A machine learning-based method for predicting the emissivity of orthogonal three-dimensional woven materials
Through machine learning-based methods, the surface emissivity of orthogonal three-way woven composite materials is quickly and accurately predicted, which solves the problem of inaccurate forecasting in the existing technology, provides new ideas for woven composite materials design, and supports structural optimization.
Patent Information
- Application Number
- CN202410713800.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-04
- Publication Date
- 2025-08-26
- Estimated Expiration
- 2044-06-04
AI Technical Summary
The prior art is difficult to predict the surface emissivity of braided composite materials quickly and accurately, limiting its application in the fields of thermal protection and electronic packaging heat exchange.
Using a machine learning-based method, by selecting the geometric structural features of orthogonal three-way woven composite materials, using Latin hypercube sampling method for data sampling, establishing feature-surface emissivity data sets, and building machine learning models such as Lasso, support vector machines, multi-layer perceptrons and random forests to predict and sensitivity analysis of surface emissivity.
A fast and accurate forecast of the surface emissivity of orthogonal three-way woven composite materials is achieved, providing new ideas for woven composite materials design, and supporting sensitivity analysis to accelerate structural optimization.
Smart Images
Figure CN118709526B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of thermal property prediction of composite materials, and in particular to a method for predicting the emissivity of orthogonal three-dimensional woven materials based on machine learning. Background Art
[0002] Due to the designability and integrity of carbon fiber braids, braided composite materials have become a rising star in the field of thermal protection materials, such as AEDPT and HEEET. Its advantage is that the precise braiding method can be changed to adapt to different detection tasks. Braided composite materials can be divided into (1) braided structures: such as plain weave, 3D 4-way, 3D 5-way, etc.; (2) woven structures: such as orthogonal 3-way, 2.5D shallow cross straight, 2.5D shallow cross bent, etc.
[0003] Rapidly predicting the surface emissivity of woven composites is a key factor limiting their application. For example, low surface emissivity is required in thermal protection applications, while higher emissivity is required in applications such as electronic packaging and heat exchange. The surface emissivity of a composite material is not only related to the surface emissivity of its components but also to its structural characteristics. Predicting the surface emissivity of a woven composite material of a given structure, or optimizing its design based on known component surface emissivities, is a current research hotspot.
[0004] Based on this, the present invention proposes an orthogonal three-dimensional woven material emissivity prediction method based on machine learning. Summary of the Invention
[0005] The purpose of the present invention is to provide a method for predicting the emissivity of orthogonal three-dimensional woven materials based on machine learning, which can quickly and accurately predict the surface emissivity of orthogonal three-dimensional woven composite materials and provide new ideas for the design of woven composite materials.
[0006] To achieve the above object, the present invention provides a method for predicting the emissivity of orthogonal three-dimensional woven materials based on machine learning, comprising the following steps:
[0007] S1. Select the geometric structure characteristics of the orthogonal three-dimensional woven composite material and determine the thickness of the calculation domain;
[0008] S2. Sampling the geometric structure feature parameters by Latin hypercube sampling method, thereby determining the parameters of each geometric structure feature;
[0009] S3. Constructing a geometric model based on the geometric structure features and calculation domain thickness in step S1 and the geometric structure feature parameters in step S2, calculating the equivalent surface emissivity of the orthogonal three-dimensional woven composite material using a numerical method, establishing a feature-surface emissivity data set, and dividing it into a training set and a test set;
[0010] S4. Build a machine learning model using the training set data and predict the surface emissivity of the test set data.
[0011] Preferably, in step S1, the geometric structure features are: warp yarn width, height and occupied space; weft yarn width, height and occupied space; normal yarn width, height and occupied space, a total of 12 types.
[0012] Preferably, the method for determining the thickness of the calculation domain is: for an orthogonal three-dimensional woven composite material with given geometric structural characteristics, a geometric model of different thicknesses is established, and the equivalent surface emissivity is calculated. When the surface emissivity changes with thickness by less than 1%, the thickness is the calculation domain thickness.
[0013] Preferably, in step S2, the amount of sampled data should meet the following conditions:
[0014] m>z×50
[0015] Among them, m is the amount of data; z is the number of types of geometric structure features.
[0016] Preferably, in step S3, the calculation method of the equivalent surface emissivity is:
[0017]
[0018] Among them, q r is the radiation heat flow; E b is the blackbody radiation energy; T is the temperature, For spatial location The radiation intensity at is the normal direction vector of the radiation intensity; σ is a constant; Ω is the solid angle; the radiation intensity I is calculated by solving the radiation transfer equation in a translucent medium without considering scattering, and the calculation method is:
[0019]
[0020] Where, κ is the absorption coefficient; For spatial location The radiation intensity at b is the blackbody radiation intensity; For spatial location The blackbody radiation intensity at ; s is the spatial position.
[0021] Preferably, the machine learning model includes Lasso method, support vector machine, multi-layer perceptron, and random forest.
[0022] Therefore, the present invention adopts the above-mentioned orthogonal three-dimensional woven material emissivity prediction method based on machine learning. By establishing a machine learning model, the surface emissivity of the orthogonal three-dimensional woven composite material is quickly predicted. The model can also be used for sensitivity analysis, thereby achieving the effect of rapid structural design.
[0023] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 This is a flow chart of a method for predicting the emissivity of orthogonal three-dimensional woven materials based on machine learning according to the present invention;
[0025] Figure 2 is the surface emissivity versus thickness graph;
[0026] Figure 3 is the geometric structure and its corresponding surface radiation intensity distribution diagram; where, Figure 3 (a) is the geometric structure; Figure 3 (b) is the corresponding surface radiation intensity distribution cloud map;
[0027] Figure 4 is the result of machine learning model training; Figure 4 (a) is the Lasso model test set and training set R 2 Score and RMSE; Figure 4 (b) is the random forest model test set and training set R 2 Score and RMSE; Figure 4 (c) is the support vector machine model test set and training set R 2 Score and RMSE; Figure 4 (d) is the multi-layer perceptron model test set and training set R 2 Score and RMSE;
[0028] Figure 5 This is the result of the overall sensitivity analysis using the machine learning model. DETAILED DESCRIPTION
[0029] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.
[0030] Unless otherwise defined, technical or scientific terms used in the present invention shall have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs.
[0031] Example 1
[0032] like Figure 1FIG. 1 is a flow chart of a method for predicting the emissivity of orthogonal three-dimensional woven materials based on machine learning according to the present invention, which specifically includes the following steps:
[0033] S1. Select the geometric structure features of the orthogonal three-dimensional woven composite material and determine the thickness of the calculation domain;
[0034] The structural characteristics are: warp yarn width, height, and occupied space; weft yarn width, height, and occupied space; normal yarn width, height, and occupied space, a total of 12 types.
[0035] The method for determining the thickness of the computational domain is as follows: for orthogonal triaxial fabrics with given geometric structural characteristics, geometric models of different thicknesses are established and equivalent surface emissivity is calculated. If the surface emissivity changes with thickness by less than 1%, the thickness is considered to be the computational domain thickness. For example, in the case of a geometric structure as shown in Table 1, the surface emissivity changes with thickness as follows Figure 2 Therefore, select Figure 2 The geometric structure and surface emissivity corresponding to the thickness of the five layers are used as one of the data for machine learning.
[0036] Table 1 Geometric structure characteristic parameters
[0037]
[0038] S2. Sampling is performed using Latin Hypercube Sampling to determine the parameters of each geometric structure feature. The amount of sampled data should meet the following conditions:
[0039] m>z×50
[0040] Among them, m is the amount of data; z is the number of types of geometric structure features;
[0041] There are 9 geometric structure features, so the data size is selected as 450.
[0042] S3. Use numerical methods to calculate the equivalent surface emissivity of orthogonal tri-directional woven composite materials and establish a characteristic-surface emissivity data set;
[0043] The calculation method of equivalent surface emissivity is:
[0044]
[0045] Among them, q r is the radiation heat flux E b is the blackbody radiation energy, T is the temperature, For spatial location The radiation intensity at is the normal direction vector of the radiation intensity, σ is a constant, Ω is the solid angle, and the radiation intensity I is calculated by solving the radiation transfer equation (RTE) in a translucent medium without considering scattering. The calculation method is:
[0046]
[0047] Where, κ is the absorption coefficient; For spatial location The radiation intensity at b is the blackbody radiation intensity. The finite element calculation results of the surface radiation intensity of the corresponding structure in Table 1 are as follows Figure 3 shown.
[0048] S4. Machine learning model training and surface emissivity prediction.
[0049] Machine learning models include Lasso method, support vector machine, multi-layer perceptron, and random forest. The 10-fold cross validation method was used to avoid overfitting of model training. 2 The model training and testing results were evaluated using RMSE.
[0050] The training results are as follows Figure 4 As shown in the figure, the comparison shows that the multilayer perceptron has the best prediction effect, with a training time of 57 seconds and a prediction time of 0.032 seconds; the finite element calculation time is 20 minutes and 43 seconds.
[0051] S5. Select the established multi-layer perceptron model and perform Sobol sensitivity analysis with the open source library Salib Library. The results of the total order sensitivity analysis are as follows: Figure 5 shown.
[0052] Therefore, the present invention adopts the above-mentioned orthogonal three-dimensional woven composite material surface emissivity prediction model based on machine learning, and its technical effects are as follows: by establishing a machine learning model, the surface emissivity of the orthogonal three-dimensional woven composite material is quickly predicted, and the model can also be used for sensitivity analysis, thereby achieving the effect of rapid structural design.
[0053] Therefore, the present invention adopts the above-mentioned orthogonal three-dimensional woven material emissivity prediction method based on machine learning, which can quickly and accurately predict the surface emissivity of orthogonal three-dimensional woven composite materials and provide new ideas for the design of woven composite materials.
[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for predicting the emissivity of orthogonal three-dimensional woven materials based on machine learning, characterized in that: The following steps are involved: S1. Select the geometric structure characteristics of the orthogonal three-dimensional woven composite material and determine the thickness of the calculation domain; S2. Sampling the parameter values of the geometric structure features by a Latin hypercube sampling method, thereby determining the parameter value of each geometric structure feature; S3. Constructing a geometric model based on the geometric structure features, the calculation domain thickness, and the parameter values of the geometric structure features in step S1 and step S2, calculating the equivalent surface emissivity of the orthogonal three-dimensional woven composite material using a numerical method, establishing a feature-surface emissivity data set, and dividing it into a training set and a test set; S4. Build a machine learning model using the training set data and predict the surface emissivity of the test set data.
2. The method for predicting the emissivity of orthogonal three-dimensional woven materials based on machine learning according to claim 1, characterized in that: In step S1, the geometric structure features are: the width, height and space occupied by the warp yarns; the width, height and space occupied by the weft yarns; and the width, height and space occupied by the normal yarns.
3. The method for predicting the emissivity of orthogonal three-dimensional woven materials based on machine learning according to claim 2, characterized in that: The method for determining the thickness of the computational domain is as follows: for an orthogonal three-dimensional woven composite material with given geometric structural characteristics, a geometric model with different thicknesses is established and the equivalent surface emissivity is calculated. When the surface emissivity changes with thickness by less than 1%, the thickness is the computational domain thickness.
4. The method for predicting the emissivity of orthogonal three-dimensional woven materials based on machine learning according to claim 3, characterized in that: In step S2, the amount of sampled data should meet the following conditions: m>z×50 Among them, m is the amount of data; z is the number of types of geometric structure features.
5. The method for predicting the emissivity of orthogonal three-dimensional woven materials based on machine learning according to claim 4, characterized in that: In step S3, the calculation method of the equivalent surface emissivity is: Among them, q r is the radiation heat flow; E b is the blackbody radiation energy; T is the temperature, For spatial location The radiation intensity at is the normal direction vector of the radiation intensity; σ is a constant; Ω is the solid angle; the radiation intensity I is calculated by solving the radiation transfer equation in a translucent medium without considering scattering, and the calculation method is: Where k is the absorption coefficient; For spatial location The radiation intensity at b is the blackbody radiation intensity; For spatial location The blackbody radiation intensity at ; s is the spatial position.
6. The method for predicting the emissivity of orthogonal three-dimensional woven materials based on machine learning according to claim 1, characterized in that: Machine learning models include Lasso method, support vector machine, multi-layer perceptron, and random forest.