Deformation memory life evaluation method of carbon nanotube fiber composite material

Through data preprocessing and timing impact analysis model, combined with deep long and short-term memory network algorithm, a deformation memory life evaluation model is constructed, which solves the accuracy and complexity of deformation memory life evaluation of carbon nanotube fiber composite materials in the existing technology, and realizes the accuracy and complexity of deformation memory life evaluation. Efficient and accurate life expectancy and evaluation.

CN119089329AInactive Publication Date: 2024-12-06金丰良
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411051539.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-01
Publication Date
2024-12-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing technology is difficult to accurately evaluate the deformation memory life of carbon nanotube fiber composite materials. Traditional mechanical testing methods cannot reflect the performance changes in the long-term use of materials. Moreover, the electrochemical and optical testing methods are complex and costly, and are susceptible to external interference. .

Method used

Using data preprocessing, timing impact analysis model and clustering hierarchy method, a deformation memory life evaluation model is constructed, and the variability characteristics and time-varying are extracted through technical means such as information weight screening, singular value decomposition, attention mechanism and fuzzy similarity relationship matrix. Features, combining deep-length short-term memory network algorithm to predict material life.

Benefits of technology

It improves the accuracy and accuracy of deformation memory life evaluation of carbon nanotube fiber composite materials, realizes automated and real-time life evaluation, adapts to different standards and material needs, is universal, saves resources, and improves work efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089329A_ABST
    Figure CN119089329A_ABST
Patent Text Reader

Abstract

The invention discloses a method for evaluating the deformation memory life of a carbon nanotube fiber composite material, which comprises the following steps of: acquiring test data and influence data of the carbon nanotube fiber composite material, and preprocessing the test data and the influence data; performing correlation screening on the influence data according to the deformation degree to obtain relational data, and inputting the relational data into a first time sequence influence analysis model to obtain a variable image characteristic; inputting the test data into a second time sequence influence analysis model to obtain time-varying features, and clustering and grading the time-varying features according to the recovery degree to obtain graded data; and constructing a deformation memory life evaluation model according to the variable shadow characteristics and the hierarchical data, inputting to-be-evaluated data into the deformation memory life evaluation model, and outputting an evaluation result. The method not only can improve the accuracy of deformation memory life evaluation of the carbon nanotube fiber composite material, but also has good interpretability, and can be directly applied to a deformation memory life evaluation system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of evaluation, and in particular to a method for evaluating the deformation memory life of a carbon nanotube fiber composite material. Background Art

[0002] With the rapid development of nanotechnology, carbon nanotubes, as a nanomaterial with excellent mechanical and electrical properties, have been widely used in aerospace, smart materials, biomedicine and other fields. Among them, carbon nanotube fiber composites have become a hot topic of research due to their high strength, high modulus, good conductivity and thermal stability.

[0003] Deformation memory is one of the important properties of smart materials, which enables materials to undergo reversible deformation under external stimuli. Carbon nanotube fiber composites have excellent deformation memory properties due to their unique nanostructure and have broad application prospects.

[0004] However, in practical applications, the deformation memory life of carbon nanotube fiber composites is a key issue. Since the material will be affected by various environmental factors during long-term use, its deformation memory performance may gradually decrease or even fail. Therefore, how to accurately evaluate the deformation memory life of carbon nanotube fiber composites is of great significance to ensure the reliability and service life of the material.

[0005] At present, there are still some problems with the deformation memory life evaluation method of carbon nanotube fiber composite materials. Although the traditional mechanical testing method can directly measure the deformation properties of the material, it cannot accurately reflect the performance changes of the material during long-term use. Some testing methods based on electrochemical and optical principles, although they have high sensitivity and resolution, are complex to operate, costly, and easily affected by external interference. Summary of the invention

[0006] The purpose of the present invention is to provide a method for evaluating the deformation memory life of a carbon nanotube fiber composite material.

[0007] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0008] The present invention comprises the following steps:

[0009] Collecting test data and impact data of the carbon nanotube fiber composite material, and preprocessing the test data and the impact data;

[0010] Perform correlation screening on the impact data according to the deformation degree to obtain relationship data, and input the relationship data into a first time series impact analysis model to obtain deformation characteristics;

[0011] Inputting the test data into a second time series impact analysis model to obtain time-varying features, and clustering and grading the time-varying features according to the recovery degree to obtain graded data;

[0012] A deformation memory life evaluation model is constructed according to the deformation characteristics and the classification data, the data to be evaluated is input into the deformation memory life evaluation model, and an evaluation result is output.

[0013] Furthermore, the method of performing relevant screening on the influence data according to the deformation degree to obtain the relationship data includes:

[0014] Calculate the information weight that affects the data:

[0015]

[0016] The information weight of the jth influencing data is The variance of the wth influencing data and the jth influencing data is The jth influencing data is k j , the mean of the w-th influencing data is The number of w-th influencing data is Impact data k j The probability of occurrence is P(k j );

[0017] The influence data with information weight greater than 0.317 are taken as key data, and the non-cotyledon node sample set is generated through the key data to calculate the similarity matrix and degree matrix:

[0018]

[0019] The similarity matrix between the a-th sample and the y-th sample at the κ-th node is The number of samples of category z is m z , the category of the ath sample is c a , the category of the yth sample is c y , the adjacency matrix of the a-th sample and the y-th sample of the κ-th node is The degree matrix of the k-th node is M κ , the kernel width is v, the distance between the a-th sample and the y-th sample is d(a,y), and the control factor is σ;

[0020] Calculate the deformation correlation matrix and goodness matrix:

[0021]

[0022] The deformation correlation matrix of the a-th sample and the y-th sample at the κ-th node is U, and the goodness matrix of the κ-th node is Q κ , the genetic number of the κth node is

[0023] Use singular value decomposition to obtain the features of relevant data and calculate the importance of the features:

[0024]

[0025] The importance of the rth feature is The rth eigenvector is h r , transposed to T, the identity matrix is ​​E, and the transposed of the rth eigenvector is

[0026] Calculate the deformation of the relevant data:

[0027]

[0028] The number of features of the relevant data is N, and the deformation degree of the jth relevant data is The number of relevant data is b, the distance between the jth relevant data and the fth relevant data is d(j,f), and the difference in deformation length before and after the relevant data changes is ΔH;

[0029] The relevant data with deformation degree greater than 0.469 are output as relational data.

[0030] Furthermore, the method of inputting the relationship data into a first timing impact analysis model to obtain the shadow variation characteristics includes:

[0031] Input the relational data into the time series impact analysis model to extract the time pattern information of the relational data;

[0032] The time pattern information is input into the attention mechanism to calculate the correlation between different dimensional relationship data and time state:

[0033]

[0034] The sliding window The number of The temporal pattern information of the relational data in the υth convolution kernel on the κth dimension is: Time is The time limit is C, The hidden state at this moment is The weight matrix is Temporal pattern information and hidden state of the κth dimension The correlation is No. The υth convolution kernel at the moment is The k-th dimension relation data The time mode information of the moment is

[0035] Compute an integer-valued vector:

[0036]

[0037] Among them The integer value vector at the moment is The activation function is sigmoid(·), and the number of dimensions is D;

[0038] Calculate the influence of relationship data:

[0039]

[0040] The influence of the ιth relation data is θ ι , the adjustment factor is ω, the influence degree is σ, the The ιth relation data n at time ι The variable is

[0041] The relationship data with an influence greater than 0.375 is regarded as the shadow characteristic.

[0042] Furthermore, the method of inputting the test data into a second timing impact analysis model to obtain time-varying characteristics includes:

[0043] Input the test data into the second timing impact analysis model to calculate the composite material life difference:

[0044]

[0045] Among them The moment and The composite material life difference at each moment is No. The composite material life at that moment is No. The life of the composite material after the test is

[0046] The test data of composite material life difference greater than 0.12 is taken as key data, and the fuzzy similarity relationship matrix is ​​established based on the key data. The expression is:

[0047]

[0048] The similarity matrix element in the lth row and ith column is The center of the pth key data in the lth row is The number of key data is The genetic factor in row l and column i is

[0049] To solve the fuzzy equivalent matrix of the fuzzy similarity relationship matrix, the expression is:

[0050]

[0051] The fuzzy equivalence matrix is ​​ε * , the fuzzy equivalent elements are The fuzzy equivalent elements are sorted in descending order, and the test data with elements 1 in the matrix are regarded as the first-class data, and the rest as the second-class data;

[0052] Sort the first-class data and the second-class data by time to obtain the first-class time data and the second-class time data.

[0053] Input the first-class time data and the second-class time data into the dynamic adjustment modal fusion model, and input the fusion result. The expression is:

[0054]

[0055] The fused data is One type of time data is The second type of time data is The exclusive OR logical operator is The fused data is projected into a low-dimensional space through a linear layer, and the expression is:

[0056]

[0057] The linear rectification function is ReLU(·), and the weight of the fusion feature is The adjustment parameters of the fused data are The low-dimensional features are Input low-dimensional features as time-varying features.

[0058] Furthermore, the method of clustering and grading the time-varying features according to the recovery degree to obtain graded data includes:

[0059] Calculate the weights of time series features:

[0060]

[0061] The weight of the xth time series feature is β x , the upper limit of observation time is q, and the xth time series feature is g x , the time series feature at the sth moment is g s , the number of time series features is n;

[0062] Calculate the similarity of neighbor time series features:

[0063]

[0064] The similarity between the xth time series feature and the uth time series feature is The number of neighboring sets of the xth time series feature and the uth time series feature is K, and the number of neighboring sets of the xth time series feature and the uth time series feature is n. k , No. The time series feature weight of a sample is No. The xth time series feature of samples is No. The u-th time series feature of a sample is The weighted distance between the xth time series feature and the uth time series feature is sd xu , the weighted distance between the xth time series feature and the tth time series feature is sd xt , the weighted distance between the uth time series feature and the tth time series feature is sd ut , the number of samples in the neighboring set K is |K|;

[0065] The time series features with similarity greater than 0.365 are divided into the initial category, and the remaining time series features are divided into the initial category with the largest similarity;

[0066] Calculate the information content of time-varying features:

[0067]

[0068] The category of the xth time series feature is The information content of the xth time series feature is fr x ,category The probability of The number of muscles of the xth time series feature is R a , the number of joints of the xth time series feature is S a ;

[0069] The time-varying feature at the center of the category is taken as the category center, and the minimum deviation error between the time-varying feature and the category center is calculated:

[0070]

[0071] The xth time series feature and category center The minimum deviation error is The number of time series features of the category is The category center of the x-th time series feature is

[0072] Compute the information score of time-varying features:

[0073]

[0074] The information score of the xth time-varying feature is No. Categories The maximum likelihood value is B max (s a ), the parameter value of the xth time-varying feature is No. Categories The contribution of

[0075] The time-varying features with information scores less than 0.385 are regarded as weak features, and the weak features are placed in the remaining categories. The information scores are recalculated, and the weak features are placed in the category with the largest information score to obtain classified data;

[0076] Calculate the recovery caused by the change of time-varying characteristics:

[0077]

[0078] The elastic weight is ζ1, the stress weight is ζ2, and the initial deformation is ΔF bg , the amount of recovery after deformation is ΔF cv , the initial stress is ΔG bg , the stress after relaxation is ΔG cv , the recovery degree of the xth time-varying feature is

[0079] The time series features with a recovery degree less than 0.27 are regarded as the first level, the time series features with a recovery degree greater than 0.27 and less than 0.53 are regarded as the second level, and the time series features with a recovery degree greater than 0.53 and less than 1 are regarded as the third level, and the classified data are output.

[0080] Furthermore, a method for constructing a deformation memory life evaluation model according to the deformation characteristics and the classification data includes:

[0081] The deformation memory life evaluation model includes time series segmentation method, empirical mode decomposition method, time series analysis algorithm, deep long short-term memory network algorithm, and genetic optimization algorithm;

[0082] The time series segmentation method divides the input data into training data and test data according to the ratio of 7:3 in chronological order;

[0083] The empirical mode decomposition method decomposes the training data signal into the intrinsic mode function, extracts the local features of the signal through the intrinsic mode function, and obtains the deformation features;

[0084] Time series analysis algorithms obtain temporal deformation features by identifying and modeling patterns and trend features in time series data and analyzing the behavior of deformation features;

[0085] The deep long short-term memory network algorithm captures the long-term dependency in the time series by learning the matching degree of the time deformation characteristics and the deformation characteristics of the composite material under different deformation and loading conditions, and predicts the deformation memory life of the composite material based on the long-term dependency and classification data. The test data is input into the deep long short-term memory network algorithm to obtain the predicted deformation memory life.

[0086] The genetic optimization algorithm adjusts the learning rate of the deformation memory life evaluation model according to the actual deformation memory life of the test data and the mean square error of the predicted deformation memory life until the mean square error is 0.17.

[0087] The beneficial effects of the present invention are:

[0088] The present invention is a method for evaluating the deformation memory life of a carbon nanotube fiber composite material. Compared with the prior art, the present invention has the following technical effects:

[0089] The present invention can improve the accuracy of deformation memory life evaluation of carbon nanotube fiber composite materials through preprocessing, correlation screening, time series impact analysis, clustering and grading, acquisition of life data and model construction steps, thereby improving the precision of deformation memory life evaluation of carbon nanotube fiber composite materials, optimizing the deformation memory life evaluation of carbon nanotube fiber composite materials, greatly saving resources and improving work efficiency, and can realize automatic evaluation of deformation memory life, and perform influence correction on deformation memory life evaluation of carbon nanotube fiber composite materials in real time, which is of great significance to the deformation memory life evaluation of carbon nanotube fiber composite materials, can adapt to the deformation memory life evaluation of carbon nanotube fiber composite materials of different standards and the deformation memory life evaluation needs of different carbon nanotube fiber composite materials, and has a certain universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1 The present invention is a flowchart of the steps of a method for evaluating the deformation memory life of a carbon nanotube fiber composite material. DETAILED DESCRIPTION

[0091] The present invention is further described below by means of specific embodiments. The illustrative embodiments and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0092] The present invention provides a method for evaluating the deformation memory life of a carbon nanotube fiber composite material, comprising the following steps:

[0093] like Figure 1 As shown, in this embodiment, the following steps are included:

[0094] Collecting test data and impact data of the carbon nanotube fiber composite material, and preprocessing the test data and the impact data;

[0095] In the actual evaluation, the test data include stress data, strain data, elastic modulus, Poisson's ratio, bending strength, bending modulus, compressive strength, compression modulus, fatigue performance data; the influencing data include fiber type, fiber modulus, strength, fracture toughness, fiber content, matrix material, matrix-fiber interface bonding, preparation process, process parameters, temperature, humidity, chemical properties, stress state, load type, working temperature;

[0096] Perform correlation screening on the impact data according to the deformation degree to obtain relationship data, and input the relationship data into a first time series impact analysis model to obtain deformation characteristics;

[0097] In the actual evaluation, the relational data are fiber type, fiber modulus, strength, fracture toughness, fiber content, temperature, humidity, chemical properties, stress state, load type, and operating temperature; the shadow characteristics are strength, fracture toughness, temperature, humidity, stress state, load type, and operating temperature;

[0098] Inputting the test data into a second time series impact analysis model to obtain time-varying features, and clustering and grading the time-varying features according to the recovery degree to obtain graded data;

[0099] In the actual evaluation, the time-varying features are stress data, strain data, elastic modulus, bending strength, bending modulus, compression strength, compression modulus, and fatigue performance data;

[0100] A deformation memory life evaluation model is constructed according to the deformation characteristics and the classification data, the data to be evaluated is input into the deformation memory life evaluation model, and an evaluation result is output.

[0101] In this embodiment, the method of performing relevant screening on the influence data according to the deformation degree to obtain the relationship data includes:

[0102] Calculate the information weight that affects the data:

[0103]

[0104] The information weight of the jth influencing data is The variance of the wth influencing data and the jth influencing data is The jth influencing data is k j , the mean of the w-th influencing data is The number of w-th influencing data is Impact data k j The probability of occurrence is P(k j );

[0105] The influence data with information weight greater than 0.317 are taken as key data, and the non-cotyledon node sample set is generated through the key data to calculate the similarity matrix and degree matrix:

[0106]

[0107] The similarity matrix between the a-th sample and the y-th sample at the κ-th node is The number of samples of category z is m z , the category of the ath sample is c a , the category of the yth sample is c y , the adjacency matrix of the a-th sample and the y-th sample of the κ-th node is The degree matrix of the k-th node is M κ , the kernel width is v, the distance between the a-th sample and the y-th sample is d(a,y), and the control factor is σ;

[0108] Calculate the deformation correlation matrix and goodness matrix:

[0109]

[0110] The deformation correlation matrix of the a-th sample and the y-th sample at the κ-th node is U, and the goodness matrix of the κ-th node is Q κ , the genetic number of the κth node is

[0111] Use singular value decomposition to obtain the features of relevant data and calculate the importance of the features:

[0112]

[0113] The importance of the rth feature is The rth eigenvector is h r , transposed to T, the identity matrix is ​​E, and the transposed of the rth eigenvector is

[0114] Calculate the deformation of the relevant data:

[0115]

[0116] The number of features of the relevant data is N, and the deformation degree of the jth relevant data is The number of relevant data is b, the distance between the jth relevant data and the fth relevant data is d(j,f), and the difference in deformation length before and after the relevant data changes is ΔH;

[0117] The relevant data with deformation degree greater than 0.469 are output as relational data.

[0118] In this embodiment, the method of inputting the relationship data into the first timing impact analysis model to obtain the shadow characteristics includes:

[0119] Input the relational data into the time series impact analysis model to extract the time pattern information of the relational data;

[0120] The time pattern information is input into the attention mechanism to calculate the correlation between different dimensional relationship data and time state:

[0121]

[0122] The sliding window The number of The temporal pattern information of the relational data in the υth convolution kernel on the κth dimension is: Time is The time limit is C, The hidden state at this moment is The weight matrix is Temporal pattern information and hidden state of the κth dimension The correlation is No. The υth convolution kernel at the moment is The k-th dimension relation data The time mode information of the moment is

[0123] Compute an integer-valued vector:

[0124]

[0125] Among them The integer value vector at the moment is The activation function is sigmoid(·), and the number of dimensions is D;

[0126] Calculate the influence of relationship data:

[0127]

[0128] The influence of the ιth relation data is θ ι , the adjustment factor is ω, the influence degree is σ, the The ιth relation data n at time ι The variable is

[0129] The relationship data with an influence greater than 0.375 is regarded as the shadow characteristic.

[0130] In this embodiment, the method of inputting the test data into the second timing impact analysis model to obtain the time-varying characteristics includes:

[0131] Input the test data into the second timing impact analysis model to calculate the composite material life difference:

[0132]

[0133] Among them The moment and The composite material life difference at each moment is No. The composite material life at that moment is No. The life of the composite material after the test is

[0134] The test data of composite material life difference greater than 0.12 is taken as key data, and the fuzzy similarity relationship matrix is ​​established based on the key data. The expression is:

[0135]

[0136] The similarity matrix element in the lth row and ith column is The center of the pth key data in the lth row is The number of key data is The genetic factor in row l and column i is

[0137] To solve the fuzzy equivalent matrix of the fuzzy similarity relationship matrix, the expression is:

[0138]

[0139] The fuzzy equivalence matrix is ​​ε * , the fuzzy equivalent elements are The fuzzy equivalent elements are sorted in descending order, and the test data with elements 1 in the matrix are regarded as the first-class data, and the rest as the second-class data;

[0140] Sort the first-class data and the second-class data by time to obtain the first-class time data and the second-class time data.

[0141] Input the first-class time data and the second-class time data into the dynamic adjustment modal fusion model, and input the fusion result. The expression is:

[0142]

[0143] The fused data is One type of time data is The second type of time data is The exclusive OR logical operator is The fused data is projected into a low-dimensional space through a linear layer, and the expression is:

[0144]

[0145] The linear rectification function is ReLU(·), and the weight of the fusion feature is The adjustment parameters of the fused data are The low-dimensional features are Input low-dimensional features as time-varying features.

[0146] In this embodiment, the method of clustering and grading the time-varying features according to the recovery degree to obtain graded data includes:

[0147] Calculate the weights of time series features:

[0148]

[0149] The weight of the xth time series feature is β x , the upper limit of observation time is q, and the xth time series feature is g x , the time series feature at the sth moment is g s , the number of time series features is n;

[0150] Calculate the similarity of neighbor time series features:

[0151]

[0152] The similarity between the xth time series feature and the uth time series feature is The number of neighboring sets of the xth time series feature and the uth time series feature is K, and the number of neighboring sets of the xth time series feature and the uth time series feature is n. k , No. The time series feature weight of a sample is No. The xth time series feature of samples is No. The u-th time series feature of a sample is The weighted distance between the xth time series feature and the uth time series feature is sd xu , the weighted distance between the xth time series feature and the tth time series feature is sd xt , the weighted distance between the uth time series feature and the tth time series feature is sd ut , the number of samples in the neighboring set K is |K|;

[0153] The time series features with similarity greater than 0.365 are divided into the initial category, and the remaining time series features are divided into the initial category with the largest similarity;

[0154] Calculate the information content of time-varying features:

[0155]

[0156] The category of the xth time series feature is The information content of the xth time series feature is fr x ,category The probability of The number of muscles of the xth time series feature is R a , the number of joints of the xth time series feature is S a ;

[0157] The time-varying feature at the center of the category is taken as the category center, and the minimum deviation error between the time-varying feature and the category center is calculated:

[0158]

[0159] The xth time series feature and category center The minimum deviation error is The number of time series features of the category is The category center of the x-th time series feature is

[0160] Compute the information score of time-varying features:

[0161]

[0162] The information score of the xth time-varying feature is No. Categories The maximum likelihood value is B max (s a ), the parameter value of the xth time-varying feature is No. Categories The contribution of

[0163] The time-varying features with information scores less than 0.385 are regarded as weak features, and the weak features are placed in the remaining categories. The information scores are recalculated, and the weak features are placed in the category with the largest information score to obtain classified data;

[0164] Calculate the recovery caused by the change of time-varying characteristics:

[0165]

[0166] The elastic weight is ζ1, the stress weight is ζ2, and the initial deformation is ΔF bg , the amount of recovery after deformation is ΔF cv , the initial stress is ΔG bg , the stress after relaxation is ΔG cv, the recovery degree of the xth time-varying feature is

[0167] The time series features with a recovery degree less than 0.27 are regarded as the first level, the time series features with a recovery degree greater than 0.27 and less than 0.53 are regarded as the second level, and the time series features with a recovery degree greater than 0.53 and less than 1 are regarded as the third level, and the classified data are output.

[0168] In this embodiment, the method for constructing a deformation memory life evaluation model according to the deformation characteristics and the classification data includes:

[0169] The deformation memory life evaluation model includes time series segmentation method, empirical mode decomposition method, time series analysis algorithm, deep long short-term memory network algorithm, and genetic optimization algorithm;

[0170] The time series segmentation method divides the input data into training data and test data according to the ratio of 7:3 in chronological order;

[0171] The empirical mode decomposition method decomposes the training data signal into the intrinsic mode function, extracts the local features of the signal through the intrinsic mode function, and obtains the deformation features;

[0172] Time series analysis algorithms obtain temporal deformation features by identifying and modeling patterns and trend features in time series data and analyzing the behavior of deformation features;

[0173] The deep long short-term memory network algorithm captures the long-term dependency in the time series by learning the matching degree of the time deformation characteristics and the deformation characteristics of the composite material under different deformation and loading conditions, and predicts the deformation memory life of the composite material based on the long-term dependency and classification data. The test data is input into the deep long short-term memory network algorithm to obtain the predicted deformation memory life.

[0174] The genetic optimization algorithm adjusts the learning rate of the deformation memory life evaluation model according to the actual deformation memory life of the test data and the mean square error of the predicted deformation memory life until the mean square error is 0.17.

[0175] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the deformation memory life of a carbon nanotube fiber composite material, characterized in that: The following steps are involved: Collecting test data and impact data of the carbon nanotube fiber composite material, and preprocessing the test data and the impact data; Perform correlation screening on the impact data according to the deformation degree to obtain relationship data, and input the relationship data into a first time series impact analysis model to obtain deformation characteristics; Inputting the test data into a second time series impact analysis model to obtain time-varying features, and clustering and grading the time-varying features according to the recovery degree to obtain graded data; A deformation memory life evaluation model is constructed according to the deformation characteristics and the classification data, the data to be evaluated is input into the deformation memory life evaluation model, and an evaluation result is output.

2. The method for evaluating the deformation memory life of a carbon nanotube fiber composite material according to claim 1, characterized in that: The method of obtaining relationship data by performing relevant screening on the influence data according to the deformation degree includes: Calculate the information weight that affects the data: The information weight of the jth influencing data is The variance of the wth influencing data and the jth influencing data is The jth influencing data is k j , the mean of the w-th influencing data is The number of w-th influencing data is n-1, and the number of influencing data k is j The probability of occurrence is P(k j ); The influence data with information weight greater than 0.317 are taken as key data, and the non-cotyledon node sample set is generated through the key data to calculate the similarity matrix and degree matrix: The similarity matrix between the a-th sample and the y-th sample at the κ-th node is The number of samples of category z is m z , the category of the ath sample is c a , the category of the yth sample is c y , the adjacency matrix of the a-th sample and the y-th sample of the κ-th node is The degree matrix of the k-th node is M κ , the kernel width is v, the distance between the a-th sample and the y-th sample is d(a,y), and the control factor is σ; Calculate the deformation correlation matrix and goodness matrix: The deformation correlation matrix of the a-th sample and the y-th sample at the κ-th node is U, and the goodness matrix of the κ-th node is Q κ , the genetic number of the κth node is Use singular value decomposition to obtain the features of relevant data and calculate the importance of the features: The importance of the rth feature is The rth eigenvector is h r , transposed to T, the identity matrix is ​​E, and the transposed of the rth eigenvector is Calculate the deformation of the relevant data: The number of features of the relevant data is N, and the deformation degree of the jth relevant data is The number of relevant data is b, the distance between the jth relevant data and the fth relevant data is d(j,f), and the difference in deformation length before and after the relevant data changes is ΔH; The relevant data with deformation degree greater than 0.469 are output as relational data.

3. The method for evaluating the deformation memory life of a carbon nanotube fiber composite material according to claim 1, characterized in that: The method of inputting the relationship data into a first timing impact analysis model to obtain the shadow characteristics includes: Input the relational data into the time series impact analysis model to extract the time pattern information of the relational data; The time pattern information is input into the attention mechanism to calculate the correlation between different dimensional relationship data and time state: The sliding window The number of The temporal pattern information of the relational data in the υth convolution kernel on the κth dimension is: Time is The time limit is C, The hidden state at this moment is The weight matrix is Temporal pattern information and hidden state of the κth dimension The correlation is No. The υth convolution kernel at the moment is The k-th dimension relation data The time mode information of the moment is Compute an integer-valued vector: Among them The integer value vector at the moment is The activation function is sigmoid(·), and the number of dimensions is D; Calculate the influence of relationship data: The influence of the ιth relation data is θ ι , the adjustment factor is ω, the influence degree is σ, the The ιth relation data n at time ι The variable is The relationship data with an influence greater than 0.375 is regarded as the shadow characteristic.

4. The method for evaluating the deformation memory life of a carbon nanotube fiber composite material according to claim 1, characterized in that: The method of inputting the test data into a second timing impact analysis model to obtain time-varying characteristics comprises: Input the test data into the second timing impact analysis model to calculate the composite material life difference: Among them The moment and The composite material life difference at each moment is No. The composite material life at that moment is No. The life of the composite material after the test is The test data of composite material life difference greater than 0.12 is taken as key data, and the fuzzy similarity relationship matrix is ​​established based on the key data. The expression is: The similarity matrix element in the lth row and ith column is The center of the pth key data in the lth row is The number of key data is The genetic factor in row l and column i is To solve the fuzzy equivalent matrix of the fuzzy similarity relationship matrix, the expression is: The fuzzy equivalence matrix is ​​ε * , the fuzzy equivalent elements are The fuzzy equivalent elements are sorted in descending order, and the test data with elements 1 in the matrix are regarded as the first-class data, and the rest as the second-class data; Sort the first-class data and the second-class data by time to obtain the first-class time data and the second-class time data. Input the first-class time data and the second-class time data into the dynamic adjustment modal fusion model, and input the fusion result. The expression is: The fused data is One type of time data is The second type of time data is The XOR logical operator is The fused data is projected into a low-dimensional space through a linear layer, and the expression is: The linear rectification function is ReLU(·), and the weight of the fusion feature is The adjustment parameters of the fused data are The low-dimensional features are Input low-dimensional features as time-varying features.

5. The method for evaluating the deformation memory life of a carbon nanotube fiber composite material according to claim 1, characterized in that: The method for clustering and grading the time-varying features according to the recovery degree to obtain graded data includes: Calculate the weights of time series features: The weight of the xth time series feature is β x , the upper limit of observation time is q, and the xth time series feature is g x , the time series feature at the sth moment is g s , the number of time series features is n; Calculate the similarity of neighbor time series features: The similarity between the xth time series feature and the uth time series feature is The number of neighboring sets of the xth time series feature and the uth time series feature is K, and the number of neighboring sets of the xth time series feature and the uth time series feature is n. k , No. The time series feature weight of a sample is No. The xth time series feature of samples is No. The u-th time series feature of a sample is The weighted distance between the xth time series feature and the uth time series feature is sd xu , the weighted distance between the xth time series feature and the tth time series feature is sd xt , the weighted distance between the uth time series feature and the tth time series feature is sd ut , the number of samples in the neighboring set K is |K|; The time series features with similarity greater than 0.365 are divided into the initial category, and the remaining time series features are divided into the initial category with the largest similarity; Calculate the information content of time-varying features: The category of the xth time series feature is The information content of the xth time series feature is fr x ,category The probability of The number of muscles of the xth time series feature is R a , the number of joints of the xth time series feature is S a ; The time-varying feature at the center of the category is taken as the category center, and the minimum deviation error between the time-varying feature and the category center is calculated: The xth time series feature and category center The minimum deviation error is The number of time series features of the category is The category center of the x-th time series feature is Compute the information score of time-varying features: The information score of the xth time-varying feature is No. Categories The maximum likelihood value is B max (s a ), the parameter value of the xth time-varying feature is No. Categories The contribution of The time-varying features with information scores less than 0.385 are regarded as weak features, and the weak features are placed in the remaining categories. The information scores are recalculated, and the weak features are placed in the category with the largest information score to obtain classified data; Calculate the recovery caused by the change of time-varying characteristics: The elastic weight is ζ1, the stress weight is ζ2, and the initial deformation is ΔF bg , the amount of recovery after deformation is ΔF cv , the initial stress is ΔG bg , the stress after relaxation is ΔG cv , the recovery degree of the xth time-varying feature is The time series features with a recovery degree less than 0.27 are regarded as the first level, the time series features with a recovery degree greater than 0.27 and less than 0.53 are regarded as the second level, and the time series features with a recovery degree greater than 0.53 and less than 1 are regarded as the third level, and the classified data are output.

6. The method for evaluating the deformation memory life of a carbon nanotube fiber composite material according to claim 1, characterized in that: The method for constructing a deformation memory life evaluation model according to the deformation characteristics and the classification data comprises: The deformation memory life evaluation model includes time series segmentation method, empirical mode decomposition method, time series analysis algorithm, deep long short-term memory network algorithm, and genetic optimization algorithm; The time series segmentation method divides the input data into training data and test data according to the ratio of 7:3 in chronological order; The empirical mode decomposition method decomposes the training data signal into the intrinsic mode function, extracts the local features of the signal through the intrinsic mode function, and obtains the deformation features; Time series analysis algorithms obtain temporal deformation features by identifying and modeling patterns and trend features in time series data and analyzing the behavior of deformation features; The deep long short-term memory network algorithm captures the long-term dependency in the time series by learning the matching degree of the time deformation characteristics and the deformation characteristics of the composite material under different deformation and loading conditions, and predicts the deformation memory life of the composite material based on the long-term dependency and classification data. The test data is input into the deep long short-term memory network algorithm to obtain the predicted deformation memory life. The genetic optimization algorithm adjusts the learning rate of the deformation memory life evaluation model according to the actual deformation memory life of the test data and the mean square error of the predicted deformation memory life until the mean square error is 0.17.