Physical data driven aircraft oil temperature fault diagnosis and prediction method and system

Abstract

The application discloses a physical data driven aircraft oil temperature fault diagnosis and prediction method and system, and belongs to the technical field of electric data processing.The method comprises the following steps: S01: calculating an oil temperature characteristic value according to engine speed and oil temperature data through a feature extraction model; S02: judging whether the oil temperature characteristic value is within a first upper limit and a lower limit threshold value; if yes, the characteristic value is normal, and then step S03 is executed, and the first upper limit and the lower limit threshold value are updated according to the characteristic value; if not, the method is ended; S03: inputting the oil temperature characteristic value into an oil temperature fault prediction model to obtain a prediction value; if the prediction value is within a second upper limit and a lower limit threshold value, it is considered that the probability of oil temperature fault occurring in future time is relatively low; if not, it is considered that the probability of oil temperature fault occurring in future time is relatively high, and a fault warning needs to be sent.The application has better interpretability and predictability, and can overcome the uncertainty and subjectivity of traditional fixed threshold diagnosis.

Description

Technical Field

[0001] This invention relates to a physical data-driven method and system for diagnosing and predicting aircraft lubricating oil temperature faults, belonging to the field of fault diagnosis and prediction technology. Background Technology

[0002] Chinese invention patent application CN117932550A discloses an online lubricating oil early warning method based on an adaptive threshold. The method includes the following steps: loading running data, constructing a differential integrated moving average autoregressive model based on a random search strategy, obtaining predicted data based on the differential integrated moving average autoregressive model and the running data, and concatenating the running data and the predicted data into concatenated data as validation data; loading initial training data and constructing an autoassociative kernel regression model with a max pooling strategy; inputting the initial training data and validation data into the autoassociative kernel regression model with a max pooling strategy to construct an adaptive threshold; and realizing a lubricating oil status early warning by judging whether the expected value of the validation data exceeds the adaptive threshold.

[0003] This invention combines a differentially integrated moving average autoregressive model and a max-pooling strategy-based autoassociative kernel regression model, enabling faster and more accurate online lubricating oil condition identification and early warning capabilities. However, this invention directly utilizes raw lubricating oil data for prediction, resulting in poor interpretability and predictability. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, the purpose of this invention is to provide a physical data-driven method and system for diagnosing and predicting aircraft lubricating oil temperature faults, which has good interpretability and predictability.

[0005] To achieve the aforementioned objective, this invention provides a physical data-driven method for diagnosing and predicting aircraft lubricating oil temperature faults, comprising the following steps:

[0006] S01: The lubricating oil temperature feature value is calculated based on engine speed and lubricating oil temperature data through the feature extraction model;

[0007] S02: Determine whether the lubricating oil temperature characteristic value is within the first upper limit threshold and the first lower limit threshold. If it is within the first upper limit threshold and the first lower limit threshold, the lubricating oil temperature characteristic value is considered normal. Then, proceed to step S03 and update the first upper limit threshold and the first lower limit threshold based on the lubricating oil temperature characteristic value. If it is not within the first upper limit threshold and the first lower limit threshold, the lubricating oil temperature characteristic value is considered abnormal. The first upper limit threshold and the first lower limit threshold are not updated, and then the process ends.

[0008] S03: Input the lubricating oil temperature characteristic value into the lubricating oil temperature fault prediction model to obtain the predicted value. The predicted value is compared with the second upper limit threshold and the second lower limit threshold. If the predicted value is within the second upper limit threshold and the second lower limit threshold, it is considered that the probability of a lubricating oil temperature failure will not occur in the future is low; otherwise, it is considered that the probability of a lubricating oil temperature failure will occur in the future is high, and a fault warning needs to be issued.

[0009] To achieve the aforementioned objective, the present invention also provides a physical data-driven aircraft oil temperature fault diagnosis and prediction system, which includes a storage medium and one or more processors. The storage medium stores a computer program, which is invoked by one or more processors to implement the aforementioned physical data-driven aircraft oil temperature fault diagnosis and prediction method.

[0010] Compared with the prior art, the present invention has the following beneficial effects:

[0011] 1. The lubricating oil temperature feature value extraction process proposed in this invention incorporates a feature extraction model. Compared to directly extracting lubricating oil temperature values, the lubricating oil temperature error values ​​used in this invention include physical model information, resulting in better interpretability and predictability.

[0012] 2. In the process of constructing the lubricating oil temperature threshold, the adaptive threshold constructed through historical data can overcome the uncertainty and subjectivity of traditional fixed threshold diagnosis. Continuous threshold updates ensure the long-term effectiveness of the adaptive threshold, which can effectively reduce the false alarm rate.

[0013] 3. The prediction method based on the autoregressive difference moving average model can take into account seasonal factors to a certain extent. Compared with traditional prediction methods, it has better accuracy and stability and can accurately provide early warning of lubricating oil temperature failure. Attached Figure Description

[0014] Figure 1 This is a flowchart of the physical data-driven aircraft lubricating oil temperature fault diagnosis and prediction method provided by the present invention.

[0015] Figure 2 This is a diagram illustrating the trailing phenomenon.

[0016] Figure 3 is a schematic diagram of the truncation phenomenon. Detailed Implementation

[0017] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0018] In this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0019] Figure 1This is a flowchart of the physical data-driven aircraft lubricating oil temperature fault diagnosis and prediction method provided by the present invention; as follows: Figure 1 As shown, the physical data-driven aircraft lubricating oil temperature fault diagnosis and prediction method provided by the present invention includes the following steps:

[0020] S01: The lubricating oil temperature feature value is calculated based on engine speed and lubricating oil temperature data through the feature extraction model;

[0021] S02: Determine whether the lubricating oil temperature characteristic value is within the first upper limit threshold and the first lower limit threshold. If it is within the first upper limit threshold and the first lower limit threshold, the lubricating oil temperature characteristic value is considered normal. Then, proceed to step S03 and update the first upper limit threshold and the first lower limit threshold based on the lubricating oil temperature characteristic value. If it is not within the first upper limit threshold and the first lower limit threshold, the lubricating oil temperature characteristic value is considered abnormal. The first upper limit threshold and the first lower limit threshold are not updated, and then the process ends.

[0022] S03: Input the lubricating oil temperature characteristic value into the lubricating oil temperature fault prediction model to obtain the predicted value. The predicted value is compared with a second upper threshold and a second lower threshold. If the predicted value is within the second upper threshold and a second lower threshold, the probability that a lubricating oil temperature failure will not occur in the future is considered low; otherwise, the probability that a lubricating oil temperature failure will occur in the future is considered high, and a fault warning needs to be issued. In this invention, .

[0023] The lubricating oil temperature feature value extraction process proposed in this invention incorporates a feature extraction model. Compared to directly extracting lubricating oil temperature values, the lubricating oil temperature error values ​​used in this invention include physical model information, resulting in better interpretability and predictability.

[0024] In this invention, the lubricating oil temperature fault prediction model is trained using an ARIMA model. The data format of the ARIMA model is as follows:

[0025] ,

[0026] In the formula, c is the constant term, p is the number of autoregressive terms, and q is the number of moving average terms; These are the coefficients of the AR model; These are the coefficients of the MR model; These are the time series values ​​of the feature values; The error sequence values ​​are used; and the eigenvalue time series values ​​are subjected to d-order differencing.

[0027] The lubricating oil temperature fault prediction model is trained from an ARIMA model. Training an ARIMA model into a lubricating oil temperature fault prediction model includes the following steps:

[0028] S1-1: Obtain historical lubricating oil temperature time series feature value dataset from the database. ;

[0029] S1-2: Extract the historical lubricating oil temperature time series feature value data based on the input vector dimension tf. Sliding segmentation into input samples , corresponding to the input sample The test label is: Input sample When input into the ARIMA model, the estimated labels output are: , where i is the sample index, tf is the input vector dimension, and tp is the output vector dimension;

[0030] S1-3: The prediction is calculated using the following formula. With test label Record the deviation MSE between them:

[0031] ,

[0032] Then, the parameters of the ARIMA model are adjusted, including the number of autoregressive terms p, the difference order d, and the number of moving average terms q. The deviation calculation and recording process is repeated to obtain the parameter values ​​of the ARIMA model at the minimum deviation value MSE as the final values.

[0033] In this invention, training the ARIMA model into a lubricating oil temperature fault prediction model includes the following steps:

[0034] S1-4: Perform autocorrelation analysis on the historical lubricating oil temperature time series characteristic value dataset and calculate the k-th order autocorrelation coefficient. For k=1,2,…,N, the calculation process is as follows:

[0035] ,

[0036] In the formula, N represents the eigenvalue data. length, The average of N feature values; It is an intermediate variable.

[0037] In this invention, training the ARIMA model into a lubricating oil temperature fault prediction model includes the following steps:

[0038] S1-5: Perform partial autocorrelation analysis on the historical lubricating oil temperature time series characteristic value dataset, and calculate the k-th order partial autocorrelation coefficient. k=1,2,…,n; k-th order partial autocorrelation coefficient From the autocorrelation coefficient It is derived from this transformation, and its calculation process is as follows:

[0039] .

[0040] When performing autocorrelation and partial autocorrelation analysis on historical lubricating oil temperature time series characteristic data, observe whether the time series autocorrelation plot and partial autocorrelation plot show, for example... Figure 2 Trailing phenomenon or for example Figure 3 The tailing phenomenon; the trailing phenomenon is that the autocorrelation coefficient or partial autocorrelation coefficient remains high after a certain order or shows obvious periodic fluctuations after a certain order; the truncation phenomenon is that the autocorrelation coefficient or partial autocorrelation coefficient approaches zero after a certain order or drops sharply after a certain order.

[0041] If the time series autocorrelation plot is tailed and the time series partial autocorrelation plot is truncated, then the difference order d and the number of moving average terms q are both 0, and the number of autoregressive terms p is the number of truncated terms in the partial autocorrelation plot; if the time series autocorrelation plot is truncated and the partial autocorrelation plot is tailed, then the number of autoregressive terms p and the difference order d are both 0, and the number of moving average terms q is the number of truncated terms in the time series autocorrelation plot; if neither of the above two conditions is met, then the values ​​of the number of autoregressive terms p, the difference order d, and the number of moving average terms q are determined through steps S1-2 and S1-3.

[0042] The prediction method based on the ARIMA model has better accuracy and stability than traditional prediction methods, and can accurately provide early warning of lubricating oil temperature failures.

[0043] In this invention, the first upper limit threshold and the first lower limit threshold are determined by a lubricating oil temperature threshold model, and the determination process includes:

[0044] S2-1: Select the latest M-1 historical feature values With lubricating oil temperature characteristic value ;

[0045] S2-2: Update the first upper limit threshold using the following formulas respectively. and the first lower threshold :

[0046] ; ,

[0047] In the formula, For adjustment coefficients;

[0048] ,

[0049] In the formula, For probability distribution; This means choosing values ​​within the range that make the probability distribution... Take the maximum value The value;

[0050] ,

[0051] In the formula, The probability distributions of the first upper bound threshold and the first lower bound threshold before the update, respectively. The mean and variance of.

[0052] The adaptive threshold constructed by this invention using historical data can overcome the uncertainty and subjectivity of traditional fixed threshold diagnosis. Continuous threshold updates ensure the long-term effectiveness of the adaptive threshold, which can effectively reduce the false alarm rate.

[0053] In this invention, the feature extraction model calculates the lubricating oil temperature feature value based on engine speed and lubricating oil temperature data, specifically including:

[0054] S4-1: Calculate the theoretical value of lubricating oil temperature according to the following formula. :

[0055] In the formula, Engine speed, For parameters;

[0056] S4-2: Calculate the characteristic value of lubricating oil temperature according to the following formula: ,

[0057] In the formula, A represents the measured value of the lubricating oil temperature, and A represents the number of data points used to calculate the characteristic value of the lubricating oil temperature.

[0058] In this invention, a steady segment extraction algorithm is used to extract the time period when the engine speed is stable.

[0059] In this invention, the stationary segment extraction algorithm includes the following steps:

[0060] S3-1: Starting from the beginning of the engine speed data during the cruise phase, select engine speed data for a window length and record the start and end positions of the window;

[0061] S3-2: Determine whether the standard deviation of the selected window data is lower than the preset value;

[0062] S3-3: If it is lower than the preset value, proceed to step S3-4; if it is not lower than the preset value, move the window starting position one position to the right and then return to step S3-2.

[0063] S5-4: Determine whether the standard deviation of the new window data is lower than a known preset value.

[0064] S3-5: If it is lower than the preset value, proceed to step S3-6; if it is not lower than the preset value, move the end of the window one position to the right and then return to step S3-2.

[0065] S3-6: Save the start and end positions of the window and terminate the program.

[0066] The present invention also provides a physical data-driven aircraft lubricating oil temperature fault diagnosis and prediction system, which includes a storage medium and one or more processors. The storage medium stores a computer program, which is called by one or more processors to implement the above-described physical data-driven aircraft lubricating oil temperature fault diagnosis and prediction method.

[0067] The present invention also provides a computer program product, which uses a computer language to compile the above-mentioned physical data-driven aircraft lubricating oil temperature fault diagnosis and prediction method into a computer program that can be called and executed by one or more processors.

[0068] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A physical data-driven method for diagnosing and predicting aircraft lubricating oil temperature faults, characterized in that, Includes the following steps: S01: The lubricating oil temperature feature value is calculated based on engine speed and lubricating oil temperature data through the feature extraction model; S02: Determine whether the lubricating oil temperature characteristic value is within the first upper limit threshold and the first lower limit threshold. If it is within the first upper limit threshold and the first lower limit threshold, determine that the lubricating oil temperature characteristic value is normal. Then execute step S03 and update the first upper limit threshold and the first lower limit threshold according to the lubricating oil temperature characteristic value. If the value is not within the first upper limit threshold and the first lower limit threshold, the lubricating oil temperature characteristic value is determined to be abnormal, the first upper limit threshold and the first lower limit threshold are not updated, and then the process ends. S03: Input the lubricating oil temperature characteristic value into the lubricating oil temperature fault prediction model to obtain the predicted value. The predicted value is compared with the second upper limit threshold and the second lower limit threshold. If the predicted value is within the second upper limit threshold and the second lower limit threshold, the probability of a lubricating oil temperature failure in the future is considered to be low; otherwise, the probability of a lubricating oil temperature failure in the future is considered to be high, and a fault warning needs to be issued.

2. The physical data-driven method for diagnosing and predicting aircraft lubricating oil temperature faults according to claim 1, characterized in that, The lubricating oil temperature fault prediction model is trained from an ARIMA model. Training an ARIMA model into a lubricating oil temperature fault prediction model includes the following steps: S1-1: Obtain historical lubricating oil temperature time series feature value dataset from the database. ; S1-2: Extract the historical lubricating oil temperature time series feature value data based on the input vector dimension tf. Sliding segmentation into input samples , corresponding to the input sample The test label is: Input sample When input into the ARIMA model, the estimated labels output are: , where i is the sample index, tf is the input vector dimension, and tp is the output vector dimension; S1-3: The prediction is calculated using the following formula. With test label Record the deviation MSE between them: ， Then, adjust the parameters of the ARIMA model and repeat the process of calculating and recording the deviation to obtain the parameter values ​​of the ARIMA model at the minimum deviation value MSE as the final values.

3. The physical data-driven method for diagnosing and predicting aircraft lubricating oil temperature faults according to claim 1, characterized in that, The algorithm for extracting stable periods of engine speed is used to extract these periods.

4. The physical data-driven method for diagnosing and predicting aircraft lubricating oil temperature faults according to claim 3, characterized in that, The stationary segment extraction algorithm includes the following steps: S3-1: Starting from the beginning of the engine speed data during the cruise phase, select engine speed data for a window length and record the start and end positions of the window; S3-2: Determine whether the standard deviation of the selected window data is lower than the preset value; S3-3: If it is lower than the preset value, proceed to step S3-4; if it is not lower than the preset value, move the window starting position one position to the right and then return to step S3-2. S3-4: Determine whether the standard deviation of the new window data is lower than a known preset value. S3-5: If it is lower than the preset value, proceed to step S3-6; if it is not lower than the preset value, move the end of the window one position to the right and then return to step S3-2. S3-6: Save the start and end positions of the window and terminate the program.

5. A physical data-driven aircraft lubricating oil temperature fault diagnosis and prediction system, characterized in that, It includes a storage medium and one or more processors, the storage medium storing a computer program, the computer program being invoked by one or more processors to implement the physical data-driven aircraft lubricating oil temperature fault diagnosis and prediction method according to any one of claims 1-4.

Images