Search

CN-121997629-A - Infrared characteristic simulation and sample generation method for aircraft surface structure damage

CN121997629ACN 121997629 ACN121997629 ACN 121997629ACN-121997629-A

Abstract

The invention belongs to the technical field of numerical simulation and damage intelligent detection, and relates to an infrared characteristic simulation and sample generation method for damage of an aircraft surface structure. The invention realizes the order-of-magnitude improvement of full-size transient thermal simulation efficiency by the design of the virtual interface layer and the secondary homogenization while ensuring the physical precision, so that batch simulation on a large number of damage working conditions is possible. Through the physical effect of the sensor of the high-fidelity analog infrared camera, the simulated image is highly consistent with the real external field data in the characteristics of noise, texture and the like, and the difference between the simulation domain and the real domain is effectively reduced. By directly correlating the simulation flow with the preset damage parameters, the automatic, accurate and zero-error generation of the true value information is realized, and the problems of high manual labeling cost and poor consistency are fundamentally solved. The invention provides a high-efficiency and reliable data generation way for training the high-precision and high-robustness outfield intelligent detection model, and has definite engineering practical value.

Inventors

  • HUANG JINZI
  • WANG XIANGYING

Assignees

  • 中国航空研究院

Dates

Publication Date
20260508
Application Date
20251219

Claims (20)

  1. 1. A method for simulating and generating infrared characteristics of damage to an aircraft surface structure is characterized by firstly constructing a microscopic and mesoscopic multi-scale physical simulation model by fusing intrinsic properties and manufacturing process parameters of a material, introducing a virtual interface layer concept to realize efficient parameterization simulation of the damage, secondly greatly reducing calculation complexity by a secondary homogenization strategy on the premise of guaranteeing the authenticity of a physical mechanism, performing transient heat conduction simulation on a full-size structure to obtain real temperature field response of the damage in an external field cooling environment, then converting the temperature field into a simulated infrared image which is highly consistent with real acquired data in noise, texture and degradation characteristics by using a high-fidelity infrared sensor imaging effect simulation, and finally automatically correlating the simulated infrared image to generate a damage true value containing category labels and space characterization information based on preset damage parameters to form a sample pair and structuring the sample library of standardized damage infrared characteristics.
  2. 2. The method for simulating and generating infrared characteristics of damaged surface structures of an aircraft according to claim 1, wherein the method specifically comprises the following steps: step 1, acquiring multi-scale physical simulation parameters, namely acquiring a multi-scale physical simulation parameter set of a target thermal protection structure according to a process file and a material performance test result; Step 2, constructing a microscopic statistical representative volume unit, namely respectively constructing a needled reinforced fiber cloth layer statistical representative volume unit model and a needled reinforced fiber net tire layer statistical representative volume unit model on a microscopic scale, and calculating equivalent thermophysical parameters of the two; Step 3, calculating microscopic equivalent thermophysical parameters, namely obtaining equivalent heat conductivity coefficients of the two statistical representative volume unit models of the needled reinforced fiber cloth layer and the needled reinforced fiber net tire layer constructed in the step 2 in the main direction of three materials through finite element simulation calculation, wherein the finite element simulation calculation is to respectively perform X, Y, Z independent steady-state thermal analysis in three orthogonal directions on each statistical representative volume unit model; Step 4, defining the attribute of the virtual interface layer, namely defining the virtual interface layer for efficiently simulating interlayer damage based on the mesoscopic equivalent thermophysical parameters obtained in the step 3; step 5, mesoscopic medium unit construction and sewing thread explicit modeling, namely constructing a mesoscopic medium unit model based on the mesoscopic equivalent thermophysical parameters obtained in the step 3 and the virtual interface layer defined in the step 4, and laying a geometric and physical foundation for subsequent integral macroscopic homogenization calculation by integrating mesoscopic equivalent properties and sewing process characteristics; And 6, calculating overall equivalent thermophysical properties of the mesoscopic units, namely performing steady-state thermal analysis on the mesoscopic units constructed in the step 5, calculating overall equivalent heat conductivity coefficients of the mesoscopic units in three main directions of materials, homogenizing the complex units into a group of orthoscopic macroscopic material parameters, providing input for subsequent macroscopic simulation, and realizing mesoscopic secondary homogenization after mesoscopic homogenization. And 7, constructing a macroscopic structure model and performing damage parameterization simulation, namely firstly, constructing a macroscopic finite element model of the whole size of the target structure based on the whole equivalent thermophysical parameters obtained by calculation in the step 6, and then, efficiently simulating various types of preset damage by parameterizing and modifying the material properties, thereby providing a calculation basis for subsequent transient thermal analysis. Step 8, transient heat conduction simulation and temperature field extraction, namely, based on the macrostructure model constructed in the step 7, performing transient heat conduction simulation for simulating the thermal load of the real external field environment, obtaining the response of the structure surface in the dynamic thermal environment, and extracting a time sequence temperature field data set of the structure surface for generating a subsequent infrared image; Step 9, infrared image generation and sensor effect simulation, namely converting the time sequence surface temperature field data set extracted in the step 8 into a high-fidelity simulation infrared image sequence, and enabling simulation data to be consistent with real external field acquisition data in characteristic by simulating an imaging physical process of a real infrared camera; The method comprises the steps of 10, generating a standardized damage infrared characteristic sample library, firstly defining a damage working condition set containing different damage types, positions and size parameters, circularly calling a parameterized damage simulation function in the step 7 according to the damage working condition set on the basis of the full-size macroscopic finite element model of the target structure constructed in the step 7, modifying material properties of corresponding areas of the model according to a set of specific damage working condition parameters in each cycle through a programmed script, efficiently generating a macroscopic finite element model with specific damage configuration, then executing transient heat conduction simulation in the step 8 on the generated macroscopic finite element model set in batches to obtain a time sequence temperature field data set corresponding to each model, then executing the infrared image generation and sensor effect simulation flow in the step 9 in batches on the time sequence temperature field data set, uniformly converting the infrared image sequence into a simulated infrared image sequence, and then carrying out automatic post-processing and integration on the simulated infrared image sequence to construct the standardized damage infrared characteristic sample library which can be directly used for machine learning model training.
  3. 3. The method for simulating and generating the infrared characteristics of the surface structure damage of the aircraft according to claim 2, wherein the step 1 parameter set comprises basic material parameters and key process parameters, wherein the basic material parameters comprise heat conductivity coefficients, specific heat capacity and density of a resin matrix and quartz fibers, and surface emissivity of the composite material in an infrared band, and are used for defining material physical properties of all scale models, the key process parameters comprise fiber reinforcement geometric parameters and Z-direction reinforcement process parameters, wherein the fiber reinforcement geometric parameters comprise surface density, single-layer thickness, weaving structure, warp and weft yarn density, yarn cross-section shape and size and fiber volume fraction of a fiber cloth layer, surface density, single-layer thickness, chopped fiber length range and fiber volume fraction of a fiber web layer, and the Z-direction reinforcement process parameters comprise needling density and needling depth of a needling process, and sewing thread materials, diameters, patterns and distances of a sewing process.
  4. 4. The method for simulating and generating infrared characteristics of structural damage to an aircraft surface according to claim 2, wherein the step 2 needled reinforced fiber cloth layer statistical representative volume unit model is composed of a resin matrix, a periodic plain weave quartz fiber yarn structure embedded in the resin matrix, and Z-direction quartz needled fibers randomly penetrating between the resin matrix and the yarns.
  5. 5. The method for simulating and generating infrared characteristics of structural damage to an aircraft surface according to claim 2, wherein the step 2 needle punched reinforced net carcass ply statistical representative volume unit model is composed of a resin matrix, quartz chopped fibers distributed in the resin matrix in random positions and random orientations, and Z-direction quartz needle punched fibers which are also randomly distributed throughout.
  6. 6. The method for simulating and generating infrared characteristics of structural damage to an aircraft surface according to claim 2, wherein the step 2 of minutely counting the representative volume unit model uses tetrahedral units for space dispersion, and the material properties of each component are defined according to the basic material parameters of the parameter set obtained in the step 1.
  7. 7. The method according to claim 2, wherein the step 2 of needling the fiber-reinforced fabric layer statistical representation volume unit model and the step 2 of needling the fiber-reinforced web layer statistical representation volume unit model are performed in a thickness direction, i.e. a Z-direction dimension, according to the single-layer thickness parameters of the fiber-fabric layer and the fiber-web layer defined in the step 1, respectively, so as to characterize a complete single layer of material.
  8. 8. The method for simulating and generating infrared characteristics of structural damage to an aircraft surface according to claim 2, wherein the statistical representative volume unit model of the needled reinforcing fiber cloth layer in step 2 has an in-plane dimension L c in the X-Y direction, and the method is characterized in that the method satisfies the periodic characterization of the woven structure and the statistical representative requirement of needled fibers at the same time, and has a calculation formula as follows: The method comprises the steps of (1) setting a weaving structure, wherein P is the basic cycle size of the weaving structure in units of millimeter, k is a cycle number coefficient which is a positive integer not smaller than 1, the typical value range of the weaving structure is 2-4 in order to ensure the stability of equivalent attributes, rho is the needling density in key process parameters obtained in the step 1, in units of needle number per square millimeter, N is the number of the needling fibers needed to be contained in the statistical representative volume unit model of the needling reinforced fiber cloth layer, and the value range of N is 3-10 in order to ensure the statistical representativeness.
  9. 9. The method for simulating and generating infrared characteristics of structural damage to an aircraft surface according to claim 2, wherein the dimension L b of the statistical representative volume unit model of the needled reinforced fiber web layer in the in-plane direction in step 2 is mainly considered to satisfy the statistical representative requirement of needled fibers, and the calculation formula is as follows: wherein ρ is the needling density in the key process parameters obtained in step 1, the unit is the number of needles per square millimeter, N 'is the number of needled fibers needed to be contained in the statistical representative volume unit model of the needled reinforced fiber web layer, and the range of the value of N' is 3 to 10 for guaranteeing statistical representativeness.
  10. 10. The method for simulating and generating the infrared characteristics of the surface structural damage of the aircraft according to claim 2 is characterized in that the step 2 is used for needling a fiber reinforced web tire layer statistics representative volume unit model, chopped fibers in the fiber reinforced web tire layer statistics representative volume unit model are realized by adopting an equivalent modeling method, quartz fiber filaments with tiny diameters and random arrangement in an actual process are equivalent to a cylindrical rod set with uniform diameters of d millimeters and random distribution of lengths and space orientations under the statistical scale of the statistics representative volume unit model, and the volume fraction of the fibers defined in the step 1 is strictly kept consistent, and d is usually 0.05-0.2 so as to solve the problem that the fiber reinforced web tire layer statistics representative volume unit model is difficult to directly and accurately model due to complex real forms.
  11. 11. The method for simulating and generating infrared characteristics of structural damage of aircraft surface according to claim 2, wherein the step 3 of steady-state thermal analysis is sequentially performed for each specific direction, the equivalent thermal conductivity of the model in the thickness direction and defined as the Z direction is calculated, a thermal flow boundary condition of constant thermal flow density q and a temperature boundary condition of constant temperature T ref are respectively applied to two opposite surfaces characterizing the direction, T ref is usually 0 ℃, the remaining four sides of the model are set as thermal insulation boundary conditions, or periodic boundary conditions are applied to simulate infinite periodic expansion of materials in the plane, after solving to obtain a steady-state temperature field, average temperature T hot of the surface to which the thermal flow boundary is applied is extracted, and average temperature difference Δt=t of the two surfaces is calculated hot T ref , according to the one-dimensional Fourier heat conduction law, the equivalent heat conduction coefficient k z of the volume unit model in the current direction, namely Z direction is calculated by the following formula: wherein H is the characteristic dimension of the model in the calculation direction, the process is repeated, the equivalent thermal conductivity coefficients k x and k y of the statistical representative volume unit model in the X and Y directions are sequentially calculated and obtained, and finally, the three material main direction equivalent thermal conductivity coefficients of the statistical representative volume unit model of the needled reinforced fiber cloth layer are output 、 、 And the needled reinforcing fiber web tire layer statistics represent three material principal direction equivalent thermal conductivities of the volumetric unit model 、 、 The parameters constitute key material properties for homogenizing the mesoscopic heterostructure, providing input for subsequent mesoscopic and macro-scale heat transfer simulation.
  12. 12. The method for simulating and generating infrared characteristics of structural damage on an aircraft according to claim 11, wherein the virtual interface layer in step 4 is a conceptual thin layer, the geometric position of the virtual interface layer is defined between the needled reinforcing fiber cloth layer and the needled reinforcing fiber web tire layer, and the equivalent thermal conductivity of the virtual interface layer in the X, Y, Z main directions of the three materials is respectively obtained by taking arithmetic average of equivalent thermophysical parameters of the adjacent fiber cloth layers and the fiber web tire layer in the main directions when the simulated structure is in an intact state in terms of material property definition, and the specific calculation formula is as follows: Wherein, the 、 、 The equivalent heat conductivity coefficients of the virtual interface layer in the main directions of the X, Y, Z materials are respectively.
  13. 13. The method for simulating and generating infrared characteristics of structural damage to an aircraft surface according to claim 12, wherein said medium unit model in step 5 is formed by periodically stacking a plurality of basic ply units in a thickness direction, i.e. a Z direction, each basic ply unit strictly following a fixed sequence of "fiber cloth layer-virtual interface layer-fiber web tire layer-virtual interface layer", wherein the fiber cloth layer and the fiber web tire layer respectively give equivalent thermal conductivity to the needled reinforced fiber cloth layer by the material thermophysical parameters thereof 、 、 Equivalent coefficient of thermal conductivity to needled reinforcing fiber net tire layer 、 、 The material properties of the virtual interface layer are based on 、 、 The application is performed.
  14. 14. The method for simulating and generating infrared characteristics of surface structure damage of an aircraft according to claim 3, wherein the dimension of the medium unit model in the step 5 is determined according to a representative principle, the dimension in the plane, namely the X-Y direction, is required to be larger than the sewing interval so as to ensure that the periodicity of a sewing pattern can be completely represented, and the thickness, namely the Z-direction dimension, is formed by stacking M basic layering units so as to reasonably represent the thickness direction characteristics of the whole layering structure, and meanwhile, the model is prevented from being excessively complicated, wherein M can be generally 4-8.
  15. 15. The method for simulating and generating infrared characteristics of structural damage to an aircraft surface according to claim 3, wherein in the step 5, the unit model is constructed by performing explicit three-dimensional geometric modeling on a sewing thread according to the sewing pattern and the sewing pitch in the acquired key process parameters, and the sewing thread is constructed as a cylinder penetrating through the whole combined unit thickness, and the spatial position and the path of the sewing thread are determined by the sewing pattern and the sewing pitch.
  16. 16. The method for simulating and generating infrared characteristics of structural damage of aircraft surface according to claim 2, wherein the steady-state thermal analysis in step 6 is to sequentially perform independent simulation in three main directions (X, Y, Z) of materials of the medium unit model, wherein the overall equivalent thermal conductivity in the Z direction is calculated by applying a constant heat flux q ' boundary condition and a constant reference temperature T ' ref boundary condition respectively on two opposite surfaces in the Z direction, i.e. the thickness direction, T ' ref is usually 0 ℃, the remaining four sides are set as adiabatic boundary conditions, extracting an average temperature T ' hot of a heat flux loading surface after obtaining a steady-state temperature field by a finite element method, and calculating a temperature difference Δt ' =t ' between the two surfaces ' hot T' ref , then calculating the overall equivalent thermal conductivity in that direction according to the one-dimensional Fourier heat conduction law Wherein H' is the characteristic dimension of the model in the calculation direction, and the equivalent heat conductivity coefficients of the medium unit in the X and Y directions are calculated and obtained sequentially by repeating the above processes And (3) with Finally obtaining a group of complete orthotropic integral equivalent thermophysical parameters , , 。
  17. 17. The method for simulating and generating infrared characteristics of damaged surface structure of an aircraft according to claim 12, wherein said step 7 macrostructure model is geometrically built up according to the dimensions of the target structure, and wherein said target structure is formed by periodically laying J sets of ply units in the thickness direction, i.e. Z direction, in terms of material definition, each set of ply units being composed of a layer of material inheriting medium cell properties and a virtual interface layer, wherein the thermophysical parameters of said layer of material directly impart the orthotropic overall equivalent thermophysical parameters outputted in step 6 , , The virtual interface layer attribute still uses the equivalent heat conductivity coefficient of the virtual interface layer in the step 4 、 、 Setting is performed.
  18. 18. The method for simulating and generating the infrared characteristics of the damage to the surface structure of the aircraft according to claim 2, wherein the step 7 of parameterized simulation of the damage is realized by programming scripts, simulating various typical damage including interlayer debonding, matrix cracking, interstitial peeling and surface ablation on the macroscopic model, positioning a set of units representing a virtual interface layer in the model for simulating the interlayer debonding damage and switching the material properties of the set of units from the good state parameters to the thermophysical parameters of air, positioning a set of units corresponding to the damaged area in the model for simulating the cracking and interstitial peeling damage and switching the material properties of the set of units corresponding to the damaged area to the thermophysical parameters of air, and modifying the material properties of the units corresponding to the damaged area for simulating the surface ablation damage.
  19. 19. The method of claim 18, wherein the modified unit material properties include equivalent thermal conductivity, specific heat capacity, and surface emissivity.
  20. 20. The method for simulating and generating the infrared characteristics of the damage to the surface structure of the aircraft according to claim 18, wherein the method for parameterizing the damage in step 7 is characterized in that the damage is simulated by changing the material property without changing the geometry and the grid of the model, and the size, the position, the depth and the type of the damage can be flexibly defined and controlled through input parameters, so that the simulation model simulating various damage can be generated efficiently.

Description

Infrared characteristic simulation and sample generation method for aircraft surface structure damage Technical Field The invention belongs to the technical field of numerical simulation and damage intelligent detection, and relates to an infrared characteristic simulation and sample generation method for damage of an aircraft surface structure. Background The surface structure of the special aircraft can generate various damages such as cracking, ablation, debonding and the like under severe thermal cycle load. These injuries not only directly threaten flight safety, but also make efficient and reliable damage detection and status assessment of the structure a necessary and burdensome safeguard after each flight mission. However, the existing detection method has limitations on efficiency and precision, so that the assessment period is long, and the quick guarantee capability of the aircraft is severely restricted. In order to break this bottleneck, the damage detection technology needs to be developed in an intelligent direction. Currently, the mainstream technical route for achieving intelligent damage detection relies on machine vision methods based on deep learning. The performance of the method is highly dependent on a training sample library which is large-scale, high-quality and accurate in labeling. However, obtaining an infrared sample aiming at damage of a thermal protection structure faces serious challenges, namely, firstly, a real damage sample is rare, the cost obtained through actual flight or real test is extremely high, and it is difficult to systematically cover all damage types, sizes, depths and combined working conditions, secondly, sample marking is difficult, characteristics of damage in an infrared image are weak and boundaries are fuzzy, fine marking needs extremely high professional cost, subjective errors exist, thirdly, data generalization is poor, and complex and changeable external field environments including different initial temperature differences, illumination, observation angles and the like are difficult to be covered by real data of a single source. Although data enhancement can be performed through traditional image transformation such as rotation, clipping and color adjustment, the method only changes the image appearance, cannot generate a new real damage characteristic on a physical mechanism, and has limited contribution to the identification and generalization capability of the improvement model on real defects. In the simulation technology, the existing numerical simulation method for the composite material is mainly focused on mechanism research on a specific damage mode or verification on an isolated example. In the prior art, even if a homogenization method based on a representative volume unit is adopted to perform multi-scale simulation on a composite material, when dealing with heterogeneous materials with complex Z-direction enhancement processes such as aircraft surface structures, including needling, sewing and the like, a built multi-level model is still complex and has high calculation cost for accurately describing a damage mechanism. This makes it difficult for traditional simulations to support high-throughput, parameterized calculations for hundreds or thousands of damage conditions, and thus cannot be directly used to construct a large-scale training sample library. In addition, the conventional simulation only outputs an idealized temperature field cloud chart, and the lack of modeling on key sensor effects such as noise, resolution limitation, non-uniformity and the like in an imaging chain of a real infrared camera leads to obvious 'domain difference' between simulation data and real acquired data, so that the performance of a model trained by the method is rapidly reduced in the practical application of an external field. In summary, in the prior art, the contradiction between the high cost and low coverage of the real sample acquisition and the high calculation consumption and low data fidelity of the traditional simulation method is increasingly prominent, which becomes a main bottleneck for restricting the development of the thermal protection structure external field rapid and intelligent infrared detection technology based on artificial intelligence. Therefore, development of a novel method capable of efficiently and automatically generating infrared damage simulation samples with accurate labeling on a large scale under the premise of guaranteeing physical authenticity is needed, and reliable data support is provided for training of intelligent detection models. Disclosure of Invention The invention aims to solve the outstanding problems of lack of high-quality training samples, difficult acquisition, high simulation calculation consumption and inconsistent data and real external field infrared image characteristics faced by the existing intelligent infrared detection method for the surface structure damage of the aircraft based on deep