A method and system for intelligent prediction of vibration of large-scale spatial deployment mechanisms

Abstract

The present invention relates to a method and system for intelligently predicting the vibration of a large-scale space deployment mechanism, comprising: obtaining data on wavelength changes in grating reflections caused by minor structural deformations due to vibration; calculating wavelength changes by receiving light waves reflected by a fiber grating sensor, converting the wavelength changes into spatial and temporal information of the vibration through signal demodulation, and analyzing and processing local vibration displacement data; using the analyzed and processed local vibration displacement data as input to a GCN-LSTM neural network; and predicting and calculating the overall vibration state of the large-scale space deployment mechanism based on a spatiotemporal coupling relationship to obtain the overall vibration displacement results of the mechanism. The present invention not only monitors the overall vibration of a large-scale space deployment mechanism in real time with high precision and high reliability, but also provides a decision-making basis for controlling the on-orbit state of a spacecraft, ensuring the high-precision attitude and stability of the spacecraft in a complex space environment.

Description

Technical Field

[0001] The present invention relates to the technical field of vibration prediction of large-scale space deployment mechanisms, and in particular to an intelligent vibration prediction method and system for large-scale space deployment mechanisms. Background Art

[0002] Large-scale deployment mechanisms are key components in the design and operation of spacecraft, widely used to deploy solar sails, antennas, and scientific instruments on satellites and probes. These mechanisms are typically composed of rigid and flexible bodies. Their structural complexity and the unique characteristics of the on-orbit environment present numerous challenges during deployment and locking. Specifically, as the mechanism unfolds and its operating state changes, complex coupled dynamics between the rigid and flexible bodies can easily lead to vibration and deformation. This vibration not only affects the stability of the mechanism but can also reduce the pointing and surface accuracy of the spacecraft, and in severe cases, even impact the success of the mission.

[0003] Traditional vibration monitoring methods mainly rely on single-point sensors, which cannot effectively capture the vibration characteristics of the entire mechanism, especially in complex space environments. Changes in the on-orbit operating environment, such as temperature, vacuum, and microgravity, bring additional complexity to the deployment process of large-scale space deployment mechanisms. How to efficiently and accurately predict the overall structural deformation information based on the local vibration displacement data collected from the sensors remains a technical challenge. To address the above problems, the present invention proposes an intelligent vibration prediction system for large-scale space deployment mechanisms. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention provides an intelligent vibration prediction method and system for large-scale space deployment mechanisms. This method addresses the problem of efficiently and accurately predicting overall structural deformation in complex space environments based on local vibration displacement data collected from sensors. This method utilizes fiber grating (FBG) sensors placed at key locations within the large-scale space deployment mechanism to monitor in real time the subtle local structural deformations caused by vibration. The FBG sensors collect local vibration displacement data based on the changes in reflection wavelength caused by vibration. During vibration monitoring, the FBG sensors accurately acquire deformation and vibration data at these locations and use this local vibration displacement data as input to a GCN-LSTM neural network. A neural network prediction algorithm utilizes the local vibration displacement data acquired by the fiber optic sensors, combined with spatiotemporal coupling, to predict and calculate the overall vibration state of the large-scale space deployment mechanism, obtaining the overall vibration displacement data. This system not only monitors the overall vibration of large-scale space deployment mechanisms in real time with high precision and reliability, but also provides a decision-making basis for on-orbit control of spacecraft, ensuring high-precision attitude and stability in complex space environments.

[0005] To solve the above technical problems, the present invention provides the following technical solution: a method for intelligently predicting vibration of a large-scale spatial deployment mechanism, comprising the following steps:

[0006] S1. Using multiple fiber Bragg grating sensors distributed at fixed positions on a large-scale spatial deployment mechanism, data on changes in the grating reflection wavelength caused by slight structural deformation due to vibration is obtained;

[0007] S2, by receiving the light waves reflected by the fiber Bragg grating sensor, calculating the wavelength change, and then converting the wavelength change into the spatial and temporal information of the vibration through signal demodulation, and analyzing and processing the local vibration displacement data;

[0008] S3. Construct a GCN-LSTM neural network combination model and use the analyzed and processed local vibration displacement data as the input of the GCN-LSTM neural network. Combined with the spatiotemporal coupling relationship, the GCN-LSTM neural network is used to establish a nonlinear relationship between the local vibration displacement data and the overall vibration displacement data. The overall vibration state of the large-scale spatial deployment mechanism is predicted and calculated to obtain the overall vibration displacement result of the mechanism.

[0009] Furthermore, in step S1, the specific process includes the following steps:

[0010] S11. Fiber Bragg grating sensors are placed at key structural points of the large-scale space deployment mechanism truss to monitor local vibrations at different locations.

[0011] S12. Fiber Bragg grating sensor transmits light into the optical fiber through a low-coherence light source. When the light wave passes through the grating in the optical fiber, the fiber Bragg grating will reflect the light of a specific wavelength back to the sensor according to its periodic structure, while the light of other wavelengths will continue to propagate, and the reflected center wavelength λ B satisfy:

[0012] λ B =2n eff Λ;

[0013] where n eff is the effective refractive index of the fiber material, and Λ is the period length of the grating.

[0014] Furthermore, in step S2, the specific process includes the following steps:

[0015] S21. Demodulate the optical signal of the fiber optic sensor into an electrical signal. The demodulated electrical signal represents the wavelength change caused by the structural vibration. The strain ε is calculated by detecting the change in the wavelength of the reflected light:

[0016]

[0017] Where Δλ is the change in reflection wavelength, λ0 is the initial reflection wavelength, and α is the strain sensitivity of the fiber material;

[0018] S22. The process of calculating the local vibration displacement L through the strain ε is as follows:

[0019] L = ε·L0;

[0020] Where L0 is the original length of the fiber Bragg grating.

[0021] Furthermore, in step S3, the local vibration displacement data is used as the input of the GCN network. The GCN network consists of multiple layers of graph convolution. Each layer performs a graph convolution operation, gradually aggregating features to a higher representation layer and outputs them through a fully connected layer for the vibration prediction task of large-scale spatial expansion mechanisms. The GCN network can be expressed by the following formula:

[0022]

[0023] Among them H (l+1) is the node feature matrix of the l+1th layer, σ is the nonlinear activation function, yes The degree matrix of , that is, a diagonal matrix, where the degree of each node is a diagonal element, It is the adjacency matrix A plus the identity matrix I to ensure that each node includes its own information, W (l) is the weight matrix of layer l, It normalizes the adjacency matrix A to prevent the eigenvalue of a node from changing with the number of neighbors, thus causing feature deviation.

[0024] At each layer, the GCN network aggregates the features of each node and passes the information of its neighboring nodes to the current node. The new features of a node v in the lth layer are It can be calculated by weighted average of neighbor node features:

[0025]

[0026] in represents the neighbor set of node v, d u and d v are the degrees of nodes u and v respectively, represents the hidden representation of node u at layer l;

[0027] In the vibration prediction of large-scale spatial deployment mechanisms, the GCN network obtains overall vibration displacement data from local vibration displacement data through layer-by-layer aggregation and convolution operations, thereby effectively capturing the local and global characteristics of the nodes and providing input data for the subsequent LSTM layer, thereby realizing time series modeling and prediction of vibration data.

[0028] Furthermore, in step S2, the input of the LSTM network consists of the overall vibration displacement data of multiple time steps. The data of each time step can be represented as a vector, which includes the displacement offset of each measurement position at the corresponding time point. In the LSTM network, the information flow is controlled by three main gating mechanisms. Based on the forget gate and the input gate, the LSTM unit state C t Will be updated, the hidden state h t The calculation combines the current unit state and the output gate. This hidden state will be used as the input of the next time step and can also be provided as an output to subsequent layers or networks.

[0029] Furthermore, the specific calculation process of the LSTM network includes the following steps:

[0030] S31, the forget gate determines which information in the current unit state needs to be forgotten, and which information should be discarded based on the current input and the previous hidden state. The gate uses the sigmoid activation function to output a value between 0 and 1, where 0 means complete discard and 1 means complete retention:

[0031] f t =σ(W f ·[h t-1 , x t ]+b f );

[0032] Among them, f t is the output of the forget gate, σ is the sigmoid activation function, W f is the weight matrix of the forget gate, h t-1 is the hidden state at the previous moment, x t is the input at the current moment, b f is the bias of the forget gate;

[0033] S32, the input gate controls the impact of the current input information on the unit state, controls the addition of the current input information, determines which information needs to be updated, and generates a new candidate value vector:

[0034] i t =σ(W i ·[h t-1 ,x t ]+b i );

[0035]

[0036] Among them, i t is the output of the input gate, is the candidate cell state, W i and W Care the weight matrices of the input gate and candidate unit state, b i and b C is the corresponding bias;

[0037] S33, the cell state is updated through the forget gate and input gate to maintain long-term and short-term dependency information:

[0038]

[0039] Among them, C t is the current cell state, C t-1 is the cell state at the previous moment.

[0040] S34, the output gate determines how the current unit state will affect the hidden state. It combines the current input and the previous hidden state and outputs a value to determine the next hidden state:

[0041] o t =σ(W o ·[h t-1 , x t ]+b o );

[0042] h t =o t *tanh(C t );

[0043] Among them, t is the output of the output gate, W o is the weight matrix of the output gate, b o is the bias of the output gate, h t is the hidden state at the current moment.

[0044] Furthermore, in step S3, a nonlinear relationship between the local vibration data and the overall vibration data is established using the GCN-LSTM neural network. The specific process includes the following steps:

[0045] S41. Assume that there are N sensors that collect local vibration displacement data at T time points. The data can be represented as an N×T matrix:

[0046]

[0047] where d ij represents the displacement of the i-th sensor at the j-th time point, 1≤i≤N,1≤j≤T;

[0048] If there is a connection between sensor i and sensor j, their connection relationship can be represented by the adjacency matrix A. ij =1, otherwise A ij= 0; Divide the data into multiple time windows so that LSTM can learn time series features. Select the time window size as W. Each input sample can contain data from W time points. Assuming that there are W time points of data as input samples, the constructed input data shape X is (N, W);

[0049] The constructed local vibration displacement data matrix D is input into the GCN layer together with the adjacency matrix A. The GCN layer uses the adjacency matrix to process the spatial features between sensors. After passing through the GCN layer, the overall vibration displacement data features obtained are input into the LSTM layer to learn the time series features. The GCN-LSTM network output layer directly predicts the displacement of the overall structure. Represents the combined displacement of multiple sensor positions:

[0050]

[0051] in represents the predicted displacement of the i-th sensor position.

[0052] Furthermore, the local vibration displacement data are set as training samples, verification samples and test samples respectively, where the data ratio of training samples, verification samples and test samples is 8:1:1. The training samples and test samples are used to train and test the GCN-LSTM neural network combination model to obtain a trained GCN-LSTM neural network combination model; the verification samples are input into the trained GCN-LSTM neural network combination model to output the prediction results.

[0053] Furthermore, a large-scale space deployment mechanism vibration intelligent prediction system is provided, comprising:

[0054] A data acquisition module, wherein the data acquisition module is used to acquire data on changes in grating reflection wavelength caused by slight structural deformation due to vibration through a plurality of fiber grating sensors distributed at fixed positions of the large spatial deployment mechanism;

[0055] A vibration monitoring module is used to receive light waves reflected by the fiber grating sensor, calculate the change in wavelength, convert the wavelength change into spatial and temporal information of vibration through signal demodulation, and perform analysis and processing of local vibration signal data;

[0056] The identification and prediction module is used to construct a GCN-LSTM neural network combination model. The analyzed and processed local vibration displacement data is used as the input of the GCN-LSTM neural network combination model. In combination with the spatiotemporal coupling relationship, the GCN-LSTM neural network is used to establish a nonlinear relationship between the local vibration displacement data and the overall vibration displacement data. The overall vibration state of the large-scale spatial deployment mechanism is predicted and calculated to obtain the overall vibration displacement result of the mechanism.

[0057] By means of the above technical solution, the present invention provides a method and system for intelligently predicting vibration of a large-scale spatial deployment mechanism, which has at least the following beneficial effects:

[0058] The present invention deploys fiber grating (FBG) sensors at key locations within a large-scale space deployment mechanism to monitor in real time the subtle local structural deformations caused by vibration. The FBG sensors collect local vibration displacement data based on the reflected wavelength changes caused by vibration. During the vibration monitoring process, the FBG sensors accurately acquire deformation and vibration data at these locations and use this local vibration displacement data as input to a GCN-LSTM neural network. A neural network prediction algorithm utilizes the local vibration displacement data acquired by the fiber optic sensors, combined with spatiotemporal coupling, to predict and calculate the overall vibration state of the large-scale space deployment mechanism, obtaining the overall vibration displacement data of the mechanism. This system not only monitors the overall vibration of the large-scale space deployment mechanism in real time with high precision and reliability, but also provides a decision-making basis for controlling the on-orbit state of the spacecraft, ensuring high-precision attitude and stability in complex space environments. This system addresses the problem of efficiently and accurately predicting overall structural deformation information in complex space environments based on local vibration displacement data collected from sensors. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0060] Figure 1 This is a structural diagram of a large-scale space deployment mechanism vibration intelligent prediction system of the present invention;

[0061] Figure 2 This is a schematic diagram of the positions of the optical fiber sensors and vibration monitoring modules of the intelligent vibration prediction system for a large-scale space deployment mechanism according to the present invention;

[0062] Figure 3 This is a schematic diagram of the network structure of the GCN-LSTM neural network of the large-scale space deployment mechanism vibration intelligent prediction system of the present invention.

[0063] In the figure, 1. Fiber Bragg grating sensor; 2. Vibration monitoring module. DETAILED DESCRIPTION

[0064] To make the above-mentioned objectives, features, and advantages of the present invention more clearly understood, the present invention is further described below in detail with reference to the accompanying drawings and specific embodiments. This will enable a full understanding of how this application uses technical means to solve technical problems and achieve technical effects, and to implement the invention accordingly.

[0065] Those skilled in the art will appreciate that all or part of the steps in the above-mentioned embodiment methods can be accomplished by instructing the relevant hardware through a program. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0066] Please refer to Figure 1-Figure 3 , shows a specific implementation of this embodiment. This embodiment uses this method to place fiber grating sensors at key locations of a large-scale space deployment mechanism to monitor in real time the tiny deformations of the local structure caused by vibration. The fiber grating sensors collect local vibration displacement data based on the changes in reflection wavelength caused by vibration. During the vibration monitoring process, the fiber grating sensors accurately acquire deformation and vibration data at local locations and use this local vibration displacement data as input to the GCN-LSTM neural network. Through a neural network prediction algorithm, the local vibration displacement data acquired by the fiber optic sensors is combined with the spatiotemporal coupling relationship to predict and calculate the overall vibration state of the large-scale space deployment mechanism, obtaining the overall vibration displacement data of the mechanism. This system not only monitors the overall vibration of the large-scale space deployment mechanism in real time with high precision and reliability, but also provides a decision-making basis for controlling the on-orbit state of the spacecraft, ensuring the high-precision attitude and stability of the spacecraft in complex space environments.

[0067] Please refer to Figure 1 This embodiment proposes a method and system for intelligent prediction of vibration of a large-scale spatial deployment mechanism, which includes the following steps:

[0068] S1. Using multiple fiber Bragg grating sensors distributed at fixed positions on a large-scale spatial deployment mechanism, data on changes in the grating reflection wavelength caused by slight structural deformation due to vibration is obtained;

[0069] As a preferred implementation of step S1, the specific process includes the following steps:

[0070] S11. Fiber Bragg grating sensors are placed at key structural points of the large-scale space deployment mechanism truss to monitor local vibrations at different locations.

[0071] S12. Fiber Bragg grating sensor transmits light into the optical fiber through a low-coherence light source. When the light wave passes through the grating in the optical fiber, the fiber Bragg grating will reflect the light of a specific wavelength back to the sensor according to its periodic structure, while the light of other wavelengths will continue to propagate, and the reflected center wavelength λ B satisfy:

[0072] λ B =2n eff Λ;

[0073] where n eff is the effective refractive index of the fiber material, and Λ is the period length of the grating.

[0074] In this embodiment, if Figure 2 As shown, the fiber Bragg grating sensor 1 is fixed at the same position of the smallest unit of the truss, and the vibration monitoring module 2 is located inside the spacecraft. The vibration monitoring module 2 supports the synchronous data acquisition and processing of multiple fiber Bragg grating sensors and can continuously monitor vibration when the deployment mechanism is deployed or locked in orbit. When the truss structure vibrates, it causes a slight deformation of the structure where the fiber Bragg grating sensor is located. This deformation changes the period length Λ of the fiber Bragg grating, resulting in a change in the reflection center wavelength λ. B The change in wavelength is proportional to the strain, and the strain ε of the next step is calculated by detecting the change in the wavelength of the reflected light. The fiber Bragg grating sensor can monitor the tiny deformation of the structure in real time, is suitable for extreme conditions in the space environment, and has the ability to resist electromagnetic interference and high temperature changes.

[0075] S2, by receiving the light waves reflected by the fiber Bragg grating sensor, calculating the wavelength change, and then converting the wavelength change into the spatial and temporal information of the vibration through signal demodulation, and analyzing and processing the local vibration displacement data;

[0076] As a preferred implementation of step S2, the specific process includes the following steps:

[0077] S21. Demodulate the optical signal of the fiber optic sensor into an electrical signal. The demodulated electrical signal represents the wavelength change caused by the structural vibration. The strain ε is calculated by detecting the change in the wavelength of the reflected light:

[0078]

[0079] Where Δλ is the change in reflection wavelength, λ0 is the initial reflection wavelength, and α is the strain sensitivity of the fiber material;

[0080] S22. The process of calculating the local vibration displacement L through the strain ε is as follows:

[0081] L = ε·L0;

[0082] Where L0 is the original length of the fiber Bragg grating.

[0083] In this embodiment, the strain change of the fiber grating sensor will continue to change with the frequency and amplitude of the vibration, and the reflection wavelength of the grating will also change periodically. The data processing module in the vibration monitoring module is responsible for demodulating the optical signal of the fiber optic sensor into an electrical signal. The demodulated electrical signal represents the wavelength change caused by the structural vibration. The signal demodulator converts the wavelength change into displacement offset and vibration frequency information, and inputs this data into the data processing unit. In the data processing module, multiple data ports of the signal acquisition unit receive electrical signals from the demodulator, representing the amplitude and frequency of the vibration at different positions. By processing these displacement signals captured by the fiber optic sensor, the displacement caused by the vibration is calculated, thereby obtaining local vibration displacement data.

[0084] S3. Construct a GCN-LSTM neural network combination model and use the analyzed and processed local vibration displacement data as the input of the GCN-LSTM neural network. Combined with the spatiotemporal coupling relationship, the GCN-LSTM neural network is used to establish a nonlinear relationship between the local vibration displacement data and the overall vibration displacement data. The overall vibration state of the large-scale spatial deployment mechanism is predicted and calculated to obtain the overall vibration displacement result of the mechanism.

[0085] As a preferred implementation of step S3, the local vibration displacement data is used as the input of the GCN network. The GCN network consists of multiple layers of graph convolution. Each layer performs a graph convolution operation, gradually aggregating features to a higher representation layer and outputs them through a fully connected layer. It is used for the vibration prediction task of large-scale spatial expansion mechanisms. The GCN network can be expressed by the following formula:

[0086]

[0087] Among them H (l+1) is the node feature matrix of the l+1th layer, σ is the nonlinear activation function, yes The degree matrix of , that is, a diagonal matrix, where the degree of each node is a diagonal element, It is the adjacency matrix A plus the identity matrix I to ensure that each node includes its own information, W (l) is the weight matrix of layer l, It normalizes the adjacency matrix A to prevent the eigenvalue of a node from changing with the number of neighbors, thus causing feature deviation.

[0088] At each layer, the GCN network aggregates the features of each node and passes the information of its neighboring nodes to the current node. The new features of a node v in the lth layer are It can be calculated by weighted average of neighbor node features:

[0089]

[0090] in represents the neighbor set of node v, d u and d v are the degrees of nodes u and v respectively; represents the hidden representation of node u at layer l;

[0091] The benefit of the GCN network in the vibration prediction of large-scale spatial deployment mechanisms is that it can effectively utilize the spatial topological information of the measurement points, obtain the overall vibration displacement data from the local vibration displacement data through layer-by-layer aggregation and convolution operations, thereby effectively capturing the local and global characteristics of the nodes, and providing input data for the subsequent LSTM layer, thereby realizing the time series modeling and prediction of vibration data.

[0092] As a preferred embodiment of step S3, the input of the LSTM network consists of the overall vibration displacement data of multiple time steps. The data of each time step can be represented as a vector, which includes the displacement offset of each measurement position at the corresponding time point. In the LSTM network, the information flow is controlled by three main gating mechanisms. Based on the forget gate and the input gate, the LSTM unit state C t Will be updated, the hidden state h t The calculation combines the current cell state with the output gate. This hidden state will be used as input for the next time step and can also be provided as output to subsequent layers or networks. LSTM networks control the flow of information through a series of gating mechanisms, effectively capturing the long-term and short-term dependencies of time series data.

[0093] More specifically, the LSTM network calculation process includes the following steps:

[0094] S31, the forget gate determines which information in the current unit state needs to be forgotten, and which information should be discarded based on the current input and the previous hidden state. The gate uses the sigmoid activation function to output a value between 0 and 1, where 0 means complete discard and 1 means complete retention:

[0095] f t =σ(W f ·[h t-1 , x t ]+b f );

[0096] Among them, f i is the output of the forget gate, σ is the sigmoid activation function, W f is the weight matrix of the forget gate, h t-1 is the hidden state at the previous moment, x t is the input at the current moment, b f is the bias of the forget gate;

[0097] S32, the input gate controls the impact of the current input information on the unit state, controls the addition of the current input information, determines which information needs to be updated, and generates a new candidate value vector:

[0098] i t =σ(W i ·[h t-1 ,x t ]+b i );

[0099]

[0100] Among them, i t is the output of the input gate, is the candidate cell state, W i and W C are the weight matrices of the input gate and candidate unit state, b i and b C is the corresponding bias;

[0101] S33, the cell state is updated through the forget gate and input gate to maintain long-term and short-term dependency information:

[0102]

[0103] Among them, C t is the current cell state, C t-1 is the cell state at the previous moment.

[0104] S34, the output gate determines how the current unit state will affect the hidden state. It combines the current input and the previous hidden state and outputs a value to determine the next hidden state:

[0105] o t =σ(W o ·[h t-1 , x t ]+b o );

[0106] h t =o t *tanh(C t );

[0107] Among them, t is the output of the output gate, W o is the weight matrix of the output gate, b o is the bias of the output gate, h t is the hidden state at the current moment.

[0108] As a preferred implementation of step S3, a GCN-LSTM neural network is used to establish a nonlinear relationship between local vibration data and overall vibration data, such as Figure 3 As shown, the specific process includes the following steps:

[0109] S41. Assume that there are N sensors that collect local vibration displacement data at T time points. The data can be represented as an N×T matrix:

[0110]

[0111] where d ij represents the displacement of the i-th sensor at the j-th time point, 1≤i≤N,1≤j≤T;

[0112] If there is a connection between sensor i and sensor j, their connection relationship can be represented by the adjacency matrix A. ij =1, otherwise A ij = 0; Divide the data into multiple time windows so that LSTM can learn time series features. Select the time window size as W. Each input sample can contain data from W time points. Assuming that there are W time points of data as input samples, the constructed input data shape X is (N, W);

[0113] The constructed local vibration displacement data matrix D is input into the GCN layer together with the adjacency matrix A. The GCN layer uses the adjacency matrix to process the spatial features between sensors. After passing through the GCN layer, the overall vibration displacement data features obtained are input into the LSTM layer to learn the time series features. The GCN-LSTM network output layer directly predicts the displacement of the overall structure. Represents the combined displacement of multiple sensor positions:

[0114]

[0115] in represents the predicted displacement of the i-th sensor position.

[0116] In this embodiment, since the vibration data obtained by the fiber grating sensor is only discrete local vibration displacement data information, it is impossible to fully display the overall vibration state of the entire large-scale space deployment mechanism. Therefore, in this embodiment, the identification and prediction module uses a vibration intelligent prediction system to establish a nonlinear coupling relationship between local vibration and overall vibration through a GCN-LSTM neural network. This neural network learns and derives the global vibration behavior of the entire flexible structure based on the local vibration displacement and time data provided by the fiber optic sensor. The identification and prediction module adopts a GCN-LSTM network algorithm to learn and derive the local vibration displacement data of a limited number of measuring points to obtain the vibration results of the entire large-scale space deployment mechanism. The vibration intelligent prediction system uses a GCN-LSTM neural network to establish a nonlinear relationship between local vibration data and overall vibration data. The GCN-LSTM network is composed of a graph convolutional network and a long short-term memory network. In the vibration intelligent prediction system, the GCN network is used to process and extract the spatial features of the local vibration displacement data, and the LSTM network is used to capture dynamic changes in time series data. The system can not only monitor the overall vibration of large space deployment mechanisms in real time with high precision and high reliability, but also provide a decision-making basis for the control of the spacecraft's on-orbit status, ensuring the spacecraft's high-precision attitude and stability in complex space environments.

[0117] More specifically, the local vibration displacement data are set as training samples, verification samples and test samples respectively, where the data ratio of training samples, verification samples and test samples is 8:1:1. The training samples and test samples are used to train and test the GCN-LSTM neural network combination model to obtain a trained GCN-LSTM neural network combination model; the verification samples are input into the trained GCN-LSTM neural network combination model to output the prediction results.

[0118] The present invention also provides a large-scale space deployment mechanism vibration intelligent prediction system, such as Figure 1 Shown, including:

[0119] A data acquisition module, wherein the data acquisition module is used to acquire data on changes in grating reflection wavelength caused by slight structural deformation due to vibration through a plurality of fiber grating sensors distributed at fixed positions of the large spatial deployment mechanism;

[0120] A vibration monitoring module is used to receive light waves reflected by the fiber grating sensor, calculate the change in wavelength, convert the wavelength change into spatial and temporal information of vibration through signal demodulation, and perform analysis and processing of local vibration signal data;

[0121] The identification and prediction module is used to construct a GCN-LSTM neural network combination model. The analyzed and processed local vibration displacement data is used as the input of the GCN-LSTM neural network combination model. In combination with the spatiotemporal coupling relationship, the GCN-LSTM neural network is used to establish a nonlinear relationship between the local vibration displacement data and the overall vibration displacement data. The overall vibration state of the large-scale spatial deployment mechanism is predicted and calculated to obtain the overall vibration displacement result of the mechanism.

[0122] More specifically, a large-scale space deployment mechanism vibration intelligent prediction device is provided, which is characterized in that it includes a processor and a storage medium; the storage medium is used to store instructions; and the processor is used to operate according to the instructions to execute the steps of the above method.

[0123] More specifically, a storage medium is provided on which a computer program is stored, wherein the computer program implements the steps of the above method when executed by a processor.

[0124] In the description of this specification, the reference terms "one embodiment," "some embodiments," "example," "specific example," or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials, or characteristics described may be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0125] The logic and / or steps represented in the flowchart or otherwise described herein may be considered, for example, as an ordered list of executable instructions for implementing logical functions, and may be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device).

[0126] The above embodiments provide a detailed introduction to the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A method for intelligent prediction of vibration of a large-scale spatial deployment mechanism, characterized in that: The following steps are involved: S1. Using multiple fiber Bragg grating sensors distributed at fixed positions on a large-scale spatial deployment mechanism, data on changes in the grating reflection wavelength caused by slight structural deformation due to vibration is obtained; S2, by receiving the light waves reflected by the fiber Bragg grating sensor, calculating the wavelength change, and then converting the wavelength change into the spatial and temporal information of the vibration through signal demodulation, and analyzing and processing the local vibration displacement data; S3. Construct a GCN-LSTM neural network combination model and use the analyzed and processed local vibration displacement data as the input of the GCN-LSTM neural network. Combined with the spatiotemporal coupling relationship, the GCN-LSTM neural network is used to establish a nonlinear relationship between the local vibration displacement data and the overall vibration displacement data. The overall vibration state of the large-scale spatial deployment mechanism is predicted and calculated to obtain the overall vibration displacement result of the mechanism.

2. The intelligent prediction method for vibration of a large-scale spatial deployment mechanism according to claim 1, characterized in that: In step S1, the specific process includes the following steps: S11. Fiber Bragg grating sensors are placed at key structural points of the large-scale space deployment mechanism truss to monitor local vibrations at different locations. S12. Fiber Bragg grating sensor transmits light into the optical fiber through a low-coherence light source. When the light wave passes through the grating in the optical fiber, the fiber Bragg grating will reflect the light of a specific wavelength back to the sensor according to its periodic structure, while the light of other wavelengths will continue to propagate, and the reflected center wavelength λ B satisfy: l B =2n eff L; where n eff is the effective refractive index of the fiber material, and Λ is the period length of the grating.

3. The intelligent prediction method for vibration of a large-scale spatial deployment mechanism according to claim 1, characterized in that: In step S2, the specific process includes the following steps: S21. Demodulate the optical signal of the fiber optic sensor into an electrical signal. The demodulated electrical signal represents the wavelength change caused by the structural vibration. The strain ε is calculated by detecting the change in the wavelength of the reflected light: Where Δλ is the change in reflection wavelength, λ0 is the initial reflection wavelength, and α is the strain sensitivity of the fiber material; S22. The process of calculating the local vibration displacement l through the strain ε is as follows: l=ε·L0; Where L0 is the original length of the fiber Bragg grating.

4. The intelligent prediction method for vibration of a large-scale spatial deployment mechanism according to claim 1, characterized in that: In step S3, the local vibration displacement data is used as the input of the GCN network. The GCN network consists of multiple layers of graph convolution. Each layer performs a graph convolution operation, gradually aggregating features to a higher representation layer and outputs them through a fully connected layer for the vibration prediction task of large-scale spatial expansion mechanisms. The GCN network is expressed by the following formula: Among them H (l+1) is the node feature matrix of the l+1th layer, H (l) represents the node feature matrix of the lth layer, σ is the nonlinear activation function, yes The degree matrix of , that is, a diagonal matrix, where the degree of each node is a diagonal element, It is the adjacency matrix A plus the identity matrix I to ensure that each node includes its own information, W (l) is the weight matrix of layer l, It normalizes the adjacency matrix A to prevent the eigenvalue of a node from changing with the number of neighbors, thus causing feature deviation. At each layer, the GCN network aggregates the features of each node and passes the information of its neighboring nodes to the current node. The new features of a node v in the lth layer are It is calculated by weighted average of neighbor node features: in represents the neighbor set of node v, ∪{v} represents the union of node v, d u and d v are the degrees of nodes u and v respectively; represents the hidden representation of node u at layer l; In the vibration prediction of large-scale spatial deployment mechanisms, the GCN network obtains overall vibration displacement data from local vibration displacement data through layer-by-layer aggregation and convolution operations, thereby effectively capturing the local and global characteristics of the nodes and providing input data for the subsequent LSTM layer, thereby realizing time series modeling and prediction of vibration data.

5. The method for intelligently predicting vibration of a large-scale spatial deployment mechanism according to claim 1, characterized in that: In step S3, the input of the LSTM network consists of the overall vibration displacement data of multiple time steps. The data of each time step is represented as a vector, which includes the displacement offset of each measurement position at the corresponding time point. In the LSTM network, the information flow is controlled by three main gating mechanisms. Based on the forget gate and the input gate, the LSTM unit state C t Will be updated, the hidden state h t The calculation combines the current unit state and the output gate. This hidden state will be used as the input of the next time step and also provided as the output to the subsequent layer or network.

6. The method for intelligently predicting vibration of a large-scale spatial deployment mechanism according to claim 5, characterized in that: The specific calculation process of the LSTM network includes the following steps: S31, the forget gate determines which information in the current unit state needs to be forgotten, and which information should be discarded based on the current input and the previous hidden state. The gate uses the sigmoid activation function to output a value between 0 and 1, where 0 means complete discard and 1 means complete retention: f t =σ(W f ·[h t-1 ,x t ]+b f ); Among them, f t is the output of the forget gate, σ is the sigmoid activation function, W f is the weight matrix of the forget gate, h t-1 is the hidden state at the previous moment, x t is the input at the current moment, b f is the bias of the forget gate; S32, the input gate controls the impact of the current input information on the unit state, controls the addition of the current input information, determines which information needs to be updated, and generates a new candidate value vector: i t =σ(W i ·[h t-1 ,x t ]+b i ); Among them, i t is the output of the input gate, is the candidate cell state, W i and W C are the weight matrices of the input gate and candidate unit state, b i and b C is the corresponding bias; S33, the cell state is updated through the forget gate and input gate to maintain long-term and short-term dependency information: Among them, C t is the current cell state, C t-1 is the unit state at the previous moment; S34, the output gate determines how the current unit state will affect the hidden state. It combines the current input and the previous hidden state and outputs a value to determine the next hidden state: the t =σ(W o ·[h t-1 ,x t ]+b o ); h t =o t *fishy(C) t ); Among them, t is the output of the output gate, W o is the weight matrix of the output gate, b o is the bias of the output gate, h t is the hidden state at the current moment.

7. The intelligent prediction method for vibration of a large-scale spatial deployment mechanism according to claim 1, characterized in that: In step S3, a nonlinear relationship between local vibration data and overall vibration data is established using the GCN-LSTM neural network. The specific process includes the following steps: S41. Assume that there are N sensors that collect local vibration displacement data at T time points. The data is represented as an N×T matrix: where d ij represents the displacement of the i-th sensor at the j-th time point, 1≤i≤N,1≤j≤T; If there is a connection between sensor i and sensor j, their connection relationship is represented by the adjacency matrix A. ij =1, otherwise A ij = 0; Divide the data into multiple time windows so that LSTM can learn time series features. Select the time window size as W, and each input sample contains data from W time points. Assuming that there are W time points of data as input samples, the constructed input data shape X is (N, W); The constructed local vibration displacement data matrix D is input into the GCN layer together with the adjacency matrix A. The GCN layer uses the adjacency matrix to process the spatial features between sensors. After passing through the GCN layer, the overall vibration displacement data features obtained are input into the LSTM layer to learn the time series features. The GCN-LSTM network output layer directly predicts the displacement of the overall structure. Represents the combined displacement of multiple sensor positions: in represents the predicted displacement of the i-th sensor position.

8. The method for intelligently predicting vibration of a large-scale spatial deployment mechanism according to claim 1, characterized in that: The local vibration displacement data are set as training samples, verification samples and test samples respectively, where the data ratio of training samples, verification samples and test samples is 8:1:

1. The training samples and test samples are used to train and test the GCN-LSTM neural network combination model to obtain a trained GCN-LSTM neural network combination model; the verification samples are input into the trained GCN-LSTM neural network combination model to output the prediction results.

9. A large-scale space deployment mechanism vibration intelligent prediction system, characterized in that: include: A data acquisition module, wherein the data acquisition module is used to acquire data on changes in grating reflection wavelength caused by slight structural deformation due to vibration through a plurality of fiber grating sensors distributed at fixed positions of the large spatial deployment mechanism; A vibration monitoring module is used to receive light waves reflected by the fiber grating sensor, calculate the change in wavelength, convert the wavelength change into spatial and temporal information of vibration through signal demodulation, and perform analysis and processing of local vibration signal data; The identification and prediction module is used to construct a GCN-LSTM neural network combination model. The analyzed and processed local vibration displacement data is used as the input of the GCN-LSTM neural network combination model. In combination with the spatiotemporal coupling relationship, the GCN-LSTM neural network is used to establish a nonlinear relationship between the local vibration displacement data and the overall vibration displacement data. The overall vibration state of the large-scale spatial deployment mechanism is predicted and calculated to obtain the overall vibration displacement result of the mechanism.

Images