Aircraft structure key part digital twinborn updating and life prediction method

Abstract

The invention belongs to the technical field of aircraft structure health monitoring, and particularly relates to an aircraft structure key part digital twin updating and life prediction method, which improves the accuracy of future crack propagation and residual life prediction of twin through a continuously updated damage evolution model so as to more accurately evaluate the structure health state. A user can be guided to make a safe and economical use and maintenance decision, reverse control of a twinborn body on a physical entity is achieved, use safety of an aircraft structure is guaranteed, and use and maintenance cost is reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of aircraft structure health monitoring, and particularly relates to a method for updating a digital twin of a key part of an aircraft structure and predicting the service life. BACKGROUND

[0002] The digital twin technology of the aircraft body structure constructs a complete mapping virtual model based on the physical model of the aircraft, uses historical data and real-time updated data of sensors to depict and reflect the whole life cycle process of the physical object, and one of the most important purposes is to realize more accurate life prediction of the body structure. Therefore, only the digital twin at the macro level of the body has little benefit, and only the digital twin at the level of the key part of the structure is meaningful, which plays an important role in improving the level of aircraft structure fault prediction and health management.

[0003] The most important thing of the digital twin technology is the data flow of each link, and the perception of the load environment data and the structural damage state data in the service process of the aircraft is crucial, and the requirement for the sensor technology is high. In recent years, although the digital twin technology has achieved numerous research results, it has not been truly applied to the aircraft body. From the aspect of data flow, the current problem of load environment data perception acquisition has been basically solved, but there is no mature sensor technology for the perception of structural damage state, which restricts the further development of the digital twin technology of the body structure, and this is one of the main reasons why the digital twin of the body is still in the stage of conceptual research and application assumption.

[0004] Among the already disclosed patents, patent CN116167153A discloses a method for real-time prediction of the maneuvering performance of an aircraft based on a digital twin. The patent constructs a digital twin model of the aircraft body, reflects the data set of the aircraft structure manufacturing stage into the digital twin, then uses the digital twin model of the body to predict the strength / stiffness characteristics of the body, further predicts the stress and strain level and the carrying capacity of the aircraft body, so as to determine the safe use conditions of the aircraft body and predict its maneuvering performance. The focus of the patent is to predict the residual carrying capacity of the aircraft through the digital twin, and further predict the maneuvering performance of the aircraft. The patent only broadly describes the implementation process in principle, does not give the specific implementation method, does not explain how to build the Bayesian network, and does not specifically give how to optimize and improve. In addition, the patent does not mention the update iteration of the damage evolution model which is crucial in the digital twin, especially how to ensure the accuracy of the update iteration of the life prediction, and only uses the theoretical damage evolution model, which is far from enough for the prediction accuracy of the complex aircraft structure under load. Under such conditions, it is difficult to truly achieve high-fidelity digital reproduction of the physical entity by the digital twin.

[0005] Patent CN111737811A discloses a helicopter moving part life management method, device and medium based on digital twinning. The patent realizes life prediction of helicopter moving parts by steps of load data acquisition, structure stress analysis, crack parameterization modeling, fracture mechanics simulation, agent model construction of fracture mechanics parameter prediction, and probability prediction of crack propagation. It supports structure risk assessment and dynamic adjustment of maintenance inspection plan. In the patent, the main technical means to realize life prediction is to obtain a large amount of structure stress and strain field and crack front stress intensity factor data under different positions and different lengths of cracks through a large number of finite element analysis and three-dimensional crack propagation analysis. Based on these data, an agent model is constructed to calculate the stress intensity factor through the stress and strain field, which is the focus of the patent. For key structures of an aircraft, such as the main load-bearing frame and beam, due to their large size and complex connection relationship, finite element analysis and calculation are time-consuming. It is not practical to use this method to obtain the data required to establish the agent model in engineering applications. In fact, the crack initiation position of the key part of the aircraft structure can be accurately determined through analysis. The patent predicts the remaining life from the visible and detectable macroscopic crack state (2mm crack in the example), which can only predict the crack propagation life. It cannot start from an invisible and undetectable equivalent initial defect, covering the crack initiation life (typical value in engineering is 0.8mm crack corresponding to life) and crack propagation life. At the same time, since the crack initiation life of the aircraft structure accounts for a larger proportion, it is more important for the digital twinning and life management of the aircraft structure. SUMMARY

[0006] Invention purpose: Based on the current technical status, the present application is based on the actual engineering application, and on the basis of overcoming the dynamic updating method of damage evolution model required by digital twinning, a digital twinning body updating and life prediction method for key parts of aircraft structures is proposed. Strain monitoring, non-destructive testing, and direct monitoring of damage sensors are used to perceive the damage state of the structure. Among them, at least strain monitoring and non-destructive testing data can be obtained to drive the update of the digital twinning body and further predict the remaining life of the structure, maximizing the value of engineering application.

[0007] Technical scheme: In order to achieve the above invention purpose, the present application proposes a digital twinning body updating and life prediction method for key parts of aircraft structures, comprising the following steps: Step S1, establish a CAE digital twinning body of the structure, arrange a plurality of virtual strain gauges on both sides of the expected crack propagation path, carry out three-dimensional crack propagation analysis of the initial crack a0, and obtain crack length and geometric correction factor (a, β) data; wherein a0 is the equivalent initial crack length of the key part, which is usually less than 0.1mm; preferably, a CAE digital twinning body of the structure is established by using finite element method Step S2, starting from a0, extract the strain response of the virtual strain gauge under the same load state at certain crack length intervals, obtain the strain-crack size dataset, establish the crack length a-strain response ε model, and obtain the "ε-a" model; preferably, use multivariate linear or nonlinear regression, neural network, or other machine learning methods for modeling; Step S3, according to the various flight state parameters recorded during the service of the aircraft, use the "flight parameter-component load" model and the "component load-stress" model developed and verified through research and test, and flight test to obtain the component load spectrum and the stress spectrum of the key parts of the structure; Step S4, from the component load spectrum in step S3, combined with the flight parameter history, select the flight state and load state of the same load state corresponding to the "ε-a" model, and extract the measured strain data and corresponding life N of the key parts near the physical entity of the aircraft structure under these load states, to obtain the measured (ε, t) dataset; Step S5, according to the ratio between the load value corresponding to the (ε, t) dataset in step S4 and the load value corresponding to the "ε-a" model, correct the (ε, t) dataset to the value under the same load value, compare it with the virtual strain gauge response value in step S2 and correct the "ε-a" model; Step S6, substitute the (ε, t) dataset into the "ε-a" model to calculate the (a, t) data based on strain monitoring; The (a, t) data predicted by the "ε-a" model remains unchanged in theory under the same load before the crack initiation at the key position of the structure, and for a long period of time, the predicted crack size is 0, and there is no available crack monitoring data to drive the damage model to update. In this case, the engineering detectable crack size a kj As a constraint, that is, when no crack is detected, the crack length predicted by the damage evolution model at time t should not be greater than a kj , to avoid the damage evolution model giving too conservative results without updating, through this constraint, a relatively conservative prediction result can be given for the crack initiation life before there is available monitoring and detection data; Step S7, select a damage evolution model, extract the model parameters corresponding to the structural material from the material performance database, and determine the prior parameters of the model; wherein the material performance data should include the crack propagation threshold value △K th , and the fracture toughness K IC ; Step S8, input the (a, β) data in step S1, the stress spectrum in step S3, and the model prior parameters in step S7, and predict the crack propagation one load cycle after another to obtain the theoretically predicted (a', t') data; Step S9: Using the (a, β) data from Step S1, the stress spectrum from Step S3, the (a, t) data from Step S6, the prior parameters of the model from Step S7, and the theoretical prediction (a', t') data from Step S8 as inputs, the damage evolution model selected in Step S7 is updated using the Bayesian method to obtain the posterior parameters of the model and complete the damage evolution model update. Step S10: Update the structural digital twin with the latest (a,t) data from step S6, and integrate the updated damage evolution model into the digital twin. Using the (a,β) data from step S1, the stress spectrum from step S3, and the model parameters from step S9 as inputs, predict the crack propagation (a,t) data and remaining life of the key parts of the aircraft structure under the current damage state. Step S11: Based on the future crack propagation and remaining life predicted by the twin, assess the structural health status, and realize the reverse control of the physical entity by the twin by proposing suggestions to adjust the aircraft's operating weight or to carry out inspection and maintenance. Step S12: As the aircraft service data and measured strain data are continuously updated, repeat steps S3 to S10 to realize the updating of the aircraft structure digital twin and life prediction.

[0008] Furthermore, in step S1, the geometric correction factor (a, β) data already has a database of typical structural details and typical crack configurations in practical applications, and can be continuously enriched through three-dimensional crack propagation simulation, and can be quickly called up in the digital twin update.

[0009] Furthermore, in step S4, the measured strain data can be obtained by testing, but is not limited to, high-reliability strain gauges and fiber Bragg grating sensors in aircraft structure applications.

[0010] Furthermore, in step S4, the load state corresponding to the “ε-a” model can be used in the field to periodically perform some standard flight take-offs and landings. Within the specified flight state parameter range, specific standard maneuvers are performed and maintained for a certain period of time (such as 1 to 3 seconds) to obtain relatively stable strain data, which facilitates the identification of strain changes under similar loads to identify structural cracks.

[0011] Furthermore, in step S6, (a,t) data can also be obtained through damage monitoring sensors or periodic non-destructive testing.

[0012] Further, in step S8, the crack propagation prediction process is as follows: for each load cycle, if the stress intensity factor range ΔK of that cycle is less than the crack propagation threshold value ΔK... thIf the crack does not propagate under the load cycle, the introduction of this criterion ensures that the crack initiation life can be predicted from the equivalent initial crack a0 (the typical value in engineering is the life corresponding to a 0.8 mm crack); if the stress intensity factor Kmax corresponding to the peak value is greater than the fracture toughness K... IC If this occurs, the structure will fracture and fail.

[0013] Further, in step S9, the damage evolution model update first requires establishing a likelihood function between the (a,t) data obtained from strain monitoring in step S6 and the theoretical prediction (a',t') data in step S8, then establishing the posterior distribution function of the model parameters, and extracting samples of the posterior parameters from the posterior distribution function.

[0014] Furthermore, the damage evolution model updates the posterior distribution of the model parameters. Since the stress intensity factor range ΔK of each load cycle under the random spectrum is not monotonically changing but random, the posterior distribution function of the model parameters is not analytical for the random spectrum. Sampling can be performed using, but is not limited to, the Markov chain Monte Carlo method.

[0015] Furthermore, the key to the Markov chain Monte Carlo method lies in establishing reasonable acceptance and rejection criteria to obtain a stable Markov chain with as few samplings as possible. Whether it is a random spectrum or a constant amplitude spectrum, the criterion for accepting the next sample is that the prediction error of the crack size of the sample is less than that of the previous sample.

[0016] Furthermore, (a,t) data acquired through non-destructive testing or damage monitoring sensors can also be used for updating the digital twin.

[0017] Technical effects: Based on the current state of technology and practical engineering applications, this invention proposes an engineering-feasible method for updating digital twins of key parts of aircraft structures and predicting their lifespan, using dynamic updating technology of damage evolution models as a foundation. This method can be applied and verified in complex ground verification tests such as full-aircraft fatigue tests. The life prediction method proposed in this invention starts from a very small equivalent initial crack. Before there are any visible, detectable, or monitorable cracks, it uses the engineering detectable crack size of key parts as a constraint for updating the damage evolution model. This enables life prediction to cover crack initiation and crack propagation, which is more comprehensive than the existing technology and can better guide the inspection, maintenance, and preventive repair of aircraft. The technical approach established by this invention is compatible with multiple methods such as strain monitoring, non-destructive testing, and direct monitoring by damage sensors to sense structural damage status, and has good scalability as sensor technology develops. This invention enables the creation of digital twins for critical components of aircraft structures. Through continuously updated damage evolution models, the accuracy of the twin in predicting future crack propagation and remaining lifespan is improved, thereby more accurately assessing the structural health status. This can guide users in making safe and economical use and maintenance decisions, enabling the twin to exert reverse control over the physical entity, ensuring the safety of aircraft structures, and reducing use and maintenance costs.

[0018] Compared to traditional life prediction methods, this invention only requires one finite element analysis to simulate crack propagation for aircraft structures to obtain the (a, β) data needed for crack propagation prediction, without the need to conduct numerous finite element analyses to establish a surrogate model for predicting stress intensity factors. In practical applications, (a, β) data for typical crack configurations of various structural details are already available, making it more feasible in engineering applications. This invention improves life prediction accuracy by dynamically updating the damage evolution model, which is significantly different from traditional methods that use surrogate models to quickly calculate stress intensity factors and Bayesian methods to update crack size distribution to improve prediction accuracy.

[0019] This invention details a method for updating damage evolution model parameters, elaborates on the method of establishing an "ε-a" model through three-dimensional crack propagation analysis, and describes how to use measured data to correct the "ε-a" model established from simulation data during aircraft service. It specifically proposes periodically performing standard takeoffs and landings and standard maneuvers to obtain relatively stable measured strain data for more accurate crack identification and monitoring; it also provides a clearer and more explicit data flow required for updating the digital twin, making it more engineering-feasible; and through dynamic updates of the damage evolution model, it achieves higher life prediction accuracy. Attached Figure Description

[0020] Figure 1 A technical flowchart of a method for updating and predicting the lifespan of digital twins of key components of an aircraft structure. Figure 2 A schematic diagram of (a-β) data for selected key areas; Figure 3 This is a schematic diagram showing the relationship between the strain of a key component and the response value of the third strain gauge in its vicinity under the same load. Figure 4 This shows how the aN curve predicted by the updated damage model changes as the amount of (a,N) monitoring data increases. Detailed Implementation

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below with reference to the accompanying drawings or specific implementation examples. It should be noted that some (but not all) of the disclosed examples are shown in the drawings. In fact, many different examples can be described, and these examples should not be construed as limited to the examples set forth herein. Rather, these examples are described to better demonstrate the positive effects of the present invention, and all aspects not detailed herein are considered to be well-known or conventional techniques in the art.

[0022] See appendix Figures 1-4 A method for updating and predicting the lifespan of digital twins of key aircraft structural components, the technical process of which is described in [link to relevant documentation]. Figure 1 The possible implementation methods are as follows: Step S1: Based on the fatigue strength analysis results of a certain aircraft structure, a CAE digital twin of the selected structure is established using the finite element method. Several virtual strain gauges are arranged on both sides of the expected crack propagation path. A three-dimensional crack propagation analysis is carried out with an initial crack of a0, and the crack length and geometric correction factor (a-β) data are obtained. See [link to relevant documentation]. Figure 2 Where a0 is the apparent equivalent initial crack length of the critical part, which is usually less than 0.1 mm. When calculating β, the maximum principal stress when there is no crack in the critical part is taken as the reference stress. The (a, β) data, in the practical application of the present invention, has formed a database of typical structural details and typical crack configurations, and can be continuously enriched through three-dimensional crack propagation simulation, and can be quickly called in the digital twin update; Step S2: Starting from a0, extract the strain response of virtual strain gauges under the same load condition at certain crack length intervals (see...). Figure 3 To obtain the strain and crack size dataset, machine learning methods such as multiple linear or nonlinear regression and neural networks are used to establish a model of the relationship between crack length a and strain response ε—the “ε-a” model. Step S3: Collect various flight status parameters recorded during the aircraft's service life, and use the "flight parameter-component load" model and "component load-stress" model, which have been verified through research and development, testing, and flight testing phases, to obtain the component load spectrum and stress spectrum of key structural parts; The component loads include the bending moment at the root of the main wing surface, the hinge moment of the control surface, the landing gear load, and the engine intersection load. The stresses are the maximum or minimum principal stresses at the key parts of each critical component. Step S4: From the component load spectrum in step S3, combined with the flight parameter history, select the flight state and load state with basically the same load size as the load state corresponding to the “ε-a” model, extract the measured strain data and corresponding lifetime N near the key parts of the aircraft structure physical entity under these load states, and obtain the measured (ε, t) dataset. The measured strain data, in aircraft structural applications, can be obtained using, but is not limited to, high-reliability strain gauges and fiber Bragg grating sensors. In this embodiment, multiple strain gauges are used to monitor crack propagation, and the relationship between strain response and crack length is as follows: Figure 3 ; The load state that needs to be determined, which is basically consistent with the load state corresponding to the “ε-a” model, can be used in the field to periodically perform some standard flight take-offs and landings. Within the specified flight state parameter range, specific standard maneuvers can be performed and maintained for a certain period of time (such as 1 to 3 seconds) to obtain relatively stable strain data, which is convenient for identifying strain changes under similar loads. Step S5: Based on the ratio between the load value corresponding to the (ε, t) data in step S4 and the load value corresponding to the “ε-a” model, the (ε, t) dataset is proportionally corrected to the value under the same load, and compared with the virtual strain gauge response value in step S2 to correct the “ε-a” model. The (ε, t) data correction is performed in the early stages of aircraft service, when there are no cracks in key parts. Based on the ratio of the actual load value of the load state selected in step S4 to the load state corresponding to the "ε-a" model, the theoretical strain value ε is corrected to eliminate the influence of strain gauge patching error and simulation analysis error. The corrected (ε, t) data is then used to update the "ε-a" model. Step S6: Substitute the (ε, t) data into the "ε-a" model to calculate the (a, t) data based on strain monitoring. Before the initiation of detectable engineering cracks in critical structural parts, theoretically, the strain response remains unchanged under the same load. Within a certain time period, the predicted crack size is theoretically 0 or a very small value, meaning there is no available (a, t) data to drive the damage model update. In this case, the detectable engineering crack size a at that location is used as the basis for the calculation. kj As a constraint, the damage evolution model should be prevented from giving overly conservative results before it has been updated. In this embodiment, the detectable crack length 'a' is given for each critical component. kj When no crack is detected, the crack length predicted by the damage evolution model at time t should not exceed a. kj If it is greater than a kj Then (a) kj The damage evolution model is updated using ,t) as a constraint to avoid overly conservative predictions; depending on the detectability of the location and the structural form, a kj The value is between 1mm and 5mm; Step S7: Select the Walker crack propagation rate formula as the damage evolution model. From the four parameters C, n, M1, and M2 of the model, determine the parameters that need to be updated: C, n, the prior values ​​of C and n, M1, M2, and the crack propagation threshold value ΔK.th Fracture toughness K IC All of these were extracted from the material properties database; Step S8: Using the (a, β) data from step S1, the stress spectrum from step S3, and the model prior parameters from step S7 as inputs, predict crack propagation cyclically to obtain the theoretical prediction (a', t') data. (See...) Figure 4 Predicted data based on prior parameters; The crack propagation prediction, for each load cycle, if the stress intensity factor range ΔK of that cycle is less than the crack propagation threshold ΔK... th If the crack does not propagate under the load cycle, this criterion ensures that the crack initiation life can be predicted from the equivalent initial crack a0 (the typical value in engineering is the life corresponding to a 0.8 mm crack); if the stress intensity factor Kmax corresponding to the peak value is greater than the fracture toughness K... IC If so, the structure will fracture and fail; Step S9: Using the (a-β) data from Step S1, the stress spectrum from Step S3, the (a,t) data from Step S6, the prior parameters of the model from Step S7, and the theoretical prediction (a',t') data from Step S8 as inputs, the damage evolution model selected in Step S7 is updated using the Bayesian method to obtain the posterior values ​​of the model parameters C and n, thus completing the damage evolution model update. The damage evolution model update first requires establishing a Gaussian likelihood function between the (a,t) data obtained from strain monitoring in step S6 and the (a',t') data predicted in step S8, and then establishing the posterior distribution function of the model parameters, and extracting samples of the posterior parameters from the posterior distribution function. The posterior distribution of the model parameters established by the damage evolution model update is not analytical for the random spectrum because the range of stress intensity factor ΔK for each load cycle under the random spectrum is not monotonically changing but random. Therefore, sampling can be performed using methods including Markov chain Monte Carlo method. The key to the Markov chain Monte Carlo method is to establish a reasonable sample acceptance and rejection criterion to obtain a stable Markov chain with as few samplings as possible. In this invention, whether it is a random spectrum or a constant amplitude spectrum, the criterion for accepting the next sample is that the prediction error of the crack size of the sample is less than that of the previous sample.

[0023] Step S10: Update the damage state and stress field of the structural digital twin using the latest (a,t) data points from Step S6, and integrate the updated damage evolution model into the digital twin. Using the (a,β) data from Step S1, the stress spectrum from Step S3, and the posterior parameters of the model from Step S8 as inputs, predict the crack propagation of key parts of the aircraft structure under the current damage state in a loop, obtaining the predicted (a”,t) data and the remaining life under the current damage state. See [link to relevant documentation]. Figure 4 Mid- and posterior parameter prediction data; For each load cycle, if the range of the stress intensity factor ΔK corresponding to the peak value is less than the crack propagation threshold ΔK... th If the crack does not propagate under this load cycle; if the stress intensity factor K corresponding to the peak value is... max Greater than fracture toughness K IC If so, the structure will break. Step S11: After acquiring new crack monitoring (a,t) data, compare it with the prediction data in step S9 to verify the prediction accuracy of the updated damage model. If the crack size prediction error of the damage model does not exceed 20% after no less than 3 tests, it is considered that the damage evolution model after multiple updates has sufficient prediction accuracy. Under this condition, based on the future crack propagation and remaining life predicted by the twin, assess the structural health status. By proposing suggestions to adjust the aircraft's usage level or to carry out inspection and maintenance, the twin can achieve reverse control of the physical entity, ensure the safety of aircraft structure use, and reduce usage and maintenance costs. Step S12: As the aircraft service data and measured strain data are continuously updated, repeat steps 3 to 10 to achieve the updating and life prediction of the aircraft structural digital twin. During this process, the periodic non-destructive testing data of the aircraft body, the non-destructive testing data during major overhauls, and the (a,t) data acquired by damage monitoring sensors can all be used for twin updating. Figure 4 It is evident that, for crack propagation prediction, as the number of monitored or detected crack data increases, the crack propagation curve predicted by the posterior parameters gradually approaches the observed value, ensuring the accuracy of twin lifetime prediction.

[0024] The proposed method for updating and predicting the lifespan of key aircraft structural components using digital twins is employed. During field operations, stable measured strain data is obtained by periodically performing standard takeoffs and landings or standard maneuvers to correct the "ε-a" model established through digital twin simulation analysis. Due to simulation errors, sensor attachment errors, and errors in force transmission simulation of complex assembly structures, this method is crucial for predicting crack size using the "ε-a" model. Given the detectable crack length for each key component, at any given time, when no crack is detected or detected, the update of the damage evolution model is constrained by the detectable crack length. That is, the crack length predicted by the updated damage evolution model at that time should not exceed 'a'. kj This ensures that the damage evolution model can be appropriately updated in a direction that tends towards reality even when no actual cracks are observed, thus having significant engineering application value.

[0025] The above specific embodiments or examples are only used to explain the technical solutions of the present invention and are not intended to limit the present application. Parts not described in detail are considered to be conventional technical means or common knowledge in the field. It can be understood by those skilled in the art that, based on the design concept of the present application, the technical solutions described in the foregoing embodiments can be adapted or some or all of the technical features can be equivalently replaced. These modifications, equivalent replacements, and adaptive improvements do not depart from the technical essence of the present invention and should all be covered within the protection scope of the present application.

Claims

1. A method for updating and predicting the lifespan of digital twins of key components of an aircraft structure, characterized in that, Includes the following steps: Step S1: Establish a CAE digital twin of the structure, arrange several virtual strain gauges on both sides of the expected crack propagation path, conduct a three-dimensional crack propagation analysis with an initial crack of a0, and obtain crack length and geometric correction factor (a, β) data; where a0 is the apparent equivalent initial crack length of the key part, and the value is less than 0.1 mm. Step S2: Starting from a0, extract the strain response of virtual strain gauges under the same load state at certain crack length intervals to obtain strain and crack size datasets, establish a relationship model between crack length a and strain response ε, and obtain the "ε-a" model. Step S3: Based on the various flight status parameters recorded during the aircraft's service life, use the "flight parameter-component load" model and the "component load-stress" model to obtain the component load spectrum and the stress spectrum of key structural parts; Step S4: From the load spectrum of the component, combined with the flight parameter history, select the flight state and the load state with the same load size as the load state corresponding to the "ε-a" model, extract the measured strain data and the corresponding lifetime N near the key parts of the aircraft structure physical entity under these load states, and obtain the measured (ε, t) dataset. Step S5: Based on the ratio between the load values ​​corresponding to the (ε, t) dataset and the load values ​​corresponding to the "ε-a" model, the (ε, t) dataset is proportionally corrected to the values ​​under the same load, and compared with the virtual strain gauge response values ​​to correct the "ε-a" model. Step S6: Substitute the (ε, t) dataset into the "ε-a" model to calculate the (a, t) data based on strain monitoring; Step S7: Select a damage evolution model, extract the model parameters corresponding to the structural materials from the material property database, and determine the prior parameters of the model; the material property data should include the crack propagation threshold value ΔK. th Fracture toughness K IC ; Step S8: Using the (a, β) data from step S1, the stress spectrum from step S3, and the model prior parameters from step S7 as inputs, predict crack propagation by loading one load cycle after another to obtain the theoretical prediction (a', t') data. Step S9: Using the (a, β) data from step S1, the stress spectrum from step S3, the (a, t) data from step S6, the prior parameters of the model from step S7, and the theoretical prediction (a', t') data from step S8 as inputs, update the damage evolution model selected in step S7 to obtain the posterior parameters of the model and complete the damage evolution model update. Step S10: Update the structural digital twin with the latest (a,t) data from step S6, and integrate the updated damage evolution model into the digital twin. Using the (a,β) data from step S1, the stress spectrum from step S3, and the model parameters from step S9 as inputs, predict the crack propagation (a,t) data and remaining life of the key parts of the aircraft structure under the current damage state. Step S11: Based on the future crack propagation and remaining life predicted by the twin, assess the structural health status, and realize the reverse control of the physical entity by the twin by proposing suggestions to adjust the aircraft's operating weight or to carry out inspection and maintenance. Step S12: As the aircraft service data and measured strain data are continuously updated, repeat steps S3 to S10 to realize the updating of the aircraft structure digital twin and life prediction.

2. The method for updating and predicting the lifespan of digital twins of key components of an aircraft structure as described in claim 1, characterized in that, In step S1, the geometric correction factor (a, β) data already has a database of typical structural details and typical crack configurations in practical applications, and can be continuously enriched through three-dimensional crack propagation simulation, and can be quickly called up in the digital twin update.

3. The method for updating and predicting the lifespan of digital twins of key components of an aircraft structure as described in claim 1, characterized in that, In step S4, the measured strain data can be obtained using high-reliability strain gauges or fiber Bragg grating sensors in aircraft structural applications.

4. The method for updating and predicting the lifespan of digital twins of key aircraft structural components as described in claim 3, characterized in that, In step S4, the load state corresponding to the "ε-a" model is used in the field to periodically perform some standard flight take-offs and landings. Within the specified flight state parameter range, specific standard maneuvers are performed and maintained for a certain period of time to obtain relatively stable strain data. The changes in strain under similar loads are then used to identify structural cracks.

5. The method for updating and predicting the lifespan of digital twins of key components of an aircraft structure as described in claim 1, characterized in that, In step S6, (a,t) data can also be obtained through damage monitoring sensors or periodic non-destructive testing.

6. The method for updating and predicting the lifespan of digital twins of key components of an aircraft structure as described in claim 1, characterized in that, In step S8, the crack propagation prediction process is as follows: For each load cycle, if the stress intensity factor range ΔK of that cycle is less than the crack propagation threshold value ΔK... th If the crack does not propagate under the load cycle, the introduction of this criterion ensures that the crack initiation life can be predicted from the equivalent initial crack a0; if the stress intensity factor Kmax corresponding to the peak value is greater than the fracture toughness K... IC If this occurs, the structure will fracture and fail.

7. The method for updating and predicting the lifespan of digital twins of key components of an aircraft structure as described in claim 1, characterized in that, In step S9, the damage evolution model is updated. First, based on the (a,t) data obtained from strain monitoring in step S6 and the (a',t') data from theoretical prediction in step S8, a likelihood function is established between the two. Then, the posterior distribution function of the model parameters is established, and samples of the posterior parameters are extracted from the posterior distribution function.

8. The method for updating and predicting the lifespan of digital twins of key components of an aircraft structure as described in claim 7, characterized in that, The posterior distribution of the model parameters established by the damage evolution model update is not analytical for the random spectrum because the range of stress intensity factor ΔK for each load cycle is not monotonically changing but random. Therefore, the Markov chain Monte Carlo method can be used for sampling.

9. The method for updating and predicting the lifespan of digital twins of key components of an aircraft structure as described in claim 8, characterized in that, The Markov chain Monte Carlo method establishes reasonable acceptance and rejection criteria to obtain a stable Markov chain with as few samplings as possible. Whether it is a random spectrum or a constant amplitude spectrum, the criterion for accepting the next sample is that the prediction error of the crack size of the sample is less than that of the previous sample.

10. The method for updating and predicting the lifespan of digital twins of key components of an aircraft structure as described in claim 1, characterized in that, (a,t) data obtained through non-destructive testing or damage monitoring sensors can also be used for updating digital twins.

Images