A fleet maintenance digital twin diagnosis and prediction method, device and storage medium based on coupling function

Abstract

The application provides a fleet maintenance digital twin diagnosis and prediction method based on coupling function, equipment and storage medium. First, the dependence of the damage state of multiple structures in the fleet is modeled by using the coupling function to obtain an approximate joint probability distribution for fleet collaborative updating, and it is deployed as a coupling function-based update step integration into the particle filter framework, thereby allowing all structures in the fleet to be updated based on observations of a single structure, thereby improving the overall prediction accuracy of structural damage in the fleet and reducing uncertainty in the process. The application relates to an electronic device and a storage medium for executing a coupling function-based method to process a fleet maintenance digital twin-based diagnosis and prediction method. The application can be integrated into a fleet maintenance digital twin, which can improve the safety of overall fleet operation without increasing costs, or effectively reduce operation and maintenance costs while maintaining safety requirements.

Description

Technical Field

[0001] This invention belongs to the fields of structural life prediction and equipment operation and maintenance, and particularly relates to a method, device and storage medium for digital twin diagnosis and prediction of fleet maintenance based on coupling functions. Background Technology

[0002] In aerospace and mechanical structures, the initiation and propagation of fatigue cracks in critical components caused by cyclic loading can lead to structural failure and compromise structural integrity. To address this challenge, various methods have been adopted, ranging from safe life and failure safety to damage tolerance. Based on these methods, traditional fleet management employs a uniform approach to ensure structural integrity, meaning each aircraft uses the same inspection and maintenance plan, with little consideration given to the differences between aircraft.

[0003] Since the 1970s, Individual Aircraft Tracking (IAT) technology has been widely used. This technology tracks the damage status of each aircraft by considering differences in load and service history. This allows for the development of individual inspection and maintenance plans for each aircraft, but the method remains deterministic. Since 2010, building on IAT, aircraft structural digital twins have improved the level of structural damage diagnosis and prediction by creating multi-physics, multi-scale, and probabilistic virtual simulation models of the system. This model integrates multiple heterogeneous and uncertain information sources from models and data to support proactive fleet maintenance decisions.

[0004] Particle filtering (PF) has been widely used for damage diagnosis and prediction in digital twins of aerospace and mechanical structures because it can simulate non-Gaussian nonlinear processes containing cognitive and stochastic uncertainties. Particle filtering uses a set of random samples to approximate the probability distribution of states and parameters, and models the changes in states and parameters in a dynamic system through a state-space model. By replacing integral operations with the sample mean, a minimum variance estimate of the system state can be obtained. In particle filtering, crack propagation is predicted in real time through analysis or empirical formulas for simple cracks, or reduced-order models for complex cracks, and an observational model is obtained based on direct crack inspection or the use of structural health monitoring techniques (such as fiber Bragg gratings or Lamb waves). The fusion of the prediction and observational models enables the diagnosis and prediction of structural crack damage propagation processes.

[0005] However, current particle filter-based digital twin methods primarily focus on individual-level diagnosis and prediction within a fleet, paying little attention to the overall interconnectedness of the fleet. Due to the similarity of missions and environments, the damage states of various structures within a fleet are correlated, providing valuable information that is not considered in traditional individual-oriented particle filtering. While particle filtering can be used to directly model multiple structures, this is a high-dimensional problem requiring a large number of particles, making it difficult to apply in practice. To effectively consider the correlations between structures within a fleet and improve the overall accuracy of fleet diagnosis and prediction, an efficient alternative method needs to be developed. Summary of the Invention

[0006] This invention proposes a coupling function-based method to address the diagnostic and prediction problems based on fleet maintenance digital twins. The coupling function models the joint multivariate distribution of multiple variables by coupling the dependency structures of their respective one-dimensional marginal distributions. This invention innovatively utilizes the coupling function to model the dependencies of damage states among multiple structures within a fleet, obtaining an approximate joint probability distribution for fleet collaborative updates. This distribution is then deployed as a coupling function-based update step integrated into particle filtering, allowing updates to all structures in the fleet based on observations of individual structures. This improves the overall prediction accuracy of structural damage within the fleet and reduces uncertainties in the process. This invention can be integrated into fleet maintenance digital twins, improving overall fleet operational safety without increasing costs, or effectively reducing maintenance costs while maintaining safety requirements. Attached Figure Description

[0007] Figure 1 A schematic diagram of digital twin diagnosis and prediction for fleet maintenance based on coupling functions;

[0008] Figure 2 A flowchart of updating the damage state of other structures based on the coupling function;

[0009] Figure 3 A comparison of diagnostic and prediction results between coupling function-based methods and particle filter methods with separate updates;

[0010] Figure 4 The result of updating the extended parameters of the method proposed in this invention;

[0011] Figure 5 The results of extended parameter updates for the traditional particle filter method. Detailed Implementation

[0012] Figure 1 This is a schematic diagram of digital twin diagnosis and prediction for fleet maintenance based on coupling functions, which associates the structural damage states of individual structure 1 and individual structure 2 during their respective service cycles through coupling functions.

[0013] Specifically, this invention proposes a method based on coupling functions to process diagnostic and prediction methods based on fleet maintenance digital twins, including the following steps.

[0014] Step 1: Construct the initial structural digital twin model

[0015] Before deploying digital twins online for structural damage diagnosis and prediction, an initial digital twin model needs to be built offline. A structural digital twin model is a multidisciplinary, multi-level probabilistic simulation model that integrates aerodynamic, structural, and fatigue models within an uncertainty analysis framework, considering the uncertainties of model parameters. The prior distribution of model parameters is determined through preliminary experiments or engineering experience, resulting in an initial structural digital twin model. This model will be continuously updated with inspection data during online deployment, thereby improving the reliability of model predictions.

[0016] Step 2: Collect usage data for each structure within the fleet

[0017] First, the operational loads need to be acquired on each structure within the fleet to serve as input for the digital twin model. There are two methods for acquiring these loads. The first method involves collecting aircraft flight data, including speed, attitude angles, and attitude angular velocities, and then using aerodynamic simulation to obtain the external loads on the structure. The second method involves deploying sensors on the structure to collect information such as strain, and then combining this data with a load inversion algorithm to obtain the external loads. For the second method, the selection and placement of sensors must be carefully considered to acquire as much information as possible.

[0018] Step 3: Predict structural damage evolution based on structural digital twin model

[0019] Based on the applied load and external structural load collected in step 2, the evolution of structural fatigue damage can be predicted using a structural digital twin model. The fatigue crack propagation process for each structure is shown below:

[0020] For the fatigue crack propagation process, the state of the system is represented by the crack length *a*. The evolution of the crack length can be expressed as:

[0021]

[0022] in, The crack length increment for each load cycle, ΔK is the range of stress intensity factor (SIF), and μ is a material parameter considering uncertainties.

[0023] The augmented state vector x is defined using the crack length 'a' and the uncertain material parameter μ. k =[μ k ,ak The complete state-space model can be modeled as follows:

[0024]

[0025] y k =a k +η k (3)

[0026] Where k is the time step, y k It is the observed value of the crack length, ω μ,k It is the material parameter μ k Evolving noise term It is noise in the crack propagation process, and follows a Gaussian distribution. and η k It is the measurement noise, which follows a zero-mean Gaussian distribution, and ΔN is the load cycle step size.

[0027] In crack propagation prediction, it is necessary to consider the state variable x from the previous time step. k-1 The state transition between two adjacent time steps is used to predict the state vector x. k ;

[0028] p(x k ∣y 1:k-1 )=∫p(x k |x k-1 )p(x k-1 ∣y 1:k-1 )dx k-1 (4)

[0029] The prediction process in Equation 4 can be performed using various uncertainty modeling methods. When using particle filtering, the specific operation involves predicting the particle state at time step k for each particle according to Equation 2, and then calculating the particle distribution and mean.

[0030] Step 4 involves inspecting any structure within the fleet to obtain damage observation data.

[0031] When a structure within the fleet is scheduled for inspection, its current damage status (y) is obtained through visual inspection or non-destructive testing. k Visual inspection yields results with significant uncertainty, while non-destructive testing methods include eddy current testing, penetrant testing, and X-ray imaging. The obtained damage observation data will be used to update the damage state and model parameters in the structural digital twin.

[0032] Step 5: Update the damage state of the current structure using particle filtering.

[0033] When the observed value y k Use y when available kUpdate state variable x k Joint probability distribution:

[0034]

[0035] Where p(x) k ∣y 1:k-1 ) is the likelihood function of the observation model, p(y) k |x k )p(y k ∣y 1:k-1 ) is a normalization constant.

[0036] In many cases, the posterior probability density function (PDF) is difficult to obtain directly using explicit methods. When using particle filtering, the posterior PDF is approximated using the following formula;

[0037]

[0038] Where, N s δ is the number of particles in the particle filter, and δ is the Dirac function. It is the i-th particle. It is the importance weight of the i-th particle.

[0039] These particles originate from an importance density This density should be similar to the desired posterior PDF and p(x) k ∣y 1:k It is easy to sample. The most commonly used importance density distribution is It simplifies the weight update equation given by equation (7).

[0040]

[0041] in, Let be the likelihood function. Then, the weights are normalized using equation (8):

[0042]

[0043] By following the steps above, the damage status and model parameters of the structural digital twin model can be updated using the observed values ​​obtained from the inspection.

[0044] Particle filtering suffers from particle degeneration, where only a few particles have significant weights, while the weights of most particles are negligible. Therefore, after weight updates, a resampling step is typically performed to address particle degeneration; the basic idea is to discard particles with low weights and duplicate those with high weights. However, the resampling process can lead to particle scarcity, where particles become identical copies after several resampling steps. This can limit the possible crack growth paths in the filter, especially for model parameters. This will significantly affect the filtering and prediction results.

[0045] To address this issue, Regularized Particle Filtering (RPF) introduces a continuous approximation p(x) for the posterior PDF using the kernel density method. k ∣y 1:k As shown in equation (9).

[0046]

[0047] Among them, K h (·) is composed of The given scaled kernel function, where h is the kernel function bandwidth, and n... x This is the dimension of the state vector. After the resampling procedure, particles are randomly drawn from successive approximations of the posterior PDF. This regularization step is used to increase particle diversity, thereby preventing the particle depletion problem.

[0048] Step 6: Update the damage state of other structures based on the coupling function.

[0049] In step 6, the relationship between the damage states of the two structures in the fleet is approximated by a coupling function, thereby updating the damage state of the uninspected structure using the observations of the inspected structure. The particle distributions of the inspected structure and the structure to be updated at the current time step are extracted from the particle filter, respectively. These particle distributions include the crack size distribution and crack propagation parameter distribution. See the flowchart below. Figure 2 .

[0050] The coupling function C is a multivariate distribution that follows a uniform distribution over a unit interval. For a given cumulative marginal distribution F1,…,F… N The N-dimensional cumulative joint distribution function F(CDF) can be written as:

[0051] F(x1,…,x N )=C(F1(x1),…,F N (x N (10)

[0052] Therefore, the coupling function C and the marginal distributions of each dimension can be used to represent N as a joint distribution.

[0053] Step 6.1 Measure the similarity between the checked structures within the fleet and other structures.

[0054] When the predicted distributions of crack length and crack propagation parameters are comparable, the potential crack distributions of two individuals are likely correlated. Therefore, the similarity between any two structures within a fleet can be determined by measuring the similarity of the predicted crack length distributions and the distributions of crack propagation parameters between the two individuals. Various similarity metrics can be used to measure the difference between two distributions, such as maximum mean difference (MMD) and KL divergence. Taking MMD as an example, the expression for maximum mean difference is as follows:

[0055]

[0056] Where, x 1 and x 2 These are two distributions that are to be measured to be similar. It is the reproducing kernel Hilbert space (RKHS) under the kernel function φ mapping, in which k(x) 1 ,x 2 )=<φ(x 1 ,x 2 )>. for Mean embedding of the distribution p in the middle, If and only if p = q.

[0057] At each time step of the particle filter, the distribution of fatigue crack length and material parameters is available, so these parameters can be extracted to calculate the MMD distance.

[0058] Step 6.2 Model the joint distribution of the two crack sizes based on the Frank coupling function.

[0059] The Frank coupling function is used to model tailless, correlated random variables, determined by only one parameter, which facilitates the measurement of the correlation between two crack length distributions. Its definition is as follows:

[0060]

[0061] Here, u1 and u2 are the cumulative distributions of the crack lengths of the two structures, respectively. θ is the correlation parameter in the Frank coupling function.

[0062] Modeling the coupling function requires the cumulative distribution functions of two marginal distributions. However, in particle filtering, the crack length distribution is represented as discrete sample values ​​x1 and x2. To address this issue, a probability distribution fitting method can be chosen to fit the probability density function distribution, and the cumulative distribution function can be obtained through integration. Taking kernel density estimation (KDE) as an example, the form of kernel density estimation is described as follows:

[0063]

[0064] Where K is the kernel (a non-negative function), h>0 is the bandwidth, and is a smoothing parameter.

[0065] After obtaining the analytical representation of the probability density distribution, the cumulative distribution function value for each sample can be calculated by integrating along the x-axis. Furthermore, using cubic spline interpolation, the inverse function of the cumulative distribution function can be determined, where the cumulative distribution value is the input and the corresponding sample value is the output.

[0066] The correlation parameter of the coupling function can be obtained from the similarity metric given in step 6.1. Due to the uniform distribution property of the cumulative distribution function, random samples of the joint cumulative distribution function can be generated by sampling from the coupling function. These samples can then be transformed into new samples using the inverse function of the cumulative distribution function.

[0067] Step 6.3 Fit and store the correlation between crack length and parameters

[0068] In each structural digital twin model, a priori distributions of crack length and propagation parameters are given during initialization. With continuous checks and updates, the correlation between crack length and model parameters gradually increases. Directly combining crack size particles with propagation parameter particles after updating the crack size distribution via a coupling function results in the loss of this correlation. Therefore, it is necessary to measure and store the correlation before updating, so as to recover it after the update.

[0069] Given the potentially high dimensionality of crack length and propagation parameters, various coupling functions can be used to measure this similarity. Among these, the Gaussian coupling function is particularly convenient because it is easy to construct and fit a multidimensional joint distribution. The Gaussian coupling function calculates the correlation between crack length and parameters before the coupling update step and recovers the correlation after the coupling update step.

[0070] The formula for the Gaussian coupling function is as follows:

[0071]

[0072] Where, Φ -1 It is the inverse function of the standard normal distribution, and R is the covariance matrix.

[0073] The specific steps are as follows: First, fit the marginal distribution of each parameter and convert it to a cumulative distribution. Then, use the inverse function of the cumulative distribution of a Gaussian distribution to transform the cumulative distribution back into a Gaussian distribution. After converting the marginal distribution of each parameter to a Gaussian distribution, the covariance between each variable can be directly calculated to generate the correlation matrix. Here, converting each marginal distribution to a Gaussian distribution is only used to fit the correlation matrix of the Gaussian coupling function.

[0074] Step 6.4 Update the damage state of the unchecked structure based on the approximate joint probability distribution

[0075] After obtaining the joint distribution of crack size distributions of the inspected structure and the structure to be updated through step 6.3, the posterior estimates of crack length distributions of unobserved individuals can be updated using the observations of the inspected structure in the same way as in step 5. First, the weight of each particle in the joint distribution is calculated based on the observations of the inspected structure. Then, the joint posterior distribution of the particles is updated through a resampling operation. Finally, the edge distribution of crack size in the structure to be updated is extracted from the joint posterior distribution.

[0076] Step 6.5 Restore the correlation between crack length and parameters after the update.

[0077] The correlation matrix calculated in step 6.3 is used to recover the correlation between crack length and parameters in the posterior particles of the structure to be updated, ensuring that they remain correlated in the posterior particles.

[0078] Step 7 continues forecasting until structural repair or replacement requirements are met.

[0079] Predict the damage evolution of each structure, develop a maintenance plan for each structure, repeat steps 3 to 6, schedule inspections for each structure, and use the inspection results to update the damage status of other structures until the structures meet the requirements for structural repair or replacement.

[0080] Example

[0081] In this embodiment, consider a simple infinite plate with a crack propagating from a hole edge. The plate is subjected to bidirectional uniform positive pressure. The formula for calculating the stress intensity factor range ΔK is:

[0082]

[0083] Where Δσ is the stress range and a is the crack length.

[0084] Paris's law is used as the crack propagation model:

[0085]

[0086] in, The crack length increment for each load cycle, C and m are material parameters in Paris's law, which are considered to be uncertain parameters.

[0087] In this embodiment, a simple heuristic method is used to convert the similarity measure obtained from the maximum mean difference into the correlation parameter in the Frank coupling function, as shown in the following formula:

[0088] θ=θ0×(d0-(αd a +(1-α)dμ (17)

[0089] Where, d a To determine the similarity between the crack lengths of two structures, d μ To assess the similarity between the crack propagation parameters of the two structures, θ0 and d0 are hyperparameters that need to be adjusted, and α is a weighting factor used to balance d. a and d μ The weight.

[0090] In this embodiment, three specimens with different initial cracks and material parameters are considered, resulting in different crack propagation histories. The actual parameters of the three specimens are shown in Table 1. The examination of each specimen was performed under different load cycles and generated through random sampling. The parameter settings of the proposed method and the traditional particle filter are shown in Table 2.

[0091] Table 1 shows the actual parameters of the three hypothesized samples.

[0092]

[0093] Table 2 Parameter settings for this embodiment

[0094]

[0095] Diagnostic and predictive outcomes such as Figure 3 As shown, the proposed method is compared with the traditional particle filtering method. It can be seen that the reduction in structural uncertainty without observation is mainly due to the update based on the coupling function, which also reduces the uncertainty throughout the crack propagation process. Table 3 also shows that the prediction accuracy has been improved.

[0096] Table 3 compares the prediction errors (RMSE) of the three samples used.

[0097]

[0098] The updated crack propagation material parameters obtained based on the method proposed in this invention and the traditional particle filtering method are as follows: Figure 4 and Figure 5 As shown, the results of the proposed method are in excellent agreement with traditional particle filtering methods. The Gaussian coupling function effectively captures the correlation between crack length and the distribution of propagation parameters during crack propagation. Therefore, the uncertainty of crack propagation parameters is significantly reduced when updated using the inspection results of the structure itself.

[0099] The coupling functions include, but are not limited to, Frank coupling functions, Clayton coupling functions, Gumbel coupling functions, Gaussian coupling functions, etc.

[0100] Similarity measures of distributions include, but are not limited to, maximum mean difference and KL divergence.

[0101] Structural inspection methods include, but are not limited to, visual inspection, eddy current testing, and penetrant testing.

[0102] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A fleet maintenance digital twin diagnosis and prediction method based on coupling functions, characterized in that, Includes the following steps, Step 1. Construct an initial structural digital twin model; Step 2. Collect usage data for each structure within the fleet; Step 3. Predict structural damage evolution based on structural digital twin model; Step 4. Inspect any structure within the fleet and obtain damage observation data; Step 5. Update the damage state of the current structure using particle filtering; Step 6. Update the damage state of other structures based on the coupling function; Step 7. Continue forecasting until structural repair or replacement is required; In step 6, the relationship between the damage states of the two structures in the fleet is approximated by the coupling function, so that the damage state of the uninspected structure is updated using the observations of the inspected structure. The particle distributions of the inspected structure and the structure to be updated at the current time step are extracted from the particle filter, and the particle distributions here include the structural crack size distribution and the crack propagation parameter distribution. Coupling function It is a multivariate distribution that follows a uniform distribution over a unit interval, and has a cumulative marginal distribution. of dimensional cumulative joint distribution function , written as The N-dimensional joint distribution is represented using a coupling function C and marginal distributions of each dimension; specifically, it includes the following steps: Step 6.1 Measure the similarity between the checked structures within the fleet and other structures; The similarity between any two structures in the fleet is determined by measuring the similarity of the predicted crack length distribution and the distribution of crack propagation parameters between two individuals. The maximum mean difference method is used to measure the difference between the two distributions. The expression for the maximum mean difference is as follows: in, and These are two distributions that are to be measured to be similar. It is a kernel function The regenerating kernel Hilbert space RKHS under mapping, in which, , for Medium distribution Mean embedding, If and only if At each time step of particle filtering, the fatigue crack length and the distribution of material parameters are extracted to calculate the MMD distance; Step 6.2 Model the joint distribution of the two crack sizes based on the Frank coupling function; The Frank coupling function is used to model tailless random variables and is determined by only one parameter, which is convenient for measuring the correlation between two crack length distributions. Its definition is as follows: in, and These are the cumulative distributions of the crack lengths in the two structures, respectively. It is the correlation parameter in the Frank coupling function; Step 6.3 Fit and store the correlation between crack length and parameters; The correlation between crack length and parameters is calculated using a Gaussian coupling function before the coupling update step, and the correlation is restored after the coupling update step. The formula for the Gaussian coupling function is as follows: in, It is the inverse function of the standard normal distribution. It is the covariance matrix; Step 6.4 Update the damage state of the unchecked structure based on the approximate joint probability distribution; After obtaining the joint distribution of crack size distribution of the inspected structure and the structure to be updated through step 6.3, the posterior estimate of crack length distribution of the unobserved individuals is updated using the observations of the inspected structure in the same way as in step 5. Step 6.5 Restore the correlation between crack length and parameters after the update; The correlation matrix calculated in step 6.3 is used to recover the correlation between crack length and parameters in the posterior particles of the structure to be updated, ensuring that they remain correlated in the posterior particles.

2. The method for fleet maintenance digital twin diagnosis and prediction based on coupling functions according to claim 1, characterized in that, Step 1 includes, In the offline phase, an initial structural digital twin model is constructed. The structural digital twin model integrates aerodynamic, structural, and fatigue models into an uncertainty analysis framework and considers the uncertainty of model parameters. The prior distribution of model parameters is determined through previous experiments or engineering experience.

3. The method for fleet maintenance digital twin diagnosis and prediction based on coupling functions according to claim 1, characterized in that, Step 2 includes acquiring the used loads and external loads on each structure within the fleet, which serve as inputs to the structural digital twin model.

4. The method for fleet maintenance digital twin diagnosis and prediction based on coupling functions according to claim 3, characterized in that, The specific implementation method of step 3 is as follows: Based on the usage load and external structural load collected in step 2, the evolution of structural fatigue damage is predicted using a structural digital twin model. The fatigue crack propagation process for each structure is shown below: For the fatigue crack propagation process, the state of the system is represented by the crack length. The evolution of crack length is represented as follows: in, The crack length increment for each load cycle. It is the range of stress intensity factor (SIF). These are material parameters that take into account uncertainties; Use crack length With uncertain material parameters Define augmented state vector The complete state-space model is modeled as follows: in, For time steps, It is the observed value of the crack length. Material parameters Evolving noise term It is noise in the crack propagation process, and follows a Gaussian distribution. ,and , It is measurement noise, which follows a zero-mean Gaussian distribution. This is the load cycle step size; In crack propagation prediction, based on the state variables of the previous time step State transitions between two adjacent time steps to predict the state vector , in, It is the likelihood function of the observation model; When using particle filtering, predict the time step for each particle according to Equation 2. The particle state is then determined, and the particle distribution and mean are calculated.

5. The method for fleet maintenance digital twin diagnosis and prediction based on coupling functions according to claim 4, characterized in that, Step 4 includes, When inspecting a structure within the fleet, the current state of damage to the structure is obtained through visual inspection or non-destructive testing. .

6. The method for fleet maintenance digital twin diagnosis and prediction based on coupling functions according to claim 5, characterized in that, The specific implementation method of step 5 is as follows: When observed value Use when available Update state variables Joint probability distribution: in, It is the likelihood function of the observation model. It is a normalization constant. When using particle filtering, the posterior PDF is approximated using the following formula: in, It is the number of particles in particle filtering. It is the Dirac function. It is the first One particle, It is the first The importance weight of each particle These particles originate from an importance density This density is similar to the desired posterior PDF and Easy to sample, importance density distribution adopts It simplifies the weight update equation given by equation (7). in, The likelihood function is then used to normalize the weights using equation (8): By following the steps above, the damage status and model parameters of the structural digital twin model can be updated using the observed values ​​obtained from the inspection.

7. The method for fleet maintenance digital twin diagnosis and prediction based on coupling functions according to claim 6, characterized in that, Step 5 also includes the following steps: Regularized Particle Filter (RPF) uses the kernel density method to introduce a continuous approximation of the posterior PDF. As shown in equation (9), in, It is by The given scaled kernel function, It is the kernel function bandwidth. It is the dimension of the state vector.

8. The method for fleet maintenance digital twin diagnosis and prediction based on coupling functions according to claim 1, characterized in that, Step 7 includes, Predict the damage evolution of each structure, develop a maintenance plan for each structure, repeat steps 3 to 6, schedule inspections for each structure, and use the inspection results to update the damage status of other structures until the structures meet the requirements for structural repair or replacement.

9. An electronic device, characterized in that... include: processor; Memory; And a program, wherein the program is stored in the memory and configured to be executed by a processor, the program comprising methods for performing any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor according to any one of claims 1-8.

Images