A method for evaluating the health state of a wind turbine bearing

Abstract

The application discloses a health state evaluation method for a fan bearing of a wind turbine generator, which utilizes evidence reasoning rules to make a fusion decision on multiple characteristic indexes, avoids the limitation of single index evaluation, and thus improves the accuracy of health state evaluation; further, in the evaluation process, the influence of external disturbances such as heavy load and high temperature on the fan bearing is considered, and thus the accuracy of health state evaluation is further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind power equipment state evaluation, and particularly relates to a health state evaluation method for a wind turbine bearing of a wind turbine. BACKGROUND

[0002] The wind turbine bearing is one of the core components of the wind turbine, and mainly functions to support a rotating shaft or other moving body, guide rotation or movement, and bear the load transmitted by the shaft and the parts on the shaft. Therefore, it is particularly important to evaluate the health state of the wind turbine bearing.

[0003] The existing health state evaluation methods for the wind turbine bearing include:

[0004] Traditional method:

[0005] The health state is mainly evaluated by observing the wear, cracks and surface quality of the bearing, and the result depends on the subjective judgment of the detection personnel, and the wind turbine needs to be shut down for maintenance.

[0006] Vibration analysis method:

[0007] The health state is evaluated by using the vibration signal of the bearing, and the wear, cracks and other problems can be highly sensitive to online monitoring. The common vibration analysis includes time domain analysis, frequency domain analysis and wavelet analysis.

[0008] The above two methods do not consider the influence of external disturbances such as heavy load and high temperature on the wind turbine bearing, and therefore the result of the health state evaluation is not accurate. SUMMARY

[0009] The present application aims to overcome the shortcomings of the prior art, and provides a health state evaluation method for a wind turbine bearing with high evaluation accuracy.

[0010] To achieve the above-mentioned purpose, the technical solution provided by the present application is:

[0011] A health state evaluation method for a wind turbine bearing, comprising:

[0012] S1, extracting a characteristic index reflecting the working state of the wind turbine bearing from the vibration signal of the wind turbine bearing;

[0013] S2, establishing an evaluation index system based on the extracted characteristic index;

[0014] S3, calculating the reliability of each characteristic index and the weight in the evaluation index system;

[0015] S4, standardizing the characteristic index data;

[0016] S5, fusing the characteristic index information and the evidence parameter by using the evidence reasoning rule to construct a health state evaluation model;

[0017] S6, using the health state evaluation model to perform bearing health state evaluation according to the standardized characteristic index data and the evidence reasoning rule to obtain a health state evaluation result;

[0018] S7, performing disturbance analysis to confirm whether the fan bearing is affected by internal and external disturbance factors; if yes, improving the health state evaluation model obtained in step S5 to obtain a disturbance-health state evaluation model and entering step S8, otherwise, the health state evaluation result obtained in step S6 is the final health state evaluation result;

[0019] S8, using the disturbance-health state evaluation model to perform bearing health state evaluation according to the standardized characteristic index data and the evidence reasoning rule to obtain a disturbance-health state evaluation result;

[0020] S9, calculating a disturbance coefficient based on the health state evaluation result obtained in step S6 and the disturbance-health state evaluation result;

[0021] S10, confirming whether the disturbance-health state evaluation model needs to be optimized based on the disturbance coefficient; if yes, returning to step S8 after optimizing the disturbance-health state evaluation model, otherwise, taking the disturbance-health state evaluation result obtained in step S8 as the final health state evaluation result.

[0022] Further, the extracted characteristic indexes include a fan bearing wear degree corresponding value, a peak value of an impact force borne by a local fault point of the fan bearing, a peak factor, an average amplitude of a waveform index of a vibration signal of the fan bearing, a kurtosis, a centroid frequency and a root mean square frequency.

[0023] Further, a calculation formula of the fan bearing wear degree corresponding value RMS is as follows:

[0024]

[0025] y(i) is a discrete-time vibration signal obtained by sampling, i=1, 2, 3,..., n s , n s is the number of vibration signals;

[0026] A calculation formula of the peak value Y p of the impact force borne by the local fault point of the fan bearing is as follows:

[0027] Y p =max(y(i)), i=1, 2,..., n s

[0028] Peak factor C f The calculation formula is as follows:

[0029]

[0030] Average amplitude S Y The calculation formula is as follows:

[0031]

[0032] The average value of the continuous time vibration signal, the calculation formula is as follows:

[0033]

[0034] The calculation formula of kurtosis Kurtosis{y(t)} is as follows:

[0035]

[0036] The calculation formula of the center frequency CF is as follows:

[0037]

[0038] f1 is the frequency value in the frequency domain, f2 is the power spectrum of the signal obtained by Fourier transform; the power spectrum P(f) is the square of the modulus of the Fourier transform of the signal:

[0039]

[0040] x(n) is a discrete signal sequence, n is a discrete signal sampling point, j is the imaginary unit of a complex number, and ω is the angular frequency;

[0041] The calculation formula of the root mean square frequency RMSF is as follows:

[0042]

[0043] Further, the calculation formula of the corresponding value RMS of the fan bearing wear degree is as follows:

[0044]

[0045] y(i) is a discrete time vibration signal obtained by sampling, i=1, 2, 3,..., n s , n s is the number of vibration signals; the peak value Y p reflecting the impact force on the local fault point of the fan bearing is calculated as follows:

[0046] Y p =max(y(i)), i=1, 2,..., ns

[0047] Peak factor C f The calculation formula is as follows:

[0048]

[0049] Average amplitude S Y The calculation formula is as follows:

[0050]

[0051] The average value of the continuous time vibration signal, the calculation formula is as follows:

[0052]

[0053] The calculation formula of kurtosis Kurtosis{y(t)} is as follows:

[0054]

[0055] The calculation formula of centroid frequency CF is as follows:

[0056]

[0057] f1 is the frequency value in the frequency domain, f2 is the power spectrum of the signal obtained by Fourier transform; The power spectrum P(f) is the square of the modulus of the Fourier transform of the signal:

[0058]

[0059] x(n) is a discrete signal sequence, n is a discrete signal sampling point, j is the imaginary unit of complex number, ω is the angular frequency;

[0060] The calculation formula of root mean square frequency RMSF is as follows:

[0061]

[0062] Further, the process of calculating the index weight includes:

[0063] There are indexes n, and the weight of the index is ω1, ω2,..., ω n ;

[0064] Calculate the standard deviation v i of the evaluation index:

[0065]

[0066] is the average value of all index data in time K; then, the variation coefficient is normalized, and the weight ω of the evaluation index is calculated i :

[0067]

[0068] Further, the standardization of the characteristic index data includes:

[0069] First, the index reference level and reference value are determined; then, based on the reference level and reference value, the index data is converted into the form of belief distribution by using the rule-based information conversion method:

[0070]

[0071] wherein h i,j is the reference value of index i at the jth level; h i,j+1 is the reference value of index i at the j+1th level; p i,j is the probability of the input value of index i falling in the jth level; p i,j+1 is the probability of the input value of index i falling in the j+1th level; p i,k is the probability of the input value of index i falling in the kth level; x i,j is the input data of index i at the jth level.

[0072] Further, the health state evaluation model is used to evaluate the bearing health state according to the standardized characteristic index data and the evidence reasoning rules, including:

[0073] Suppose a node collects data segments T, and each information segment has an indicator; the input index data is x i , e i is the evidence representation, i=1,...,I, I is the number of indexes; the identified framework is composed of N evaluation levels H n , n=1,...,N, N is the total number of evaluation levels, i.e. Θ={H1,...,H N}; after data standardization, the evidence representation is in the following belief distribution form:

[0074] e i ={(H n ,p n,i ),n=1,...,N;(Θ,p Θ,i )}

[0075] p n,i is the belief degree of the standardized input data of index i in the evaluation level H n ; Θ is the identified framework including all evaluation levels; p Θ,iIt is the degree of belief relative to the identification framework Θ;

[0076] The reliability of the evidence is r i It satisfies 0≤r i ≤1; the weight of evidence is ω i After normalization, 0 ≤ ω can be satisfied. i ≤1; with reliable evidence e i The weighted belief distribution is as follows:

[0077]

[0078] P(Θ) is the power set, For the evaluation level H n The mixed probability quality of index i satisfies:

[0079]

[0080]

[0081] c rw,i =1 / (1+ω) i -r i ) represents the regularization coefficient; m n,i For the evaluation level H n The basic probability quality of index i. It is an empty set, satisfying m n,i =ω i p n,i ;

[0082] Each indicator describes the characteristics of the wind turbine bearing vibration signal from different perspectives; the data of each indicator are standardized; for any two pieces of evidence e i e j Evaluate their joint belief support p n,e(2) for:

[0083]

[0084] For e i and e j After combination, it is assigned to the evaluation level H n The non-standardized combinatorial probability quality; For e i and e j The non-standardized combination probability mass e assigned to evaluation level D after combination i e j Both A and B are rating levels H. n The non-empty subset;

[0085] Comprehensive belief p n,e(I)is determined by the formula of I evidences:

[0086]

[0087] wherein k = 3, 4,..., I, m n,e(k-1) is the normalized combination probability mass assigned to the class H A,e(k-1) and the class A after the combination of the initial indicators k-1; n is the unnormalized combination probability mass assigned to the power set after the fusion of the first indicator k, m p(Θ),e(k-1) is the normalized probability mass assigned to the power set after the fusion of the first indicator k-1; are the unnormalized combination probability masses of the first k indicators after the combination assigned to the evaluation classes H n and D, respectively; n,e(k) is the belief degree of the first indicator k after the fusion to the evaluation class H n , and satisfies m n,e(1) = m n,1 , m p(Θ),e(1) = m p(Θ),1 ; through iteration, the comprehensive evaluation result is obtained:

[0088] e(I) = {(H n , p n,e(I) ), n = 1,..., N, (Θ, p Θ,e(I) )}

[0089] Let the utility of the evaluation class H n be u(H n ), and the expected utility of the evaluation result is obtained:

[0090]

[0091] u(e(I)) is the expected utility of the evaluation, which is used to evaluate the health state of the motor bearing, that is, the PM(x i ) corresponding to the obtained is obtained; PM(x i ) is the health state evaluation result under the input indicator data x i , and u(e(I)) is the health state evaluation result of the comprehensive utility considering all input parameters and evidences, which are equivalent in results.

[0092] Further, when the bearing health state is evaluated according to the standardized feature indicator data and the evidence reasoning rules using the perturbation-health state evaluation model, the identification framework of the evaluation model is Θ = {H1,..., H N}, and the evidences are e i , i = 1,..., I, and the confidence of the corresponding area of each evidence is {r'1, r'2,..., r'​I} and weights {ω'1, ω'2,..., ω' I} and the reference value is h i,j , i = 1, 2,..., I, j = 1, 2,..., J, the index evidence e i under the disturbance condition is calculated as follows:

[0093]

[0094] k≠l,l+1,h I and h1 correspond to the maximum reference value and the minimum reference value respectively, x i is the input index data, h l+1 ≤x i ≤h l ; σ i is the disturbance intensity, Δx i is the disturbance variable.

[0095] The rule-based information transformation method is used to convert it into the belief distribution form, the evidence reasoning rule is used to fuse all indexes, and the disturbance- health state evaluation result of the bearing is obtained, that is, the disturbance- health state evaluation result PM(x i under the input index data x i +σ i Δx i ) is obtained.

[0096] Further, the disturbance coefficient S i is obtained based on the health state evaluation result and the disturbance- health state evaluation result obtained in step S6, and the calculation formula is as follows:

[0097]

[0098] PM(x i +σ i Δx i ) is the disturbance- health state evaluation result, PM(x i ) is the health state evaluation result; x i is the input index data; σ i is the disturbance intensity, Δx i is the disturbance variable.

[0099] Further, if the disturbance coefficient S i ≤|ε|, ε is the maximum error of the disturbance coefficient, then the disturbance- health state evaluation model does not need to be optimized, otherwise the disturbance- health state evaluation model needs to be optimized.

[0100] Compared with the prior art, the technical scheme has the following principles and advantages:

[0101] This technical solution utilizes evidence-based reasoning rules to fuse multiple feature indicators for decision-making, avoiding the limitations of single-indicator assessment and thus improving the accuracy of health status assessment. Furthermore, the assessment process considers the susceptibility of wind turbine bearings to external disturbances such as heavy loads and high temperatures, thereby further improving the accuracy of health status assessment. Attached Figure Description

[0102] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the services required in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0103] Figure 1 This is a flowchart illustrating the principle of a method for assessing the health status of wind turbine bearings according to the present invention. Detailed Implementation

[0104] The present invention will be further described below with reference to specific embodiments:

[0105] like Figure 1 As shown in this embodiment, a method for assessing the health status of wind turbine bearings includes:

[0106] S1. Extract characteristic indicators reflecting the working status from the vibration signal of the wind turbine bearing;

[0107] The extracted feature indicators include the wear value of the wind turbine bearing, the peak value and peak factor reflecting the impact force on the local fault point of the wind turbine bearing, and the average amplitude, kurtosis, centroid frequency and root mean square frequency reflecting the waveform index of the wind turbine bearing vibration signal.

[0108] The formula for calculating the RMS wear value of wind turbine bearings is as follows:

[0109]

[0110] y(i) is a discrete-time vibration signal obtained through sampling, i = 1, 2, 3, ..., n s n s It is the number of vibration signals;

[0111] The peak value Y reflects the impact force experienced at a local fault point in the wind turbine bearing. p The calculation formula is as follows:

[0112] Y p =max(y(i)), i = 1, 2, ..., n s

[0113] Peak factor C f The calculation formula is as follows:

[0114]

[0115] Average amplitude S Y The calculation formula is as follows:

[0116]

[0117] The average value of the continuous time vibration signal, the calculation formula is as follows:

[0118]

[0119] The calculation formula of kurtosis Kurtosis{y(t)} is as follows:

[0120]

[0121] The calculation formula of centroid frequency CF is as follows:

[0122]

[0123] f1 is the frequency value in the frequency domain, f2 is the power spectrum of the signal obtained by Fourier transform; the power spectrum P(f) is the square of the modulus of the Fourier transform of the signal:

[0124]

[0125] x(n) is a discrete signal sequence, n is a discrete signal sampling point, j is the imaginary unit of a complex number, and ω is the angular frequency;

[0126] The calculation formula of root mean square frequency RMSF is as follows:

[0127]

[0128] S2, an evaluation index system is established based on the extracted feature indexes;

[0129] S3, the reliability of each feature index and the weight in the evaluation index system are calculated;

[0130] The process of calculating the reliability of the feature index includes:

[0131] Suppose there are indexes n, and the reliability of the index is r1, r2,..., r n Based on the distance-based method:

[0132]

[0133] x i(k) the monitoring data X at the indicator time k i , is the average value of all indicator data within the time K segment, is the distance between x i (k) and ; the average distance of all test data within the time K segment is as follows:

[0134]

[0135] The reliability of the indicator is defined as follows:

[0136]

[0137] is the maximum value of , characterizes the fluctuation of each indicator.

[0138] The process of calculating the weight of the indicator includes:

[0139] There are indicators n, and the weight of the indicator is ω1, ω2,..., ω n ;

[0140] The standard deviation v i of the evaluation indicator is calculated:

[0141]

[0142] is the average value of all indicator data within the time K segment; then, the coefficient of variation is normalized, and the weight ω i of the evaluation indicator is calculated:

[0143]

[0144] S4, standardize the characteristic indicator data;

[0145] Standardizing the characteristic indicator data includes:

[0146] First, determine the indicator reference level and reference value; then, based on the reference level and reference value, use a rule-based information conversion method to convert the indicator data into the form of belief distribution:

[0147]

[0148] where h i,j is the reference value of indicator i at the jth level; h i,j+1 is the reference value of indicator i at the j+1th level; p i,j is the probability that the input value of indicator i falls within level j; p i,j+1 is the probability of the input value of indicator i falling in the grade j+1; p i,k is the probability of the input value of indicator i falling in the grade k; x i,j is the input data of indicator i in the jth grade.

[0149] S5, fusing the characteristic indicator information and the evidence parameter by using the evidence reasoning rule to construct a health state evaluation model z(t) = Γ[x(t), ω, r];

[0150] where Γ(·) is a nonlinear function corresponding to the evidence reasoning rule, z(t) is the health state evaluation result without disturbance, r is the indicator reliability, and ω is the indicator weight.

[0151] S6, using the health state evaluation model to perform bearing health state evaluation according to the standardized characteristic indicator data and the evidence reasoning rule to obtain a health state evaluation result;

[0152] The process of this step includes:

[0153] includes:

[0154] Suppose that a node collects data pieces T, and each information piece has an indicator; the input indicator data is x i , e i is an evidence representation, i = 1,..., I, I is the number of indicators; the identified framework is composed of N evaluation levels H n , n = 1,..., N, N is the total number of evaluation grades, that is, Θ = {H1,..., H N}; after data standardization, the evidence representation is in the following belief distribution form:

[0155] e i = {(H n , p n,i ), n = 1,..., N; (Θ, p Θ,i )}

[0156] p n,i is the belief degree of the standardized input data of indicator i in the evaluation level H n ; Θ is an identified framework including all evaluation levels; p Θ,i is the belief degree of the indicator relative to the identified framework Θ;

[0157] The reliability of the evidence is r i , which satisfies 0 ≤ r i ≤ 1; the evidence weight is ω i , which satisfies 0 ≤ ω i ≤ 1 after normalization; the weighted belief distribution of the evidence e i with reliability is:

[0158]

[0159] P(Θ) where P is the power set, is the evaluation grade H n The mixed probability mass of the lower index i satisfies:

[0160]

[0161] c rw,i =1 / (1+ω i -r i ) is the regularization coefficient; m n,i is the evaluation grade H n The basic probability mass of the lower index i, is the empty set, satisfying m n,i =ω i p n,i ;

[0162] Each indicator describes the characteristics of the fan bearing vibration signal from different angles; the data of each index is standardized; for any two evidences e i , e j , their joint belief support p n,e(2) is:

[0163]

[0164]

[0165] is the non-standardized combination probability mass assigned to the evaluation grade H i after e j and e n are combined; is the non-standardized combination probability mass assigned to the evaluation grade D after e i and e j are combined e i , e j ; A and B are both non-empty subsets of the evaluation grade H n ;

[0166] The comprehensive belief p n,e(I) is determined by the formula of I evidences:

[0167]

[0168] where k=3,4,...,I, m n,e(k-1) , m A,e(k-1) is the normalized combination probability mass assigned to the grade H n and the grade A after the initial index k-1 is combined; m is the non-normalized probability mass assigned to the power set after fusion of the first indicator k p(Θ),e(k-1) m is the normalized probability mass assigned to the power set after fusion of the first indicator k-1 are the non-normalized combined probability masses after combination of the first k indicators assigned to the evaluation grades H n and D respectively; p n,e(k) is the belief degree of the first indicator k after fusion to the evaluation grade H n , and satisfies m n,e(1) = m n,1 , m p(Θ),e(1) = m p(Θ),1 ; through iteration, the comprehensive evaluation result is obtained:

[0169] e(I) = {(H n , p n,e(I) ), n = 1,..., N, (Θ, p Θ,e(I) )}

[0170] Let the utility of the evaluation grade H n be u(H n ), and the expected utility of the evaluation result is obtained:

[0171]

[0172] u(e(I)) is the expected utility of the evaluation, which is used to evaluate the health state of the motor bearing, that is, the PM(x i ) corresponding to the obtained is obtained; PM(x i ) is the health state evaluation result under the input indicator data x i , and u(e(I)) is the comprehensive utility health state evaluation result after comprehensively considering all input parameters and evidence, and the two are equivalent in results.

[0173] S7, perform disturbance analysis to confirm whether the fan bearing is affected by internal and external disturbance factors; if so, improve the health state evaluation model obtained in step S5 to obtain a disturbance-health state evaluation model, and proceed to step S8, otherwise the health state evaluation result obtained in step S6 is the final health state evaluation result;

[0174] S8, use the disturbance-health state evaluation model to perform bearing health state evaluation according to the standardized characteristic indicator data and evidence reasoning rules to obtain a disturbance-health state evaluation result;

[0175] When using the disturbance-health state evaluation model to perform bearing health state evaluation according to the standardized characteristic indicator data and evidence reasoning rules, the identification framework of the evaluation model is Θ = {H1,..., H N}, and the evidence is e iLet i = 1, ..., I, and the confidence level of each evidence region be {r'1, r'2, ..., r'}. I} and weights {ω'1,ω'2,...,ω' I}, the reference value is h i,j Let i = 1, 2, ..., I, j = 1, 2, ..., J, and the index evidence e under perturbation conditions. i The distribution of beliefs is calculated as follows:

[0176]

[0177] k≠l,l+1,h I h1 and h2 correspond to the maximum and minimum reference values, respectively, x i To input indicator data, h l+1 ≤x i ≤h l ;σ i Let Δx be the disturbance intensity. i For disturbance variables;

[0178] A rule-based information transformation method is used to convert the data into a belief distribution form. Evidence-based reasoning rules are then used to fuse all indicators to obtain the bearing's disturbance-health status assessment result, which is also the corresponding input indicator data x. i The perturbation-health status assessment results PM(x) i +σ i Δx i ).

[0179] S9. Based on the health status assessment results and disturbance-health status assessment results obtained in step S6, calculate the disturbance coefficient S. i The calculation formula is as follows:

[0180]

[0181] PM(x i +σ i Δx i ) represents the disturbance-health status assessment result, PM(x) i ) represents the health status assessment result; x i Input indicator data; σ i Let Δx be the disturbance intensity. i This is a disturbance variable.

[0182] S10. Based on the disturbance coefficient, determine whether the disturbance-health status assessment model needs to be optimized. If so, optimize the disturbance-health status assessment model and return to step S8. Otherwise, take the disturbance-health status assessment result obtained in step S8 as the final health status assessment result.

[0183] In this step, if the disturbance coefficient S i ≤ |ε|, ε is the maximum error of the disturbance coefficient, then the disturbance- health state evaluation model does not need to be optimized, otherwise the disturbance- health state evaluation model needs to be optimized.

[0184] The above-described embodiments are only preferred embodiments of the present application, and are not intended to limit the scope of the present application. Any changes made in the shape and principle of the present application should be covered within the scope of protection of the present application.

Claims

1. A method of assessing the health of a wind turbine bearing, the method comprising: The method comprises the following steps: S1, extracting a characteristic index reflecting the working state of the fan bearing from the vibration signal of the fan bearing; S2, establishing an evaluation index system based on the extracted characteristic index; S3, calculating the reliability of each characteristic index and the weight in the evaluation index system; S4, standardizing the characteristic index data; S5, fusing the characteristic index information and the evidence parameter by using the evidence reasoning rule to construct a health state evaluation model; S6, using the health state evaluation model to perform bearing health state evaluation according to the standardized characteristic index data and the evidence reasoning rule, and obtaining a health state evaluation result; S7, performing disturbance analysis to confirm whether the fan bearing is affected by internal and external disturbance factors; if yes, improving the health state evaluation model constructed in step S5 to obtain a disturbance-health state evaluation model, and entering step S8; otherwise, the health state evaluation result obtained in step S6 is the final health state evaluation result; S8, using the disturbance-health state evaluation model to perform bearing health state evaluation according to the standardized characteristic index data and the evidence reasoning rule, and obtaining a disturbance-health state evaluation result; S9, calculating a disturbance coefficient based on the health state evaluation result obtained in step S6 and the disturbance-health state evaluation result; S10, confirming whether the disturbance-health state evaluation model needs to be optimized based on the disturbance coefficient; if yes, optimizing the disturbance-health state evaluation model and returning to step S8; otherwise, taking the disturbance-health state evaluation result obtained in step S8 as the final health state evaluation result. The extracted characteristic indexes include a fan bearing wear degree corresponding value, a peak value reflecting the impact force on a local fault point of the fan bearing, a peak factor, an average amplitude of a waveform index reflecting the vibration signal of the fan bearing, a kurtosis, a center frequency, and a root mean square frequency.

2. A method of assessing the health of a wind turbine bearing according to claim 1, wherein, The calculation formula of the fan bearing wear degree corresponding value RMS is as follows: y(i) is a discrete-time vibration signal obtained by sampling, i = 1, 2, 3,..., n s , n s is the number of vibration signals; The peak value Y of the impact force on the local fault point of the bearing of the fan p The calculation formula is as follows: Y p = max(y(i)), i = 1, 2,..., n s Peak factor C f The formula for calculating C is as follows: Average amplitude S Y The formula for calculating S is as follows: is the average value of the continuous-time vibration signal, and its calculation formula is as follows: The calculation formula of the kurtosis Kurtosis{y(t)} is as follows: The calculation formula of the center frequency CF is as follows: f1 is a frequency value in the frequency domain, f2 is a power spectrum of the signal obtained by Fourier transform; the power spectrum P(f) is the square of the modulus of the Fourier transform of the signal: x(n) is a discrete signal sequence, n is a discrete signal sampling point, j is the imaginary unit of a complex number, and ω is an angular frequency; The calculation formula of the root mean square frequency RMSF is as follows:

3. A method of assessing the health of a wind turbine bearing according to claim 1, wherein, The process of calculating the reliability of the characteristic index includes: There are indexes n, and the reliability of the index is r1, r2,..., r n The distance-based method obtains: x i (k) is the monitoring data X at time k for the indicator i , is the average of all indicator data over the time K segment, is the distance between x i (k) and ; the average distance of all test data over the time K segment is as follows: The reliability of the index is defined as follows: is the maximum value of is the maximum value of characterizes the fluctuation of each indicator.

4. A method of assessing the health of a wind turbine bearing according to claim 1, wherein, The process of calculating the index weight includes: There is an index n, and the weight of the index is ω1, ω2,..., ω n ; Standard deviation v of the evaluation index i : is the average value of all index data in the time K section; then, the coefficient of variation is normalized, and the weight ω of the evaluation index is calculated i :

5. A method of assessing the health of a wind turbine bearing according to claim 1, wherein, The standardization of the characteristic index data includes: First, the reference level and reference value of the index are determined; then, based on the reference level and reference value, a rule-based information conversion method is used to convert the index data into the form of belief distribution: where h i,j is the reference value for indicator i at the jth level; h i,j+1 is the reference value for indicator i at the j+1th level; p i,j is the probability that the input value for indicator i falls in level j; p i,j+1 is the probability that the input value for indicator i falls in level j+1; p i,k is the probability that the input value for indicator i falls in level k; x i,j is the input data for indicator i at the jth level.

6. A method of assessing the health of a wind turbine bearing according to claim 1, wherein, Using the health state evaluation model to perform bearing health state evaluation according to the standardized characteristic index data and the evidence reasoning rule includes: Let a node collect data pieces T, and each information piece has an indicator; the input indicator data is x i , e i As evidence representation, i = 1,..., I, I is the number of indicators; the identified framework is composed of N evaluation levels H n , n = 1,..., N, N is the total number of evaluation levels, that is, Θ = {H1,..., H N}; after data standardization, the evidence representation is in the following belief distribution form: e i = {(H n ,p n,i ), n = 1,..., N; (Θ, p Θ,i )} p n,i is the belief degree of the standardized input data for indicator i at evaluation level H n ; Θ is the identification framework comprising all evaluation levels; p Θ,i is the belief degree of the indicator with respect to the identification framework Θ. The reliability of the evidence is r i , which satisfies 0≤r i ≤1; the weight of the evidence is ω i , which satisfies 0≤ω i ≤1 after normalization; and the weighted belief distribution of the evidence e i with the reliability is: P(Θ) where P is the power set, is the evaluation grade H n The mixing probability mass of the lower index i satisfies: c rw,i = 1 / (1 + ω i -r i ) is a regularization coefficient; m n,i is the evaluation rating H n the basic probability mass of the lower index i, is empty, satisfying m n,i = ω i p n,i ; Each indicator describes the characteristics of the fan bearing vibration signal from different angles; the data of each indicator is standardized; for any two evidences e i , e j , the joint belief support degree p n,e(2) of them is: for e i and e j non-normalized combined probability mass assigned to evaluation grade H n after combination; for e i and e j non-normalized combined probability mass e assigned to evaluation grade D after combination i , e j ; A, B are both non-empty subsets of evaluation grades H n ; Synthetic belief p n,e(I) Determined from the formula of I evidence: where k = 3, 4,..., I, m n,e(k-1) , m A,e(k-1) is the normalized combination probability mass assigned to the rating H n and the rating A after the combination of the initial indicators k-1; is the unnormalized combination probability mass assigned to the power set after the fusion of the first indicator k, m p(Θ),e(k-1) is the normalized combination probability mass assigned to the power set after the fusion of the first indicator k-1; are the unnormalized combination probability masses assigned to the evaluation ratings H n and D after the combination of the first k indicators, respectively; p n,e(k) is the belief degree of the first indicator k to the evaluation rating H n after the fusion, and satisfies m n,e(1) = m n,1 , m p(Θ),e(1) = m p(Θ),1 ; through iteration, the comprehensive evaluation result is obtained: e(I) = {(H n ,p n,e(I) ), n = 1,..., N, (Θ, p Θ,e(I) )} Let the utility of rating level H n be u(H n ), and the expected utility of obtaining the rating result be: u(e(I)) is the expected utility of evaluation, used to evaluate the health state of the motor bearing, i.e. corresponds to PM(x i ); PM(x i ) is the health state assessment result under the input indicator data x i , u(e(I)) is the comprehensive utility health state assessment result after considering all input parameters and evidence, the two are equivalent in results.

7. A method of assessing the health of a wind turbine bearing according to claim 6, wherein, Using the disturbance-healthy state evaluation model, the bearing healthy state is evaluated according to the standardized characteristic index data and the evidence reasoning rules, the identification framework of the evaluation model is Θ = {H1,...,H N}, the evidence is e i , i = 1,...,I, the confidence of each evidence area is {r'1,r'2,...,r' I}, the weight is {ω'1,ω'2,...,ω' I}, the reference value is h i,j , i = 1,2,...,I, j = 1,2,...,J, and the belief distribution of the index evidence e i under the disturbance condition is calculated as follows: k≠l, l+1, h I and hi correspond to the maximum and minimum reference values, respectively, x i is the input indicator data, h l+1 ≤x i ≤h l ; σ i is the perturbation strength, Δx i is the perturbation variable; The rule-based information transformation method is used to convert the input index data x into the belief distribution form, and the evidence reasoning rule is used to fuse all indexes to obtain the disturbance-health state evaluation result PM(x i of the bearing under the input index data x i +σ i Δx i .

8. A method of assessing the health of a wind turbine bearing according to claim 7, wherein, Based on the health state evaluation result and the disturbance-health state evaluation result obtained in step S6, the disturbance coefficient S is calculated i The calculation formula is as follows: PM(x i +σ i Δx i ) is the disturbance-healthy state evaluation result, PM(x i ) is the healthy state evaluation result; x i is the input index data; σ i is the disturbance intensity, Δx i is the disturbance variable.

9. A method of assessing the health of a wind turbine bearing according to claim 8, wherein, If the perturbation coefficient S i ≤ |ε|, ε is the maximum error of the perturbation coefficient, then the perturbation- health state evaluation model does not need to be optimized, otherwise the perturbation- health state evaluation model needs to be optimized.

Images