rolling bearing performance data column data, determine the number K of groupable bearing performance data intervals;

[0018] Rolling bearing performance data group spacing p i :

[0019]

[0020] In the formula, p i is the rolling bearing performance data group interval; i is the time series number, m is the number of time series; K is the number of groups that the rolling bearing performance data series can be grouped into;

[0021] The rolling bearing performance range (a i ,b i ) is divided into K intervals:

[0022] (a i ,a i +p i ],(a i +p i ,a i +2p i ]…,(a i +(K-1)p i ,b i ]; i = 1, 2, …, m

[0023] Among them, a i 、b i are the critical values ​​of statistical intervals; p i is the rolling bearing performance data group interval; i is the time series number, m is the number of time series; K is the number of groups that the rolling bearing performance data series can be grouped into;

[0024] Step 5: Frequency of rolling bearing performance data:

[0025] According to the formula (a i ,a i +p i ],(a i +p i ,a i +2p i ]…,(a i +(K-1)p i ,b i ]; i = 1, 2, ..., m, rolling bearing performance interval and initial time series X0 and time series X i , calculate the initial time series X0 and time series X in the rolling bearing performance range i The number of data n 0k and n ik , where k = 1, 2, ..., K;

[0026] Step 6. Frequency of rolling bearing performance data:

[0027] According to the frequency of rolling bearing performance data and the initial time series X0 and time series X i The number of data, calculation formula (a i ,a i +p i ],(a i +p i ,a i +2p i ]…,(a i +(K-1)p i ,b i ]; i = 1, 2, ..., m, the initial time series X0 and time series X of the rolling bearing performance interval i The frequency of the data is f0k 、f ik :

[0028]

[0029] In the formula, f 0k is the frequency of the kth interval of the initial time series; k is the frequency number of the initial time series, K is the number of time series frequencies; N is the number of time series data; n 0k is the frequency of the kth interval of the initial time series;

[0030]

[0031] In the formula, f ik is the frequency of the kth interval of the i-th time series; i is the time series number, m is the number of time series; k is the time series frequency number, K is the number of time series frequencies; N is the number of time series data; n ik is the frequency of the kth interval of the i-th time series;

[0032] Step 7. Cumulative frequency of rolling bearing performance data:

[0033] The cumulative frequency of the initial time series data of rolling bearing performance is

[0034]

[0035] In the formula, F 0v is the vth cumulative frequency of the initial time series; v is the cumulative frequency sequence number of the initial time series, K is the number of cumulative frequencies of the initial time series; f 0k is the frequency of the kth interval of the initial time series; k is the frequency sequence number of the initial time series, and K is the cumulative number of frequencies of the initial time series;

[0036] The performance cumulative frequency of the rolling bearing i-th time series data is

[0037]

[0038] In the formula, F iv is the vth cumulative frequency of the i-th time series; i is the time series number, m is the number of time series; v is the cumulative frequency number of the time series, K is the number of cumulative frequencies of the time series; f ik is the frequency of the kth interval of the time series; k is the frequency serial number of the initial time series, and K is the cumulative frequency number of the initial time series;

[0039] Step 8. Rolling bearing performance statistics:

[0040] Rolling bearing performance statistics D ik :

[0041] D ik =|F 0k -F ik |;i=1,2,…,m;k=1,2,…,K

[0042] Where D ik is the absolute difference between the kth frequency of the initial time series of the rolling bearing and the i-th time series; F 0k is the cumulative frequency of the kth interval of the initial time series; F ik is the cumulative frequency of the kth interval of the i-th time series; i is the time series number, m is the number of time series; k is the time series frequency number, K is the number of time series frequencies;

[0043] The absolute difference D between the initial time series frequency of the rolling bearing and the i-th time series frequency ik As a statistic, D ik represents the similarity between the distribution of the initial time series of the rolling bearing and the i-th time series data; D ik The smaller it is, the more similar the data distribution of the initial time series is to that of the ith time series. ik The larger the value is, the more dissimilar the distribution of the initial time series is to the i-th time series; the more similar the data distributions are, the closer their performances are, that is, the same; otherwise, the performances are different, which is used to analyze the retention of rolling bearing performance;

[0044] Step 9. Critical value D of rolling bearing performance statistics 2N,α :

[0045] The critical value of the statistic is D 2N,α , where N is the number of time series data; α is the significance level; this value can be found in the critical value table of Kolmogorov test;

[0046] Step 10: Rolling bearing performance maintains reliability:

[0047] In order to analyze the similarity between the initial time series and the i-th time series of the rolling bearing, find the absolute difference D between the frequency of the initial time series and the frequency of the i-th time series. ik The maximum value maxD ik , if maxD ik Less than the critical value of the statistic D 2N,α , indicating that the data distribution of the initial time series is similar to that of the i-th time series, that is, the performance of the two is consistent. Continue to compare the i=i+1th time series until the p=i=i+1th time series, maxD ik Greater than the critical value of the statistic D 2N,α , indicating that the data distribution of the initial time series is not similar to that of the i-th time series, that is, the performance is different, the rolling bearing performance has changed, and the intervals are merged into Q i, the performance of the rolling bearing is kept reliable;

[0048] Step 11: Find the order statistic Q i The median M i , average value A i , median variance SM i , mean square error SA i ,

[0049]

[0050]

[0051] Where M min is the median M of different stages i The minimum value of the median M at different time stages is calculated i With the average value A i The absolute difference R i And the variation characteristic parameter C i :

[0052] R i =|M i -A i | (21)

[0053]

[0054] In the formula, the absolute difference R i Indicates the variability of the vibration data of the main shaft bearing of the desulfurization pump. The larger the value, the more serious the variation. The variation characteristic parameter C i Indicates the degree of variation in the performance of the desulfurization pump main shaft bearing. The larger the value, the worse the performance.

[0055] The dynamic performance evaluation method of the desulfurization pump main shaft bearing based on statistical robust theory, when the median estimation in the second step is combined with the Huber M estimation method to obtain the robust data of the desulfurization pump main shaft bearing vibration, is as follows:

[0056] First, arrange the order statistics Y in ascending order i ,

[0057] Y i ={y i (1),y i (2),…,y i (k),…,y i (n)}, i=1, ……, q; n=1, ……, N (2)

[0058] In the formula, y i(k) is the kth vibration data of the i-th time stage; i is the time stage, there are q time stages; k is the kth moment, there are N moments in total;

[0059] According to statistics, find the sequence Y i median β i

[0060]

[0061] Where:

[0062] β i ——Sequence y i median;

[0063] y i ——The rolling bearing vibration sequence in which the absolute values ​​of the rolling bearing vibration data at the i-th time stage are arranged in ascending order;

[0064]

[0065] Where:

[0066] β i ——Sequence Y i median;

[0067] y i ——The rolling bearing vibration sequence in which the absolute values ​​of the rolling bearing vibration data at the i-th time stage are arranged in ascending order;

[0068] Data left sequence, right sequence

[0069] Assume y i (b) and y i (e) is the data sequence y i The b and e experimental data values ​​in the data are b and e, respectively, where b and e are two values ​​in data 1, 2, ..., N, and y i (b)≤β i , β i ≤y i (e);

[0070] When the data column y i Chinese: When y i (n)≤y i (b)

[0071] y i (n) = y i (b); i=1,2,…,m; n=1,2,…,N (5)

[0072] When i (b)≤y i (n) time:

[0073] y i (n) = y i (e); i=1,2,…,m; n=1,2,…,N (6)

[0074] Where:

[0075] y i (n)——the nth data of the absolute value of the rolling bearing vibration data in the i-th time period arranged in ascending order;

[0076] Get the left data sequence: y i (b),…,β i ,y i The number of data in (b) is n1;

[0077] Get the right data sequence: β i ,…,y i (e), y i The number of data in (e) is n2;

[0078] According to the left and right data sequences, the improved data sequence Z can be obtained i :

[0079] Z i ={z i (n)}={y i (b),…,β i ,…,y i (e)};

[0080] i=1,2,…,m; n=1,2,…,N (7)

[0081] Where:

[0082] z i (n)——the nth data of the rolling bearing at the i-th time stage;

[0083] y i (b) y i (e)——Data sequence Y i The two experimental data values ​​b and e in ;

[0084] b, e——two values ​​in data 1, 2, …, N;

[0085] β i ——Sequence Y i median;

[0086] Calculated Improved Data Column Z i The average value η i (n1,n2)

[0087]

[0088] Where:

[0089] η i (n1,n2)——Robust data column Z i The average value of

[0090] i——time phase number;

[0091] m——number of time stages;

[0092] n1——y i (b) The number of data;

[0093] n2——y i (e) the number of data;

[0094] n——data sequence number, there are N in total;

[0095] i——time phase number;

[0096] m——number of time stages;

[0097] Calculate the median beta i and improved data column Z i The average value η i The absolute difference D of (n1,n2) i

[0098] D i =|β i -η i (n1,n2); i=1,2,…,m (9)

[0099] Where:

[0100] D i ——Median β i and improved data column Z i The average value η i The absolute difference of (n1,n2);

[0101] β i ——Median;

[0102] η i (n1,n2) – average value in the robust data column;

[0103] Calculation of n1, n2:

[0104] In this paper, according to the robust characteristics of the median in modern statistics, n1 and n2 are equal, that is, n1 = n2.

[0105]

[0106] Where:

[0107] n1——y i (b) The number of data;

[0108] n2——y i (e) the number of data;

[0109] N——number of data;

[0110] N——odd number, use formula (13);

[0111] N——even number, use formula (14);

[0112] Calculate the critical value K of the distribution function i1 , K i2

[0113] K i1 , K i2 is the critical value of the vibration data in the i-th time stage. Formula (7-9) obtains a series of difference values ​​D according to different n1 and n2. i , where D imin D i The minimum value of y i (b) and y i (e) correspond to K i1 and K i2 ,Right now

[0114] K i1 =y i (b); i=1,2,…,m (13)

[0115] K i2 =y i (e); i=1,2,…,m (14)

[0116] Where:

[0117] K i1 ——the minimum value of the distribution function;

[0118] K i2 ——the maximum value of the distribution function;

[0119] y i (b) y i (e)——Data sequence Y i The two experimental data values ​​b and e in ;

[0120] b, e——two values ​​in data 1, 2, …, N;

[0121] According to K i1 , K i2 The eigenvalue interval of the experimental data in this time period is obtained [Ki1 ,K i2 ], K i1 is the minimum critical value of the eigenvalue interval, K i2 is the maximum critical value of the distribution function;

[0122] According to the formula [K min1 ,K min2 ] and replace the data in the eigenvalue interval with the critical value of the eigenvalue interval, and the robust distribution function P can be obtained. i (n):

[0123]

[0124] i=1,2,…,m, n=1,2,…,N (15).

[0125] By adopting the above technical solution, the present invention has the following advantages:

[0126] The present invention is based on statistical robust theory, collects vibration data of rolling bearings of special sensors, and divides time stages by using robust theory and non-parametric hypothesis test fusion method; analyzes vibration data of different time stages by using median estimation and Huber M estimation fusion method, and obtains the intrinsic interval, median, mean value and variation characteristic coefficient of desulfurization pump main shaft bearing vibration data; finally, the dynamic performance and service status of the desulfurization pump main shaft bearing are judged according to the variation characteristic coefficient. The median and Huber M estimation fusion method involved in the present invention can determine the robustness of unknown distribution time series, and obtain the boundary value and confidence level of robust data; the fusion of non-parametric Smirnov hypothesis test and robust theory can effectively reduce the impact of unrobust data and reduce the risk caused by confidence level selection. The present invention does not need to know the distribution type and confidence level of the data, and classifies the rolling bearing performance data to evaluate the dynamic performance of the rolling bearing, providing an effective method for dynamic analysis of unknown distribution time series data. BRIEF DESCRIPTION OF THE DRAWINGS

[0127] Figure 1 is the vibration of the rolling bearing during time periods 1-4 in the embodiment of the present invention;

[0128] Figure 2 is the vibration variation characteristics of the rolling bearing at different time stages in the embodiment of the present invention;

[0129] Figure 3 is a time series of vibration signals of a 7008AC / P2 rolling bearing in an embodiment of the present invention;

[0130] Figure 4 7008AC / P2 rolling bearing variation coefficient change diagram in the embodiment of the present invention. DETAILED DESCRIPTION

[0131] The present invention can be explained in more detail by the following examples, but the present invention is not limited to the following examples;

[0132] Before the present invention was proposed, the main shaft bearing of the high-power desulfurization pump was a combination of tapered bearings and cylindrical bearings selected according to load, speed and environment, and there was little data specifically studying the performance of the main shaft bearing of the desulfurization pump. At present, the performance analysis of bearings mainly focuses on the performance research and fault diagnosis of rolling bearings.

[0133] The desulfurization pump operates in an alkaline and neutral environment. Its load (due to blade wear) and temperature are constantly changing, which leads to the nonlinear, diverse and complex performance of the main shaft bearing. That is, there is no fixed distribution type or it cannot be reflected by a certain distribution in statistics. This is an important part of the lack of information theory. There are two main characteristics: first, the data sequence distribution is not uniform and cannot be described by a certain distribution; second, the change trend of the data sequence is unknown, and its change trend has no certain rules and is random, which belongs to the category of lack of information.

[0134] For a long time, the performance evaluation of bearings has been mainly based on the statistical analysis of fatigue life and reliability, with Weibull distribution as the basic assumption. However, before the desulfurization pump main shaft bearings suffer from fatigue damage such as spalling, pitting, and plastic deformation, they have already been stuck, sintered, and fractured. These new characteristics and reliability concepts pose new challenges to traditional bearing evaluation theories, because the probability distribution of these performance data is unknown and the characteristic data is scarce, especially the nonlinear dynamic contact and collision between the internal parts of the bearing, the unknown distribution of nonlinear viscosity-temperature-viscosity-pressure characteristics of the lubricating medium, and the uncertainty and diversity of unknown distribution nonlinear characteristics between the precision / quality loss and the performance data, which makes the bearing performance and change trend change with time and service conditions, making it difficult to establish a comprehensive, reasonable, and effective theoretical model.

[0135] In recent years, many scholars have used different methods to study the nonlinearity, stability and influencing factors of rolling bearing performance, and have achieved many results. The results show that the bearing performance is nonlinear, variable, diverse and the influencing factors are complex. This can lead to the conclusion that rolling bearings have unknown distribution characteristics. The reason is that there are many influencing factors. At the same time, these factors have different effects under different conditions and are coupled with each other, resulting in unknown distribution characteristics of bearing performance.

[0136] The research on bearing failure mainly focuses on three aspects: signal analysis and processing technology, pattern recognition methods, and performance evaluation and prediction.

[0137] (1) Bearing fault signal analysis and processing technology mainly includes traditional signal analysis methods and modern signal analysis and processing technology. Traditional signal analysis methods are based on the assumption that the signal is linear, stable and has minimum phase. They use averaging, power spectrum, refined spectrum, envelope analysis and factor discrimination methods to give the statistical average results of the signal from the time domain or frequency domain. This method will ignore some important features in the signal, especially some subtle features, which can easily lead to missed diagnosis and misjudgment. Modern signal analysis and processing technology starts from nonlinear, non-causal, non-minimum phase system, non-Gaussian, non-stationary, non-integer dimensional signal and non-white additive noise, bringing new vitality to mechanical vibration signal analysis. Wavelet analysis technology has been widely used in the field of bearing faults due to its algorithm diversity, information sensitivity and information re-extractability; the Hilbert spectrum obtained by data adaptiveness through empirical mode decomposition has a clear physical meaning and is widely used in the extraction of bearing fault diagnosis features; the combination of cyclic statistics theory and wavelet theory integrates the advantages of the two methods and is widely used in rolling bearings and gears with obvious modulation characteristics.

[0138] (2) Fault diagnosis is achieved by using pattern recognition methods based on signal analysis and processing. Artificial Neural Network (ANN) has the advantages of nonlinearity, self-organization, adaptability, self-learning, large-scale parallel processing and associative ability, and can establish the connection between signal characteristics and faults; Support Vector Machine (SVM) can solve the problem of small sample learning, and the established classifier has good generalization ability; Hidden Markov Model (HMM) has the properties of a double random process and has been widely used in the identification of drill wear status.

[0139] (3) SOM and other methods are rolling bearing performance degradation assessment methods. They use data under normal conditions to establish an assessment model and use the minimum quantization error as a performance degradation indicator. Some scholars use neural networks to study rolling bearing failures and remaining life. Non-statistical analysis of bearing damage performance degradation studies the relationship between statistical parameters in the form of bearing case analysis, but no consistent results have been obtained. Chaos is used to study bearing performance degradation, and it is found that the correlation dimension plays an important role in rolling bearing performance degradation. Fuzzy mathematics and grayscale theory are used to study rolling bearing performance changes. These methods study the performance degradation of rolling bearings from different aspects and have achieved many research results, but there are also many shortcomings. These methods require the rolling bearing to be in a normal state. The performance of rolling bearings is affected by many factors. The normal state is changing and it is difficult to obtain accurately. Some methods solve the initial state of the rolling bearing, but require the confidence level of the data. Therefore, the above methods fail to solve the problems of unknown distribution and unknown confidence level in the process of rolling bearing performance degradation, and fail to evaluate the service condition of rolling bearings well.

[0140] From the perspective of statistical theory, the performance data characteristics of the main shaft bearing of the desulfurization pump show unknown distribution characteristics. Traditional bearing performance evaluation is based on classical statistical theory, which requires a large amount of data, given trends and probability distribution. The dynamic performance evaluation of the main shaft bearing of the desulfurization pump does not meet this demanding requirement, which is the fundamental difficulty of classical statistics. It has become a scientific frontier topic in the field of bearing performance analysis and information analysis, and is a key issue that urgently needs to be solved in the development of related disciplines and basic research on bearing industrial applications.

[0141] Based on this, an invention patent for a dynamic evaluation system for the main shaft bearing of a desulfurization pump was disclosed. By online analyzing the vibration data of the desulfurization pump bearing, the operating status of the main shaft bearing is evaluated in real time, aiming to improve the efficiency and reliability of the desulfurization pump and reduce maintenance costs.

[0142] Combined with Figures 1 to 4 The present invention discloses a method for evaluating the dynamic performance of a desulfurization pump main shaft bearing based on statistical robust theory. The method constructs a desulfurization pump main shaft bearing performance evaluation model with three coefficients: confidence level, median, and median variance of vibration data at different time stages to evaluate the operating status of the desulfurization pump main shaft bearing. The specific method is as follows:

[0143] The first step is to measure the vibration of the main shaft bearing of the desulfurization pump through the sensor and collect data. The method is used to divide different time stages. The data of each time stage constitutes the statistic X i ,

[0144] X i ={x i (1),x i (2),…,x i (k),…,x i(n)}, i=1, ……, q; n=1, ……, N (1)

[0145] In the formula, x i (k) is the kth vibration data of the i-th time stage; i is the time stage, there are q time stages; k is the kth moment, there are N moments in total;

[0146] The second step is to calculate the absolute value of the vibration data, and obtain the robust data of the desulfurization pump main shaft bearing vibration according to the median estimation and Huber M estimation fusion method (for details, please refer to the Chinese patent, publication number CN105631238B);

[0147] During implementation, the median estimation and Huber M estimation fusion method are used to obtain the robust data of the desulfurization pump main shaft bearing vibration, as follows:

[0148] First, arrange the order statistics Y in ascending order i ,

[0149] Y i ={y i (1),y i (2),…,y i (k),…,y i (n)}, i=1, ……, q; n=1, ……, N (2)

[0150] In the formula, y i (k) is the kth vibration data of the i-th time stage; i is the time stage, there are q time stages; k is the kth moment, there are N moments in total;

[0151] According to statistics, find the sequence Y i median β i

[0152]

[0153] Where:

[0154] β i ——Sequence y i median;

[0155] y i ——The rolling bearing vibration sequence in which the absolute values ​​of the rolling bearing vibration data at the i-th time stage are arranged in ascending order;

[0156]

[0157] Where:

[0158] β i ——Sequence Y i median;

[0159] y i ——The rolling bearing vibration sequence in which the absolute values ​​of the rolling bearing vibration data at the i-th time stage are arranged in ascending order;

[0160] Data left sequence, right sequence

[0161] Assume y i (b) and y i (e) is the data sequence y i The b and e experimental data values ​​in the data are b and e, respectively, where b and e are two values ​​in data 1, 2, ..., N, and y i (b)≤β i , β i ≤y i (e);

[0162] When the data column y i Chinese: When y i (n)≤y i (b)

[0163] y i (n) = y i (b); i=1,2,…,m; n=1,2,…,N (5)

[0164] When i (b)≤y i (n) time:

[0165] y i (n) = y i (e); i=1,2,…,m; n=1,2,…,N (6)

[0166] Where:

[0167] y i (n)——the nth data of the absolute value of the rolling bearing vibration data in the i-th time period arranged in ascending order;

[0168] Get the left data sequence: y i (b),…,β i ,y i The number of data in (b) is n1;

[0169] Get the right data sequence: β i ,…,y i (e), y i The number of data in (e) is n2;

[0170] According to the left and right data sequences, the improved data sequence Z can be obtained i :

[0171] Zi ={z i (n)}={y i (b),…,β i ,…,y i (e)};

[0172] i=1,2,…,m; n=1,2,…,N (7)

[0173] Where:

[0174] z i (n)——the nth data of the rolling bearing at the i-th time stage;

[0175] y i (b) y i (e)——Data sequence Y i The two experimental data values ​​b and e in ;

[0176] b, e——two values ​​in data 1, 2, …, N;

[0177] β i ——Sequence Y i median;

[0178] Calculated Improved Data Column Z i The average value η i (n1,n2)

[0179]

[0180] Where:

[0181] η i (n1,n2)——Robust data column Z i The average value of

[0182] i——time phase number;

[0183] m——number of time stages;

[0184] n1——y i (b) The number of data;

[0185] n2——y i (e) the number of data;

[0186] n——data sequence number, there are N in total;

[0187] i——time phase number;

[0188] m——number of time stages;

[0189] Calculate the median beta i and improved data column Z iThe average value η i The absolute difference D of (n1,n2) i

[0190] D i =|β i -η i (n1,n2); i=1,2,…,m (9)

[0191] Where:

[0192] D i ——Median β i and improved data column Z i The average value η i The absolute difference of (n1,n2);

[0193] β i ——Median;

[0194] η i (n1,n2) – average value in the robust data column;

[0195] Calculation of n1, n2:

[0196] In this paper, according to the robust characteristics of the median in modern statistics, n1 and n2 are equal, that is, n1 = n2.

[0197]

[0198] Where:

[0199] n1——y i (b) The number of data;

[0200] n2——y i (e) the number of data;

[0201] N——number of data;

[0202] N——odd number, use formula (13);

[0203] N——even number, use formula (14);

[0204] Calculate the critical value K of the distribution function i1 , K i2

[0205] K i1 , K i2 is the critical value of the vibration data in the i-th time stage. Formula (7-9) obtains a series of difference values ​​D according to different n1 and n2. i , where D imin D i The minimum value of y i (b) and yi (e) correspond to K i1 and K i2 ,Right now

[0206] K i1 =y i (b); i=1,2,…,m (13)

[0207] K i2 =y i (e); i=1,2,…,m (14)

[0208] Where:

[0209] K i1 ——the minimum value of the distribution function;

[0210] K i2 ——the maximum value of the distribution function;

[0211] y i (b) y i (e)——Data sequence Y i The two experimental data values ​​b and e in ;

[0212] b, e——two values ​​in data 1, 2, …, N;

[0213] According to K i1 , K i2 The eigenvalue interval of the experimental data in this time period is obtained [K i1 ,K i2 ], K i1 is the minimum critical value of the eigenvalue interval, K i2 is the maximum critical value of the distribution function;

[0214] According to the formula [K min1 ,K min2 ] and replace the data in the eigenvalue interval with the critical value of the eigenvalue interval, and the robust distribution function P can be obtained. i (n):

[0215]

[0216] i=1,2,…,m, n=1,2,…,N (15).

[0217] Step 3: According to the statistical robustness theory, the absolute difference between the median and the mean is the smallest and the most robust to obtain the confidence level ɑ i ;

[0218] Step 4: Rolling bearing performance range (a i ,b i ):

[0219] a i 、b i are the critical values ​​of the statistical interval, where a i Slightly smaller than y min1 ; b i Slightly larger than y min2 , thus obtaining the rolling bearing performance interval (a i ,b i );y i (1) is the minimum value of the order statistics of the initial series and time series of bearing performance, y i (2N) is the maximum value of the order statistics of the initial series and time series of bearing performance;

[0220] According to the number N of rolling bearing performance data column data, determine the number K of groupable bearing performance data intervals;

[0221] Rolling bearing performance data group spacing p i :

[0222]

[0223] In the formula, p i is the rolling bearing performance data group interval; i is the time series number, m is the number of time series; K is the number of groups that the rolling bearing performance data series can be grouped into;

[0224] The rolling bearing performance range (a i ,b i ) is divided into K intervals:

[0225] (a i ,a i +p i ],(a i +p i ,a i +2p i ]…,(a i +(K-1)p i ,b i ]; i = 1, 2, …, m

[0226] Among them, a i 、b i are the critical values ​​of statistical intervals; p i is the rolling bearing performance data group interval; i is the time series number, m is the number of time series; K is the number of groups that the rolling bearing performance data series can be grouped into;

[0227] Step 5: Frequency of rolling bearing performance data:

[0228] According to the formula (a i ,a i +pi ],(a i +p i ,a i +2p i ]…,(a i +(K-1)p i ,b i ]; i = 1, 2, ..., m, rolling bearing performance interval and initial time series X0 and time series X i , calculate the initial time series X0 and time series X in the rolling bearing performance range i The number of data n 0k and n ik , where k = 1, 2, ..., K;

[0229] Step 6. Frequency of rolling bearing performance data:

[0230] According to the frequency of rolling bearing performance data and the initial time series X0 and time series X i The number of data, calculation formula (a i ,a i +p i ],(a i +p i ,a i +2p i ]…,(a i +(K-1)p i ,b i ]; i = 1, 2, ..., m, the initial time series X0 and time series X of the rolling bearing performance interval i The frequency of the data is f 0k 、f ik :

[0231]

[0232] In the formula, f 0k is the frequency of the kth interval of the initial time series; k is the frequency number of the initial time series, K is the number of time series frequencies; N is the number of time series data; n 0k is the frequency of the kth interval of the initial time series;

[0233]

[0234] In the formula, f ik is the frequency of the kth interval of the i-th time series; i is the time series number, m is the number of time series; k is the time series frequency number, K is the number of time series frequencies; N is the number of time series data; n ik is the frequency of the kth interval of the i-th time series;

[0235] Step 7. Cumulative frequency of rolling bearing performance data:

[0236] The cumulative frequency of the initial time series data of rolling bearing performance is

[0237]

[0238] In the formula, F 0v is the vth cumulative frequency of the initial time series; v is the cumulative frequency sequence number of the initial time series, K is the number of cumulative frequencies of the initial time series; f 0k is the frequency of the kth interval of the initial time series; k is the frequency sequence number of the initial time series, and K is the cumulative number of frequencies of the initial time series;

[0239] The performance cumulative frequency of the rolling bearing i-th time series data is

[0240]

[0241] In the formula, F iv is the vth cumulative frequency of the i-th time series; i is the time series number, m is the number of time series; v is the cumulative frequency number of the time series, K is the number of cumulative frequencies of the time series; f ik is the frequency of the kth interval of the time series; k is the frequency serial number of the initial time series, and K is the cumulative frequency number of the initial time series;

[0242] Step 8. Rolling bearing performance statistics:

[0243] Rolling bearing performance statistics D ik :

[0244] D ik =|F 0k -F ik |;i=1,2,…,m;k=1,2,…,K

[0245] Where D ik is the absolute difference between the kth frequency of the initial time series of the rolling bearing and the i-th time series; F 0k is the cumulative frequency of the kth interval of the initial time series; F ik is the cumulative frequency of the kth interval of the i-th time series; i is the time series number, m is the number of time series; k is the time series frequency number, K is the number of time series frequencies;

[0246] The absolute difference D between the initial time series frequency of the rolling bearing and the i-th time series frequency ik As a statistic, D ik represents the similarity between the distribution of the initial time series of the rolling bearing and the i-th time series data; D ikThe smaller it is, the more similar the data distribution of the initial time series is to that of the ith time series. ik The larger the value is, the more dissimilar the distribution of the initial time series is to the i-th time series; the more similar the data distributions of the two are, the closer their performances are, that is, the same performances; otherwise, the performances of the two are different, which is used to analyze the retention of rolling bearing performance;

[0247] Step 9. Critical value D of rolling bearing performance statistics 2N,α :

[0248] The critical value of the statistic is D 2N,α , where N is the number of time series data; α is the significance level; this value can be found in the critical value table of Kolmogorov test;

[0249] Step 10: Rolling bearing performance maintains reliability:

[0250] In order to analyze the similarity between the initial time series and the i-th time series of the rolling bearing, find the absolute difference D between the frequency of the initial time series and the frequency of the i-th time series. ik The maximum value maxD ik , if maxD ik Less than the critical value of the statistic D 2N,α , indicating that the data distribution of the initial time series is similar to that of the i-th time series, that is, the performance of the two is consistent. Continue to compare the i=i+1th time series until the p=i=i+1th time series, maxD ik Greater than the critical value of the statistic D 2N,α , indicating that the data distribution of the initial time series is not similar to that of the i-th time series, that is, the performance is different, the rolling bearing performance has changed, and the intervals are merged into Q i , the performance of the rolling bearing is kept reliable;

[0251] Step 11: Find the order statistic Q i The median M i , average value A i , median variance SM i , mean square error SA i ,

[0252]

[0253] Where M min is the median M of different stages i The minimum value of the median M at different time stages is calculated i With the average value A i The absolute difference R i And the variation characteristic parameter C i :

[0254] R i=|M i -A i | (21)

[0255]

[0256] In the formula, the absolute difference R i Indicates the variability of the vibration data of the main shaft bearing of the desulfurization pump. The larger the value, the more serious the variation. The variation characteristic parameter C i Indicates the degree of variation in the performance of the desulfurization pump main shaft bearing. The larger the value, the worse the performance.

[0257] The specific embodiments of the present invention are as follows:

[0258] Rolling bearing vibration data from Case Western Reserve University Center website

[0259] According to the bearing L10 life, the reliability of rolling bearings is 90%. Therefore, it is considered that under normal working conditions of rolling bearings, the significance level is 0 to 0.1. In GB / T24607-2009 "Rolling Bearing Life and Reliability Test Procedure", it is stipulated that the fatigue spalling area should not be less than 0.5mm 2 When the bearing fails, the spalling diameter is about 0.798mm, and the bearing performance fails. To this end, the bearing performance degradation problem is studied, and the normal operation condition of the vibration data of the channel damage diameter 0mm is selected; the channel damage diameter 0.7112mm (reaching 89% of the failure diameter) is the critical condition of the failure data, and its vibration data law is studied. At the same time, two channel damage diameters of 0.1778mm and 0.5334mm are inserted as transition states.

[0260] The rolling bearing vibration data from the Case Western Reserve University Center website in the United States was used, and the vibration data was measured from the rolling bearing fault simulation experimental platform. The rolling bearing fault simulation experimental platform consists of a power source motor, a torque sensor / decoder (sensor), and a power test circuit. SKF6205 is the tested bearing, which supports the rotating shaft of the motor. The bearing vibration is measured using an acceleration sensor in V. The speed of SKF6205 is 1797r / min, and the sampling frequency is 12kHz. The rolling bearing vibration was measured when the inner ring groove damage diameter was 0mm, 0.1778mm, 0.5334mm, and 0.7112mm, respectively, and the measurement data was 1600.

[0261] Since the rolling bearing vibration is obtained under four conditions of groove damage diameter, it can be simulated as the vibration data obtained in four time stages X i; i = 1, 2, 3, 4; each data sequence has N = 1600 data. Figure 1 (a) to (d). It can be seen that as the service time of the bearing increases, the larger the vibration range of the rolling bearing, the worse its performance. Therefore, by analyzing the vibration performance of the rolling bearing, the service status at different times can be evaluated.

[0262] According to the above method Figure 1 In the rolling vibration data processing, when the number of data is 200, 400, and 800, according to the above method, the data can be divided into 1-4 service conditions, which is completely consistent with the rolling bearing experiment. The specific evaluation results of the four experimental data are shown in Tables 1 and Figure 2 result.

[0263] Table 1 Rolling bearing vibration performance data processing results

[0264]

[0265] According to Table 1 and Figure 2 When the inner ring groove is not damaged (the diameter is 0mm), the variation characteristic coefficient is 16.7%, indicating that the service performance of the rolling bearing is normal, which is used as the reference value of the rolling bearing variation characteristic; when the diameter of the inner ring groove damage of the rolling bearing is small (the diameter is 0.1778mm), the variation characteristic coefficient is 134.1%, and the vibration performance is significantly aggravated, indicating that the vibration performance of the rolling bearing is sensitive, and the service performance changes at the same time, and the bearing operation condition should be closely monitored; when the inner ring groove damage of the rolling bearing is large (0.5334mm), the variation characteristic coefficient is 315.8%, indicating that the service condition of the rolling bearing has obviously begun to deteriorate, and it may fail at any time, and the bearing needs to be replaced; when the inner ring groove damage diameter continues to increase (0.7112mm), the variation characteristic coefficient quickly reaches 1015.7%, which increases by an order of magnitude, indicating that the service performance of the rolling bearing has failed and a major safety accident may occur. At the same time, it shows that the rolling bearing variation characteristic coefficient has changed significantly, which can indicate the deterioration of the performance of the rolling bearing.

[0266] Desulfurization pump main shaft bearing performance degradation experiment

[0267] (1) Experimental subjects

[0268] In order to further verify the effectiveness of the above method and evaluation system, the vibration data of 7008AC / P2 rolling bearing was taken as the research object to evaluate the performance and service status of precision bearings.

[0269] (2) Experimental equipment

[0270] The vibration of bearing 7008AC / P2 was carried out on the ABLT-1A bearing life enhancement testing machine.

[0271] This experiment is a rolling bearing fatigue life strengthening (fast) experiment conducted on the ABLT-1A bearing life strengthening test machine of Hangzhou Bearing Test Research Center. The test machine is mainly composed of a test head, a test head seat, a transmission system, a loading system, a lubrication system, a computer control system, etc. Since the test equipment is an existing equipment and is not the point of protection of the present invention, its specific structure will not be described. The bearing model used in the experiment is 7008AC / P2, and the number is 4 sets.

[0272] (3) Experimental parameters

[0273] The rated dynamic load (Cr) of precision rolling bearing 7008AC / P2 is 19KN. When oil lubrication is sufficient, its maximum speed is 12000r / min, and P / C=0.3.

[0274] The axial load (Fa) of the strengthening test is 4.58KN, the radial load (Fr) is 4.17KN, and the speed is 4000r / min. The vibration data is collected once per minute, the sampling frequency is 300Hz, and 8670 data are collected.

[0275] Under the above test conditions, the vibration data of precision bearing 7008AC / P2 was collected. Figure 3 .

[0276] According to the above method Figure 3 In the data processing, taking 200 vibration data as a unit, the above method is used for analysis. The vibration data of the main shaft rolling bearing of the 7008AC / P2 desulfurization pump can be divided into 4 groups, namely 4 service conditions. The first type is 1-2800 vibration data, the second type is 2800-5000 vibration data, the third type is 5000-7000 vibration data, and the fourth type is 7000-1670 vibration data. The specific results of these 4 service conditions are shown in Tables 2 and Figure 4 result.

[0277] Table 2.7008AC / P2 rolling bearing service performance evaluation

[0278]

[0279] According to Table 2 and Figure 4 It can be seen that during the strengthening process of the 7008AC / P2 rolling bearing, the median value and coefficient of variation of the vibration changed significantly, with the median value of the vibration increasing from 9.3m / s at the beginning to 2 18.8m / s to failure 2 , and experienced 12.2m / s in the middle 2 、12.1m / s 2transition period; the coefficient of variation characteristics changed significantly from 0.56% at the beginning to 28.6% at the failure period, with fluctuations of 2.11% and 0.42% in the middle. It also reflects the information of the median and average vibration values, and can well show the changes in the dynamic performance of the rolling bearing, which can be used as a basis for evaluating the dynamic performance of the desulfurization pump.

[0280] The present invention proposes to use a fusion method of median estimation and Huber M estimation and a fusion method of Smirnov non-parametric hypothesis test to reasonably divide the service intervals of rolling bearing performance data and evaluate the service status of rolling bearings.

[0281] Firstly, the vibration data of rolling bearings at different time stages were processed by the fusion method of median estimation and Huber M estimation, and the robust data boundary values ​​and significance levels of rolling bearings at different time stages were obtained. Secondly, the similarity of the distribution of vibration data at different time stages was analyzed by Smirnov non-parametric hypothesis test and significance level, and the data at different time stages were merged to obtain the service status data of rolling bearings. Finally, the service status of rolling bearings was constructed by parametric-non-parametric fusion analysis, and the service performance of rolling bearings was evaluated in real time.

[0282] The coefficient of variation in the parameter-nonparameter evaluation system comprehensively reflects the information characteristics of the intrinsic interval, median, and average value, and can determine the service condition of the rolling bearing. This method and evaluation system do not require knowing the distribution type and confidence level of the data, and are used to evaluate the dynamic performance of rolling bearings, providing an effective method for dynamic analysis of unknown distribution data.

[0283] At present, the performance evaluation method of rolling bearings generally analyzes the performance data as normal distribution or three-parameter Weibull distribution; some non-statistical methods such as fuzzy mathematics, membership function, and chaos theory methods require thresholds or confidence levels for good evaluation, which are difficult to meet the characteristics of unknown distribution and unknown confidence level of bearing performance, and fail to solve the data evaluation problem of unknown distribution and unknown confidence level. The method provided in this paper can solve the problem of unknown distribution and unknown confidence level of rolling bearing vibration data, and provide a way of thinking for the performance evaluation of unknown distribution time series.

[0284] It should be understood that parts not elaborated in detail in this specification belong to the prior art.

[0285] It should be understood that the above description of the implementation examples is relatively detailed and cannot be regarded as limiting the scope of patent protection of the present invention. Under the guidance of the present invention, ordinary technicians in this field can also make substitutions or modifications without departing from the scope of protection of the claims of the present invention, which all fall within the scope of protection of the present invention. The scope of protection requested for the present invention shall be based on the attached claims.

Claims

1. A method for evaluating the dynamic performance of a desulfurization pump main shaft bearing based on statistical robust theory, characterized by: The evaluation method constructs a desulfurization pump main shaft bearing performance evaluation model based on the three coefficients of the confidence level, median, and median variance of vibration data at different time stages to evaluate the operating status of the desulfurization pump main shaft bearing. The specific method is as follows: The first step is to measure the vibration of the main shaft bearing of the desulfurization pump through the sensor and collect data. The method is used to divide different time stages. The data of each time stage constitutes the statistic X i , X i ={x i (1),x i (2),…,x i (k),…,x i (n)},i=1, ……, q; n=1, ……, N (1) In the formula, x i (k) is the kth vibration data of the i-th time stage; i is the time stage, there are q time stages; k is the kth moment, there are N moments in total; The second step is to calculate the absolute value of the vibration data, and obtain the robust data of the desulfurization pump main shaft bearing vibration according to the fusion method of median estimation and Huber M estimation; Step 3: According to the statistical robustness theory, the absolute difference between the median and the mean is the smallest and the most robust to obtain the confidence level ɑ i ; Step 4: Rolling bearing performance range (a i ,b i ): a i 、b i are the critical values ​​of the statistical interval, where a i Slightly smaller than y min1 ; b i Slightly larger than y min2 , thus obtaining the rolling bearing performance interval (a i ,b i );y i (1) is the minimum value of the order statistics of the initial series and time series of bearing performance, y i (2N) is the maximum value of the order statistics of the initial series and time series of bearing performance; According to the number N of rolling bearing performance data column data, determine the number K of groupable bearing performance data intervals; Rolling bearing performance data group spacing p i : In the formula, p i is the rolling bearing performance data group interval; i is the time series number, m is the number of time series; K is the number of groups that the rolling bearing performance data series can be grouped into; The rolling bearing performance range (a i ,b i ) is divided into K intervals: (a i ,a i +p i ],(a i +p i ,a i +2p i ]…,(a i +(K-1)p i ,b i ];i=1,2,…,m Among them, a i 、b i are the critical values ​​of statistical intervals; p i is the rolling bearing performance data group interval; i is the time series number, m is the number of time series; K is the number of groups that the rolling bearing performance data series can be grouped into; Step 5: Frequency of rolling bearing performance data: According to the formula (a i ,a i +p i ],(a i +p i ,a i +2p i ]…,(a i +(K-1)p i ,b i ]; i = 1, 2, ..., m, rolling bearing performance interval and initial time series X0 and time series X i , calculate the initial time series X0 and time series X in the rolling bearing performance range i The number of data n 0k and n ik , where k = 1, 2, ..., K; Step 6. Frequency of rolling bearing performance data: According to the frequency of rolling bearing performance data and the initial time series X0 and time series X i The number of data, calculation formula (a i ,a i +p i ],(a i +p i ,a i +2p i ]…,(a i +(K-1)p i ,b i ]; i = 1, 2, ..., m, the initial time series X0 and time series X of the rolling bearing performance interval i The frequency of the data is f 0k 、f ik : In the formula, f 0k is the frequency of the kth interval of the initial time series; k is the frequency number of the initial time series, K is the number of time series frequencies; N is the number of time series data; n 0k is the frequency of the kth interval of the initial time series; In the formula, f ik is the frequency of the kth interval of the i-th time series; i is the time series number, m is the number of time series; k is the time series frequency number, K is the number of time series frequencies; N is the number of time series data; n ik is the frequency of the kth interval of the i-th time series; Step 7. Cumulative frequency of rolling bearing performance data: The cumulative frequency of the initial time series data of rolling bearing performance is In the formula, F 0v is the vth cumulative frequency of the initial time series; v is the cumulative frequency sequence number of the initial time series, K is the number of cumulative frequencies of the initial time series; f 0k is the frequency of the kth interval of the initial time series; k is the frequency sequence number of the initial time series, and K is the cumulative number of frequencies of the initial time series; The performance cumulative frequency of the rolling bearing i-th time series data is In the formula, F iv is the vth cumulative frequency of the i-th time series; i is the time series number, m is the number of time series; v is the cumulative frequency number of the time series, K is the number of cumulative frequencies of the time series; f ik is the frequency of the kth interval of the time series; k is the frequency serial number of the initial time series, and K is the cumulative frequency number of the initial time series; Step 8. Rolling bearing performance statistics: Rolling bearing performance statistics D ik : D ik =|F 0k -F ik |;i=1,2,…,m;k=1,2,…,K Where D ik is the absolute difference between the kth frequency of the initial time series of the rolling bearing and the i-th time series; F 0k is the cumulative frequency of the kth interval of the initial time series; F ik is the cumulative frequency of the kth interval of the i-th time series; i is the time series number, m is the number of time series; k is the time series frequency number, K is the number of time series frequencies; The absolute difference D between the initial time series frequency of the rolling bearing and the i-th time series frequency ik As a statistic, D ik represents the similarity between the distribution of the initial time series of the rolling bearing and the i-th time series data; D ik The smaller it is, the more similar the data distribution of the initial time series is to that of the i-th time series. ik The larger the value is, the more dissimilar the distribution of the initial time series is to the i-th time series; the more similar the data distributions of the two are, the closer their performances are, that is, the same performances; otherwise, the performances of the two are different, which is used to analyze the retention of rolling bearing performance; Step 9. Critical value D of rolling bearing performance statistics 2N,α : The critical value of the statistic is D 2N,α , where N is the number of time series data; α is the significance level; this value can be found in the critical value table of Kolmogorov test; Step 10: Rolling bearing performance maintains reliability: In order to analyze the similarity between the initial time series and the i-th time series of the rolling bearing, find the absolute difference D between the frequency of the initial time series and the frequency of the i-th time series. ik The maximum value maxD ik , if maxD ik Less than the critical value of the statistic D 2N,α , indicating that the data distribution of the initial time series is similar to that of the i-th time series, that is, the performance of the two is consistent. Continue to compare the i=i+1th time series until the p=i=i+1th time series, maxD ik Greater than the critical value of the statistic D 2N,α , indicating that the data distribution of the initial time series is not similar to that of the i-th time series, that is, the performance is different, the rolling bearing performance has changed, and the intervals are merged into Q i , the performance of the rolling bearing is kept reliable; Step 11: Find the order statistic Q i The median M i , average value A i , median variance SM i , mean square error SA i , Where M min is the median M of different stages i The minimum value of the median M at different time stages is calculated i With the average value A i The absolute difference R i And the variation characteristic parameter C i : R i =|M i -A i | (21) In the formula, the absolute difference R i Indicates the variability of the vibration data of the main shaft bearing of the desulfurization pump. The larger the value, the more serious the variation. The variation characteristic parameter C i Indicates the degree of variation in the performance of the desulfurization pump main shaft bearing. The larger the value, the worse the performance.

2. The method for evaluating the dynamic performance of the main shaft bearing of a desulfurization pump based on statistical robust theory according to claim 1 is characterized by: When the median estimation in the second step is combined with the Huber M estimation method to obtain the robust data of the desulfurization pump main shaft bearing vibration, the details are as follows: First, arrange the order statistics Y in ascending order i , AND i ={and i (1),and i (2),…,and i (k),…,and i (n)},i=1, ……, q; n=1, ……, N (2) In the formula, y i (k) is the kth vibration data of the i-th time stage; i is the time stage, there are q time stages; k is the kth moment, there are N moments in total; According to statistics, find the sequence Y i median β i Where: β i ——Sequence y i median; y i ——The rolling bearing vibration sequence in which the absolute values ​​of the rolling bearing vibration data at the i-th time stage are arranged in ascending order; Where: β i ——Sequence Y i median; y i ——The rolling bearing vibration sequence in which the absolute values ​​of the rolling bearing vibration data at the i-th time stage are arranged in ascending order; Data left sequence, right sequence Assume y i (b) and y i (e) is the data sequence y i The b and e experimental data values ​​in the data are b and e, respectively, where b and e are two values ​​in data 1, 2, ..., N, and y i (b)≤β i , β i ≤y i (e); When the data column y i Chinese: When y i (n)≤y i (b) y i (n)=y i (b); i=1,2,…,m; n=1,2,…,N (5) When i (b)≤y i (n) time: y i (n)=y i (e); i=1,2,…,m; n=1,2,…,N (6) Where: y i (n)——the nth data of the absolute value of the rolling bearing vibration data in the i-th time period arranged in ascending order; Get the left data sequence: y i (b),…,β i ,y i The number of data in (b) is n1; Get the right data sequence: β i ,…,y i (e), y i The number of data in (e) is n2; According to the left and right data sequences, the improved data sequence Z can be obtained i : Z i {z i (n)}={y i (b),…,β i ,…,y i (e)}? i=1,2,…,m; n=1,2,…,N (7) Where: z i (n)——the nth data of the rolling bearing at the i-th time stage; y i (b) y i (e)——Data sequence Y i The two experimental data values ​​b and e in ; b, e——two values ​​in data 1, 2, …, N; β i ——Sequence Y i median; Calculated Improved Data Column Z i The average value η i (n1,n2) Where: η i (n1,n2)——Robust data column Z i The average value of i——time phase number; m——number of time stages; n1——y i (b) The number of data; n2——y i (e) the number of data; n——data sequence number, there are N in total; i——time phase number; m——number of time stages; Calculate the median beta i and improved data column Z i The average value η i The absolute difference D of (n1,n2) i D i =|β i -or i (n1,n2); i=1,2,…,m (9) Where: D i ——Median β i and improved data column Z i The average value η i The absolute difference of (n1,n2); β i — Median; η i (n1,n2) – average value in the robust data column; Calculation of n1, n2: In this paper, according to the robust characteristics of the median in modern statistics, n1 and n2 are equal, that is, n1 = n2. Where: n1——y i (b) The number of data; n2——y i (e) the number of data; N——number of data; N——odd number, use formula (13); N——even number, use formula (14); Calculate the critical value K of the distribution function i1 , K i2 K i1 , K i2 is the critical value of the vibration data in the i-th time stage. Formula (7-9) obtains a series of difference values ​​D according to different n1 and n2. i , where D imin D i The minimum value of y i (b) and y i (e) correspond to K i1 and K i2 ,Right now K i1 =y i (b); i=1,2,…,m (13) K i2 =y i (e); i=1,2,…,m (14) Where: K i1 ——the minimum value of the distribution function; K i2 ——the maximum value of the distribution function; y i (b) y i (e)——Data sequence Y i The two experimental data values ​​b and e in ; b, e——two values ​​in data 1, 2, …, N; According to K i1 , K i2 The eigenvalue interval of the experimental data in this time period is obtained [K i1 ,K i2 ], K i1 is the minimum critical value of the eigenvalue interval, K i2 is the maximum critical value of the distribution function; According to the formula [K min1 ,K min2 ] and replace the data in the eigenvalue interval with the critical value of the eigenvalue interval, and the robust distribution function P can be obtained. i (n):

Citation Information

Cited By

  • Self-balancing multistage pump axial force stabilizing device and method based on discharge control

    CN120720237A