Method for predicting residual service life of aero-engine based on data and model hybrid drive

Abstract

The invention discloses an aero-engine remaining service life prediction method based on data and model hybrid drive, and relates to the field of aero-engine service life prediction, and the method comprises the steps: collecting the operation data of an aero-engine; preprocessing the data to obtain a training sample; inputting the training sample into an auto-encoder based on a time convolutional network, taking the deviation between time windows as a health index of the aero-engine, and ensuring the effectiveness of the constructed health index by using three evaluation indexes; and carrying out degradation modeling and RUL prediction by using a random effect Wiener process model with an acceleration factor to obtain an RUL predicted value of the aero-engine and a corresponding probability density function thereof. According to the hybrid driving method, the advantages of a statistical data driving method and a deep learning method are combined, and the problems that current deep learning lacks transparency and interpretability and is relatively limited in uncertainty quantification capability, and the statistical data driving method is limited in capability of processing and monitoring big data are solved.

Description

Technical Field

[0001] The present invention relates to the field of aircraft engine life prediction, and in particular to a method for predicting the remaining service life of an aircraft engine based on a hybrid drive of data and models. Background Art

[0002] Aircraft engines are the core power source of aircraft, and their operating status is directly related to the flight performance and safety of aircraft. Therefore, it is particularly important to carry out intelligent fault prognostics and health management (PHM) for aircraft engines. PHM technology is a new type of equipment comprehensive support technology that not only improves equipment safety and reliability but also reduces maintenance support costs. Under the PHM technology framework, equipment maintenance activities such as condition-based maintenance and spare parts management are based on the remaining predicted life (RUL) prediction. Therefore, accurately predicting the RUL of aircraft engines can help airlines better manage and optimize maintenance plans, reduce operating costs, and improve the safety and reliability of aircraft operations.

[0003] In recent years, data-driven methods have gradually become the mainstream technology in the field of RUL prediction. Commonly used data-driven methods are mainly divided into two categories: one is machine learning methods, and the other is statistical data-driven methods. Machine learning methods are divided into shallow machine learning and deep learning methods. Shallow machine learning relies heavily on complex feature engineering technology, while deep learning has a powerful ability to automatically extract deep features of raw data and is widely used. However, it usually runs in a black box manner, lacks transparency and interpretability, and due to its deterministic structure, the uncertainty quantification ability of this method is relatively limited. Statistical data-driven methods mainly include Wiener process, Gamma process, inverse Gaussian process, and Markov process. The Wiener process can model non-monotonic degradation data and is widely used, but its prediction accuracy depends largely on the trend of system degradation characteristics and its ability to process monitoring big data is limited.

[0004] Therefore, it is necessary to combine the advantages of statistical data-driven methods and deep learning methods to improve the RUL prediction performance. Summary of the invention

[0005] In order to solve the problem that deep learning lacks transparency and explainability and has relatively limited uncertainty quantification capabilities, the prediction accuracy of statistical data driven methods depends largely on the trend of system degradation characteristics and has limited ability to process monitoring big data. The present invention provides a method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models, the method comprising:

[0006] S1. Collecting the operating data of aircraft engines;

[0007] S2, preprocess the data to obtain training samples;

[0008] S3. Input the training samples into the autoencoder based on the time convolution network. The deviation between each time window is used as the health indicator of the aircraft engine. Three evaluation indicators are used to ensure the effectiveness of the constructed health indicator.

[0009] S4. Apply the random effect Wiener process model with acceleration factor to perform degradation modeling and RUL prediction on the health indicators of the constructed aircraft engine, and obtain the RUL prediction value of the aircraft engine and its corresponding probability density function.

[0010] The principle of the invention is as follows: an autoencoder based on a time convolutional network is used to construct unsupervised health indicators; three indicators of monotonicity, robustness and trend are used to ensure that the constructed health indicators can accurately characterize the degradation process of aircraft engines; a random effect Wiener process model with an acceleration factor is used to perform degradation modeling and RUL prediction on the constructed health indicators of aircraft engines. The introduction of the acceleration factor clearly reveals the dependency between the degradation rate and volatility, thereby explaining the fact that the greater the degradation rate of many products, the greater the volatility; by combining the statistical data driven method with the deep learning method, both the powerful ability of the deep learning method to automatically extract the deep features of the original data and the ability of the statistical data driven method to quantify uncertainty are utilized, thereby providing a guiding scheme for the formulation and decision-making of subsequent predictive maintenance strategies.

[0011] Furthermore, in S1, the specific steps for collecting the operating data of the aircraft engine include: correctly installing and connecting temperature, pressure, speed and other sensors on the aircraft engine system; configuring the data transmission mode, sampling frequency and data storage format parameters of the sensors; ensuring the normal operation of the sensors, recording the operating data of the aircraft engine, and storing the collected sensor data.

[0012] Furthermore, in S2, the specific steps of preprocessing the data include: normalizing the collected operating data based on the Min-max standardization method to eliminate the influence of the data dimension; screening out sensor data that can accurately characterize the degradation process based on the change trend of the aircraft engine operating data; accurately capturing the correlation between adjacent time points and converting multiple sensor data into the required sample size based on the sliding window processing technology;

[0013] Furthermore, in S3, the specific steps of constructing the unsupervised health index based on the autoencoder of the time convolution network include: In The data are sequentially input into the autoencoder based on the time convolutional network for reconstruction; the deviation between each time window is calculated; the Savitzky-Golay filter is used to reduce the interference caused by data noise; and three evaluation indicators are used to ensure the effectiveness of the constructed health indicators.

[0014] Furthermore, in S4, the specific steps of using the random effect Wiener process model with acceleration factor to perform degradation modeling and RUL prediction on the health indicators of the constructed aircraft engine include: constructing a random effect Wiener process model with acceleration factor; deriving RUL and the corresponding probability density formula using the first arrival time; and using a stepwise parameter estimation method to solve the unknown parameters in the random effect Wiener process model with acceleration factor.

[0015] One or more technical solutions provided by the present invention have at least the following technical effects or advantages:

[0016] 1. The data and model hybrid-driven aircraft engine remaining useful life prediction method of the present invention can effectively combine the advantages of both statistical data driven methods and deep learning methods, and can quantify the uncertainty of the prediction while realizing the RUL prediction of the aircraft engine.

[0017] 2. The method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models of the present invention can construct the health index of the aircraft engine in an unsupervised manner and can accurately characterize the operating status of the aircraft engine.

[0018] 3. The method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models of the present invention can clearly reveal the dependency between degradation rate and volatility by introducing an acceleration factor into the Wiener process model, thereby explaining the fact that the greater the degradation rate of many products, the greater the volatility.

[0019] 4. The data and model-based hybrid-driven aircraft engine remaining useful life prediction method of the present invention can provide maintenance teams with greater flexibility, optimize maintenance plans, choose to repair or replace parts at the most appropriate time, and minimize downtime and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation on the embodiments of the present invention;

[0021] Figure 1 It is a flow chart of the method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models in the present invention;

[0022] Figure 2 It is a schematic diagram of an autoencoder based on a time convolutional network in the method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and a model in the present invention;

[0023] Figure 3 It is a structural diagram of a temporal convolutional network in the method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models in the present invention;

[0024] Figure 4 It is a process for solving unknown parameters of a random effect Wiener process model with an acceleration factor in the method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models in the present invention;

[0025] Figure 5 It is a flow chart of data preprocessing in the method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models in the present invention;

[0026] Figure 6 It is a schematic diagram of the sliding window processing process in the remaining useful life prediction method of an aircraft engine based on data and model hybrid drive in the present invention.

[0027] Figure 7 It is a prediction effect diagram on FD001 of the method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models in the present invention;

[0028] Figure 8 It is a prediction effect diagram on FD004 of the method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models in the present invention; DETAILED DESCRIPTION

[0029] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0030] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those within the scope of this description. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0031] Embodiment 1

[0032] refer to Figure 1-Figure 8 This embodiment provides a method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models, the method comprising:

[0033] S1. Collecting the operating data of aircraft engines;

[0034] S2, preprocess the data to obtain training samples;

[0035] The specific steps of preprocessing the aircraft engine data to obtain training samples include: clustering the collected operating data according to the operating conditions, normalizing the data according to the operating conditions to eliminate the influence of the data dimension; screening sensor data that can accurately characterize the degradation process according to the change trend of the aircraft engine operating data; using the sliding window technology to process the screened sensor data to construct training samples;

[0036] S3. The training samples are input into the autoencoder based on the temporal convolutional network. The deviation between each time window is used as the health indicator of the aircraft engine. Three evaluation indicators are used to ensure the effectiveness of the constructed health indicator.

[0037] The specific steps of constructing the unsupervised health index for the autoencoder based on the time convolution network include: In The data are sequentially input into the autoencoder based on the time convolution network for reconstruction; the deviation between each time window is calculated; the Savitzky-Golay filter is used to reduce the interference caused by data noise; three evaluation indicators are used to ensure the effectiveness of the constructed health indicators;

[0038] S4. Apply the random effect Wiener process model with acceleration factor to perform degradation modeling and RUL prediction on the health indicators of the constructed aircraft engine, and obtain the RUL prediction value of the aircraft engine and its corresponding probability density function.

[0039] Among them, the specific steps of using the random effect Wiener process model with acceleration factor to perform degradation modeling and RUL prediction on the health indicators of the constructed aircraft engine include: constructing a random effect Wiener process model with acceleration factor; deriving RUL and the corresponding probability density formula using the first arrival time; and using a stepwise parameter estimation method to solve the unknown parameters in the random effect Wiener process model with acceleration factor.

[0040] This embodiment proposes a method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models, preprocesses the data to obtain data that can accurately characterize the degradation process; constructs unsupervised health indicators through an autoencoder based on a time convolutional network; ensures the effectiveness of the constructed health indicators through three indicators: monotonicity, robustness, and trend; and uses a random effect Wiener process model with an acceleration factor to perform degradation modeling and RUL prediction on the constructed aircraft engine health indicators, so that it can realize the RUL prediction of the aircraft engine while quantifying the uncertainty of the prediction, providing guidance for the formulation and decision-making of subsequent predictive maintenance strategies.

[0041] Embodiment 2

[0042] Based on the above embodiment 1, in this embodiment, the calculation formula of Min-max normalization is:

[0043]

[0044] in, and X i,j Represent the data after and before normalization, i and j represent the i-th data point and j-th sensor, respectively. and Represent the minimum and maximum values ​​of the jth sensor respectively;

[0045] The calculation formulas for the three indicators of monotonicity, robustness and trend are as follows:

[0046]

[0047] Among them, Mon(X), Rob(X), and Tre(X) represent the monotonicity value, robustness value, and trend value, respectively. N represents the length of the health indicator. diff＞0 and N diff＜0 They represent the number of differences greater than 0 and less than 0, respectively, h i Indicates HI at t i The value of the moment, and represent the average trend term and random term of HI respectively;

[0048] The calculation formula of Mean Absolute Error (MAE) is:

[0049]

[0050] The calculation formula of Root Mean Square Error (RMSE) is:

[0051]

[0052] Among them, MAE and RMSE represent the mean absolute error and root mean square error respectively, K represents the number of prediction samples, and y pre and act Represent the predicted value and the true value respectively;

[0053] The expression of the random effect Wiener process model with acceleration factor is:

[0054] X(t)=uΛ(t)+σ B B(ξΛ(t))

[0055] Where u represents the drift coefficient u = rξ and σ B represents the diffusion coefficient, ξ=|e u | represents the acceleration factor, Λ(t) represents the degradation trend term;

[0056] The RUL and corresponding probability density formula derived from the first arrival time are as follows:

[0057] T=inf{t>0:X(t)≥ω|X(0)<ω}

[0058]

[0059] Among them, ω represents the threshold, σ B represents the diffusion coefficient, Λ(t) represents the degradation trend term;

[0060] The calculation formula for the unknown parameters in the random effect Wiener process model with acceleration factor is as follows:

[0061]

[0062] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0063] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for predicting the remaining useful life of an aircraft engine based on a hybrid drive of data and models, characterized in that: The method comprises: S1. Collecting the operating data of aircraft engines; S2, preprocess the data to obtain training samples; S3. Input the training samples into the autoencoder based on the time convolution network. The deviation between each time window is used as the health indicator of the aircraft engine. Three evaluation indicators are used to ensure the effectiveness of the constructed health indicator. S4. Apply the random effect Wiener process model with acceleration factor to perform degradation modeling and RUL prediction on the health indicators of the constructed aircraft engine, obtain the RUL prediction value of the aircraft engine and its corresponding probability density function, and evaluate the model.

2. The method for predicting the remaining useful life of an aircraft engine based on data and model hybrid drive according to claim 1, characterized in that: In the above S1, collecting the operating data of the aircraft engine in step S1 includes: S11. Correctly install and connect temperature, pressure, speed and other sensors on the aircraft engine system; S12, configuring the data storage mode, sampling frequency and data storage format parameters of the sensor; S13, ensuring the normal operation of the sensor, recording the operation data of the aircraft engine, and storing the collected sensor data.

3. The method for predicting the remaining useful life of an aircraft engine based on data and model hybrid drive according to claim 1, characterized in that: In S2, the specific steps of preprocessing the data include: S21. Normalize the collected operating data according to the working conditions to eliminate the influence of the data dimension; use the Min-max normalization method to make the sensor value fluctuate within the range of [-1,1]; in, and X i,j Represent the data after and before normalization, respectively. i and j represent the i-th data point and the j-th sensor, respectively. and Represent the minimum and maximum values ​​of the jth sensor respectively; S22. Filter out sensor data that can accurately describe the degradation process according to the change trend of the aircraft engine operation data; S23. Use sliding window processing technology to process the filtered sensor data and construct training samples.

4. The method for predicting the remaining useful life of an aircraft engine based on data and model hybrid drive according to claim 1, characterized in that: The specific steps of constructing the health index of the aircraft engine include: S31. For a given training sample in Will Input them into the autoencoder based on the time convolutional network in sequence, and get S32. As the equipment runs, its health status will gradually deteriorate. The health index can be calculated by calculating the deviation between each time window using the following formula: Among them, N represents The number of elements in , represents the current degradation state, and Respectively represent The maximum and minimum values ​​of Represents current health indicators; S33, using Savitzky-Golay filter to reduce the interference caused by data noise; S34. Three evaluation indicators are used to ensure the effectiveness of the constructed health indicators. The indicator calculation formula is as follows: Among them, N represents the length of the health indicator, N diff＞0 and N diff＜0 They represent the number of differences greater than 0 and less than 0, respectively, h i Indicates HI at t i The value of the moment, and represent the average trend term and random term of HI respectively.

5. The method for predicting the remaining useful life of an aircraft engine based on data and model hybrid drive according to claim 1, characterized in that: The random effect Wiener process model with acceleration factor is used for modeling and RUL prediction, specifically including: S41. Construct a random effect Wiener process model with an acceleration factor. The expression is as follows: X(t)=uΛ(t)+σ B B(ξΛ(t)) Where u represents the drift coefficient u = rξ and σ B represents the diffusion coefficient, ξ=|e u | represents the acceleration factor, Λ(t) represents the degradation trend term; S42. Use the first arrival time to derive the RUL and the corresponding probability density formula. The calculation formula is as follows: T=inf{t>0:X(t)≥ω|X(0)<ω} Among them, ω represents the threshold, σ B represents the diffusion coefficient, Λ(t) represents the degradation trend term; S43. A stepwise parameter estimation method is used to solve the unknown parameters in the random effect Wiener process model with acceleration factor. The calculation formula is as follows: S44. The performance indicators used in the evaluation of the remaining useful life prediction method of aircraft engines based on a hybrid drive of data and models are as follows: Among them, K represents the number of prediction samples, y pre and act represent the predicted value and the true value respectively.

Images