Method for predicting residual life of rotating machinery of aero-engine based on digital twinning

Abstract

The invention provides an aero-engine rotating machinery residual life prediction method based on digital twinning, which comprises the following steps: establishing an aero-engine rotating machinery performance degradation digital twinning framework, and realizing a bidirectional mapping and dynamic interaction mechanism between a physical entity and a virtual model; establishing a model for predicting the residual life of the rotary machinery of the digital twin aero-engine; a virtual-real interaction mechanism of a physical space and a digital space is established, the physical space transmits measured data of the aero-engine rotating machinery to the digital space, and the virtual space maps degradation state data into health factors; and performing real-time prediction on the residual life of the aero-engine rotating machinery according to the prediction model, and outputting a prediction result. The real-time operation capability of the method can also meet the requirements in practical engineering, the problem of online prediction of the rotating machinery is effectively solved, and the overall prediction efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of aero-engines, and in particular, to a method for predicting the remaining life of aero-engine rotating machinery based on digital twin. Background Art

[0002] As a key component during the operation of an aircraft, the working environment of an aero-engine is complex and changeable. The failure of an aero-engine usually not only causes heavy cost losses, but also is one of the major factors leading to casualties. Therefore, the prediction of the remaining life of aero-engine rotating machinery has become an important safety research. However, in the process of using common remaining life prediction methods, the real-time computing ability cannot meet the requirements of actual engineering, resulting in low prediction efficiency and affecting the high efficiency during the overall use process. Summary of the Invention

[0003] Aiming at the defects in the prior art, the purpose of the present invention is to provide a method for predicting the remaining life of aero-engine rotating machinery based on digital twin. The real-time computing ability of the method of the present invention can also meet the requirements of actual engineering, effectively solve the problem of online prediction of rotating machinery, and improve the overall prediction efficiency.

[0004] To solve the above problems, the technical solution of the present invention is as follows:

[0005] A method for predicting the remaining life of aero-engine rotating machinery based on digital twin includes the following steps:

[0006] Establish a digital twin framework for the performance degradation of aero-engine rotating machinery to realize the two-way mapping and dynamic interaction mechanism between the physical entity and the virtual model;

[0007] Establish a model for predicting the remaining life of digital twin aero-engine rotating machinery;

[0008] Establish a virtual-real interaction mechanism between the physical space and the digital space. The physical space transmits the measured data of the aero-engine rotating machinery to the digital space, and the virtual space then maps the degradation state data into health factors;

[0009] Perform real-time prediction of the remaining life of the aero-engine rotating machinery according to the prediction model and output the prediction result.

[0010] Preferably, in the step of establishing a digital twin framework for the performance degradation of aero-engine rotating machinery and realizing the bidirectional mapping and dynamic interaction mechanism between the physical entity and the virtual model, the physical entity is the actual performance degradation process of the rotating machinery, the virtual model is the digital twin model describing the performance degradation of the rotating machinery, and the interaction between the physical entity and the virtual model is that the rotating machinery entity dynamically updates the virtual model through twin data, and the virtual model evaluates and predicts the degradation state of the rotating machinery entity through twin data.

[0011] Preferably, the twin data includes measured data, degradation state data, model parameter data, and evaluation and prediction data of the performance degradation state of aero-engine rotating machinery. The digital twin model includes four components: a common characterization model, an individual characterization model, a dynamic evolution model, and a degradation tracking and prediction model.

[0012] Preferably, the common characterization model is used to describe the general law in the performance degradation process. According to the fatigue crack failure mechanism of the rotating machinery, a performance degradation index of the turbine disk is constructed, and a common characterization model representing the general law of the performance degradation of the rotating machinery is established using the Paris model.

[0013] Preferably, the individual characterization model is used to describe the individual differences in the performance degradation process, analyze the sources of uncertainty in the performance degradation process of the rotating machinery, and use uncertainty to characterize the individual differences of the rotating machinery, reflecting the personalized customization characteristics of the digital twin model.

[0014] Preferably, the dynamic evolution model is used to describe the variation law of the degradation state and parameters with time and the transmission process of model uncertainty. A dynamic evolution model of the performance degradation of the rotating machinery is established through the DBN model.

[0015] Preferably, the degradation tracking and prediction model aims to enable the twin model to have the ability to evaluate and predict the degradation state, which is realized through the particle filter algorithm, thus establishing a complete digital twin model of performance degradation.

[0016] Preferably, the steps of establishing a model for predicting the remaining life of a digital twin aero-engine rotating machinery specifically include:

[0017] Set the input as x i , and the output as y i , M is the number of data, and the degradation process is described as follows:

[0018]

[0019] Among them, is the regression value of y(x i , ω i ), x is the input vector, ξ i~N(0,σ 2 ) is independent and identically distributed noise, ω is the weight vector, Φ is the design matrix, and K(x, x i ) is the kernel function;

[0020]

[0021] Among them, η is the kernel parameter to be optimized, and the hyperparameter α i is introduced. Assuming that ω i follows a Gaussian distribution with a mean of 0 and a variance of ;

[0022]

[0023] Among them, α = [α0, α1, …, α M ;

[0024] The variance V of the weight ω = [V0, V1, …, V M , and the mean μ = [μ0, μ1, …, μ M are calculated as follows:

[0025] V = (σ -2 Φ T Φ + A) -1

[0026] μ = σ -2 VΦ T y T

[0027] Among them, A = diag(α0, α1, …, α M ), and α and σ 2 are the parameters to be updated. The update process is as follows:

[0028]

[0029] Among them, γ i = 1 - α i V ii , and V ii is the diagonal element of V.

[0030] Preferably, the step of performing real-time prediction of the remaining life of the aero-engine rotating machinery according to the prediction model and outputting the prediction result specifically includes the following steps:

[0031] Collect the real-time vibration signal of the rotating machinery, and perform preprocessing operations such as noise reduction, effective signal interception, and normalization;

[0032] Input the preprocessed vibration signal into the convolutional autoencoder, and calculate the reconstruction error between the original signal and the reconstructed signal;

[0033] Map the reconstruction error to a health factor;

[0034] Update the Weibull reliability function parameters according to the health factor;

[0035] Let E1, E2, …, E q , …, E n be n prediction error samples, and the kernel density estimation form is:

[0036]

[0037] where: represents the probability distribution function, E represents any prediction error sample, h represents the bandwidth, and K(·) represents the kernel function satisfying the following constraints:

[0038] ρ τ (u) = τuI(u ≥ 0) + (τ - 1)uI(u < 0);

[0039]

[0040] where I is the indicator function. Given a certain sample, the probability density function is mainly affected by the kernel function K(·) and the bandwidth h;

[0041]

[0042] Compared with the prior art, the method of the present invention can reflect the general degradation trend of the equipment, and its prediction results can also be used as an auxiliary decision-making basis for the maintenance and replacement time to a certain extent. At the same time, the real-time operation ability of the method of the present invention can also meet the requirements in actual engineering, effectively solve the problem of online prediction of rotating machinery, improve the overall prediction efficiency, and provide a more rapid and efficient guarantee for people's safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] By reading the following detailed description of the non-limiting embodiments with reference to the accompanying drawings, other features, objects, and advantages of the present invention will become more apparent:

[0044] Figure 1 is a flow block diagram of the method for predicting the remaining life of aero-engine rotating machinery based on digital twin of the present invention;

[0045] Figure 2 is a diagram of the dynamic evolution model of the performance degradation of aero-engine rotating machinery of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] The present invention will be described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that for those of ordinary skill in the art, several changes and improvements can be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0047] Specifically, the present invention provides a method for predicting the remaining life of an aero-engine rotating machinery based on digital twin, as Figure 1 and Figure 2 shown, the method includes the following steps:

[0048] S1: Establish a digital twin framework for the performance degradation of an aero-engine rotating machinery to achieve a two-way mapping and dynamic interaction mechanism between the physical entity and the virtual model;

[0049] Specifically, perform digital twin modeling on the performance degradation of an aero-engine rotating machinery. The physical entity is the actual performance degradation process of the rotating machinery, and the virtual model is the digital twin model describing the performance degradation of the rotating machinery. The interaction between the physical entity and the virtual model is that the rotating machinery entity dynamically updates the virtual model through twin data, and the virtual model evaluates and predicts the degradation state of the rotating machinery entity through twin data;

[0050] The twin data mainly includes measured data, degradation state data, model parameter data, evaluation and prediction data of the performance degradation state, etc. of the aero-engine rotating machinery. The structural framework of the digital twin model includes four components: a common characterization model, an individual characterization model, a dynamic evolution model, and a degradation tracking and prediction model, etc.

[0051] In a preferred embodiment, the common characterization model is used to describe the general law in the performance degradation process. According to the fatigue crack failure mechanism of the rotating machinery, a performance degradation index of the turbine disk is constructed, and a common characterization model representing the general law of the performance degradation of the rotating machinery is established by using the Paris model.

[0052] In a preferred embodiment, the individual characterization model is used to describe the individual differences in the performance degradation process. Analyze the sources of uncertainty in the performance degradation process of the rotating machinery, and use uncertainty to characterize the individual differences of the rotating machinery, reflecting the characteristics of personalized customization of the digital twin model.

[0053] In a preferred embodiment, the dynamic evolution model is used to describe the change law of the degradation state and parameters over time and the transmission process of model uncertainty. A dynamic evolution model of the performance degradation of the rotating machinery is established through the DBN model.

[0054] In a preferred embodiment, the degradation tracking and prediction model aims to endow the twin model with the ability to evaluate and predict the degradation state, which is realized by the particle filter algorithm, so as to establish a complete digital twin model for performance degradation.

[0055] S2: Establish a model for predicting the remaining life of a digital twin aero-engine rotating machinery;

[0056] Specifically, the steps for establishing the prediction model are as follows:

[0057] Set the input as x i , and the output as y i , M is the number of data, and the degradation process is described as follows:

[0058]

[0059] Among them, is the regression value of y(x i , ω i ), x is the input vector, ξ i ~N(0,σ 2 ) is independent and identically distributed noise, ω is the weight vector, Φ is the design matrix, and K(x,x i ) is the kernel function.

[0060]

[0061] Among them, η is the kernel parameter to be optimized, introduce the hyperparameter α i , assume that ω i obeys a Gaussian distribution with a mean of 0 and a variance of α i -1 .

[0062]

[0063] Among them, α = [α0,α1,…,α M .

[0064] The variance V of the weight ω = [V0,V1,…,V M , and the calculation formulas for the mean μ = [μ0,μ1,…,μ M are:

[0065] V = (σ -2 Φ T Φ + A) -1 μ = σ -2 VΦ T y T

[0066] Among them, A = diag(α0,α1,…,α M ), and α and σ 2is the parameter to be updated, and the update process is as follows:

[0067]

[0068] where γ i = 1 - α i V ii and V ii are the diagonal elements of V.

[0069] S3: Establish a virtual-real interaction mechanism between the physical space and the digital space. The physical space transmits the measured data of the aero-engine rotating machinery to the digital space, and the virtual space then maps the degradation state data into health factors;

[0070] Specifically, use the remaining life prediction mapping function of the aero-engine rotating machinery to map the reconstruction error in the interval [0, +∞) into a health factor in the interval [0, 1]. The health factor is used to quantitatively describe the health state of the rotating machinery. A health factor of 1 indicates that the aero-engine rotating machinery has not started to degrade, and a health factor of 0 indicates that the aero-engine rotating machinery has completely degraded and failed.

[0071] According to the definitions of the reconstruction error, health factor, and the degradation law of the rotating machinery, the mapping function should meet the following requirements:

[0072] The domain is [0, +∞), and the range is [0, 1];

[0073] It is monotonically decreasing and smoothly differentiable on the domain for convenient calculation;

[0074] It has saturation. When the independent variable approaches 0, the function value approaches 1, and when the independent variable approaches +∞, the function value approaches 0. The commonly used Sigmoid function in neural networks has monotonicity and saturation;

[0075] However, the Sigmoid function is a monotonically increasing function, and its domain is (-∞, +∞), and the range is (0, 1). In order to make the Sigmoid function meet the above requirements, the Sigmoid function is improved to obtain an exponential mapping function:

[0076] In the formula: h represents the exponential mapping function of the reconstruction error and the health factor; k is the shape coefficient; b is the bias constant. In a preferred embodiment, the bias constant b ensures that the health factor approaches 1 when the reconstruction error is small, and the shape coefficient k is used to adjust the decreasing rate of the mapping function.

[0077] The health factor has the same definition as the reliability, and both are indicators characterizing the health state of the rotating machinery and can be used interchangeably. Therefore, directly replace the reliability with the calculated health factor and substitute it into the reliability function for corresponding fitting construction and parameter update.

[0078] S4: Perform real-time prediction of the remaining life of the aero-engine rotating machinery according to the trained prediction model, and output the prediction result.

[0079] Specifically, the real-time prediction of the remaining life of the aero-engine rotating machinery includes the following steps:

[0080] (1) Collect the real-time vibration signals of the rotating machinery, and perform preprocessing operations such as noise reduction, effective signal interception, and normalization.

[0081] (2) Input the preprocessed vibration signals into the convolutional autoencoder, and calculate the reconstruction error between the original signal and the reconstructed signal.

[0082] (3) Map the reconstruction error to a health factor.

[0083] (4) Update the parameters of the Weibull reliability function according to the health factor.

[0084] (5) Let E1, E2, …, E q , …, E n be n prediction error samples, and the kernel density estimation form is:

[0085]

[0086] Where: represents the probability distribution function, E represents any prediction error sample, h represents the bandwidth, and K(·) represents the kernel function that satisfies the following constraints:

[0087] ρ τ (u) = τuI(u ≥ 0) + (τ - 1)uI(u < 0);

[0088] ρ τ (u) = τu, u ≥ 0

[0089] ρ τ (u) = (τ - 1)u, u < 0

[0090] Where I is the indicator function. Given a certain sample, the probability density function is mainly affected by the kernel function K(·) and the bandwidth h.

[0091]

[0092] Use the integral method to calculate the cumulative distribution function, select the 5% quantile of the cumulative distribution function, calculate the fluctuation interval of the prediction error under 95% confidence, and superimpose the 95% confidence error fluctuation interval on the prediction result of the model to obtain the prediction interval of the rotating machinery life.

[0093] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A method for predicting the remaining useful life of aero-engine rotating machinery based on digital twin, characterized in that, The method includes the following steps: Establish a digital twin framework for the performance degradation of aero-engine rotating machinery to achieve a two-way mapping and dynamic interaction mechanism between the physical entity and the virtual model; Establish a model for predicting the remaining useful life of the digital twin aero-engine rotating machinery; Establish a virtual-real interaction mechanism between the physical space and the digital space. The physical space transmits the measured data of the aero-engine rotating machinery to the digital space, and the virtual space then maps the degradation state data into health factors; Perform real-time prediction of the remaining useful life of the aero-engine rotating machinery according to the prediction model and output the prediction results.

2. The method for predicting the remaining life of an aero-engine rotating machinery based on digital twin according to claim 1, wherein, In the step of establishing a digital twin framework for the performance degradation of aero-engine rotating machinery to achieve a two-way mapping and dynamic interaction mechanism between the physical entity and the virtual model, the physical entity is the actual performance degradation process of the rotating machinery, the virtual model is the digital twin model describing the performance degradation of the rotating machinery, and the interaction between the physical entity and the virtual model is that the rotating machinery entity dynamically updates the virtual model through twin data, and the virtual model evaluates and predicts the degradation state of the rotating machinery entity through twin data.

3. The method for predicting the remaining life of an aero-engine rotating machinery based on digital twin according to claim 2, wherein The twin data includes the measured data, degradation state data, model parameter data, and evaluation and prediction data of the performance degradation state of the aero-engine rotating machinery. The digital twin model includes four components: a common characterization model, an individual characterization model, a dynamic evolution model, and a degradation tracking and prediction model.

4. The method for predicting the remaining life of an aeroengine rotating machinery based on digital twin according to claim 3, wherein The common characterization model is used to describe the general law in the performance degradation process. According to the fatigue crack failure mechanism of the rotating machinery, a performance degradation index of the turbine disk is constructed, and a common characterization model representing the general law of the performance degradation of the rotating machinery is established using the Paris model.

5. The method for predicting the remaining life of an aero-engine rotating machinery based on digital twin according to claim 3, characterized in that, The individual characterization model is used to describe the individual differences in the performance degradation process, analyze the sources of uncertainty in the performance degradation process of the rotating machinery, and use uncertainty to characterize the individual differences of the rotating machinery, reflecting the personalized customization characteristics of the digital twin model.

6. The method for predicting the remaining life of an aero-engine rotating machinery based on digital twin according to claim 3, wherein, The dynamic evolution model is used to describe the variation law of the degradation state and parameters over time and the transmission process of model uncertainty. A dynamic evolution model of the performance degradation of the rotating machinery is established through the DBN model.

7. The method for predicting the remaining life of an aero-engine rotating machinery based on digital twin according to claim 3, wherein The degradation tracking and prediction model aims to enable the twin model to have the ability to evaluate and predict the degradation state, which is achieved through the particle filter algorithm, thereby establishing a complete digital twin model of performance degradation.

8. The method for predicting the remaining life of an aero-engine rotating machinery based on digital twin according to claim 1, wherein The step of establishing a model for predicting the remaining useful life of the digital twin aero-engine rotating machinery specifically includes: Set the input as x i , and the output is y i , where M is the number of data, and the degradation process is described as follows: Among them, is the regression value of y(x i , ω i ), x is the input vector, ξ i ~N(0, σ 2 ) is independent and identically distributed noise, ω is the weight vector, Φ is the design matrix, K(x, x i ) is the kernel function; Among them, η is the kernel parameter to be optimized, and the hyperparameter α is introduced i , assuming ω i follows a Gaussian distribution with a mean of 0 and a variance of ​ where α = [α0, α1, …, α M ; The variance V of the weight ω = [V0, V1, …, V M , and the calculation formula for the mean μ = [μ0, μ1, …, μ M is as follows: V = (σ -2 Φ T Φ + A) -1 μ = σ -2 VΦ T y T where A = diag(α0, α1, …, α M ), α and σ 2 are parameters to be updated, and the update process is as follows: where γ i = 1 - α i V ii and V ii are the diagonal elements of V.

9. The method for predicting the remaining useful life of an aeroengine rotating machinery based on digital twin according to claim 1, characterized in that The step of performing real-time prediction of the remaining useful life of the aero-engine rotating machinery according to the prediction model and outputting the prediction results specifically includes the following steps: Collect the real-time vibration signal of the rotating machinery and perform preprocessing operations such as noise reduction, effective signal interception, and normalization; Input the preprocessed vibration signal into the convolutional autoencoder and calculate the reconstruction error between the original signal and the reconstructed signal; Map the reconstruction error into a health factor; Update the parameters of the Weibull reliability function according to the health factor; Let \(E_1, E_2, \ldots, E\) q , \ldots, E n be \(n\) predicted error samples, and the kernel density estimation form is: Wherein: represents a probability distribution function, E represents any predicted error sample, h represents a bandwidth, and K(·) represents a kernel function satisfying the following constraints: ρ τ (u) = τuI(u ≥ 0) + (τ - 1)uI(u < 0); Where I is the index function. After a certain sample is given, the probability density function is mainly affected by the kernel function K(·) and the bandwidth h;

Images