A method for predicting aging of wind turbine gearbox

Abstract

The present invention discloses a method for predicting the aging of a wind turbine gearbox, the method comprising acquiring pre-collected gearbox bearing temperature and ambient temperature data; performing mixed modeling on the gearbox bearing temperature and ambient temperature data through a PCA algorithm to obtain a one-dimensional temperature data set; inputting the one-dimensional temperature data set into a pre-constructed gearbox performance degradation model to calculate the remaining life of the gearbox; and correcting the fault threshold through real-time feedback of pre-acquired abnormal operating condition fault samples to obtain a final remaining life. The present invention uses a PCA algorithm to perform mixed modeling on the bearing temperature and the ambient temperature, fully extracts the connection characteristics between the temperatures, and makes the prediction model more accurate. The abnormal operating condition sample data is used to correct the fault threshold in real time, and the remaining life probability density result is adjusted, thereby correcting the maximum value of the gearbox life limit, obtaining a more accurate gearbox remaining life, and finally predicting the aging trend of the gearbox.

Description

Technical Field

[0001] The invention relates to a method for predicting aging of a fan gearbox, and belongs to the technical field of fan gearboxes. Background Art

[0002] The general prediction method for fan gearbox aging only extracts data on bearing temperature and speed, and uses the performance degradation Wiener model algorithm to predict the remaining life of the fan gearbox. The existing prediction method for the remaining life of the fan gearbox can reflect the aging trend of the gearbox to a certain extent, but the method is too single and only considers the bearing temperature, without considering the impact of ambient temperature on the aging trend. At the same time, the existing method only predicts the remaining life of the gearbox under normal working conditions, without considering the impact of abnormal working conditions on aging, so there is a certain error in predicting the remaining life of the gearbox. Summary of the invention

[0003] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method for predicting the aging of a wind turbine gearbox. The PCA algorithm is used to perform mixed modeling of the bearing temperature and the ambient temperature, and the connection characteristics between the temperatures are fully extracted to make the prediction model more accurate.

[0004] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0005] In a first aspect, the present invention provides a method for predicting aging of a wind turbine gearbox, comprising:

[0006] Obtain the pre-collected gearbox bearing temperature and ambient temperature data;

[0007] The gearbox bearing temperature and the ambient temperature data are mixed and modeled using a PCA algorithm to obtain a one-dimensional temperature data set;

[0008] Inputting the one-dimensional temperature data set into a pre-built gearbox performance degradation model to calculate the remaining life of the gearbox;

[0009] The fault threshold is corrected through real-time feedback of abnormal operating condition fault samples acquired in advance to obtain the final remaining life.

[0010] Further, the obtaining of the pre-collected gearbox bearing temperature and ambient temperature data includes:

[0011] Gearbox bearing temperature time pairs (t0, C0), (t1, C1), (t2, C2), … (t n ,C n ), ambient temperature time pairs (t0,C'0), (t1,C'1), (t2,C'2),…(t n ,C' n) and speed time pairs (t0, R0), (t1, R1), (t2, R2), … (t n ,R n ), where time is t0 <t1<t2<…<t n Arranged in order, C n Indicates t n Bearing temperature at time, C' n Indicates t n Ambient temperature at the moment, R n Indicates t n The speed of the moment.

[0012] Furthermore, the gearbox bearing temperature and ambient temperature data are mixed and modeled by the PCA algorithm to obtain a one-dimensional temperature data set, including:

[0013] De-mean the gearbox bearing temperature and ambient temperature data;

[0014] Calculate the covariance matrix of gearbox bearing temperature and ambient temperature data;

[0015] Calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix;

[0016] The original features are projected onto the selected feature vector to obtain new one-dimensional features after dimensionality reduction.

[0017] Further, the de-averaging of the gearbox bearing temperature and ambient temperature data includes:

[0018] The bearing temperature is averaged and the calculation formula is as follows:

[0019]

[0020] In formula 1 is the average value of the bearing temperature, n is the logarithm of the bearing temperature over time, C i is the bearing temperature at the i-th moment;

[0021] The ambient temperature is averaged and the calculation formula is as follows:

[0022]

[0023] In formula 2 is the average ambient temperature, n is the logarithm of ambient temperature over time, C' i is the ambient temperature at the i-th moment.

[0024] Furthermore, the covariance matrix of the gearbox bearing temperature and ambient temperature data obtained by the calculation includes:

[0025] Find the covariance matrix B, the formula is as follows:

[0026]

[0027] in The solution formula is as follows:

[0028]

[0029] In formula 4 is the average value of the bearing temperature, n is the logarithm of the bearing temperature over time, C i is the bearing temperature at the i-th moment;

[0030] The other calculation methods in Formula 3 are similar, and the covariance matrix B under these n samples can be finally calculated.

[0031] Further, the calculation of the eigenvalues ​​and corresponding eigenvectors of the covariance matrix includes:

[0032] Find the eigenvalue λ of the covariance matrix B and the corresponding eigenvector u, Bu = λu, there will be n eigenvalues ​​λ, each λ i Corresponding to a feature vector u i , sort the eigenvalues ​​λ in descending order, select the largest first k, and extract their corresponding k eigenvectors to obtain a set of {(λ1,u1),(λ2,u2),...(λ k ,u k )}.

[0033] Furthermore, the projecting of the original features onto the selected feature vector to obtain a new one-dimensional feature after dimensionality reduction includes:

[0034] The original characteristics of the bearing temperature and ambient temperature samples are (C i ,C' i ) T , the new feature after projection is (y i ) T ,i ranges from 0,1,…n.

[0035] Furthermore, the one-dimensional temperature data set is input into a pre-built gearbox performance degradation model to calculate the remaining life of the gearbox, including:

[0036] The temperature y and speed after hybrid dimension reduction are quantified, and the relative temperature is expressed as F n =|y n | / R n ; Such quantified data may have certain errors, so the average of m consecutive data values ​​is processed. n At time t, the temperature trend The temperature trend xn The initial value of the relative temperature F is obtained. The gearbox performance degradation trend expression x is obtained through the Wiener process model. t =F+μt+σB(t), where x t is the performance degradation trend at time t, μ is the drift coefficient, σ is the diffusion coefficient, and B(t) is the standard Wiener process Brownian motion;

[0037] Due to the relative temperature increase Δx of the gearbox j =μΔt j +σΔB(Δt j ), and in the Wiener process model, ΔB(Δt j ) is normally distributed in N(0,Δt j ), so Δx j Normal distribution in N(μΔt j ,σ2Δt j ), so the maximum likelihood estimation function of the temperature sample is expressed as L(μ,σ)=f(Δx1)f(Δx2)...f(Δx n ), where f(Δx n ) indicates that at t n Mixed temperature trend x at the moment n The probability density function of L(μ,σ) is obtained by partial differentiation of the function L(μ,σ) to obtain the maximum likelihood estimates μ', σ'. The specific solution formula is as follows:

[0038]

[0039]

[0040] Substitute the above maximum likelihood estimates μ' and σ' into the probability density function of gearbox life, set the fault threshold η, and obtain the upper limit of gearbox life T MAX and the remaining life of the gearbox T RUL ; Assume that the gearbox fault threshold η=F max / R max , when the probability density f(t i ) is larger, the gearbox i The greater the probability density of the point reaching the fault value η, the greater the probability density will be; therefore, the maximum probability density f max (t), the corresponding time T MAX As the upper limit of gearbox life, the probability density function of gearbox life is as follows:

[0041]

[0042] Furthermore, the real-time feedback correction of the fault threshold through the abnormal operating condition fault samples acquired in advance to obtain the final remaining life includes:

[0043] The data under abnormal working conditions are collected from the fan SCADA system to form fault samples. The above-mentioned method of hybrid modeling of bearing temperature and ambient temperature is used to obtain the time queue of relative temperature and speed. The temperature value is divided by the speed to obtain a data set. The data set is sequentially compared with the fault threshold η=F max / R max Make difference calculation;

[0044] The calculated results are averaged to obtain the average value of the difference η' avg , so the corrected fault threshold η is:

[0045] Substituting the fault threshold into the above life probability density function, we can get the time T corresponding to the upper limit of life: MAX , so the remaining life of the gearbox is T RUL =T MAX -t, t indicates the current running time.

[0046] In a second aspect, the present invention provides a wind turbine gearbox aging prediction device, comprising:

[0047] An acquisition unit, used for acquiring the pre-collected gearbox bearing temperature and ambient temperature data;

[0048] A hybrid modeling unit, used for performing hybrid modeling on the gearbox bearing temperature and the ambient temperature data through a PCA algorithm to obtain a one-dimensional temperature data set;

[0049] A calculation unit, used for inputting the one-dimensional temperature data set into a pre-built gearbox performance degradation model to calculate the remaining life of the gearbox;

[0050] The correction unit is used to correct the fault threshold in real time through the abnormal operating condition fault samples acquired in advance, so as to obtain the final remaining life.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] The present invention provides a method for predicting the aging of a wind turbine gearbox. The PCA algorithm is used to perform mixed modeling of the bearing temperature and the ambient temperature, and the connection characteristics between the temperatures are fully extracted to make the prediction model more accurate. The maximum likelihood estimation function is used to solve the maximum estimation value, and it is back-substituted into the life density function. At the same time, a threshold is set. When the probability density is larger, the probability density of the gearbox reaching the fault value at a certain point is also larger. Therefore, the probability density is used to reflect the life of the gearbox. When the probability density is the largest, the time at this time is the fault limit. The abnormal working condition sample data is used to correct the fault threshold in real time, and the remaining life probability density result is adjusted, thereby correcting the maximum value of the gearbox life limit, obtaining a more accurate remaining life of the gearbox, and finally predicting the aging trend of the gearbox. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flow chart of a method for predicting aging of a wind turbine gearbox provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0055] Example 1

[0056] This embodiment introduces a method for predicting aging of a wind turbine gearbox, including:

[0057] Obtain the pre-collected gearbox bearing temperature and ambient temperature data;

[0058] The gearbox bearing temperature and the ambient temperature data are mixed and modeled using a PCA algorithm to obtain a one-dimensional temperature data set;

[0059] Inputting the one-dimensional temperature data set into a pre-built gearbox performance degradation model to calculate the remaining life of the gearbox;

[0060] The fault threshold is corrected through real-time feedback of abnormal operating condition fault samples acquired in advance to obtain the final remaining life.

[0061] The wind turbine gearbox aging prediction method provided in this embodiment specifically involves the following steps:

[0062] Step 1: Establish a hybrid model: (1) Collect the gearbox bearing temperature-time pairs (t0, C0), (t1, C1), (t2, C2), … (t n ,C n ), ambient temperature time pairs (t0,C'0), (t1,C'1), (t2,C'2),…(t n ,C' n ) and speed time pairs (t0, R0), (t1, R1), (t2, R2), … (t n ,R n ), where time is t0 <t1<t2<…<t n Arranged in order, C n Indicates t n Bearing temperature at time, C' n Indicates t n Ambient temperature at the moment, R n Indicates t n The speed of the moment.

[0063] (2) Since the gearbox bearing temperature is correlated with the ambient temperature, the PCA algorithm is used to map the two-dimensional data into a low-dimensional space to maximize the amount of information extracted. The operation steps are as follows:

[0064] 2.a) Center all features (remove the mean)

[0065] The bearing temperature is averaged and the calculation formula is as follows:

[0066]

[0067] In formula 1 is the average value of the bearing temperature, n is the logarithm of the bearing temperature over time, C i is the bearing temperature at the i-th moment.

[0068] The ambient temperature is averaged and the calculation formula is as follows:

[0069]

[0070] In formula 2 is the average ambient temperature, n is the logarithm of ambient temperature over time, C' i is the ambient temperature at the i-th moment.

[0071] 2.b) Find the covariance matrix B

[0072]

[0073] in The solution formula is as follows:

[0074]

[0075] In formula 4 is the average value of the bearing temperature, n is the logarithm of the bearing temperature over time, C i is the bearing temperature at the i-th moment.

[0076] The other calculation methods in Formula 3 are similar, and the covariance matrix B under these n samples can be finally calculated.

[0077] 2.c) Find the eigenvalues ​​and corresponding eigenvectors of the covariance matrix B

[0078] Using matrix knowledge, we can find the eigenvalue λ and the corresponding eigenvector u of the covariance matrix B, Bu = λu. There will be n eigenvalues ​​λ, and each λ i Corresponding to a feature vector u i , sort the eigenvalues ​​λ in descending order, select the largest first k, and take out their corresponding k eigenvectors, we will get a set of {(λ1,u1),(λ2,u2),...(λk ,u k )}.

[0079] 2.d) Project the original features onto the selected feature vector to obtain a new one-dimensional feature after dimensionality reduction

[0080] The original characteristics of the bearing temperature and ambient temperature samples are (C i ,C' i ) T , the new feature after projection is (y i ) T ,i ranges from 0,1,…n.

[0081] Step 2: Substitute the data into the performance degradation model and solve the probability density value:

[0082] (1) The temperature y and speed after hybrid dimensionality reduction are quantified, and the relative temperature is expressed as F n =|y n | / R n ; Such quantified data may have certain errors, so the average of m consecutive data values ​​is processed. n At time t, the temperature trend Through the temperature trend x n The initial value of the relative temperature F is obtained. The gearbox performance degradation trend expression x is obtained through the Wiener process model. t =F+μt+σB(t), where x t is the performance degradation trend at time t, μ is the drift coefficient, σ is the diffusion coefficient, and B(t) is the standard Wiener process Brownian motion.

[0083] (2) Due to the relative temperature increment Δx of the gearbox j =μΔt j +σΔB(Δt j ), and in the Wiener process model, ΔB(Δt j ) is normally distributed in N(0,Δt j ), so Δx j Normal distribution in N(μΔt j ,σ2Δt j ), so the maximum likelihood estimation function of the temperature sample is expressed as L(μ,σ)=f(Δx1)f(Δx2)...f(Δx n ), where f(Δx n ) indicates that at t n Mixed temperature trend x at the moment n The probability density function of L(μ,σ) is obtained by partial differentiation of the function L(μ,σ) to obtain the maximum likelihood estimates μ', σ'. The specific solution formula is as follows:

[0084]

[0085]

[0086] (3) Substitute the above maximum likelihood estimates μ' and σ' into the probability density function of gearbox life, set the fault threshold η, and obtain the upper limit of gearbox life T MAX and the remaining life of the gearbox T RUL Assume that the gearbox fault threshold η=F max / R max , when the probability density f(t i ) is larger, the gearbox i The greater the probability density of the point reaching the fault value η, the greater the probability density will be. max (t), the corresponding time T MAX As the upper limit of gearbox life, the probability density function of gearbox life is as follows:

[0087]

[0088] Step 3 uses the real-time feedback of abnormal operating condition fault samples to correct the fault threshold and obtain the final remaining life:

[0089] (1) Collect data under abnormal conditions from the fan SCADA system to form fault samples. Use step 1 to perform mixed modeling on the bearing temperature and ambient temperature to obtain a time queue of relative temperature and speed. Divide the temperature value by the speed to obtain a data set. The data set is sequentially compared with the fault threshold η = F max / R max Calculate the difference.

[0090] (2) The calculated results are averaged to obtain the average value η' of the difference avg , so the corrected fault threshold η is:

[0091] (3) Substitute the fault threshold into the life probability density function in step 2 to obtain the time T corresponding to the upper limit of life MAX , so the remaining life of the gearbox is T RUL =T MAX -t, t indicates the current running time.

[0092] This embodiment uses the PCA algorithm to perform mixed modeling of the bearing temperature and the ambient temperature, fully extracting the connection characteristics between the temperatures to make the prediction model more accurate. The sample maximum likelihood estimation function is used to solve the maximum still estimation value, and then back-substituted into the life density function. At the same time, a threshold is set. When the probability density is larger, the probability density of the gearbox reaching the fault value at a certain point is also larger. Therefore, the probability density is used to reflect the life of the gearbox. When the probability density is the largest, the time at this time is the fault limit. Fully consider the impact of the fault threshold on life prediction, use abnormal working condition sample data, correct the fault threshold in real time, adjust the remaining life probability density result, and correct the maximum value of the gearbox life limit to obtain a more accurate remaining life of the gearbox, and finally predict the aging trend of the gearbox.

[0093] Example 2

[0094] This embodiment provides a wind turbine gearbox aging prediction device, comprising:

[0095] An acquisition unit, used for acquiring the pre-collected gearbox bearing temperature and ambient temperature data;

[0096] A hybrid modeling unit, used for performing hybrid modeling on the gearbox bearing temperature and the ambient temperature data through a PCA algorithm to obtain a one-dimensional temperature data set;

[0097] A calculation unit, used for inputting the one-dimensional temperature data set into a pre-built gearbox performance degradation model to calculate the remaining life of the gearbox;

[0098] The correction unit is used to correct the fault threshold in real time through the abnormal operating condition fault samples acquired in advance, so as to obtain the final remaining life.

[0099] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting aging of a wind turbine gearbox, characterized in that: include: Obtain the pre-collected gearbox bearing temperature and ambient temperature data; The gearbox bearing temperature and the ambient temperature data are mixed and modeled using a PCA algorithm to obtain a one-dimensional temperature data set; Inputting the one-dimensional temperature data set into a pre-built gearbox performance degradation model to calculate the remaining life of the gearbox; The fault threshold is corrected by real-time feedback of abnormal operating condition fault samples acquired in advance to obtain the final remaining life; The one-dimensional temperature data set is input into a pre-built gearbox performance degradation model to calculate the remaining life of the gearbox, including: The temperature y and speed after hybrid dimension reduction are quantified, and the relative temperature is expressed as F n =|y n | / R n ; Such quantified data may have certain errors, so the average of m consecutive data values ​​is processed. n At time t, the temperature trend Through the temperature trend x n Get the initial value of relative temperature F; get the gearbox performance degradation trend expression x through the Wiener process model t =F+μt+σB(t), where x t is the performance degradation trend at time t, μ is the drift coefficient, σ is the diffusion coefficient, and B(t) is the standard Wiener process Brownian motion; Due to the relative temperature increase Δx of the gearbox j =μΔt j +σΔB(Δt j ), and in the Wiener process model, ΔB(Δt j ) is normally distributed in N(0,Δt j ), so Δx j Normal distribution in N(μΔt j ,σ2Δt j ), so the maximum likelihood estimation function of the temperature sample is expressed as L(μ,σ)=f(Δx1)f(Δx2)...f(Δx n ), where f(Δx n ) indicates that at t n Mixed temperature trend x at the moment n The probability density function of L(μ,σ) is obtained by partial differentiation of the function L(μ,σ) to obtain the maximum likelihood estimates μ', σ'; the specific solution formula is as follows: Substitute the above maximum likelihood estimates μ' and σ' into the probability density function of gearbox life, set the fault threshold η, and obtain the upper limit T of gearbox life MAX and the remaining life of the gearbox T RUL ; Assume that the gearbox fault threshold η=F max / R max , when the probability density f(t i ) is larger, the gearbox i The greater the probability density of the point reaching the fault value η, the greater the probability density will be; therefore, the maximum probability density f max (t), the corresponding time T MAX As the upper limit of gearbox life, the probability density function of gearbox life is as follows: The method of correcting the fault threshold by real-time feedback of the abnormal operating condition fault samples acquired in advance to obtain the final remaining life includes: The data under abnormal working conditions are collected from the fan SCADA system to form fault samples. The above-mentioned method of hybrid modeling of bearing temperature and ambient temperature is used to obtain the time queue of relative temperature and speed. The temperature value is divided by the speed to obtain a data set. The data set is sequentially compared with the fault threshold η = F max / R max Make difference calculation; The calculated results are averaged to obtain the average value of the difference η' avg , so the corrected fault threshold η is Substituting the failure threshold into the above life probability density function, we can get the time T corresponding to the upper limit of life: MAX , so the remaining life of the gearbox is T RUL =T MAX -t, t indicates the current running time.

2. The wind turbine gearbox aging prediction method according to claim 1, characterized in that: The obtaining of the pre-collected gearbox bearing temperature and ambient temperature data includes: Gearbox bearing temperature time pairs (t0, C0), (t1, C1), (t2, C2), … (t n ,C n ), ambient temperature time pairs (t0,C'0), (t1,C'1), (t2,C'2),…(t n ,C' n ) and speed time pairs (t0, R0), (t1, R1), (t2, R2), … (t n ,R n ), where time is t0 <t1<t2<…<t n Arranged in order, C n Indicates t n Bearing temperature at time, C' n Indicates t n Ambient temperature at the moment, R n Indicates t n The speed of the moment.

3. The wind turbine gearbox aging prediction method according to claim 2 is characterized in that: The gearbox bearing temperature and ambient temperature data are mixed and modeled by the PCA algorithm to obtain a one-dimensional temperature data set, including: De-mean the gearbox bearing temperature and ambient temperature data; Calculate the covariance matrix of gearbox bearing temperature and ambient temperature data; Calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix; The original features are projected onto the selected feature vector to obtain new one-dimensional features after dimensionality reduction.

4. The wind turbine gearbox aging prediction method according to claim 3 is characterized in that: The de-averaging of the gearbox bearing temperature and the ambient temperature data includes: The bearing temperature is averaged and the calculation formula is as follows: In formula 1 is the average value of the bearing temperature, n is the logarithm of the bearing temperature over time, C i is the bearing temperature at the i-th moment; The ambient temperature is averaged and the calculation formula is as follows: In formula 2 is the average ambient temperature, n is the logarithm of ambient temperature over time, C' i is the ambient temperature at the i-th moment.

5. The wind turbine gearbox aging prediction method according to claim 4 is characterized in that: The covariance matrix of the gearbox bearing temperature and ambient temperature data obtained by the calculation includes: Find the covariance matrix B, the formula is as follows: in The solution formula is as follows: In formula 4 is the average value of the bearing temperature, n is the logarithm of the bearing temperature over time, C i is the bearing temperature at the i-th moment; The other calculation methods in Formula 3 are similar, and the covariance matrix B under these n samples can be finally calculated.

6. The wind turbine gearbox aging prediction method according to claim 5, characterized in that: The step of calculating the eigenvalues ​​and corresponding eigenvectors of the covariance matrix includes: Find the eigenvalue λ of the covariance matrix B and the corresponding eigenvector u, Bu = λu, there will be n eigenvalues ​​λ, each λ i Corresponding to a feature vector u i , sort the eigenvalues ​​λ in descending order, select the largest first k, and extract their corresponding k eigenvectors to obtain a set of {(λ1,u1),(λ2,u2),...(λ k ,u k )}.

7. The wind turbine gearbox aging prediction method according to claim 6, characterized in that: The projecting of the original features onto the selected feature vector to obtain a new one-dimensional feature after dimensionality reduction includes: The original characteristics of the bearing temperature and ambient temperature samples are (C i ,C' i ) T , the new feature after projection is (y i ) T ,i ranges from 0,1,…n.

8. A wind turbine gearbox aging prediction device, characterized in that: include: An acquisition unit, used for acquiring the pre-collected gearbox bearing temperature and ambient temperature data; A hybrid modeling unit, used for performing hybrid modeling on the gearbox bearing temperature and the ambient temperature data through a PCA algorithm to obtain a one-dimensional temperature data set; A calculation unit, used for inputting the one-dimensional temperature data set into a pre-built gearbox performance degradation model to calculate the remaining life of the gearbox; A correction unit is used to correct the fault threshold in real time through the abnormal operating condition fault samples acquired in advance, so as to obtain the final remaining life; In the calculation unit, the one-dimensional temperature data set is input into a pre-built gearbox performance degradation model to calculate the remaining life of the gearbox, including: The temperature y and speed after hybrid dimension reduction are quantified, and the relative temperature is expressed as F n =|y n | / R n ; Such quantified data may have certain errors, so the average of m consecutive data values ​​is processed. n At time t, the temperature trend Through the temperature trend x n Get the initial value of relative temperature F; get the gearbox performance degradation trend expression x through the Wiener process model t =F+μt+σB(t), where x t is the performance degradation trend at time t, μ is the drift coefficient, σ is the diffusion coefficient, and B(t) is the standard Wiener process Brownian motion; Due to the relative temperature increase Δx of the gearbox j =μΔt j +σΔB(Δt j ), and in the Wiener process model, ΔB(Δt j ) is normally distributed in N(0,Δt j ), so Δx j Normal distribution in N(μΔt j ,σ2Δt j ), so the maximum likelihood estimation function of the temperature sample is expressed as L(μ,σ)=f(Δx1)f(Δx2)...f(Δx n ), where f(Δx n ) indicates that at t n Mixed temperature trend x at the moment n The probability density function of L(μ,σ) is obtained by partial differentiation of the function L(μ,σ) to obtain the maximum likelihood estimates μ', σ'; the specific solution formula is as follows: Substitute the above maximum likelihood estimates μ' and σ' into the probability density function of gearbox life, set the fault threshold η, and obtain the upper limit T of gearbox life MAX and the remaining life of the gearbox T RUL ; Assume that the gearbox fault threshold η=F max / R max , when the probability density f(t i ) is larger, the gearbox i The greater the probability density of the point reaching the fault value η, the greater the probability density will be; therefore, the maximum probability density f max (t), the corresponding time T MAX As the upper limit of gearbox life, the probability density function of gearbox life is as follows: In the correction unit, the fault threshold is corrected in real time through the abnormal operating condition fault samples acquired in advance, and the final remaining life is obtained, including: The data under abnormal working conditions are collected from the fan SCADA system to form fault samples. The above-mentioned method of hybrid modeling of bearing temperature and ambient temperature is used to obtain the time queue of relative temperature and speed. The temperature value is divided by the speed to obtain a data set. The data set is sequentially compared with the fault threshold η = F max / R max Make difference calculation; The calculated results are averaged to obtain the average value of the difference η' avg , so the corrected fault threshold η is Substituting the failure threshold into the above life probability density function, we can get the time T corresponding to the upper limit of life: MAX , so the remaining life of the gearbox is T RUL =T MAX -t, t indicates the current running time.

Images