Intelligent Identification Method for Fatigue Damage of Spacecraft Structures Based on Fiber Optic Sensing and Ultrasonic Guided Waves

Abstract

The present invention provides an intelligent identification method for spacecraft structural fatigue damage based on optical fiber sensing and ultrasonic guided waves. The method comprehensively acquires spacecraft structural status data by combining optical fiber sensing and ultrasonic guided waves. Based on this, a deep network model is constructed using multiple noise reduction autoencoders (DAEs). The method adaptively extracts structural damage features directly from the spacecraft structural status data, and realizes the integrated processing of feature extraction and damage identification. This method gets rid of the dependence on manual experience and a large number of signal processing technologies, improves the damage identification ability and generalization performance of the model, and realizes the autonomous and accurate identification of spacecraft structural fatigue damage.

Description

Technical Field

[0001] The present invention relates to the technical field of spacecraft structure health monitoring, and in particular to an intelligent identification method for spacecraft structure fatigue damage based on optical fiber sensing and ultrasonic guided waves. Background Art

[0002] Structural fatigue damage refers to the process by which a structure, under cyclic stress or strain, gradually develops localized permanent damage at one or several locations, leading to cracks or sudden complete fracture after a certain number of cycles. Statistics show that fatigue damage is the primary cause of structural failure and retirement in metal materials, accounting for approximately 80%-90% of fracture failures in metal components of mechanical equipment. Spacecraft operate in the harsh space environment, where mechanical structures such as solar panels and flexible robotic arms are subjected to various alternating loads, such as heat and force, over a long period of time, resulting in fatigue damage and fracture. Fatigue damage and fracture are the primary cause of failure in metal structures, and fatigue cracks are the initial stage of fatigue damage and fracture. Although small cracks in a spacecraft may not significantly impact structural performance, they can continue to propagate under the cyclic effects of external alternating loads. If such damage remains undetected for an extended period, it can lead to a precipitous decline in structural performance, resulting in irreversible damage and potential hazards, posing significant risks to space flight and even mission failure. Therefore, identifying fatigue damage in spacecraft structures and monitoring fatigue damage in spacecraft structures in real time are of great significance for ensuring the safe and stable operation of spacecraft in orbit.

[0003] Structural damage identification can be treated as a pattern recognition problem. Different damage patterns, damage locations, and damage degrees have different damage characteristics. Traditional structural damage identification methods mainly adopt the "signal processing-based feature extraction + machine learning model" model. The quality of feature extraction largely determines the performance of damage identification. Common feature extraction methods include statistical analysis, autoregressive models, wavelet analysis, and empirical mode decomposition. However, these methods require a deeper understanding of the extracted signals and the use of a variety of signal processing techniques. At the same time, traditional structural damage identification methods use shallow models (such as support vector machines) to characterize the complex mapping relationship between structural state signals and structural damage. The model's damage identification ability and generalization performance are obviously insufficient.

[0004] In recent years, deep learning, with its powerful automatic feature extraction capabilities, has achieved significant success in fields such as image and speech recognition. Deep learning aims to simulate the brain's learning process, building deep models and combining them with massive amounts of training data to learn the implicit features of the data. This allows it to capture the rich intrinsic information of the data and improve the accuracy of classification or prediction. Deep learning can also play a significant role in the field of spacecraft structural health monitoring. Summary of the Invention

[0005] The present invention provides an intelligent identification method for spacecraft structural fatigue damage based on optical fiber sensing and ultrasonic guided waves. The method comprehensively acquires spacecraft structural status data by combining optical fiber sensing and ultrasonic guided waves. Based on this, a deep fatigue damage identification model is established to adaptively extract structural damage features directly from the spacecraft structural status data, thus breaking away from the reliance on manual experience and a large number of signal processing technologies, and realizing autonomous and accurate identification of spacecraft structural fatigue damage.

[0006] The present invention provides a method for intelligently identifying fatigue damage of spacecraft structures based on optical fiber sensing and ultrasonic guided waves, comprising the following steps:

[0007] Step 1: Installing fiber Bragg grating sensors and ultrasonic guided wave piezoelectric sensors on the surface of the spacecraft structure;

[0008] Step 2: Using the fiber Bragg grating sensor to monitor changes in the spacecraft structure to determine whether the spacecraft structural state is abnormal, and using the ultrasonic guided wave piezoelectric sensor to emit a guided wave signal according to a preset time period to actively detect the spacecraft structure to determine whether the spacecraft structural state is abnormal;

[0009] Step 3: If it is determined that the spacecraft structure is abnormal, the structural strain data measured by the fiber Bragg grating sensor and the structural ultrasonic guided wave response signal measured by the ultrasonic guided wave piezoelectric sensor are obtained, and the structural ultrasonic guided wave response signal is subjected to FFT transformation and normalization processing to obtain normalized structural ultrasonic guided wave response signal spectrum data;

[0010] Step 4: Input the structural strain data and the normalized structural ultrasonic guided wave response signal spectrum data into a pre-trained fatigue damage identification model to obtain the damage location and damage extent of the spacecraft.

[0011] Optionally, the fatigue damage identification model includes an input layer, a hidden layer and an output layer, the input layer is used to receive the structural strain data and the normalized structural ultrasonic guided wave response signal spectrum data, the hidden layer is composed of a plurality of noise reduction autoencoders stacked together, and is used to extract the structural fatigue damage features contained in the structural strain data and the normalized structural ultrasonic guided wave response signal spectrum data, and the output layer contains a classifier, and is used to classify the structural damage location and damage degree using the structural fatigue damage features extracted by the hidden layer.

[0012] Furthermore, the denoising autoencoder includes an encoding network and a decoding network, wherein the encoding network adds noise to the input data in the high-dimensional space and converts the input data in the high-dimensional space into an encoding vector in the low-dimensional space, and the decoding network reconstructs the encoding vector in the low-dimensional space into the input data in the high-dimensional space.

[0013] Furthermore, before step 4, the following steps are also included:

[0014] Fiber Bragg grating sensors are used to collect structural strain sample data of non-destructive spacecraft specimens and fatigue-damaged spacecraft specimens. Ultrasonic guided wave piezoelectric sensors are used to collect structural ultrasonic guided wave response signal spectrum sample data of non-destructive spacecraft specimens and fatigue-damaged spacecraft specimens. The structural strain sample data and the structural ultrasonic guided wave response signal spectrum sample data are combined into a training sample data set.

[0015] Determine the number of input layer nodes of the fatigue damage identification model to be trained according to the dimensions of the structural strain sample data and the structural ultrasonic guided wave response signal spectrum sample data, determine the number of output layer nodes of the fatigue damage identification model to be trained according to the output result, and select the activation function, optimizer, learning rate and model parameters of the fatigue damage identification model to be trained based on experience;

[0016] The training sample data set is used to train the fatigue damage recognition model to be trained. The training process mainly includes a pre-training stage and a fine-tuning stage:

[0017] In the pre-training stage, the model is trained layer by layer using unsupervised learning. First, the training sample data x m Train the first denoising autoencoder DAE1 and transform x m Encoded as

[0018]

[0019] Where θ1 is the parameter of DAE1;

[0020] Then, It is trained as the input of the second denoising autoencoder DAE2 and encoded as Repeat the above process until the Nth denoising autoencoder DAE is completed N training, and input x m The final code is

[0021]

[0022] After completing the pre-training, the supervised learning method is used to fine-tune the parameters of the entire model through the error back propagation algorithm. The output of the model is expressed as,

[0023]

[0024] Where θ N+1 is the parameter of the output layer;

[0025] x m The actual classification is d m , the model minimizes φ SDAE (Θ) Complete fine-tuning,

[0026]

[0027] Where Θ is the parameter set of the model, Θ={θ1,θ2,…,θ N+1}, L is the reconstruction error function, usually using the traditional variance cost function,

[0028] L(y m ,d m )=||d m -y m || 2 ;

[0029] After the fatigue damage recognition model training is completed, the model training time and training accuracy indicators are judged. If the training indicators do not meet the requirements, the model activation function, optimizer, learning rate and model structure parameters are adjusted to continue training until all training indicators meet the requirements, and the trained fatigue damage recognition model is obtained.

[0030] Furthermore, before step 1, the method further includes:

[0031] Finite element analysis software is used to perform numerical simulation of spacecraft fatigue damage signals, and the layout positions of fiber Bragg grating sensors and ultrasonic guided wave piezoelectric sensors are determined based on the simulation results.

[0032] The present invention provides an intelligent identification method for fatigue damage of spacecraft structures based on optical fiber sensing and ultrasonic guided waves. Compared with the existing technology, the present invention has the following beneficial effects: (1) the spacecraft structure status data is comprehensively acquired by combining passive monitoring of optical fiber sensing with active detection of ultrasonic guided waves; (2) in terms of feature extraction, the method gets rid of the dependence on a large number of signal processing technologies and manual experience, directly and adaptively extracts damage features from the spacecraft structure status data, and realizes the integrated processing of feature extraction and damage identification; (3) a deep network model is established to characterize the complex mapping relationship between structural status and health status, thereby improving the damage identification ability and generalization performance of the model and meeting the practical application requirements of fatigue damage identification of on-orbit spacecraft structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0034] Figure 1 This is a schematic diagram of the dimensions of a spacecraft structure fatigue specimen;

[0035] Figure 2 Flowchart for the preparation of fatigue test specimens for spacecraft structures;

[0036] Figure 3 This is the layout diagram of the fiber Bragg grating sensor;

[0037] Figure 4 This is the layout diagram of the ultrasonic guided wave piezoelectric sensor;

[0038] Figure 5 Installation location diagram for spacecraft structural fatigue test specimens;

[0039] Figure 6 This is a working process diagram of the fatigue damage identification model;

[0040] Figure 7 This is the network structure diagram of the automatic encoder;

[0041] Figure 8 This is a flow chart of the intelligent identification method for spacecraft structure fatigue damage based on fiber optic sensing and ultrasonic guided waves. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0043] refer to Figure 1-8 In this embodiment, an aluminum alloy standard specimen is used to simulate a spacecraft for structural fatigue damage identification. First, a spacecraft structural fatigue specimen is prepared according to the actual structure of the spacecraft. The specific steps are as follows:

[0044] The spacecraft structure fatigue test specimens were prepared using aluminum alloy sheet of model 6061-T6, with a density of ρ = 2700 kg / m 3 , elastic modulus E = 75.6 GPa, Poisson's ratio v = 0.33, specimen dimensions are as follows Figure 1As shown, the thickness is 5mm. A 5mm long opening is prefabricated at the middle edge of the specimen. The stress at the top of the prefabricated opening is more concentrated, which is conducive to the initiation and propagation of fatigue cracks.

[0045] Using a fatigue testing machine, fatigue specimens are prepared by high cycle fatigue loading. The preparation process is as follows: Figure 2 As shown in the figure, 13 specimens were selected during the preparation process, numbered 0 to 12. Specimen 0 was a non-destructive specimen and was not fatigue loaded. Ultrasonic guided wave and optical fiber strain detection were subsequently performed directly, and the results were normalized as the benchmark values of the other fatigue specimens. The other specimens were subjected to tensile fatigue loading while ensuring that the experimental conditions were exactly the same. Specimens with different degrees of damage were prepared by loading different numbers of fatigue cycles within the fatigue life.

[0046] Then, sensors are placed on the spacecraft structural fatigue specimen. The specific steps are as follows:

[0047] Use ABAQUS finite element analysis software to perform numerical simulations of spacecraft fatigue damage signals, explore the propagation characteristics of ultrasonic guided waves in fatigue specimens, and explore the strain field information of crack propagation in fatigue specimens, providing a basis for sensor placement.

[0048] The fiber Bragg grating sensor is arranged as follows Figure 3 As shown, each sensor measures 25mm*5.5mm. Six fiber Bragg grating sensors are pasted in two rows and symmetrically distributed about the crack in the horizontal direction to monitor the relationship between the difference in crack propagation direction and the strain field. Three sensors are evenly spaced in the vertical direction to monitor the relationship between crack extension and the strain field.

[0049] Ultrasonic guided wave piezoelectric sensor arrangement as follows Figure 4 As shown, the sensor layout uses a single-sided pitch-catch pattern. In this pattern, the driving and receiving sensors are located on either side of the crack in the specimen. The response signals collected by the sensors include both scattered and transmitted waves from the crack. To reduce aliasing between scattered or transmitted sound waves from the damage and reflected waves from the right boundary of the structure, the distance between the receiving sensor and the right boundary of the specimen should be kept within a certain range. Furthermore, the distance between the driving and receiving sensors should be appropriately selected.

[0050] Next, we use the prepared spacecraft structure fatigue specimens to collect model training sample data and build and train the fatigue damage identification model. The specific steps are as follows:

[0051] The spacecraft structure fatigue specimens No. 0-12 were installed at positions 1 and 2 of the spacecraft simulation device to simulate the fatigue damage of the solar panels and the fatigue damage of the robotic arm. Figure 5As shown in the figure, a fiber Bragg grating sensor and a fiber Bragg grating sensor demodulator were used to collect the structural strain data of the non-destructive specimen and the fatigue damaged specimen. A piezoelectric ultrasonic guided wave detector was used to synchronously collect the structural ultrasonic guided wave response signals of the corresponding specimens, and FFT transformation and normalization were performed. The structural strain data and structural ultrasonic guided wave response signals of the non-destructive specimen and the fatigue damaged specimen were used to form a training sample data set, which was divided into a training set and a test set in a ratio of 9:1.

[0052] Multiple denoising autoencoders (DAEs) are used to build a fatigue damage recognition model, such as Figure 6 As shown in Figure 1, the fatigue damage identification model includes an input layer, a hidden layer, and an output layer. The input layer is used to receive structural strain data and normalized structural ultrasonic guided wave response signal spectrum data; the hidden layer is composed of a stack of multiple denoising autoencoders (DAEs) to extract the structural fatigue damage features contained in the structural strain data and the normalized structural ultrasonic guided wave response signal spectrum data. The autoencoder (AE) includes an encoding network and a decoding network, as shown in Figure 1. Figure 7 As shown in the figure, the encoding network converts the input data in high-dimensional space into encoding vectors in low-dimensional space, and the decoding network reconstructs the encoding vectors in low-dimensional space into input data in high-dimensional space. The denoising autoencoder (DAE) is an improvement of AE. Its core idea is that the encoding network adds noise with certain statistical characteristics to the input data and then encodes the input data. The decoding network then estimates the original form of the input data from the noisy data, thereby learning more robust features from the noisy input data and reducing the model's sensitivity to small random perturbations. The output layer contains a classifier, which is used to classify the structural damage location and damage degree using the structural fatigue damage features extracted by the hidden layer. The number of input layer nodes of the fatigue damage recognition model to be trained is determined according to the dimension of the structural strain sample data and the structural ultrasonic guided wave response signal spectrum sample data. The number of output layer nodes of the fatigue damage recognition model to be trained is determined according to the output results (crack location, crack length). The activation function, optimizer, learning rate and model parameters (such as the number of hidden layers, the number of hidden layer neurons and the number of iterations) of the fatigue damage recognition model to be trained are selected based on experience.

[0053] The training sample data set is used to train the fatigue damage recognition model to be trained. The training process mainly includes the pre-training stage and the fine-tuning stage. The pre-training stage uses an unsupervised learning method to train the model layer by layer to effectively mine the fatigue damage characteristics in the input data. First, the training sample data x m Train the first denoising autoencoder DAE1 and transform x m Encoded as

[0054]

[0055] Where θ1 is the parameter of DAE1;

[0056] Then, It is trained as the input of the second denoising autoencoder DAE2 and encoded as Repeat the above process until the Nth denoising autoencoder DAE is completed N training, and input x m The final code is

[0057]

[0058] After completing the pre-training, in order to realize the identification of structural damage, an output layer with classification function is added, and the supervised learning method is used to fine-tune the parameters of the entire model through the error back propagation algorithm. The output of the model is expressed as,

[0059]

[0060] Where θ N+1 is the parameter of the output layer;

[0061] x m The actual classification is d m , the model minimizes φ SDAE (Θ) Complete fine-tuning,

[0062]

[0063] Where Θ is the parameter set of the model, Θ={θ1,θ2,…,θ N+1}, L is the reconstruction error function, usually using the traditional variance cost function,

[0064] L(y m ,d m )=||d m -y m || 2 ;

[0065] After the fatigue damage recognition model training is completed, the model training time and training accuracy indicators are judged. If the training indicators do not meet the requirements, the model activation function, optimizer, learning rate and model structure parameters are adjusted to continue training until all training indicators meet the requirements. The trained fatigue damage recognition model is obtained and tested.

[0066] Finally, test the recognition accuracy of the trained fatigue damage recognition model:

[0067] The trained fatigue damage recognition model was tested using the test set as input. The test results showed that the recognition accuracy rate was ≥90%, which is capable of autonomous and accurate identification of fatigue damage in spacecraft structures.

[0068] In practical applications, such as Figure 8 As shown in Figure 2, the specific steps for identifying fatigue damage in spacecraft structures are as follows:

[0069] Step 1: Installing fiber Bragg grating sensors and ultrasonic guided wave piezoelectric sensors on the surface of the spacecraft structure;

[0070] Step 2: Using the fiber Bragg grating sensor to monitor changes in the spacecraft structure to determine whether the spacecraft structural state is abnormal, and using the ultrasonic guided wave piezoelectric sensor to emit a guided wave signal according to a preset time period to actively detect the spacecraft structure to determine whether the spacecraft structural state is abnormal;

[0071] A fiber Bragg grating sensor is set on the surface of the spacecraft structure. Its parameters will change due to the strain caused by structural fatigue damage. The spacecraft structure strain data is obtained by obtaining the changes in the grating echo spectral characteristics. The fiber Bragg grating sensor works continuously and performs long-term passive monitoring of the spacecraft structure status.

[0072] At the same time, ultrasonic guided wave piezoelectric sensors are installed on the surface of the spacecraft structure for active ultrasonic guided wave detection. Once the fiber optic sensing technology detects an abnormal structural state, the ultrasonic guided wave driver sensor is triggered to generate a guided wave signal. The receiving sensor then collects the ultrasonic guided wave response signal from the spacecraft structure, actively detecting the spacecraft structural state.

[0073] In addition, in order to avoid the omission of structural abnormalities due to fiber optic sensing technology when the structure is damaged (for example, if the crack area is far away from the fiber Bragg grating sensor, the fiber Bragg grating sensor will not produce a strain response to the structural crack), the ultrasonic guided wave device will periodically generate guided wave signals according to predetermined settings to actively detect the structural status of the spacecraft.

[0074] Step 3: If it is determined that the spacecraft structure is abnormal, the structural strain data measured by the fiber Bragg grating sensor and the structural ultrasonic guided wave response signal measured by the ultrasonic guided wave piezoelectric sensor are obtained, and the structural ultrasonic guided wave response signal is subjected to FFT transformation and normalization processing to obtain normalized structural ultrasonic guided wave response signal spectrum data;

[0075] Step 4: Input the structural strain data and the normalized structural ultrasonic guided wave response signal spectrum data into a pre-trained fatigue damage identification model to obtain the damage location and damage extent of the spacecraft.

[0076] The damage location and damage degree may specifically be a crack location and a crack length.

[0077] Through the above-mentioned intelligent identification method of spacecraft structural fatigue damage based on optical fiber sensing and ultrasonic guided waves, the spacecraft structural status data is comprehensively acquired by combining passive monitoring of optical fiber sensing with active detection of ultrasonic guided waves. A deep network model is constructed using multiple noise reduction autoencoders (DAEs) to adaptively extract damage features from the spacecraft structural status data, and realize the integrated processing of feature extraction and damage identification, thereby improving the damage identification ability and generalization performance of the model and meeting the practical application needs of fatigue damage identification of on-orbit spacecraft structures.

[0078] Although the present invention has been disclosed above by way of embodiments, they are not intended to limit the present invention. Any person skilled in the art may make slight changes and modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the claims.

Claims

1. An intelligent identification method for fatigue damage of spacecraft structures based on optical fiber sensing and ultrasonic guided waves, characterized in that: The following steps are involved: Step 1: Installing fiber Bragg grating sensors and ultrasonic guided wave piezoelectric sensors on the surface of the spacecraft structure; Step 2: Using the fiber Bragg grating sensor to monitor changes in the spacecraft structure to determine whether the spacecraft structural state is abnormal, and using the ultrasonic guided wave piezoelectric sensor to emit a guided wave signal according to a preset time period to actively detect the spacecraft structure to determine whether the spacecraft structural state is abnormal; Step 3: If it is determined that the spacecraft structure is abnormal, the structural strain data measured by the fiber Bragg grating sensor and the structural ultrasonic guided wave response signal measured by the ultrasonic guided wave piezoelectric sensor are obtained, and the structural ultrasonic guided wave response signal is subjected to FFT transformation and normalization processing to obtain normalized structural ultrasonic guided wave response signal spectrum data; Step 4: Input the structural strain data and the normalized structural ultrasonic guided wave response signal spectrum data into a pre-trained fatigue damage identification model to obtain the damage location and damage extent of the spacecraft.

2. The intelligent identification method for spacecraft structural fatigue damage based on optical fiber sensing and ultrasonic guided waves according to claim 1 is characterized in that: The fatigue damage identification model includes an input layer, a hidden layer, and an output layer. The input layer is used to receive the structural strain data and normalized structural ultrasonic guided wave response signal spectrum data. The hidden layer is composed of a stack of multiple noise reduction autoencoders and is used to extract structural fatigue damage features contained in the structural strain data and the normalized structural ultrasonic guided wave response signal spectrum data. The output layer contains a classifier for classifying the structural damage location and damage degree using the structural fatigue damage features extracted by the hidden layer.

3. The intelligent identification method for spacecraft structural fatigue damage based on optical fiber sensing and ultrasonic guided waves according to claim 2 is characterized in that: The denoising autoencoder includes an encoding network and a decoding network. The encoding network adds noise to input data in a high-dimensional space and converts the input data in the high-dimensional space into an encoding vector in a low-dimensional space. The decoding network reconstructs the encoding vector in the low-dimensional space into input data in the high-dimensional space.

4. The intelligent identification method for spacecraft structural fatigue damage based on optical fiber sensing and ultrasonic guided waves according to claim 3 is characterized in that: The step 4 also includes: Fiber Bragg grating sensors are used to collect structural strain sample data of non-destructive spacecraft specimens and fatigue-damaged spacecraft specimens. Ultrasonic guided wave piezoelectric sensors are used to collect structural ultrasonic guided wave response signal spectrum sample data of non-destructive spacecraft specimens and fatigue-damaged spacecraft specimens. The structural strain sample data and the structural ultrasonic guided wave response signal spectrum sample data are combined into a training sample data set. Determine the number of input layer nodes of the fatigue damage identification model to be trained according to the dimensions of the structural strain sample data and the structural ultrasonic guided wave response signal spectrum sample data, determine the number of output layer nodes of the fatigue damage identification model to be trained according to the output result, and select the activation function, optimizer, learning rate and model parameters of the fatigue damage identification model to be trained based on experience; The fatigue damage recognition model to be trained is trained using the training sample data set, and the training process mainly includes a pre-training stage and a fine-tuning stage; In the pre-training stage, the model is trained layer by layer using unsupervised learning. First, the training sample data x m Train the first denoising autoencoder DAE1 and transform x m Encoded as Where θ1 is the parameter of DAE1; Then, It is trained as the input of the second denoising autoencoder DAE2 and encoded as Repeat the above process until the Nth denoising autoencoder DAE is completed N training, and input x m The final code is After pre-training, supervised learning is used to fine-tune the entire model parameters through the error back propagation algorithm. The output of the model is expressed as Where θ N+1 is the parameter of the output layer; x m The actual classification is d m , the model minimizes φ SDAE (Θ) Complete fine-tuning, Where Θ is the parameter set of the model, Θ={θ1,θ2,…,θ N+1 }, L is the reconstruction error function, usually using the traditional variance cost function, L(y m ,d m )=||d m -y m || 2 ; After the fatigue damage recognition model training is completed, the model training time and training accuracy indicators are judged. If the training indicators do not meet the requirements, the model activation function, optimizer, learning rate and model structure parameters are adjusted to continue training until all training indicators meet the requirements, and the trained fatigue damage recognition model is obtained.

5. The intelligent identification method for spacecraft structural fatigue damage based on optical fiber sensing and ultrasonic guided waves according to any one of claims 1 to 4, characterized in that: The step 1 also includes: Finite element analysis software is used to perform numerical simulation of spacecraft fatigue damage signals, and the layout positions of fiber Bragg grating sensors and ultrasonic guided wave piezoelectric sensors are determined based on the simulation results.

Images