Flight safety assessment method and device based on icing condition inversion and uncertainty quantification

Abstract

The invention relates to a flight safety assessment method and device based on icing condition inversion and uncertainty quantification. The method comprises the following steps: acquiring icing observation data on the surface of an aircraft; generating ice type prediction data based on the forward prediction model; defining an observation difference index through the icing observation data and the ice type prediction data; modeling an icing condition inversion problem as a Bayesian inversion problem for solving the flight environment parameter posterior probability distribution; performing adaptive iterative sampling in a parameter space through an adaptive Monte Carlo sampling algorithm to generate posterior probability distribution of the flight environment parameters; based on the posterior probability distribution, constructing an uncertainty quantification result of the icing condition; and generating a flight safety assessment result through the uncertainty quantification result. According to the method, the reliability and robustness of aircraft icing simulation and prediction can be remarkably improved, and the method has wide application value in aircraft design, airworthiness pre-analysis and icing test planning.

Description

Technical Field

[0001] This application relates to the field of computer information processing, and more specifically, to a flight safety assessment method and apparatus based on icing condition inversion and uncertainty quantification. Background Technology

[0002] During flight, when the outside air temperature is low and contains supercooled water droplets, icing can easily occur on parts of the aircraft, such as the wings, tail, and engine intakes. Icing not only alters the aerodynamic shape of the wings, disrupts airflow adhesion, reduces lift, and increases drag, but can also lead to engine failure, instrument errors, and even flight accidents. Therefore, accurately predicting and analyzing the aircraft icing process and its uncertainties is of great significance for ensuring flight safety and designing anti-icing and de-icing systems.

[0003] Currently, research on aircraft icing mainly includes two directions: the forward problem and the inverse problem.

[0004] The direct problem refers to predicting the morphology and thickness distribution of ice formation based on known environmental conditions (such as droplet diameter, water content, angle of attack, velocity, and temperature) using physical models or numerical simulation methods. Common methods include ice formation simulation calculations based on heat balance equations, droplet collision models, and freezing fraction models.

[0005] The inverse problem involves inferring environmental parameters that might lead to a given icing pattern based on known icing results (such as experimentally observed ice formations or images of icing). This process involves solving high-dimensional nonlinear equations, which present challenges such as multiple solutions, noise sensitivity, and uncertainty propagation.

[0006] Existing inverse problem research often relies on deterministic optimization methods (such as least squares inversion, gradient descent, and evolutionary algorithms). These methods typically depend on initial values, are prone to getting trapped in local optima, and struggle to obtain the probability distribution information of input parameters, failing to fully reflect the sources and propagation paths of uncertainty in icing conditions. Furthermore, traditional icing models exhibit significant predictive volatility under environmental parameter perturbations, lack robustness and reliability assessment mechanisms, making it difficult to quantify the stability of models during validation. In recent years, uncertainty quantification methods based on Bayesian inference have been introduced into complex engineering systems. However, they still cannot solve the shortcomings of existing technologies in aircraft icing condition inversion and uncertainty analysis, such as insufficient accuracy, low model stability, and difficulty in quantifying uncertainty.

[0007] The information disclosed in the background section is only intended to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0008] In view of this, this application provides a flight safety assessment method and apparatus based on icing condition inversion and uncertainty quantification, which can significantly improve the reliability and robustness of aircraft icing simulation and prediction, and has broad application value in aircraft design, airworthiness pre-analysis and icing test planning.

[0009] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0010] According to one aspect of this application, a flight safety assessment method based on icing condition inversion and uncertainty quantification is proposed. The method includes: acquiring icing observation data of the aircraft surface; generating icing pattern prediction data based on a forward prediction model, wherein the forward prediction model generates the icing pattern prediction data through denoising diffusion; defining an observation difference index between the icing observation data and the icing pattern prediction data; modeling the icing condition inversion problem as a Bayesian inversion problem of solving the posterior probability distribution of the flight environment parameters, wherein the observation difference index serves as the likelihood function of the Bayesian inversion problem; generating the posterior probability distribution of the flight environment parameters through adaptive iterative sampling in the parameter space using an adaptive Monte Carlo sampling algorithm; constructing an uncertainty quantification result of the icing conditions based on the posterior probability distribution; and generating a flight safety assessment result using the uncertainty quantification result.

[0011] In one exemplary embodiment of this application, the method further includes: constructing the forward prediction model based on a physics-driven conditional diffusion generation model; wherein the inputs to the forward prediction model are flight environment parameters and wing reference profiles, the flight environment parameters including: flight altitude, angle of attack, airflow speed, droplet diameter, droplet water content and ambient temperature, and the output of the forward prediction model is ice type prediction data.

[0012] In one exemplary embodiment of this application, generating ice pattern prediction data based on a forward prediction model includes: generating initial sampled values ​​of flight environment parameters; and inputting the initial sampled values ​​into the forward prediction model to generate the ice pattern prediction data.

[0013] In one exemplary embodiment of this application, defining an observation difference index by means of the icing observation data and the ice pattern prediction data includes: defining an image-level observation difference index by means of the icing observation data and the ice pattern prediction data, wherein the observation difference index includes: RMSE, SSIM, and DICE coefficient.

[0014] In one exemplary embodiment of this application, the icing condition inversion problem is modeled as a Bayesian inversion problem of solving the posterior probability distribution of the flight environment parameters, including: defining the objective function of the Bayesian inversion problem as:

[0015] P(θ∣D)∝P(D∣θ)P(θ)

[0016] Where θ represents the flight environment parameter to be inverted, D is the icing observation data, P(D|θ) is the likelihood function, and P(θ) represents the prior distribution of the flight environment parameter.

[0017] In one exemplary embodiment of this application, the posterior probability distribution of the flight environment parameters is generated by adaptive iterative sampling in the parameter space using an adaptive Monte Carlo sampling algorithm, including: iterative sampling in the parameter space using an adaptive Monte Carlo sampling algorithm; dynamically updating the mean and variance of the proposed distribution using historical sample distributions to achieve dynamic adaptive adjustment of the sampling region; and obtaining the posterior probability distribution of the flight environment parameters when the adaptive Monte Carlo sampling process meets the convergence condition.

[0018] In one exemplary embodiment of this application, when the adaptive Monte Carlo sampling process meets the convergence condition, the posterior probability distribution of the flight environment parameters is obtained, including: calculating the weight of each sampling point according to the likelihood function and performing normalization processing; and resampling high-confidence region samples according to the sample weights to gradually approximate the true posterior probability distribution.

[0019] In one exemplary embodiment of this application, constructing the uncertainty quantification result of icing conditions based on the posterior probability distribution includes: based on the posterior probability distribution, statistically analyzing the mean, variance, and confidence interval of each parameter in the flight environment parameters to construct the uncertainty quantification result of aircraft icing conditions.

[0020] In one exemplary embodiment of this application, generating flight safety assessment results through the uncertainty quantification results includes: generating a flight route planning map through the uncertainty quantification results; and / or generating a weather avoidance route through the uncertainty quantification results; and / or generating anti-icing structure optimization data through the uncertainty quantification results; and / or generating icing condition judgment indicators through the uncertainty quantification results.

[0021] According to one aspect of this application, a flight safety assessment device based on icing condition inversion and uncertainty quantification is proposed. The device includes: an observation module for acquiring icing observation data of the aircraft surface; a prediction module for generating icing pattern prediction data based on a forward prediction model, wherein the forward prediction model generates the icing pattern prediction data through denoising diffusion; an index module for defining an observation difference index between the icing observation data and the icing pattern prediction data; a modeling module for modeling the icing condition inversion problem as a Bayesian inversion problem of solving the posterior probability distribution of the flight environment parameters, wherein the observation difference index serves as the likelihood function of the Bayesian inversion problem; an iteration module for generating the posterior probability distribution of the flight environment parameters through adaptive iterative sampling in the parameter space using an adaptive Monte Carlo sampling algorithm; a quantification module for constructing an uncertainty quantification result of the icing conditions based on the posterior probability distribution; and an evaluation module for generating a flight safety assessment result based on the uncertainty quantification result.

[0022] According to one aspect of this application, an electronic device is provided, comprising: one or more processors; a storage device for storing one or more programs; and, when the one or more programs are executed by the one or more processors, causing the one or more processors to implement the method as described above.

[0023] According to one aspect of this application, a computer-readable medium is provided having a computer program stored thereon that, when executed by a processor, implements the method described above.

[0024] According to the flight safety assessment method and apparatus based on icing condition inversion and uncertainty quantification of this application, the following steps are taken: Icing observation data of the aircraft surface is acquired; ice pattern prediction data is generated based on a forward prediction model, wherein the forward prediction model generates the ice pattern prediction data through denoising diffusion; an observation difference index is defined between the icing observation data and the ice pattern prediction data; the icing condition inversion problem is modeled as a Bayesian inversion problem to solve the posterior probability distribution of the flight environment parameters, wherein the observation difference index is used as the likelihood function of the Bayesian inversion problem; adaptive iterative sampling is performed in the parameter space using an adaptive Monte Carlo sampling algorithm to generate the posterior probability distribution of the flight environment parameters; uncertainty quantification results of icing conditions are constructed based on the posterior probability distribution; and flight safety assessment results are generated from the uncertainty quantification results. This method can significantly improve the reliability and robustness of aircraft icing simulation and prediction, and has broad application value in aircraft design, airworthiness pre-analysis, and icing test planning.

[0025] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0026] The above and other objects, features, and advantages of this application will become more apparent from the detailed description of exemplary embodiments with reference to the accompanying drawings. The drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0027] Figure 1 This is a flowchart illustrating a flight safety assessment method based on icing condition inversion and uncertainty quantification, according to an exemplary embodiment.

[0028] Figure 2 This is a schematic diagram illustrating the result of the posterior distribution of parameters during the inversion process, according to another exemplary embodiment.

[0029] Figure 3 This is a statistical table of inversion errors for various environmental parameters, illustrated according to another exemplary embodiment, used to quantify the accuracy of uncertainty analysis.

[0030] Figure 4 This is a convergence diagram of an adaptive Monte Carlo sampling algorithm according to another exemplary embodiment.

[0031] Figure 5 This is a comparison result of predicted ice type and measured ice type, and a table of main evaluation indicators, shown according to another exemplary embodiment.

[0032] Figure 6 This is a block diagram illustrating a flight safety assessment device based on icing condition inversion and uncertainty quantification according to an exemplary embodiment.

[0033] Figure 7 This is a block diagram illustrating an electronic device according to an exemplary embodiment.

[0034] Figure 8 This is a block diagram illustrating a computer-readable medium according to an exemplary embodiment. Detailed Implementation

[0035] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the embodiments set forth herein; rather, they are provided so that this application will be thorough and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted.

[0036] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0037] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0038] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0039] It should be understood that although the terms first, second, third, etc., may be used herein to describe various components, these components should not be limited by these terms. These terms are used to distinguish one component from another. Therefore, the first component discussed below may be referred to as the second component without departing from the teachings of this application. As used herein, the term "and / or" includes all combinations of any one and more of the associated listed items.

[0040] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of exemplary embodiments, and the modules or processes in the drawings are not necessarily essential for implementing this application, and therefore cannot be used to limit the scope of protection of this application.

[0041] The relevant technical terms used in this application are explained below:

[0042] Aircraft ice shape inversion refers to the process of inferring environmental conditions (such as angle of attack, air temperature, liquid water content, ice crystal size, etc.) from observed ice shape maps (e.g., binary ice shape maps) and wing geometry. This application embeds a high-fidelity conditional diffusion generation model into an adaptive Bayesian inversion framework, achieving not only parameter inversion but also outputting its complete posterior distribution, thereby quantifying the sources and propagation paths of environmental condition uncertainties.

[0043] The Forward Ice Shape Model (FEM) generates ice shape predictions from an input environmental parameter vector θ and a wing reference profile using a conditional diffusion generative model. By learning from large-scale real / simulated ice shape data, this model can generate pixel-level accurate, structurally continuous ice shape images that strictly adhere to wing geometric constraints. Compared to traditional numerical simulations, this model boasts high computational efficiency and strong nonlinear fitting capabilities, making it a stable and reliable forward mapping module in Bayesian inversion.

[0044] Adaptive Monte Carlo Sampling: A stochastic sampling method for solving high-dimensional parametric inverse problems. This application adaptively adjusts the covariance matrix of the proposal distribution, making the sampling more concentrated in the high posterior probability region, thereby accelerating convergence and improving inversion accuracy. Simultaneously, by calculating the posterior distribution and confidence interval of the sampling results, the uncertainty of icing conditions is quantified.

[0045] Likelihood Function: Used to measure the degree of matching between the predicted ice pattern and the observed ice pattern. In this application, pixel-level RMSE or squared error is used to calculate the difference in ice patterns, and it is converted into a Gaussian likelihood function to provide a basis for posterior probability calculation.

[0046] Prior Distribution: Physically probable assumptions about the ice type input parameters before inversion. In this application, the prior distribution can be a Gaussian distribution or a uniform distribution, combined with historical meteorological data or physical constraints to ensure that sampling is carried out within a reasonable parameter range.

[0047] Posterior Distribution: The parameter distribution obtained through inversion, which integrates prior information and the likelihood function, is used to quantify the uncertainty of environmental parameters. This application generates a large number of posterior samples through adaptive Monte Carlo sampling and performs statistical analysis on them to obtain the mean, variance, and confidence intervals.

[0048] Ice Shape Metrics: To evaluate the inversion accuracy, this application uses RMSE, SSIM, PSNR and DICE coefficients to compare the predicted ice shape with the observed ice shape at the pixel level and structure level, and can be used to guide the weight adjustment during the sampling process.

[0049] Figure 1 This is a flowchart illustrating a flight safety assessment method based on icing condition inversion and uncertainty quantification according to an exemplary embodiment. The flight safety assessment method 10 based on icing condition inversion and uncertainty quantification includes at least steps S102 to S214.

[0050] like Figure 1 As shown, in S102, icing observation data of the aircraft surface is acquired. Icing observation data may be, for example, image data, and image data will be used as a specific example in the following description. It is understood that icing observation data may also be other types of data, and this application is not limited thereto.

[0051] In S104, ice pattern prediction data is generated based on a forward prediction model, which generates the ice pattern prediction data through denoising diffusion. For example, initial sampled values ​​of flight environment parameters can be generated; these initial sampled values ​​are then input into the forward prediction model to generate the ice pattern prediction data.

[0052] In one embodiment, the method further includes: constructing the forward prediction model based on a physics-driven conditional diffusion generation model; wherein the inputs to the forward prediction model are flight environment parameters and wing reference profile, the flight environment parameters including: flight altitude, angle of attack, airflow speed, droplet diameter, droplet water content and ambient temperature, and the output of the forward prediction model is ice type prediction data.

[0053] Based on the known mapping relationship between physical parameters and ice type results, a positive prediction model is constructed. Where θ = {V, α, T, MVD, LWC, H, Time} represents the set of input parameters. This represents a binary image indicating the predicted ice type.

[0054] A binary image is defined as:

[0055]

[0056] A conditional diffusion generation model based on the U-Net architecture is adopted as the forward prediction model. This model takes the normalized environmental parameter vector and the initial contour map of the wing surface as dual conditional inputs. By learning the distribution of large-scale real or simulated ice pattern data, it gradually generates pixel-level ice pattern images that conform to physical laws during the denoising process.

[0057] After the model is trained, input the given set of environmental parameters θ = {V, α, T, MVD, LWC, H, Time} into the model to obtain the predicted ice type.

[0058] The ice pattern prediction images can be used for flight safety assessments, identifying lift loss risks in advance by observing the coverage of ice patterns on the airfoil leading edge; secondly, they can be used for aircraft aerodynamic design optimization, assisting engineers in determining the layout and geometric sensitivity of anti-icing systems; in addition, they can be used for airworthiness verification preparation, serving as pre-analysis data to reduce wind tunnel testing costs.

[0059] In another specific embodiment, a positive mapping relationship from icing environment parameters to ice pattern distribution is established based on a diffusion model prediction network. These environmental parameters include, but are not limited to, flight altitude, angle of attack, airflow speed, droplet diameter, droplet water content, and ambient temperature; the output is the icing morphology on the wing surface.

[0060] In S106, an observation difference index is defined using the icing observation data and the ice pattern prediction data. For example, an image-level observation difference index can be defined using the icing observation data and the ice pattern prediction data. This observation difference index includes RMSE, SSIM, and the DICE coefficient. Alternatively, for example, ice pattern observation data can be collected from icing wind tunnel experiments or actual flight tests, and the positive model prediction results can be compared with the measured results; image difference indices (such as RMSE, SSIM, and the DICE coefficient) can be defined to characterize the error features between the model prediction and the observation results.

[0061] In one specific embodiment, a binary image of the wing icing pattern can be obtained from wind tunnel experiments or simulations. obs Define the error measurement function:

[0062]

[0063] RMSE measures pixel-level differences

[0064] This error function is used to define the posterior likelihood function:

[0065]

[0066] In S108, the icing condition inversion problem is modeled as a Bayesian inversion problem of solving the posterior probability distribution of the flight environment parameters, wherein the observation difference index is used as the likelihood function of the Bayesian inversion problem.

[0067] In a specific embodiment, the problem of icing parameter inversion is transformed into a Bayesian framework:

[0068] P(θ∣I obs )∝P(I obs |θ)P(θ);

[0069] Where θ represents the icing environment parameters to be inverted, D represents the observed ice type data, P(D|θ) is the likelihood function, and P(θ) is the prior distribution of the parameters.

[0070]

[0071] Where σ represents the standard deviation of the error term, used to control the likelihood width.

[0072] In S110, an adaptive iterative sampling algorithm is used in the parameter space to generate the posterior probability distribution of the flight environment parameters. For example, iterative sampling is performed in the parameter space using an adaptive Monte Carlo sampling algorithm; the mean and variance of the proposed distribution are dynamically updated using historical sample distributions to achieve dynamic adaptive adjustment of the sampling region; and the posterior probability distribution of the flight environment parameters is obtained when the adaptive Monte Carlo sampling process meets the convergence condition.

[0073] An adaptive Monte Carlo algorithm is used to sample in the parameter space. By dynamically adjusting the mean and covariance matrix of the proposal distribution, the sampling direction is gradually moved closer to the high probability density region, thereby improving sampling efficiency and convergence speed. The update formula is as follows:

[0074] Σ t+1 =(1-α)Σ t +αCov(θ 1:t )

[0075] Where, Σ t Let α be the covariance matrix of the current proposed distribution, and α be the adaptive adjustment coefficient.

[0076] The weight of each sample is calculated based on the likelihood function:

[0077]

[0078] Posterior probability estimation is achieved through normalization, and a resampling mechanism is used to enhance the distribution density of samples in high-confidence regions in order to obtain more accurate posterior approximation results.

[0079] More specifically, when the adaptive Monte Carlo sampling process meets the convergence condition, the posterior probability distribution of the flight environment parameters is obtained, including: calculating the weight of each sampling point according to the likelihood function and performing normalization processing; resampling high-confidence region samples according to the sample weights to gradually approximate the true posterior probability distribution.

[0080] In one specific embodiment, the Adaptive Monte Carlo (MCMC) algorithm is used to sample the posterior distribution. In the initial stage, N sets of parameter samples θ are randomly generated. i And calculate its corresponding likelihood value.

[0081] In the t-th iteration, the proposal distribution is dynamically adjusted based on the covariance matrix of the samples from the previous stage:

[0082] Σ t+1 =(1-α)Σ t +α·Cov(θ 1:t )

[0083] Where α is the adaptive coefficient, which is generally taken as 0.01≤α≤0.1.

[0084] Each iteration is based on the acceptance probability:

[0085]

[0086] Perform an accept or reject update operation.

[0087] When the sample weight variance exceeds a set threshold, a resampling mechanism is triggered to maintain sample diversity. Systematic resampling or stratified resampling methods are employed to improve the stability of the posterior estimate.

[0088] In S112, the uncertainty quantification result of icing conditions is constructed based on the posterior probability distribution. The uncertainty quantification result of aircraft icing conditions can be constructed by statistically analyzing the mean, variance, and confidence interval of each parameter in the flight environment parameters based on the posterior probability distribution.

[0089] For example, by statistically obtaining the posterior sample distribution, the mean, variance, and confidence interval of each parameter can be calculated to quantitatively characterize the sensitivity of the input parameters to the ice pattern results; at the same time, the comparison results between the posterior predicted ice pattern map and the measured ice pattern map can be output to generate an uncertainty visualization image and stability assessment index.

[0090] Based on the final sample set {θ i ,w i} Calculate the posterior mean μ for each parameter. θ ,variance And confidence intervals. Generate corresponding predicted ice type maps for all samples. Calculate the average ice type With standard deviation distribution σ I This enables visualization of the uncertainty in ice-type results.

[0091] The output includes:

[0092] (1) Posterior distribution of parameters (such as probability density of velocity, angle of attack, etc.);

[0093] (2) Average ice type prediction map;

[0094] The uncertainty quantification results can be used to determine the most unfavorable weather conditions, supporting worst-case scenario analysis required for airworthiness assessments, as airworthiness evaluations often explore the worst ice type, such as maximum thickness and maximum surface area. Uncertainty quantification can provide upper and lower bounds and confidence intervals for the ice type, helping to identify the "most unfavorable conditions." Secondly, the maximum ice thickness protection capability assessment can be used to evaluate the sensitivity of various parameters, thereby helping engineers determine which physical quantities must be controlled more precisely and which can be relaxed, improving the cost-effectiveness of testing.

[0095] In S114, flight safety assessment results are generated using the uncertainty quantification results. The prediction models under different icing conditions are validated by combining the inverted posterior distribution; the stability and robustness of the models are evaluated through variance analysis, confidence interval overlap, and sensitivity indices, thereby achieving a comprehensive uncertainty analysis of the icing process.

[0096] For example, the uncertainty quantification results can be used to generate flight route planning maps; for example, the uncertainty quantification results can be used to generate weather avoidance routes; for example, the uncertainty quantification results can be used to generate anti-icing structure optimization data; for example, the uncertainty quantification results can be used to generate icing condition judgment indicators.

[0097] Based on the posterior distribution of the parameters obtained from sampling, a set of ice type prediction results is generated.

[0098] 1. Calculate the average ice type probability map:

[0099]

[0100] 2. Reflects the probability of ice forming at each pixel (0-1);

[0101] 3. Binarization yields the final predicted ice pattern map:

[0102]

[0103] 4. Draw an ice-type uncertainty map (variance heatmap for each pixel):

[0104]

[0105] Visualize the uncertainty of the periglacial region and reflect the sensitivity of the icing boundary.

[0106] The final output includes:

[0107] (1) Binary image of ice formation;

[0108] (2) Ice-type probability diagram and uncertainty heatmap;

[0109] The results can be used to back-calculate wind tunnel test parameters, such as guiding the setting of spray volume, nozzle temperature, and flight speed, thereby reducing the number of repeated tests. Secondly, it can be used for environmental factor analysis of abnormal icing events. Given an abnormal icing pattern (observed during flight testing), the most likely LWC, MVD, AOA, and other conditions obtained through inversion can be used for accident analysis and fault diagnosis. In addition, it can be used for real-time flight condition diagnosis. If airborne sensors detect the initial icing pattern on the surface, the current weather conditions (MVD, LWC) can be estimated through the inversion model, serving as a supplement to weather radar and weather forecasts, and helping pilots decide whether to leave the icing area.

[0110] More specifically, the ice type prediction results and the parameter distribution obtained by inversion not only have theoretical rigor, but can also be used for flight safety risk identification, analysis of icing sensitivity under different operating conditions, and auxiliary setting of icing test conditions in wind tunnels or flight tests.

[0111] In operational decision support, the optimal estimates and confidence intervals of parameters can upgrade the traditional "qualitative icing warning" to a "probabilistic risk assessment," used for refined route planning and weather avoidance. The uncertainty quantification results can support the determination of the most unfavorable icing conditions in airworthiness assessment and provide a reference for anti-icing structure optimization in the aerodynamic design phase. This invention achieves a unification of high-precision generation capability and uncertainty quantification capability through deep fusion of conditional diffusion generation model and Bayesian inversion mechanism, significantly improving the reliability and robustness of aircraft icing simulation and prediction, and possessing broad application value in aircraft design, airworthiness pre-analysis, and icing test planning.

[0112] According to the flight safety assessment method based on icing condition inversion and uncertainty quantification of this application, the following steps are taken: Icing observation data of the aircraft surface is acquired; ice pattern prediction data is generated based on a forward prediction model, wherein the forward prediction model generates the ice pattern prediction data through denoising diffusion; an observation difference index is defined between the icing observation data and the ice pattern prediction data; the icing condition inversion problem is modeled as a Bayesian inversion problem of solving the posterior probability distribution of the flight environment parameters, wherein the observation difference index is used as the likelihood function of the Bayesian inversion problem; adaptive iterative sampling is performed in the parameter space using an adaptive Monte Carlo sampling algorithm to generate the posterior probability distribution of the flight environment parameters; uncertainty quantification results of icing conditions are constructed based on the posterior probability distribution; and flight safety assessment results are generated through the uncertainty quantification results. This method can significantly improve the reliability and robustness of aircraft icing simulation and prediction, and has broad application value in aircraft design, airworthiness pre-analysis, and icing test planning.

[0113] It should be clearly understood that this application describes how specific examples are formed and used, but the principles of this application are not limited to any details of these examples. Rather, based on the teachings of the disclosure of this application, these principles can be applied to many other embodiments.

[0114] Compared with the prior art, the beneficial effects of this application are:

[0115] 1. Sampling accuracy and parameter estimation stability are significantly improved.

[0116] The adaptive Monte Carlo algorithm proposed in this application introduces a dynamic step size adjustment and likelihood adaptive weighting mechanism in parameter estimation, effectively reducing sampling bias caused by random walks. Experimental results show that in typical parameter inversion tasks, the average relative error of this method is reduced by approximately 47% compared to the traditional Monte Carlo algorithm. Specifically, the error in liquid water content (LWC) is reduced from 9.05% to 4.13%, the error in median particle size (MVD) is reduced from 10.62% to 8.18%, and the height error is reduced to 3.98%, significantly improving estimation accuracy and algorithm stability.

[0117] 2. Significant improvements in convergence speed and acceptance rate.

[0118] By introducing an adaptive acceptance threshold function, the algorithm exhibits greater exploratory activity in the early stages and automatically converges to a stable state in the later stages, making the sampling process more efficient. Compared to traditional algorithms, the acceptance rate is improved by over 20%, effectively reducing invalid sampling and enhancing overall computational efficiency.

[0119] 3. Enhanced robustness to complex parameter spaces

[0120] This application utilizes a multi-scale perturbation strategy and an adaptive variance adjustment mechanism to enhance the stability of the algorithm in nonlinear, multi-peak parameter spaces. In the estimation of highly sensitive parameters such as temperature and angle of attack (AOA), traditional methods have errors as high as 15%, while this application controls the error to within 10%, indicating that the method has stronger global optimization ability and noise resistance under complex conditions.

[0121] 4. The algorithm has strong versatility and scalability.

[0122] The algorithm presented in this application features a simple and highly modular structure, allowing it to be used in conjunction with various posterior estimation models and non-Gaussian noise environments. It is applicable to multiple fields, including flight meteorological parameter inversion, environmental monitoring, and engineering uncertainty analysis. Furthermore, because the sampling distribution can be dynamically adjusted, this algorithm can be transferred to different scenarios without resetting the initial step size or noise distribution, demonstrating superior generalization performance.

[0123] 5. Strong technological integration and innovation

[0124] This application combines adaptive probability correction, dynamic acceptance criteria, local perturbation search, and global likelihood guidance mechanism to form a self-closing adaptive Monte Carlo framework of "sampling-update-convergence". Compared with existing fixed-step Monte Carlo sampling methods, this application achieves breakthrough improvements in theoretical convergence, estimation accuracy, and computational efficiency, providing a new technical route for parameter inversion of complex systems.

[0125] 6. Deeply integrate high-fidelity generation with uncertainty quantification to achieve paradigm updates.

[0126] This application, for the first time, uses a conditional diffusion generation model based on the U-Net architecture as a differentiable, high-fidelity forward operator, and embeds an adaptive Bayesian inversion framework into the system, breaking through the limitations of "inefficient physical simulation" or "black-box neural networks" in traditional methods. This diffusion model generates physically consistent, pixel-level accurate ice-shaped images under dual conditions of environmental parameters and wing profile, providing a high signal-to-noise ratio benchmark for likelihood function construction. The adaptive Monte Carlo mechanism then efficiently explores the posterior distribution of parameters, achieving uncertainty quantification. The deep integration of these two approaches forms a closed-loop analysis paradigm of "high-precision generation—probabilistic inversion—uncertainty visualization," achieving breakthrough improvements in inversion accuracy, physical consistency, and reliability assessment capabilities compared to existing technologies.

[0127] In summary, this application overcomes the problems of traditional Monte Carlo sampling, such as being prone to getting trapped in local optima, slow convergence, and unstable parameters, through system optimization in four dimensions: accuracy, efficiency, robustness, and generalization. It significantly improves the accuracy and computational reliability of parameter inversion for complex systems and has high engineering application value and promising prospects for promotion.

[0128] The following specific embodiments illustrate the practical application of this application.

[0129] In one example, to address the difficulty in accurately inferring aircraft icing conditions and the uncertainty in icing pattern image prediction results, we first collect aircraft wing icing observation data obtained from experiments or numerical simulations, including binary images of icing patterns on the wing surface. obs And the corresponding set of environmental parameters θ={V,α,T,MVD,LWC,H,Time}.

[0130] 1. Data preprocessing:

[0131] Data normalization: Environmental parameters are normalized to map parameters of different dimensions to a unified numerical range [0,1], in order to ensure the numerical stability of subsequent neural network inputs and the efficiency of sampling algorithms.

[0132] Image size standardization: The size of the binary image of the ice pattern on the wing surface and the reference wing outline is adjusted so that all images are uniformly 136×136 pixels to meet the input requirements of the neural network.

[0133] Image tensor quantization: Converts the normalized image into a tensor format and performs channel adjustments (such as single-channel grayscale) to facilitate input into a convolutional neural network for ice-type image prediction.

[0134] Reference contour processing: The wing surface contour map is binarized to extract wing boundary information for use in forward physical model calculation and inversion prediction.

[0135] Through the above steps, a unified and standardized input dataset was established, ensuring that each environmental parameter vector corresponds to the wing icing pattern image and contour. Figure 1 This one-to-one correspondence lays the foundation for subsequent positive prediction model training and adaptive Monte Carlo inversion sampling, while improving the accuracy and stability of ice type inversion results.

[0136] System Configuration:

[0137] This application may include the following configuration and parameter settings:

[0138] (1) Basic Model:

[0139] Forward icing prediction model: A well-trained conditional diffusion generation model is used as the base model for forward icing prediction, which is used to obtain environmental parameter vector θ and wing reference profile diagram I. ref As a condition-guided generation of ice pattern prediction I pred .

[0140] Input conditions: The ice-shaped image is uniformly adjusted to 136×136 pixels, the wing reference outline is adjusted to 142×142 pixels, and converted to a single-channel grayscale image as a shape prior.

[0141] (2) Data preprocessing parameters:

[0142] Normalization: Each dimension of the environment parameter vector is normalized to the [0,1] interval using MinMaxScaler.

[0143] Binarization: Both the forward prediction results and the observed ice pattern images are binarized with a threshold of 0.5 to generate ice pattern images.

[0144] (3) Adaptive Monte Carlo sampling parameters:

[0145] Initial sample generation: from the prior Gaussian distribution N(E[θ]) i ],σ 2 ) Sample 500 initial samples to ensure that the sampled values ​​are within the physical boundary of the parameter [θ] min ,θmax ].

[0146] Number of iterations: Total number of MCMC samplings num_samples = M + N = 1500 (where M = 500 are the initial samples and N = 1000 are the valid samples).

[0147] Proposed distribution: Adaptive Gaussian distribution, initial standard deviation σ prop =1, and the standard deviation is dynamically adjusted based on the historical sample covariance during the iteration process:

[0148]

[0149] Acceptance rate calculation: Using Metropolis-Hastings acceptance rate:

[0150]

[0151] (4) Posterior distribution and error calculation

[0152] Likelihood function: Pixel-level RMSE calculation based on ice-shaped images:

[0153]

[0154] Prior distribution: based on parametric Gaussian distribution

[0155]

[0156] Posterior density: The posterior probability is posterior(θ) = L(I) pred ∣θ)·P(θ).

[0157] (5) Evaluation parameters

[0158] Pixel-level error: RMSE

[0159] Structural similarity: SSIM

[0160] Signal-to-noise ratio: PSNR

[0161] Ice type consistency coefficient: DICE

[0162] Confidence interval calculation: 95% confidence level, calculated based on later valid samples:

[0163]

[0164] (6) Visual parameters

[0165] Boundary drawing: Use skimage.measure.find_contours to extract ice-shaped boundaries.

[0166] Overlay display: The predicted ice shape map is overlaid with the observed ice shape map boundary. Red represents the observed boundary and blue represents the predicted boundary, which facilitates comparison and verification of the inversion results.

[0167] Image size: Display window size 6×6 inches, coordinate axes removed, maintaining a clear display of the binary ice map.

[0168] 3. Specific Implementation Steps

[0169] Step 1: Initialization

[0170] Input observation data on wing surface icing, including binary ice pattern image I. obs And the initial environment parameter vector θ0. Adjust the ice-shaped image to 136×136 pixels and convert it to a single-channel tensor to construct the initial state s0=(θ0,I ref ), where I ref For reference, a binary image of the wing profile is used. The environmental parameter vector is normalized to ensure that the values ​​of each dimension are in the range [0,1].

[0171] Step 2: Forward diffusion generation of ice-shaped training and prediction

[0172] This step constructs a high-fidelity, conditionally controllable forward ice-type prediction model, which serves as the core forward operator of the entire inversion framework. The model employs a conditional diffusion generation model, using environmental variables θ and the wing reference profile I. ref As a dual-condition input, combined with binarization processing (threshold 0.5), an ice-shaped image that meets the physical conditions is generated for comparison and error calculation in subsequent inversion steps.

[0173] Model infrastructure design

[0174] The base generator employs a U-Net-based denoising diffusion probability model, featuring multi-scale feature extraction and skip connections to preserve ice edge details. Six-dimensional icing environment parameters are mapped into high-dimensional embedding vectors using a dedicated multilayer perceptron projector. The wing reference contour map serves as a shape prior and is stitched together with the noisy image to form a dual-channel input, ensuring that the generated ice contour position remains consistent with the wing position.

[0175] diffusion model process

[0176] For the forward noise addition process, Gaussian noise is gradually added to the real ice pattern image within T=1000 steps. The specific noise addition formula is as follows:

[0177]

[0178] in This is the cumulative noise scheduling coefficient.

[0179] The U-Net is trained using a reverse process to predict the noisy residual at each step. The loss function uses the standard mean squared error (MSE):

[0180]

[0181] Dice loss can be optionally added during training to enhance the consistency of ice edge structures. The optimizer uses AdamW with a batch size of 32.

[0182] 200 rounds of training.

[0183] Inference and prediction process

[0184] Given environmental parameters and wing profile, from pure Gaussian noise Initially, 1000 steps of denoising are performed to generate continuous ice pattern images, which are then binarized with a threshold of 0.5 to obtain the final predicted ice pattern mask.

[0185] Step 3: Initial Sampling and Posterior Calculation

[0186] Based on the target inversion parameter θ in the initial environment parameter vector i The expected value E[θ i The initial adaptive Monte Carlo sampling set is generated using the prior standard deviation σ and σ. Ensure that the sampled values ​​are within the preset physical range (such as the extreme value range of airfoil parameters).

[0187] For each sampled sample:

[0188] Substitute the parameters into the forward model to generate an ice pattern prediction map.

[0189] Calculate the pixel-level error (RMSE) between the predicted ice pattern map and the observed ice pattern map.

[0190] Calculate the likelihood function based on the error and sample variance.

[0191] Multiply the likelihood value by the prior distribution to calculate the initial posterior probability density.

[0192] The sample with the highest posterior density is selected as the initial value θ. init Entering the MCMC iteration.

[0193] Step 4: Adaptive Monte Carlo Inversion Iteration

[0194] With initial value θ init As the current parameter, perform M+N adaptive Monte Carlo sampling operations.

[0195] Each iteration calculates the sampling covariance based on historical samples and dynamically adjusts the standard deviation of the proposed distribution to achieve adaptive sampling.

[0196] Generate proposal samples θ prop The Metropolis-Hastings acceptance rate formula is used to calculate whether to accept the proposal:

[0197]

[0198] If accepted, update the current sample to θ. prop Otherwise, retain θ. current .

[0199] Record the samples and corresponding posterior expected values ​​for each iteration to analyze convergence and sample stability.

[0200] Step 5: Posterior Analysis and Parameter Estimation

[0201] Take the effective samples from the later N iterations as the inversion result and calculate the sample mean. and standard deviation σ θ .

[0202] Calculate the confidence interval of the inversion parameters

[0203] Substituting the mean parameters into the forward physics model generates the final ice shape diagram. Binarization processing yields pixel-level ice-shaped distributions.

[0204] Step Six: Output Results and Visualization

[0205] Output the environmental parameters obtained from the final inversion. Confidence intervals and sample variances are used to quantify the uncertainty of freezing conditions.

[0206] Output ice type Figure 2 Value images, including ice thickness distribution and icing morphology on the wing surface.

[0207] Optional features include generating overlay visualizations of ice shape boundaries, comparing predicted ice shape boundaries with observed ice shape boundaries for experimental or flight safety assessments.

[0208] Step 7: Error and Indicator Evaluation

[0209] Calculate the pixel-level RMSE between the predicted ice pattern map and the observed ice pattern map.

[0210] Calculate the structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR).

[0211] Calculate the DICE coefficient of the binary map to evaluate the consistency between the predicted ice type and the actual observation.

[0212] In typical aircraft icing parameter inversion tasks, the icing uncertainty analysis method based on adaptive Monte Carlo inversion in this application achieved significant performance improvements in multiple sample tests. Specific results are as follows: Figures 3-5 As shown.

[0213] 4. Implementation Results

[0214] The method proposed in this application significantly improves the accuracy of inversion of key meteorological and dynamic parameters compared to the traditional Monte Carlo method.

[0215] like Figure 2 As shown, under the same initial conditions, the relative errors of the algorithm in this application on parameters such as liquid water content (LWC), median particle size (MVD), height, angle of attack (AOA), temperature, and velocity are reduced by approximately 54.4%, 22.6%, 10.6%, 33.1%, 48.3%, and 51.6%, respectively, and the overall mean square error is reduced by approximately 40%, indicating that the convergence stability and numerical accuracy of the algorithm in parameter estimation are significantly optimized.

[0216] Furthermore, compared with traditional fixed-step sampling methods, this algorithm can obtain a stable and reliable posterior distribution with fewer sampling times, and its posterior sample distribution is concentrated with small variance, proving the convergence and reliability of the estimation results; Figure 3 As shown, the adaptive Monte Carlo sampling algorithm proposed in this application introduces dynamic step size adjustment and likelihood adaptive weighting mechanism during the sampling process, achieving smooth and fast convergence.

[0217] Furthermore, in the validation of ice shape prediction (see...) Figure 4 The ice profile predicted by this application based on inversion parameters shows a high degree of consistency with the ice profile measured in the wind tunnel. Among the main evaluation indicators, the average shape matching error is less than 5%, and the boundary deviation is controlled within 2 mm. Compared with traditional inversion results, which have problems such as boundary discontinuity and thickness deviation, the ice profile generated by this method is more in line with physical laws and can effectively support the analysis of icing mechanism and the design of anti-icing system.

[0218] Comprehensive test results show that this application outperforms existing technologies in terms of parameter inversion accuracy, convergence efficiency, and consistency of ice type prediction, demonstrating the following advantages:

[0219] (1) The algorithm has higher parameter estimation accuracy and stability;

[0220] (2) The sampling process converges faster and has a higher acceptance rate;

[0221] (3) It has strong robustness in complex nonlinear parameter spaces;

[0222] (4) The method has good scalability and can be extended to other uncertainty inversion tasks.

[0223] Therefore, the adaptive Monte Carlo inversion method of this application is not only significantly superior to traditional techniques in terms of accuracy and efficiency, but also has reliable physical interpretability and engineering application value, and can provide a new and effective technical means for high-precision inversion and uncertainty assessment of aircraft icing characteristic parameters.

[0224] Those skilled in the art will understand that all or part of the steps of the above embodiments are implemented as a computer program executed by a CPU. When the computer program is executed by the CPU, it performs the functions defined by the method provided in this application. The program can be stored in a computer-readable storage medium, such as a read-only memory, a magnetic disk, or an optical disk.

[0225] Furthermore, it should be noted that the above figures are merely illustrative representations of the processes included in the method according to exemplary embodiments of this application, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0226] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0227] Figure 6 This is a block diagram illustrating a flight safety assessment device based on icing condition inversion and uncertainty quantification, according to an exemplary embodiment. Figure 6 As shown, the flight safety assessment device 60 based on icing condition inversion and uncertainty quantification includes: observation module 602, prediction module 604, index module 606, modeling module 608, iteration module 610, quantification module 612, and assessment module 614. The flight safety assessment device 60 based on icing condition inversion and uncertainty quantification may also include: construction module 616.

[0228] The observation module 602 is used to acquire icing observation data on the aircraft surface;

[0229] The prediction module 604 is used to generate ice pattern prediction data based on a forward prediction model, wherein the forward prediction model generates the ice pattern prediction data through denoising diffusion; the prediction module 604 is also used to generate initial sampled values ​​of flight environment parameters; and input the initial sampled values ​​into the forward prediction model to generate the ice pattern prediction data.

[0230] The indicator module 606 is used to define an observation difference index based on the icing observation data and the ice shape prediction data; the indicator module 606 is also used to define an image-level observation difference index based on the icing observation data and the ice shape prediction data, the observation difference index including: RMSE, SSIM, and DICE coefficient.

[0231] Modeling module 608 is used to model the icing condition inversion problem as a Bayesian inversion problem of solving the posterior probability distribution of the flight environment parameters, wherein the observation difference index is used as the likelihood function of the Bayesian inversion problem; modeling module 608 is also used to define the objective function of the Bayesian inversion problem as:

[0232] P(θ∣D)∝P(D∣θ)P(θ)

[0233] Where θ represents the flight environment parameter to be inverted, D is the icing observation data, P(D|θ) is the likelihood function, and P(θ) represents the prior distribution of the flight environment parameter.

[0234] The iteration module 610 is used to perform adaptive iterative sampling in the parameter space using an adaptive Monte Carlo sampling algorithm to generate the posterior probability distribution of the flight environment parameters; the iteration module 610 is also used to perform iterative sampling in the parameter space using an adaptive Monte Carlo sampling algorithm; dynamically update the mean and variance of the proposed distribution through the historical sample distribution to achieve dynamic adaptive adjustment of the sampling area; and obtain the posterior probability distribution of the flight environment parameters when the adaptive Monte Carlo sampling process meets the convergence condition.

[0235] The quantization module 612 is used to construct the uncertainty quantification result of icing conditions based on the posterior probability distribution; the quantization module 612 is also used to construct the uncertainty quantification result of aircraft icing conditions by statistically analyzing the mean, variance and confidence interval of each parameter in the flight environment parameters based on the posterior probability distribution.

[0236] The evaluation module 614 is used to generate flight safety evaluation results based on the uncertainty quantification results. The evaluation module 614 is also used to generate route planning maps based on the uncertainty quantification results; to generate weather avoidance routes based on the uncertainty quantification results; to generate anti-icing structure optimization data based on the uncertainty quantification results; and to generate icing condition judgment indicators based on the uncertainty quantification results.

[0237] The construction module 616 is used to construct the forward prediction model based on a physics-driven conditional diffusion generation model; wherein, the input of the forward prediction model is flight environment parameters and wing reference profile, the flight environment parameters include: flight altitude, angle of attack, airflow speed, droplet diameter, droplet water content and ambient temperature, and the output of the forward prediction model is ice type prediction data.

[0238] The flight safety assessment device based on icing condition inversion and uncertainty quantification according to this application acquires icing observation data of the aircraft surface; generates icing pattern prediction data based on a forward prediction model, wherein the forward prediction model generates the icing pattern prediction data through denoising diffusion; defines an observation difference index between the icing observation data and the icing pattern prediction data; models the icing condition inversion problem as a Bayesian inversion problem of solving the posterior probability distribution of the flight environment parameters, wherein the observation difference index is used as the likelihood function of the Bayesian inversion problem; generates the posterior probability distribution of the flight environment parameters through adaptive iterative sampling in the parameter space using an adaptive Monte Carlo sampling algorithm; constructs the uncertainty quantification result of icing conditions based on the posterior probability distribution; and generates flight safety assessment results through the uncertainty quantification result. This method can significantly improve the reliability and robustness of aircraft icing simulation and prediction, and has broad application value in aircraft design, airworthiness pre-analysis, and icing test planning.

[0239] Figure 7 This is a block diagram illustrating an electronic device according to an exemplary embodiment.

[0240] The following reference Figure 7 To describe an electronic device 700 according to this embodiment of the present application. Figure 7 The electronic device 700 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0241] like Figure 7 As shown, the electronic device 700 is presented in the form of a general-purpose computing device. The components of the electronic device 700 may include, but are not limited to: at least one processing unit 710, at least one storage unit 720, a bus 730 connecting different system components (including storage unit 720 and processing unit 710), a display unit 740, etc.

[0242] The storage unit stores program code that can be executed by the processing unit 710, causing the processing unit 710 to perform the steps described in this specification according to various exemplary embodiments of this application. For example, the processing unit 710 can perform actions such as... Figure 2 , Figure 3 , Figure 4 The steps are shown in the figure.

[0243] The storage unit 720 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 7201 and / or a cache storage unit 7202, and may further include a read-only memory unit (ROM) 7203.

[0244] The storage unit 720 may also include a program / utility 7204 having a set (at least one) program module 7205, such program module 7205 including but not limited to: an operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0245] Bus 730 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0246] Electronic device 700 can also communicate with one or more external devices 700' (e.g., keyboard, pointing device, Bluetooth device, etc.), enabling users to communicate with devices that interact with electronic device 700, and / or any device (e.g., router, modem, etc.) that allows electronic device 700 to communicate with one or more other computing devices. This communication can be performed via input / output (I / O) interface 750. Furthermore, electronic device 700 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 760. Network adapter 760 can communicate with other modules of electronic device 700 via bus 730. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 700, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0247] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software, or by combining software with necessary hardware. Therefore, as... Figure 8 As shown, the technical solution according to the embodiments of this application can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, mobile hard drive, etc.) or on a network, and includes several instructions to cause a computing device (such as a personal computer, server, or network device, etc.) to execute the above-described method according to the embodiments of this application.

[0248] The software product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections with one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0249] The computer-readable storage medium may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0250] Program code for performing the operations of this application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0251] The aforementioned computer-readable medium carries one or more programs, which, when executed by a device, cause the computer-readable medium to perform the following functions: acquire icing observation data of the aircraft surface; generate icing pattern prediction data based on a forward prediction model, wherein the forward prediction model generates the icing pattern prediction data through denoising diffusion; define an observation difference index based on the icing observation data and the icing pattern prediction data; model the icing condition inversion problem as a Bayesian inversion problem of solving the posterior probability distribution of the flight environment parameters, wherein the observation difference index serves as the likelihood function of the Bayesian inversion problem; generate the posterior probability distribution of the flight environment parameters through adaptive iterative sampling in the parameter space using an adaptive Monte Carlo sampling algorithm; construct an uncertainty quantification result of the icing conditions based on the posterior probability distribution; and generate a flight safety assessment result based on the uncertainty quantification result.

[0252] Those skilled in the art will understand that the above modules can be distributed in the device as described in the embodiments, or they can be modified accordingly and placed in one or more devices that are unique to this embodiment. The modules in the above embodiments can be combined into one module, or they can be further divided into multiple sub-modules.

[0253] Through the description of the above embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, mobile terminal, or network device, etc.) to execute the methods according to the embodiments of this application.

[0254] Exemplary embodiments of this application have been specifically shown and described above. It should be understood that this application is not limited to the detailed structures, arrangements, or implementation methods described herein; rather, this application is intended to cover various modifications and equivalent arrangements contained within the spirit and scope of the appended claims.

Claims

1. A flight safety assessment method based on icing condition inversion and uncertainty quantification, characterized in that, include: Acquire observation data on icing on the aircraft surface; Ice pattern prediction data is generated based on a positive prediction model, which generates the ice pattern prediction data through denoising diffusion. An observation difference index is defined by comparing the icing observation data with the ice type prediction data. The icing condition inversion problem is modeled as a Bayesian inversion problem to solve the posterior probability distribution of the flight environment parameters, wherein the observation difference index is used as the likelihood function of the Bayesian inversion problem. The posterior probability distribution of the flight environment parameters is generated by adaptive iterative sampling in the parameter space using an adaptive Monte Carlo sampling algorithm. The uncertainty quantification result of the icing conditions is constructed based on the posterior probability distribution; Flight safety assessment results are generated based on the uncertainty quantification results.

2. The method as described in claim 1, characterized in that, Also includes: The positive prediction model is constructed based on a physics-driven conditional diffusion generation model; The inputs to the forward prediction model are flight environment parameters and wing reference profile. The flight environment parameters include: flight altitude, angle of attack, airflow speed, droplet diameter, droplet water content, and ambient temperature. The output of the forward prediction model is ice type prediction data.

3. The method as described in claim 1, characterized in that, Ice pattern prediction data is generated based on a positive prediction model, including: Generate initial sampled values ​​for flight environment parameters; The initial sampled values ​​are input into the forward prediction model to generate the ice pattern prediction data.

4. The method as described in claim 1, characterized in that, An observation difference index is defined by comparing the icing observation data with the ice type prediction data, including: Image-level observation difference indices are defined by comparing the icing observation data with the ice type prediction data. These indices include RMSE, SSIM, and DICE coefficients.

5. The method as described in claim 1, characterized in that, The icing condition inversion problem is modeled as a Bayesian inversion problem of solving the posterior probability distribution of the flight environment parameters, including: Define the objective function of the Bayesian inversion problem as: P(θ∣D)∝P(D∣θ)P(θ) Where θ represents the flight environment parameter to be inverted, D is the icing observation data, P(D|θ) is the likelihood function, and P(θ) represents the prior distribution of the flight environment parameter.

6. The method as described in claim 1, characterized in that, Adaptive iterative sampling in the parameter space using an adaptive Monte Carlo sampling algorithm to generate the posterior probability distribution of the flight environment parameters includes: Iterative sampling in the parameter space is performed using an adaptive Monte Carlo sampling algorithm; By dynamically updating the mean and variance of the proposed distribution through historical sample distribution, the sampling area can be dynamically and adaptively adjusted. When the adaptive Monte Carlo sampling process meets the convergence condition, the posterior probability distribution of the flight environment parameters is obtained.

7. The method as described in claim 6, characterized in that, When the adaptive Monte Carlo sampling process meets the convergence condition, the posterior probability distribution of the flight environment parameters is obtained, including: Each sampling point has a weight calculated based on the likelihood function, and then normalized. Resample high-confidence region samples based on sample weights to gradually approximate the true posterior probability distribution.

8. The method as described in claim 1, characterized in that, The uncertainty quantification result of the icing conditions is constructed based on the posterior probability distribution, including: Based on the posterior probability distribution, the mean, variance, and confidence interval of each parameter in the flight environment parameters are statistically analyzed to construct the uncertainty quantification result of aircraft icing conditions.

9. The method as described in claim 1, characterized in that, The flight safety assessment results are generated based on the uncertainty quantification results, including: A route planning map is generated based on the uncertainty quantification results; and / or The uncertainty quantification results are used to generate weather avoidance routes; and / or The anti-icing structure optimization data is generated based on the uncertainty quantification results; and / or The uncertainty quantification results are used to generate icing condition judgment indicators.

10. A flight safety assessment device based on icing condition inversion and uncertainty quantification, characterized in that, include: The observation module is used to acquire observation data on icing on the aircraft surface; The prediction module is used to generate ice pattern prediction data based on a positive prediction model, wherein the positive prediction model generates the ice pattern prediction data through denoising diffusion. The indicator module is used to define an observation difference index based on the icing observation data and the ice type prediction data; The modeling module is used to model the icing condition inversion problem as a Bayesian inversion problem of solving the posterior probability distribution of the flight environment parameters, wherein the observation difference index is used as the likelihood function of the Bayesian inversion problem. An iterative module is used to perform adaptive iterative sampling in the parameter space using an adaptive Monte Carlo sampling algorithm to generate the posterior probability distribution of the flight environment parameters; The quantization module is used to construct the uncertainty quantification result of the icing conditions based on the posterior probability distribution; The assessment module is used to generate flight safety assessment results based on the uncertainty quantification results.

Images