A threshold change-based online monitoring failure residual service life estimation method

Abstract

This invention relates to a method for estimating the remaining service life of online monitoring faults based on threshold changes, belonging to the field of intelligent operation and maintenance of industrial equipment. Based on online monitoring data of industrial equipment, this invention obtains early warning datasets at different times under different thresholds by changing thresholds, clarifies the relationship between thresholds and statistical parameters, and then predicts the failure state and remaining service life under the fault threshold using statistical distribution methods. This invention has the advantages of clear principle, scientific logic, accurate calculation, and high efficiency. It can connect industrial equipment operating status data, fault failure data, and remaining service life, and predict the remaining service life of different types of industrial equipment based on online monitoring data, providing an effective technical means for industrial equipment condition assessment.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance of industrial equipment, specifically to a method for estimating the remaining service life of industrial equipment through online monitoring of faults. Background Technology

[0002] Industrial equipment operates in harsh environments, with unpredictable operating conditions and complex failure mechanisms. Online monitoring points can acquire big data on industrial equipment operation, and threshold-based methods are often used for early warning and alarms based on the monitoring data, providing decision support for unit shutdown and maintenance.

[0003] Online operational data is highly complex, making it difficult to obtain health status information and assess remaining service life through online monitoring data. This is because most online monitoring data for industrial equipment indicates a normal state, with very little fault data. Therefore, predicting health status and remaining service life based on monitoring data under normal conditions has always been a challenge in the field of intelligent operation and maintenance of industrial equipment.

[0004] Therefore, there is an urgent need to develop a method for estimating the remaining service life of equipment based on online monitoring faults. This method needs to consider the relationship between normal online monitoring data, health status, and remaining service life of industrial equipment, establish a connection between online monitoring data, fault failure data, and remaining service life, and be able to estimate the remaining service life based on normal online monitoring data of the unit. This will provide accurate decision support for condition-based maintenance of the unit and avoid serious economic losses caused by accidents. Summary of the Invention

[0005] The purpose of this invention is to propose a method for predicting the remaining service life of a generating unit based on normal online monitoring data and for predicting the remaining service life of an online monitoring fault based on threshold changes. The technical solution of this invention includes the following steps:

[0006] S1. Determine the average value of historical time series data for continuous online monitoring of industrial equipment;

[0007] S2. Determine the fault threshold of the online monitoring data according to the standard provisions and relevant requirements;

[0008] S3. Determine the change threshold based on the average value of the online monitoring data and the fault threshold of the online monitoring data obtained above;

[0009] S4. Compare the online monitoring time series historical data with each change threshold to obtain the threshold exceedance data matrix;

[0010] S5. Establish data patterns among parameters exceeding different threshold values;

[0011] S6. Establish the relationship between the coefficients in the parameters that exceed the threshold under different changing thresholds;

[0012] S7. Establish the relationship between the power exponents in the over-threshold parameter patterns under different changing thresholds;

[0013] S8. Determine the power function distribution parameters corresponding to the fault threshold;

[0014] S9. Establish the fault time series corresponding to the fault threshold.

[0015] S10. Fit the fault time series to a Weibull distribution to obtain the remaining service life corresponding to the monitoring data.

[0016] In the above-mentioned method for estimating the remaining service life of online monitoring faults based on threshold changes, step S1 further includes: using historical data of continuous online monitoring of industrial equipment for no less than 30 days, with N data points, namely a1, a2, ..., a j …、a N The average value of the online monitoring data is obtained as θ0=Σa j / N.

[0017] In the above-mentioned method for estimating the remaining service life of online monitoring faults based on threshold changes, step S3 further includes: based on the average value θ0 of the obtained online monitoring data and the fault threshold θ of the online monitoring data. s Determine the threshold values ​​θ1, θ2, ... θ for K linear increments ∆θ. i 、…、θ K , satisfying: θ K =θ K-1 +∆θ and θ1<θ K ≤η(θ s +θ0), where η is the range coefficient, 0.2≤η≤0.6.

[0018] In the above-mentioned online monitoring fault remaining service life prediction method based on threshold change, step S4 further includes: processing the online monitoring time series historical data a1, a2, ..., a... j …、a N , respectively with each change threshold θ1, θ2, ... θ i 、…、θ K In comparison, values ​​not less than the threshold are 0, and values ​​greater than the threshold are 1. Using the horizontal axis to represent K different threshold variations and the vertical axis to represent N different time points, a threshold exceedance data matrix is ​​established for this online monitoring under different threshold variations:

[0019]

[0020] In the formula: H is the data matrix exceeding the threshold; h ij This indicates whether the j-th time series data is greater than the i-th change threshold; if it is, it is 1, otherwise it is 0.

[0021] In the above-mentioned online monitoring fault remaining service life prediction method based on threshold change, step S5 further includes: using time series data as the independent variable and each row H(i,:) in the threshold-exceeding data matrix H as the dependent variable, fitting and solving its power function distribution to obtain the power function distribution parameters corresponding to each set of threshold-exceeding data: coefficient c i And the power index m i .

[0022] In the above-mentioned online monitoring fault remaining service life prediction method based on threshold change, step S6 further includes: using different changing thresholds θ i -θ0 is the independent variable, and c under different threshold values. i Using the dependent variable as an example, we refit and solve for its power function distribution to obtain the power function coefficients c. i The corresponding power function distribution parameters are: coefficient c' and exponent m'.

[0023] In the above-mentioned online monitoring fault remaining service life estimation method based on threshold change, step S7 further includes: using different changing thresholds θ i -θ0 is the independent variable, and m under different threshold values. i As the dependent variable, we fit and solve for its exponential function distribution to obtain the power exponent m of the power function. i The corresponding exponential function distribution parameters are: total coefficient d' and exponential coefficient g'.

[0024] In the above-mentioned online monitoring fault remaining service life prediction method based on threshold change, step S8 further includes: setting the fault threshold θ s Substituting the power function distribution parameters c' and m' into the power function, the fault threshold θ is... s Substituting the distribution parameters d' and g' of the exponential function into the exponential function, we obtain the fault threshold θ. s The corresponding power function distribution parameters: coefficient c s And the power index m s .

[0025] In the above-mentioned online monitoring fault remaining service life prediction method based on threshold change, step S9 further includes: selecting the median z of K changing thresholds to obtain θ. z The corresponding over-threshold data sequence 1: b is used as the independent variable sequence X, and the fault threshold power function distribution parameter c is used. s and m s Substituting into the power function, we obtain the corresponding dependent variable fault time series Y, i.e., Y1, Y2, ..., Y... b .

[0026] In the above-mentioned online monitoring fault remaining service life prediction method based on threshold change, step S10 further includes: fitting the fault time series Y with a Weibull distribution to obtain the fault threshold θ.s Based on the corresponding Weibull distribution size and shape parameters, and under the specified failure probability of the online monitoring reaching the fault threshold, the corresponding remaining service life is obtained:

[0027]

[0028] In the formula: T R α represents the remaining service life; γ represents the Weibull distribution scaling parameter; γ represents the failure probability; and β represents the Weibull distribution shape parameter.

[0029] The beneficial effects of this invention are:

[0030] This invention proposes an online fault remaining service life prediction method based on threshold changes. This innovative method achieves the following technical effects:

[0031] 1. Exploring statistical patterns in data exceeding threshold times. Based on actual online monitoring data of the generating units, data patterns were analyzed, statistical methods were used to mine data relationships, and a general power function relationship for data exceeding threshold times was established, achieving good results;

[0032] 2. Establish the relationship between normal online monitoring data and remaining service life. Using a threshold variation method, obtain the over-threshold data matrix of online monitoring data at different thresholds, extract the statistical distribution patterns, establish the mathematical relationship between data at different thresholds, and then predict the remaining service life. Attached Figure Description

[0033] Figure 1 This is a flowchart of an online fault remaining service life prediction method based on threshold changes. Detailed Implementation

[0034] The present application will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention, and should not be construed as limiting the scope of protection of the present application.

[0035] The process of the online monitoring fault remaining service life prediction method based on threshold changes is as follows: Figure 1 As shown, the specific implementation method one includes the following steps:

[0036] S1. Determine the average value of historical time series data for continuous online monitoring of industrial equipment;

[0037] S2. Determine the fault threshold of the online monitoring data according to the standard provisions and relevant requirements;

[0038] S3. Determine the change threshold based on the average value of the online monitoring data and the fault threshold of the online monitoring data obtained above;

[0039] S4. Compare the online monitoring time series historical data with each change threshold to obtain the threshold exceedance data matrix;

[0040] S5. Establish data patterns among parameters exceeding different threshold values;

[0041] S6. Establish the relationship between the coefficients in the parameters that exceed the threshold under different changing thresholds;

[0042] S7. Establish the relationship between the power exponents in the over-threshold parameter patterns under different changing thresholds;

[0043] S8. Determine the power function distribution parameters corresponding to the fault threshold;

[0044] S9. Establish the fault time series corresponding to the fault threshold.

[0045] S10. Fit the fault time series to a Weibull distribution to obtain the remaining service life corresponding to the monitoring data.

[0046] In this embodiment, an online fault monitoring and remaining service life prediction method was established according to the process. Over-threshold data was established based on different threshold changes, the relationship between over-thresholds under different thresholds was determined, and the remaining service life corresponding to the fault threshold was obtained, which has strong versatility.

[0047] Specific implementation method two: such as Figure 1 As shown, this embodiment further defines S1 as described in Specific Embodiment 1. In this embodiment, S1 is based on historical time series data of industrial equipment continuous online monitoring for no less than 30 days, with N data points, namely a1, a2, ..., a j …、a N The average value of the online monitoring data is obtained as θ0=Σa j / N.

[0048] In this embodiment, an online fault monitoring and remaining service life prediction method was established according to the process. Over-threshold data was established based on different threshold changes, the relationship between over-thresholds under different thresholds was determined, and the remaining service life corresponding to the fault threshold was obtained, which has strong versatility.

[0049] Specific implementation method three: such as Figure 1 As shown, this embodiment further defines S3 as described in Specific Embodiment 1. In this embodiment, in S3, the average value θ0 of the obtained online monitoring data and the fault threshold θ of the online monitoring data are used... s Determine the threshold values ​​θ1, θ2, ... θ for K linear increments ∆θ. i 、…、θ K , satisfying: θ K =θ K-1+∆θ and θ1<θ K ≤η(θ s +θ0), where η is the range coefficient, 0.2≤η≤0.6.

[0050] In this embodiment, different change thresholds are obtained by linear increment. Other increment methods can also be implemented in this way.

[0051] Specific implementation method four: such as Figure 1 As shown, this embodiment further defines S4 as described in Specific Embodiment 1. In this embodiment, in S4, the online monitoring time series historical data a1, a2, ..., a j …、a N , respectively with each change threshold θ1, θ2, ... θ i 、…、θ K In comparison, values ​​not less than the threshold are 0, and values ​​greater than the threshold are 1. Using the horizontal axis to represent K different threshold variations and the vertical axis to represent N different time points, a threshold exceedance data matrix is ​​established for this online monitoring under different threshold variations:

[0052]

[0053] In the formula: H is the data matrix exceeding the threshold; h ij This indicates whether the j-th time series data is greater than the i-th change threshold; if it is, it is 1, otherwise it is 0.

[0054] In this embodiment, by establishing an over-threshold data matrix, multiple sets of over-threshold data corresponding to different thresholds are established, thus laying a data foundation for over-threshold and fault analysis.

[0055] Specific implementation method five: such as Figure 1 As shown, this embodiment further defines S5 as described in Specific Embodiment 1. In this embodiment, in S5, time series data is used as the independent variable, and each row H(i, :) in the threshold data matrix H is used as the dependent variable. The power function distribution is fitted and solved to obtain the power function distribution parameters corresponding to each set of threshold data: coefficient c. i And the power index m i .

[0056] In this embodiment, the correlation coefficient and power exponent are extracted according to the power function distribution, and the parameter relationship between different threshold data is established to obtain the parameter pattern, providing a basis for the corresponding parameters of other thresholds.

[0057] Specific implementation method six: such as Figure 1 As shown, this embodiment further defines S6 as described in Specific Embodiment 1. In this embodiment, in S6, different changing thresholds θ are used. i-θ0 is the independent variable, and c under different threshold values. i Using the dependent variable as an example, we refit and solve for its power function distribution to obtain the power function coefficients c. i The corresponding power function distribution parameters are: coefficient c' and exponent m'.

[0058] In this embodiment, the relationship between the power function coefficients and the changing threshold is obtained, which also conforms to the power function distribution and has universality.

[0059] Specific implementation method seven: such as Figure 1 As shown, this embodiment further defines S7 as described in Specific Embodiment 1. In this embodiment, in S7, different variation thresholds θ are used. i -θ0 is the independent variable, and m under different threshold values. i As the dependent variable, we fit and solve for its exponential function distribution to obtain the power exponent m of the power function. i The corresponding exponential function distribution parameters are: total coefficient d' and exponential coefficient g'.

[0060] In this embodiment, the relationship between the power exponent of the power function and the changing threshold is obtained, which also conforms to the power function distribution and has universality.

[0061] Specific implementation method eight: such as Figure 1 As shown, this embodiment further defines S8 as described in Specific Embodiment 1. In this embodiment, in S8, the fault threshold θ is... s Substituting the power function distribution parameters c' and m' into the power function, the fault threshold θ is... s Substituting the distribution parameters d' and g' of the exponential function into the exponential function, we obtain the fault threshold θ. s The corresponding power function distribution parameters: coefficient c s And the power index m s .

[0062] In this embodiment, the statistical function relationship of the fault threshold is established by using the obtained power function of the over-threshold data, and the power function distribution parameters and statistical distribution of the fault threshold are obtained.

[0063] Specific implementation method nine: as follows Figure 1 As shown, this embodiment further defines S9 as described in Specific Embodiment 1. In this embodiment, in S9, the median z of K changing thresholds is selected to obtain θ. z The corresponding over-threshold data sequence 1: b is used as the independent variable sequence X, and the fault threshold power function distribution parameter c is used. s and m s Substituting into the power function, we obtain the corresponding dependent variable fault time series Y, i.e., Y1, Y2, ..., Y... b .

[0064] In this embodiment, the corresponding fault time is established under fault conditions to provide basic data for subsequent extraction of statistical probability distribution parameters.

[0065] Specific implementation method ten: as follows Figure 1 As shown, this embodiment further defines S10 as described in Specific Embodiment 1. In this embodiment, in S10, the fault time series Y is fitted with a Weibull distribution to obtain the fault threshold θ. s Based on the corresponding Weibull distribution size and shape parameters, and under the specified failure probability of the online monitoring reaching the fault threshold, the corresponding remaining service life is obtained:

[0066]

[0067] In the formula: T R α represents the remaining service life; γ represents the Weibull distribution scaling parameter; γ represents the failure probability; and β represents the Weibull distribution shape parameter.

[0068] In this embodiment, by using the threshold change method and the fault threshold as the basis for remaining service life analysis, the remaining service life of online monitoring data with different test types and installation methods can be estimated, which is applicable to the remaining service life assessment of various industrial equipment.

[0069] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. However, these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the remaining service life of faults based on threshold changes in online monitoring. Its features include the following steps: S1. Determine the average value of historical time series data for continuous online monitoring of industrial equipment; S2. Determine the fault threshold of the online monitoring data according to the standard provisions and relevant requirements; S3. Determine the change threshold based on the average value of the online monitoring data and the fault threshold of the online monitoring data obtained above; S4. Compare the online monitoring time series historical data with each change threshold to obtain the threshold exceedance data matrix; S5. Establish data patterns among parameters exceeding different threshold values; S6. Establish the relationship between the coefficients in the parameters that exceed the threshold under different changing thresholds; S7. Establish the relationship between the power exponents in the over-threshold parameter patterns under different changing thresholds; S8. Determine the power function distribution parameters corresponding to the fault threshold; S9. Establish the fault time series corresponding to the fault threshold. S10. Fit the fault time series to a Weibull distribution to obtain the remaining service life corresponding to the monitoring data.

2. The online monitoring fault remaining service life prediction method based on threshold change as described in claim 1, characterized in that, The S1 further comprises: according to a certain industrial equipment continuous online monitoring time series historical data for no less than 30 days, the data amount is N, respectively a1, a2, …, a j … N , the average value θ0=Σa j / N of the online monitoring data is obtained.

3. The online monitoring fault remaining service life prediction method based on threshold change as described in claim 1, characterized in that, S3 further includes: based on the average value θ0 of the obtained online monitoring data and the fault threshold θ of the online monitoring data. s Determine the threshold values ​​θ1, θ2, ... θ for K linear increments ∆θ. i 、…、θ K , satisfying: θ K =θ K-1 +∆θ and θ1<θ K ≤η(θ s +θ0), where η is the range coefficient, 0.2≤η≤0.

6.

4. The online monitoring fault remaining service life prediction method based on threshold change as described in claim 1, characterized in that, S4 further includes: storing the online monitoring time series historical data a1, a2, ..., a j …、a N , respectively with each change threshold θ1, θ2, ... θ i 、…、θ K In comparison, values ​​not less than the threshold are 0, and values ​​greater than the threshold are 1. Using the horizontal axis to represent K different threshold variations and the vertical axis to represent N different time points, a threshold exceedance data matrix is ​​established for this online monitoring under different threshold variations: In the formula: H is the data matrix exceeding the threshold; h ij This indicates whether the j-th time series data is greater than the i-th change threshold; if it is, it is 1, otherwise it is 0.

5. The online monitoring fault remaining service life prediction method based on threshold change as described in claim 1, characterized in that, S5 further includes: using time series data as the independent variable and each row H(i,:) in the threshold data matrix H as the dependent variable, fitting and solving its power function distribution to obtain the power function distribution parameters corresponding to each set of threshold data: coefficient c i And the power index m i .

6. The online monitoring fault remaining service life prediction method based on threshold change as described in claim 1, characterized in that, S6 further includes: using different varying threshold values ​​θ i -θ0 is the independent variable, and c under different threshold values. i Using the dependent variable as an example, we refit and solve for its power function distribution to obtain the power function coefficients c. i The corresponding power function distribution parameters are: coefficient c' and exponent m'.

7. The online monitoring fault remaining service life prediction method based on threshold change as described in claim 1, characterized in that, S7 further includes: using different variation thresholds θ i -θ0 is the independent variable, and m under different threshold values. i As the dependent variable, we fit and solve for its exponential function distribution to obtain the power exponent m of the power function. i The corresponding exponential function distribution parameters are: total coefficient d' and exponential coefficient g'.

8. The online monitoring fault remaining service life prediction method based on threshold change as described in claim 1, characterized in that, S8 further includes: setting the fault threshold θ s Substituting the power function distribution parameters c' and m' into the power function, the fault threshold θ is... s Substituting the distribution parameters d' and g' of the exponential function into the exponential function, we obtain the fault threshold θ. s The corresponding power function distribution parameters: coefficient c s And the power index m s .

9. The online monitoring fault remaining service life prediction method based on threshold change as described in claim 1, characterized in that, S9 further includes: selecting the median z of K change thresholds to obtain θ. z The corresponding over-threshold data sequence 1: b is used as the independent variable sequence X, and the fault threshold power function distribution parameter c is used. s and m s Substituting into the power function, we obtain the corresponding dependent variable fault time series Y, i.e., Y1, Y2, ..., Y... b .

10. The online monitoring fault remaining service life prediction method based on threshold change as described in claim 1, characterized in that, S10 further includes: fitting the fault time series Y with a Weibull distribution to obtain the fault threshold θ. s Based on the corresponding Weibull distribution size and shape parameters, and under the specified failure probability of the online monitoring reaching the fault threshold, the corresponding remaining service life is obtained: In the formula: T R α represents the remaining service life; γ represents the Weibull distribution scaling parameter; γ represents the failure probability; and β represents the Weibull distribution shape parameter.

Images