Aircraft surface temperature prediction and regulation method based on time sequence diagram neural network

Abstract

The invention belongs to the technical field of aircraft thermal protection monitoring and control, and discloses an aircraft surface temperature prediction and regulation method based on a time sequence diagram neural network, and the method comprises the steps: arranging a high-temperature-resistant fiber grating sensing array in a modular heat insulation tile, and monitoring the temperature of an upper layer plate and a lower layer plate of the heat insulation tile in real time; abstracting all heat insulation tiles on the surface of the aircraft and the adjacency relation thereof into a graph structure; constructing a node time sequence feature sequence based on temperature and flight attitude data; using a time sequence diagram neural network model to predict a future thermal state and discriminate a health level; and when a prediction result reaches a risk or failure threshold value, implementing feedforward attitude regulation and control to reduce the thermal load of the target area by optimizing and adjusting the attack angle and the roll angle of the aircraft. According to the invention, real-time sensing, advanced prediction and active regulation and control of the thermal state are realized, and the safety, reliability and intelligent level of the thermal protection system are improved.

Description

Technical Field

[0001] This application belongs to the field of aircraft thermal protection monitoring and control technology, specifically involving a method for predicting and controlling aircraft surface temperature based on a time-series graph neural network. Background Technology

[0002] Hypersonic vehicles endure intense aerodynamic heating during reentry and high Mach number cruise, resulting in high heat flux density, strong thermal gradients, and significant spatiotemporal nonuniform thermal loads on their outer surfaces. Influenced by changes in flight attitude, airflow disturbances, and differences in the thermal properties of insulation materials, the surface temperature of the vehicle exhibits complex dynamic evolution characteristics. To ensure structural safety and reusability, modular thermal insulation tile systems made of ceramic-based materials are typically installed on the outer surface of the vehicle to achieve aerodynamic thermal protection.

[0003] However, existing modular thermal insulation tile systems primarily rely on passive protection and generally lack the ability to perceive, analyze, and provide feedback on complex thermal environments in real time. During flight, traditional thermal insulation tiles struggle to continuously acquire surface and back temperature information, making it difficult for the system to promptly identify and warn of failures such as localized detachment, crack propagation, or abnormal ablation. Current thermal protection control largely depends on post-event monitoring or threshold triggering mechanisms, with attitude adjustments typically lagging behind the occurrence of thermal anomalies, making it difficult to proactively avoid potential thermal failure risks. Furthermore, thermal response modeling of thermal insulation tiles still relies heavily on physical and empirical models, with limited integration with artificial intelligence technologies, making it difficult to fully leverage time-series temperature data and multi-source operational information during flight.

[0004] In terms of operation and maintenance and lifespan management, the health status assessment of thermal insulation tiles mainly relies on offline methods such as manual inspection or infrared imaging. This method has low detection efficiency, long maintenance cycles, and makes it difficult to quantitatively characterize thermal fatigue accumulation, material aging, and progressive ablation behavior, thus failing to achieve reliable prediction of remaining lifespan. In addition, the heat conduction process of thermal insulation tiles under high-temperature conditions has significant spatial coupling and time accumulation effects, and traditional methods based on single-point or instantaneous measurements are difficult to accurately reflect their true thermal response characteristics.

[0005] Due to the lack of high spatiotemporal resolution and continuous reliable temperature response data, and insufficient integration of thermal insulation tile systems with artificial intelligence algorithms, digital twins, and intelligent diagnostic technologies, existing thermal protection systems are unable to achieve autonomous prediction and feedforward control of thermal states. This restricts the development of thermal protection systems from passive protection to intelligent perception, active prediction, and adaptive control, and makes it difficult to meet the safety, reliability, and intelligence requirements of next-generation hypersonic vehicles. Summary of the Invention

[0006] To address the issue of lag in control caused by the lack of real-time global temperature sensing and spatiotemporal coupling modeling capabilities in existing thermal protection systems, this application provides a method for predicting and controlling aircraft surface temperature based on a temporal graph neural network. By constructing a prediction framework that integrates distributed fiber optic monitoring, graph structure spatial modeling, and temporal neural networks, it achieves advanced prediction of the future thermal state of each thermal insulation tile and drives the feedforward adjustment of flight attitude accordingly. This transforms the thermal protection mode from a reactive, post-event response to an active, pre-event intervention, improving the system's safety, reliability, and autonomous decision-making capabilities under extreme thermal environments.

[0007] To achieve the above technical objectives, this application specifically adopts the following technical solution: In one aspect of this application, a method for predicting and controlling the surface temperature of an aircraft based on a time-series graph neural network is provided, comprising the following steps: S1. A high-temperature resistant fiber optic grating sensor array is installed inside the modular heat insulation tile to perform real-time distributed monitoring of the temperature of the upper and lower layers of the heat insulation tile and obtain temperature data. S2. Model all modular thermal insulation tiles on the outer surface of the aircraft as a graph structure. , where the set of nodes Each node in Corresponding to a heat insulation tile; edge set The edge in Represents a node and The corresponding heat insulation tiles are adjacent in physical space; S3. Based on the temperature data and the aircraft attitude parameters received in real time from the flight control system, within the time interval... Each node Built on continuous Temporal feature sequences at each time step:

[0008] Among them, time step Node feature vectors Defined as:

[0009] in, Temperature of the upper layer of the heat insulation tile; Temperature of the lower layer of the insulation tile; The temperature difference between the upper and lower shelves; This represents the rate of change of temperature difference. The current angle of attack of the aircraft; This is the aircraft's current roll angle; S4. The time-series feature sequence The input is fed into a time-series graph neural network model to predict the future time window for each thermal insulation tile. Internal temperature or thermal risk indicators ; S5. Based on the prediction results of step S4 With the preset threshold set Compare and determine the health status of each insulation tile. ,in, ; S6. When at least one heat insulation tile reaches a preset risk or failure level, construct and solve the attitude feedforward control optimization problem, and change the aerodynamic heat flow distribution on the outer surface by adjusting the aircraft's angle of attack and / or roll angle to reduce the thermal load on the target heat insulation tile.

[0010] In one embodiment, in step S1, the modular heat insulation tile includes a ceramic substrate shell designed as a hollow regular hexagonal prism. The ceramic substrate shell includes an upper plate and a lower plate. The upper plate and the lower plate have three optical fiber embedding channels distributed along the center lines of opposite sides. The high-temperature resistant fiber grating sensing array includes high-temperature resistant sapphire optical fibers arranged in the optical fiber embedding channels. The sapphire optical fibers have multiple equally spaced fiber grating nodes engraved on them.

[0011] In one embodiment, the diameter of the fiber embedding channel is 0.4-0.6 mm; the spacing between the fiber Bragg grating nodes is 10-20 mm; and the sapphire fiber is fixed in the fiber embedding channel by an inorganic adhesive with a temperature resistance greater than 1500°C and a coefficient of linear expansion matching that of the ceramic substrate shell.

[0012] In one implementation, during the assembly of modular thermal insulation tiles, a structural bonding and composite filling process is used between adjacent thermal insulation tiles; specifically including: A fast-curing ceramic adhesive film is applied to the surface of the aircraft skin to bond the heat insulation tiles; A flexible ceramic fiber sealing strip is installed at the joint between adjacent heat insulation tiles, and the outside of the flexible ceramic fiber sealing strip is filled with fast-curing high-temperature resistant ceramic adhesive; Apply a quick-drying, high-temperature erosion-resistant coating to the outer surface of the joint.

[0013] In one implementation, in step S4, the time-series graph neural network model includes: The graph neural network module is used to aggregate the features of nodes at each time step to obtain spatial embedding features that reflect the spatial thermal coupling relationship between the node and its neighboring nodes; the graph neural network module is a graph convolutional network, a graph attention network, or a message passing neural network. The temporal neural network module is used to model the temporal evolution of spatially embedded feature sequences at multiple consecutive time steps and output the prediction results within a preset future time window; the temporal neural network module is a long short-term memory network, a gated recurrent unit, or a temporal graph neural network.

[0014] In one implementation, step S5 specifically involves: Set multiple thresholds corresponding to the predicted temperature or thermal risk index of the insulation tile; The predicted result for each heat insulation tile is compared with the threshold, and its health status is divided into four levels: normal, warning, risk, and failure.

[0015] In one implementation, in step S6, the objective function of the attitude feedforward control optimization problem is to minimize the predicted temperature or thermal risk index of the target heat insulation tile, the optimization variables are the aircraft angle of attack and roll angle, and the constraints are flight dynamics and control constraints; the optimization problem is solved using model predictive control, rule-driven control, or learning-based policy mapping methods.

[0016] In one implementation, the method further includes state visualization and interaction: It can receive and display the temperature data, health status, and aircraft attitude parameters of each heat insulation tile in real time. The layout of the heat insulation tiles on the surface of the aircraft is displayed graphically, and different colors are used to map their health status. Dynamically plot the curves of temperature, temperature difference, and attitude parameters of each heat insulation tile over time; Record and display control commands, posture adjustment actions, and health status switching events.

[0017] In another aspect of this application, a modular heat-insulating tile is provided for implementing the aforementioned method for predicting and controlling aircraft surface temperature based on a time-series graph neural network, comprising: The ceramic-based shell is constructed as a hollow regular hexagonal prism shell, consisting of an upper plate, a lower plate, and a sidewall connecting the upper plate and the lower plate; optical fiber embedding channels are provided in the upper plate and the lower plate. A lattice support layer and a filling layer are disposed within the internal space of the ceramic substrate shell; the lattice support layer comprises a plurality of closely arranged permeable lattice units; the filling layer is made of a lightweight material with low thermal conductivity and fills the spaces between adjacent permeable lattice units and the internal cavities of each permeable lattice unit. The real-time temperature monitoring module comprises a high-temperature resistant sapphire fiber grating sensor array embedded in the optical fiber channel, and an optical signal processing unit that is communicatively connected to the sapphire fiber grating sensor array.

[0018] The beneficial effects of this application are as follows: This application enables distributed real-time monitoring of the upper and lower surface temperatures of modular thermal insulation tiles, acquiring complete thermal state information within the tile body. By abstracting the thermal insulation tile array into a graph structure and combining it with a temporal neural network, the spatial thermal coupling and temporal evolution between tiles are effectively characterized, thereby achieving advanced and high-precision prediction of future thermal states. Based on the prediction results, the aircraft attitude is fed forward and actively optimized to optimize surface heat flow distribution, avoiding the lag of traditional post-event control and significantly improving the system's safety margin. The integrated state visualization module provides operators with global situational awareness and decision support, enhancing the system's maintainability and task analysis capabilities. The modular sensing and interface design reduces maintenance complexity, and the multi-layer composite sealing process effectively suppresses heat leakage at joints, jointly improving the system's engineering applicability and long-term reliability under extreme thermal environments. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the aircraft surface temperature prediction and control method based on a time-series graph neural network according to this application. Figure 2 This is a schematic diagram of the structure of the modular ceramic matrix composite heat insulation tile according to an embodiment of this application; Figure 3 This is a schematic diagram of a high-temperature resistant sapphire optical fiber according to an embodiment of this application; Figure 4 This is a schematic diagram of the bonding of modular ceramic matrix composite heat insulation tiles according to an embodiment of this application; Figure 5 This is a schematic diagram of the hexagonal laying of modular heat insulation tiles on the outer surface of an aircraft, according to an embodiment of this application. Figure 6 This is a schematic diagram of a graph structure constructed based on the adjacency relationship of the insulation tiles in an embodiment of this application; Figure 7 This is a schematic diagram showing the stacking of the node features and temporal features of the thermal insulation tile in an embodiment of this application; Figure 8 This is a block diagram of the temporal graph neural network structure (GNN + LSTM / GRU) according to an embodiment of this application. Figure 9 This is a schematic diagram of the closed-loop process of prediction-health assessment-attitude feedforward control in an embodiment of this application; Figure 10 This is a visual interface diagram of an embodiment of this application; In the diagram: 1-Upper layer, 2-Lower layer, 3-Fiber optic embedding channel, 4-Sapphire fiber, 5-Fiber optic connector terminal, 6-Adhesive film, 7-Thermal isolation structure. Detailed Implementation

[0020] The technical solution of this application will be clearly and completely described below with reference to specific embodiments. However, those skilled in the art will understand that the embodiments described below are only some embodiments of this application, not all embodiments, and are only used to illustrate this application, and should not be regarded as limiting the scope of this application. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] This application addresses the technical bottlenecks of existing thermal protection systems, such as the inability to sense the internal temperature field of the tile in real time, the difficulty in modeling the spatial thermal coupling and temporal evolution between multiple tiles, and the response lag caused by reliance on ex-post threshold control. It proposes a collaborative technical solution that integrates distributed optical fiber sensing, graph structure system modeling, and intelligent predictive feedforward control. By embedding a high-temperature resistant fiber grating array inside the modular thermal insulation tile, synchronous real-time monitoring of the upper and lower surface temperatures is achieved, acquiring information on the thermal state and heat flow direction inside the tile. The physical adjacency relationship of all thermal insulation tiles on the aircraft surface is abstracted into a graph structure, thereby transforming the continuously distributed thermal protection system into a discretized model that can characterize spatial thermal coupling. Based on the monitored temperature data and flight attitude parameters, a temporal feature sequence of each thermal insulation tile is constructed, and a temporal graph neural network is used to simultaneously capture the spatial correlation and temporal evolution trend of the temperature field, enabling prediction of future thermal states. Based on the prediction results, the health status of the thermal insulation tiles is classified and judged. Before the actual occurrence of thermal risks, the aircraft attitude is proactively adjusted by solving optimization problems to change the aerodynamic heat flow distribution, thus forming an intelligent closed loop from real-time perception to prediction to feedforward control, enabling the thermal protection system to proactively avoid risks.

[0022] Specifically, refer to Figure 1 As shown, the aircraft surface temperature prediction and control method based on time-series graph neural networks in this application is implemented through the following steps: S1. A high-temperature resistant fiber optic grating sensor array is installed inside the modular heat insulation tile to perform real-time distributed monitoring of the temperature of the upper and lower layers of the heat insulation tile and obtain temperature data. In some embodiments, the modular heat insulation tile employs a hollow hexagonal prism-shaped ceramic substrate shell, which includes an upper plate and a lower plate. Each plate contains three pre-embedded fiber optic channels distributed along opposite midlines. A high-temperature resistant fiber grating sensing array is composed of high-temperature resistant sapphire fibers arranged within these channels. Multiple equally spaced fiber grating nodes are formed on the sapphire fibers using femtosecond laser writing technology.

[0023] In some embodiments, the structure of the modular heat-insulating tile follows the basic design of the applicant's published patent CN120308326B, comprising a ceramic substrate shell, a lattice support layer, and a filling layer. The improvement of this application lies in the integration of a real-time temperature monitoring module within the tile body. This module consists of a sapphire fiber optic grating sensor array and a corresponding optical signal processing unit, used to synchronously collect and feedback temperature data from the upper and lower layers of the heat-insulating tile.

[0024] The modular heat insulation tile specifically includes the following components: The ceramic-based outer shell is a hollow regular hexagonal prism shell, consisting of an upper plate, a lower plate, and sidewalls connecting the upper and lower plates. Optical fiber embedding channels for laying sensing optical fibers are respectively formed in the upper and lower plates.

[0025] A lattice support layer and a filling layer are disposed within the internal space of the ceramic substrate shell; the lattice support layer comprises a plurality of closely arranged permeable lattice units; the filling layer is made of a lightweight material with low thermal conductivity and fills the spaces between adjacent permeable lattice units and the internal cavities of each permeable lattice unit.

[0026] The real-time temperature monitoring module comprises a high-temperature resistant sapphire fiber grating sensor array embedded in the optical fiber channel, and an optical signal processing unit that is communicatively connected to the sapphire fiber grating sensor array.

[0027] In some embodiments, the fiber embedding channels in the upper and lower plates of the ceramic substrate housing are designed with a diameter of 0.4-0.6 mm to accommodate and fix the sapphire fiber. This size range provides the necessary space margin for smooth embedding and final fixation of the sapphire fiber, while also ensuring that the overall structural strength of the ceramic substrate housing itself is not significantly weakened after the channels are opened, and is also conducive to the effective transmission of optical signals in the fiber.

[0028] In some embodiments, multiple fiber Bragg grating nodes etched on the sapphire fiber are distributed at equal intervals along the fiber's length. The specific value of this spacing is determined based on the size of a single thermal insulation tile and the system's spatial resolution requirements for temperature measurement, for example, set between 10-20 mm. This aims to obtain sufficiently dense temperature measurement points while avoiding excessive redundant data, thus balancing measurement accuracy with subsequent data processing efficiency.

[0029] In some embodiments, sapphire optical fibers are deployed in the fiber embedding channel using a specialized inorganic adhesive. After curing, the inorganic adhesive firmly bonds and encapsulates the sapphire optical fibers within the channel. The coefficient of linear expansion of the cured inorganic adhesive matches the ceramic material constituting the ceramic substrate shell, ensuring that when the entire thermal insulation tile assembly operates under high-temperature conditions, the difference in deformation between the inorganic adhesive and the ceramic substrate shell due to thermal expansion is minimized. This maintains the long-term stability of the bonding interface and prevents the sapphire optical fibers from loosening or detaching due to thermal stress, or from causing attenuation of optical signal transmission quality.

[0030] In some embodiments, the end of the sapphire optical fiber is led out from the rear of the heat insulation tile substrate through a ceramic encapsulation tube and connected to a standardized optical signal interface. The encapsulation tube and optical signal interface allow individual heat insulation tile modules to be easily plugged and replaced from the sensor network, and also facilitate the rapid construction and expansion of the entire monitoring system.

[0031] In some embodiments, the assembly process of modular thermal insulation tiles employs structural bonding and composite filling techniques to achieve the joining of thermal insulation tiles and the connection of thermal insulation tiles to the aircraft skin.

[0032] Specifically, during assembly, a layer of fast-curing ceramic adhesive film is first uniformly coated onto the surface of the aircraft skin. This adhesive film possesses strong initial tack at room temperature, with an initial shear strength greater than 0.5 MPa. Utilizing this characteristic, operators directly attach the heat insulation tiles to the skin surface coated with the adhesive film. The initial adhesion provided by the adhesive film ensures the heat insulation tiles remain accurately positioned before complete curing, eliminating the need for additional external fixing forces. After the initial positioning of the heat insulation tiles on the skin is completed, the joints between adjacent heat insulation tiles are sealed. Flexible ceramic fiber sealing strips are placed at the joints, with the initial shape and thickness of the flexible ceramic fiber sealing strips matching the designed gaps between the tiles. When the heat insulation tiles are controlled and pressed into the final assembly position, the flexible ceramic fiber sealing strips undergo reversible compression deformation, filling the joint gaps and forming a physical barrier with thermal barrier and initial airtightness.

[0033] In some embodiments, to further enhance the sealing reliability, mechanical strength, and long-term durability of the joint, a fast-curing, high-temperature resistant ceramic adhesive is injected onto the outside of the compressed flexible ceramic fiber sealing strip. This ceramic adhesive, in its liquid state, has sufficient viscosity to ensure uniform distribution and filling of the narrow space within the joint. After curing, this ceramic adhesive layer serves as a second sealing layer, working in conjunction with the inner flexible ceramic fiber sealing strip to enhance the overall integrity of the joint area.

[0034] In some embodiments, after the composite filling structure of the joint has cured, a fast-drying, high-temperature erosion-resistant coating is sprayed onto the outer surface. This fast-drying, high-temperature erosion-resistant coating can quickly reach a surface-dry state at room temperature and, after subsequent curing, forms a robust protective layer. This protective layer can withstand the aerodynamic heat flow and particle erosion generated during high Mach number flight, providing external protection for the joint area.

[0035] S2. Model all modular heat insulation tiles on the outer surface of the aircraft as a graph structure; where each heat insulation tile corresponds to a node in the graph structure, and an edge is established between the nodes corresponding to adjacent heat insulation tiles in the physical space.

[0036] Specifically, to facilitate intelligent analysis, the entire thermal protection system is modeled as a graph structure, denoted as... , where the set of nodes Include Each node, i.e. Each node This uniquely corresponds to a specific modular heat-insulating tile on the outer surface of the aircraft. (Side set) It is used to express the connection relationship between nodes, and is a set of node pairs, that is... If and only if two heat-insulating tiles and When the physical spatial layout of the aircraft skin surfaces is directly adjacent, at the corresponding nodes and Establish an undirected edge between them Since thermal insulation tiles are typically laid in a regular hexagonal shape, the resulting abstract graph structure exhibits a regular hexagonal grid topology. This transforms the continuously distributed physical thermal protection system into discrete, structured graph data, thereby formally describing and characterizing the spatial thermal coupling relationships between different thermal insulation tiles.

[0037] S3. Based on the temperature data obtained in step S1 and the aircraft attitude parameters obtained by the flight control system, a time-series feature sequence is constructed for each node over multiple consecutive time steps; the time-series feature sequence includes at least the upper plate temperature, lower plate temperature, temperature difference between the upper and lower plates, temperature difference change rate, aircraft angle of attack, and aircraft roll angle of the corresponding heat insulation tile in each time step.

[0038] For any node At a specific time t, define an eigenvector. eigenvectors It integrates sensor data and flight status parameters, and its specific composition is as follows:

[0039] in, The temperature of the upper layer of the heat insulation tile is derived from real-time measurement data of the fiber Bragg grating sensor array embedded inside the heat insulation tile. The temperature of the lower layer of the heat insulation tile is derived from real-time measurement data of the fiber Bragg grating sensor array embedded inside the heat insulation tile. The temperature difference between the upper and lower shelves; This represents the rate of change of temperature difference. The aircraft's current angle of attack is provided online by the flight control system; The current roll angle of the aircraft is provided online by the flight control system.

[0040] To capture the evolution of thermal state over time, the temporal characteristics of nodes are obtained. For nodes... Selecting from historical moments up to the current moment For a continuous time period, the feature vectors of each moment within that time period are arranged and stacked in chronological order to form a node temporal feature sequence. , is represented as: .

[0041] S4. Input the time-series feature sequence into the time-series graph neural network model to predict the temperature or thermal risk index of each heat insulation tile within a preset time window in the future.

[0042] The obtained multi-timestep node feature sequences are input into the temporal graph neural network model, including: At any time step t Using a temporal graph neural network to analyze node feature vectors Spatial aggregation is performed to obtain the spatial embedding features of the nodes. :

[0043] in, For nodes The set of neighboring nodes; The edge weights of the graph; This is the trainable parameter matrix for a graph neural network; It is a non-linear activation function.

[0044] Continuous L The spatial features of each time step are input into a time-series graph neural network to obtain the future preset time window. Prediction results within:

[0045] in, For time-series graph neural network functions; To predict the time step length; This is a predictive indicator of the future temperature or thermal risk of the insulation tile.

[0046] The obtained multi-timestep node temporal feature sequence The data is input into a time-series graph neural network model to predict the future thermal state of the insulation tile.

[0047] In some embodiments, the time-series graph neural network model includes at least the following two functional modules: Module 1, Graph Neural Network Module (GNN) At each time step t Using Graph Convolutional Networks (GCN), Graph Attention Networks (GAT), or Message Passing Neural Networks (MPNN) to analyze a given time step. t Node feature vectors Spatial aggregation is performed to extract the spatial thermal coupling characteristics between the insulation tiles. The basic calculation form is as follows:

[0048] in, For nodes Spatial embedding features at time step t For nodes The set of neighboring nodes, For adjacency weight, For a trainable parameter matrix, It is a non-linear activation function.

[0049] Module 2, Temporal Neural Network Module Spatial embedding features of continuous time steps The input is fed into a Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), or Temporal GNN network to model the temporal evolution of the temperature of the insulation tile and output the predicted temperature or thermal risk index within a preset future time window.

[0050] S5. Based on the prediction results of step S4 With the preset threshold set Compare and determine the health status of each insulation tile. ,in, .

[0051] Prediction results based on the output of the time-series graph neural network model and a predefined set of thresholds The future health status of each modular insulation tile is assessed and determined.

[0052] gather It includes a series of critical values ​​corresponding to temperature or thermal risk indicators. Threshold The numerical criteria were determined based on the material tolerance limit of the heat insulation tile, engineering safety specifications, and historical operating data, and met the following requirements. < < < The relationship.

[0053] The specific judgment process is as follows: for each heat insulation tile (corresponding to a node) ), and in the future Predicted value With the threshold set The various thresholds in the data are compared sequentially. Based on the comparison results, the heat insulation tile is then... Health status at all times It falls into one of the following four distinct levels: normal, warning, risk, or failure.

[0054] When the predicted value Less than or equal to the threshold When the predicted value is within the normal range, it is considered a "normal" state; when the predicted value is within the normal range, it is Above the threshold But below or equal to the threshold When the predicted value is in a "warning" state, it is determined to be in an "early warning" state; when the predicted value Above the threshold But below or equal to the threshold When the predicted value is [value], it is classified as a "risk" state; when the predicted value is [value], it is classified as a "risk" state. Above the threshold When the temperature or thermal stress exceeds the material's safe operating limit, it is considered to be in a "failure" state. A failure state typically means that the predicted temperature or thermal stress has exceeded the material's safe operating limit, posing an imminent risk of structural damage.

[0055] S6. When at least one heat insulation tile reaches a preset risk or failure level, construct and solve the attitude feedforward control optimization problem, and change the aerodynamic heat flow distribution on the outer surface by adjusting the aircraft's angle of attack and / or roll angle to reduce the thermal load on the target heat insulation tile.

[0056] Based on the determination result of step S5, if it is confirmed that at least one modular insulation tile has a predicted health status... When the preset "risk" or "failure" level is reached, the attitude feedforward control mechanism is automatically activated. The core of the attitude feedforward control mechanism is to construct and solve an optimization control problem with the aircraft attitude angle as the adjustment variable. It aims to adjust the spatial distribution of aerodynamic heat flux on the outer surface of the aircraft by actively changing the flight attitude, thereby reducing the thermal load borne by the target heat insulation tile in a high-risk or failure warning state.

[0057] Specifically, the attitude feedforward control is formalized as an optimization problem. The objective function of this optimization problem is set to minimize all target heat insulation tiles that require intervention (their indices form a set). The sum of predicted temperatures or thermal risk indicators at future times, mathematically expressed as:

[0058] The optimization variable is the angle of attack of the aircraft. and roll angle .

[0059] During the solution process, it is ensured that the adjusted attitude angles satisfy the engineering conditions such as the aircraft flight dynamics equations, control stability requirements, structural load limitations, and mission path constraints, which together constitute the constraints of the optimization problem.

[0060] In some embodiments, to solve the constrained optimization problem and generate the final attitude control commands, methods such as model predictive control, rule-driven control, or learning-based policy mapping are employed. Model predictive control solves the optimization problem in the finite-time domain online in each control cycle and applies the first control variable to the aircraft. Rule-driven control issues commands based on a pre-defined logical rule base that correlates thermal state with attitude adjustment. Learning-based policy mapping utilizes a trained neural network or other function approximator to directly map the recommended attitude adjustment based on the current thermal state prediction results. Through any of these methods, the thermal state information output by the prediction module is transformed into forward-looking attitude control actions, thus forming a closed loop from perception to prediction to decision-control, achieving proactive, feedforward management of thermal loads.

[0061] In some embodiments, the aircraft surface temperature prediction and control method further includes system status visualization and human-machine interaction functions. This is achieved through a separate data processing and graphics display module, which establishes a communication connection with the aforementioned heat-insulating tile temperature sensing module, time-series neural network prediction module, and flight control system.

[0062] The system status visualization and human-machine interaction module receives and processes data from each link in real time, including: real-time temperature measurement data of each heat insulation tile (such as top surface temperature and bottom surface temperature), the health status label of the heat insulation tile determined by step S5, and real-time attitude parameters from the flight control system (such as angle of attack and roll angle).

[0063] In terms of graphical display, the system status visualization and human-computer interaction module generates a schematic diagram of the thermal insulation tile layout that is consistent with or highly similar to the actual tile layout on the aircraft's outer surface. In this diagram, the position and shape of each thermal insulation tile correspond to its physical location. Based on the current health status of each thermal insulation tile (e.g., normal, warning, risk, failure), different predefined colors (e.g., gray, yellow, orange, and red) are used to fill or highlight the graphic of that tile, allowing the operator to intuitively and quickly identify abnormal areas and their severity within the entire thermal protection system.

[0064] Simultaneously, the system status visualization and human-computer interaction module dynamically plots and updates curves showing the changes of key parameters over time. This includes at least: time-series curves of the top and bottom surfaces of a specific heat-insulating tile, time-series curves of the temperature difference between the upper and lower surfaces, and the temperature rise rate curve, to demonstrate the evolution trend of its thermal response; it also includes time-series curves of aircraft attitude parameters (such as angle of attack and roll angle) to analyze the correlation between attitude changes and thermal load distribution.

[0065] In addition, the system status visualization and human-machine interaction module records events and manages logs throughout the entire system's operation. It automatically records and displays key events generated within the control cycle, such as: details of attitude adjustment commands issued by the control system, actual attitude angle changes performed by the aircraft, and the time and specific information of any change in the health status of any heat insulation tile (e.g., from "normal" to "warning").

[0066] Example 1 To verify the feasibility and engineering adaptability of the modular thermal insulation tile structure with intelligent sensing function proposed in this application, the configuration of a typical thermal insulation tile module is described below, combining specific structural design and fiber optic sensing embedding technology. (Refer to...) Figure 2 As shown.

[0067] In this embodiment, the overall structure of the heat-insulating tile still consists of a ceramic substrate shell, a lattice support layer, and a filling layer. Its basic structural form is based on the modular ceramic heat-insulating tile described in the applicant's patent CN120308326B. However, a real-time temperature monitoring module is added to achieve synchronous acquisition of the temperatures of the upper plate 1 and the lower plate 2 of the tile body. In this embodiment, a sapphire fiber grating (FBG) sensor array is integrated inside the tile body, combined with an external optical signal processing unit to form a complete temperature measurement system.

[0068] In terms of overall structural design, the heat insulation tile body consists of a ceramic matrix shell, a lattice support layer, and a lightweight filling layer. The ceramic matrix shell is made of SiC-based ceramic material with a density of 2.85 g / cm³, a room temperature strength of 350 MPa, and a high temperature strength of 220 MPa at 1200℃, which meets the strength requirements for thermal protection of the aircraft's outer surface. The ceramic matrix shell is machined into a hollow regular hexagonal prism shell with a side length of 100 mm and a total thickness of 20 mm. The upper plate 1 and the lower plate 2 of the ceramic matrix shell are both 2 mm thick. Three optical fiber embedding channels 3 are prefabricated along the centerline of opposite sides of the upper plate 1 and the lower plate 2. The three embedding channels 3 are all arranged along the centerline of opposite sides of the hexagonal prism. The cross-sectional diameter of each embedding channel 3 is 0.5 mm, and the roughness of the optical fiber embedding channel 3 is controlled to Ra≤1.6 μm to ensure stable fiber bonding and uniform subsequent adhesion. After processing, the fiber optic embedding channel 3 undergoes deburring and ceramic microcrack repair to ensure the stability of the optical signal after the fiber is installed. The three fiber optic embedding channels 3 are evenly arranged on the center line of the upper and lower plates of the tile body, so that the temperature monitoring points are evenly covered on the upper plate 1 and the lower plate 2 of the tile. Based on the temperature difference between the two layers, the direction and gradient change of local heat flux can be deduced, forming a complete thermal response monitoring capability.

[0069] In terms of embedded design of fiber Bragg grating sensing arrays ( Figure 3 High-temperature resistant sapphire fiber 4 is selected as the sensing carrier, and high-reflectivity FBG nodes are fabricated in the grating area using femtosecond laser writing technology. In this embodiment, the outer diameter of a single sapphire fiber 4 is 0.25 mm, the length is 173.2 mm, and a total of 10 fiber grating nodes are set with a spacing of 17.32 mm to achieve precise capture of the temperature distribution at different locations of the heat insulation tile. The reflection peak intensity stability of the sapphire fiber 4 is better than ±0.5 dB, and the temperature measurement accuracy is approximately ±0.3℃. The sapphire fiber 4 is laid along the prefabricated fiber embedding channel 3 during embedding, and is fixed by a high-temperature resistant inorganic adhesive. The inorganic adhesive is an alumina sol containing ZrO2 ceramic adhesive system with a temperature resistance of 1700℃ and a thermal conductivity of 0.5 W / (m·K). The coefficient of linear expansion after curing matches the ceramic shell.

[0070] In the specific operation, a 0.1mm thick adhesive layer is first applied to the bottom of the fiber embedding channel 3. The sapphire fiber 4 is then gently pressed into the fiber, allowing it to naturally adhere to the bottom of the fiber embedding channel 3. A second 0.15mm encapsulating adhesive layer is then applied over the sapphire fiber 4, completely covering the fiber. The inorganic adhesive is pre-cured at 120–180℃ for 20–30 minutes and finally cured at 450–600℃ for 25–40 minutes. This process ensures the fiber remains stable during flight vibrations and thermal cycling, preventing relative slippage or localized stress concentration. After the inorganic adhesive cures, the tail end of the sapphire fiber 4 is led out through a ceramic encapsulation tube to the interface area on the side of the heat insulation tile. The interface area is equipped with quick-plug fiber optic connection terminals 5, forming a standardized interconnection with adjacent tiles, facilitating the replacement of tile components and the rapid deployment of the testing network.

[0071] Example 2 This embodiment illustrates the specific implementation method of modular intelligent thermal insulation tiles in the actual assembly and layout optimization process. Figure 4 , Figure 5 This section illustrates the interconnection method of sapphire optical fiber 4, the rapid curing process of adhesive film 6, and the thermal insulation structure 7 between tiles. Through reasonable structural design and material selection, the modular thermal insulation tile array can maintain reliable thermal encapsulation and stable temperature measurement links even under high temperature, vibration, and strong airflow scouring environments.

[0072] During the assembly of modular thermal insulation tiles, a 1.5mm thick, rapidly curing ceramic adhesive film 6 containing SiC micropowder is first uniformly coated onto the surface of the aircraft skin. This adhesive film 6 uses a water-based modified alumina sol system with a solid content of 55–65 wt%, of which the SiC micropowder mass fraction is 8–12%. This adhesive film 6 exhibits strong initial tack at room temperature, with an initial shear strength reaching 0.8 MPa. This allows the thermal insulation tiles to achieve stable initial adhesion without external force after bonding, ensuring the tiles maintain precise spatial positioning before curing. The adhesive film 6 develops sufficient adhesion after being left at room temperature for 20–40 seconds, making it suitable for rapid assembly of large-area arrays.

[0073] To achieve scalability of the temperature monitoring network between tiles, this embodiment features fiber optic leads on the sides of the insulation tiles, and interconnects the fiber optic links during tile assembly using quick-plug fiber optic connectors 5. The fiber optic leads achieve modular networking via quick-plug fiber optic connectors 5, with a link insertion loss of less than 0.8dB. The internal optical path supports 1550nm wavelength fiber Bragg grating signal transmission and can be directly connected to the array bus optical path without additional splicing, thus significantly improving assembly efficiency and reducing maintenance complexity.

[0074] To prevent the joints between tiles from becoming potential heat conduction channels, this embodiment incorporates a thermal insulation structure 7 between the tiles. First, a flexible ceramic fiber sealing strip is arranged. This strip is made of Al2O3-SiO2 fiber with a bulk density of 0.18 g / cm³ and an initial thickness of 1.8 mm. During tile pressing, the flexible ceramic fiber sealing strip automatically undergoes 20–35% reversible compression deformation, ensuring tight filling and initial thermal barrier and airtightness in the joint area during assembly. The thermal conductivity of the compressed flexible ceramic fiber sealing strip is approximately 0.08 W / (m·K) at 800°C, effectively preventing heat coupling between adjacent tiles. On the outside of the flexible ceramic fiber sealing strip, this embodiment further introduces a layer of rapidly curing, high-temperature resistant ceramic adhesive to form a second sealing structure. This high-temperature resistant ceramic adhesive uses a ceramic-organic composite system containing ZrO2 toughening powder, with a viscosity controlled at 22000 mPa·s, ensuring uniform flow and filling in narrow joints. The colloid provides an initial strength of 2 MPa at room temperature and can be cured within 30 seconds under mild heating at 120°C, enabling the joint to acquire thermal shock resistance, erosion resistance, and vibration resistance in a very short time. The cured ceramic adhesive can withstand temperatures up to 1650°C, and its shear strength can be stably maintained at 12 MPa.

[0075] After the joint filling and curing are completed, a fast-drying, high-temperature erosion-resistant coating can be directly sprayed onto the outer surface of the joint. This coating uses a modified silicate-SiC microparticle reinforcement system, with a coating thickness of 115μm. It can form a surface dry layer within 60 seconds at room temperature and complete curing at 300℃. Its resistance to airflow erosion can withstand the total pressure and thermal flow impact under Mach 7 flight conditions, forming the final external protective layer for the joint area.

[0076] By employing the aforementioned structural layout and material system, this embodiment achieves highly reliable thermal isolation and high sealing performance of the thermal insulation tile array in high-temperature environments. Simultaneously, a scalable fiber optic interconnect system enables real-time, continuous tile-level temperature monitoring. The system can expand its sensor network without altering the tile structure as the array size increases, providing aircraft thermal protection systems with stable and rapid-response networked temperature monitoring capabilities in high-enthalpy environments.

[0077] Example 3 This embodiment takes a modular thermal protection system for hypersonic vehicles as the application object and presents a human-machine collaborative intelligent thermal protection implementation scheme based on graph structure modeling, time-series graph neural network prediction, attitude feedforward control, and system state visualization. It is used to realize real-time perception, early prediction, and active control of the thermal state of the insulation tile.

[0078] The method in this embodiment can be deployed on an airborne computing platform or a ground simulation verification system. Its functional flow includes: thermal protection system graph modeling and perception data construction, time-series graph neural network prediction, thermal insulation tile health status judgment, attitude feedforward control, and system status visualization and human-computer interaction, forming a complete prediction-control closed loop.

[0079] In this embodiment, as Figure 5 As shown, the modular heat-insulating tiles on the outer surface of the aircraft are laid in a regular hexagonal pattern, with adjacent tiles forming a regular spatial adjacency relationship through physical contact. Based on this, the modular heat-insulating tile system can be abstracted into a graph structure, as shown below. Figure 6 As shown, each heat insulation tile corresponds to a node in the graph, and graph edges are established between adjacent heat insulation tiles to characterize the spatial thermal coupling relationship between them. First, a graph structure model of the modular heat insulation tile system on the outer surface of the aircraft is performed. The modular heat insulation tile system on the outer surface of the aircraft is abstracted as an undirected graph:

[0080] Among them, the node set Each node in Corresponding to a single independent heat insulation tile; edge set Used to indicate the physical adjacency between heat insulation tiles, when heat insulation tiles and When directly adjacent in space, at nodes and Establish graph edges between them.

[0081] In this embodiment, modular heat insulation tiles are laid in a hexagonal pattern on the outer surface of the aircraft. Therefore, the constructed graph structure has a hexagonal topology, which is used to accurately depict the spatial thermal coupling relationship between the heat insulation tiles.

[0082] After completing the graph structure modeling, the distributed temperature data of the upper and lower layers of the heat insulation tile are acquired based on the fiber Bragg grating (FBG) sensor array deployed inside the heat insulation tile; at the same time, the flight control system acquires the aircraft attitude parameters in real time.

[0083] For any node At time step t Construct node feature vectors:

[0084] in, Temperature of the upper layer of the heat insulation tile; Temperature of the lower layer of the insulation tile; The temperature difference between the upper and lower shelves; This represents the rate of change of temperature difference. The current angle of attack of the aircraft; This represents the aircraft's current roll angle.

[0085] Node In continuous time interval The feature vectors within the node are stacked in chronological order to form a node temporal feature sequence: ; This serves as the input data for subsequent temporal graph neural networks. The feature vectors of each node across multiple consecutive time steps are stacked sequentially to form a node temporal feature sequence, as shown in the diagram. Figure 7 As shown.

[0086] like Figure 8 As shown, this embodiment uses a temporal graph neural network to predict the future thermal state of the thermal insulation tile. The model consists of a graph neural network module for extracting spatial thermal coupling features and a temporal neural network module for modeling the temperature-time evolution. In this embodiment, the constructed multi-time-step node feature sequence is input into the temporal graph neural network model to predict the future thermal state of the thermal insulation tile.

[0087] At any time step t Graph neural networks are used to spatially aggregate node features. Taking a graph convolutional network as an example, its computational form is as follows:

[0088] in: For nodes The set of neighboring nodes; Adjacency weight; It is a trainable parameter matrix; It is a non-linear activation function.

[0089] Through the above calculations, the time step of each heat insulation tile is obtained. t Spatial embedding features This is used to characterize the spatial thermal coupling relationship between it and adjacent insulating tiles.

[0090] Continuous L Spatial embedding feature sequence at each time step:

[0091] The input is fed into a time-series graph neural network for time evolution modeling. The time-series graph neural network can be a Long Short-Term Memory (LSTM) network, a gated recurrent unit (GRU), or another type of time-series graph neural network, and its output is a future preset time window. Prediction results within:

[0092] Among them, the prediction results This includes the future temperature of the insulation tiles or comprehensive thermal risk indicators.

[0093] In this embodiment, based on the predicted temperature or thermal risk index, and combined with a preset set of thresholds:

[0094] The health status of thermal insulation tiles is divided into four levels: normal, warning, risk, and failure, and the health status label is dynamically updated for each thermal insulation tile.

[0095] When the prediction indicates that at least one insulation tile will enter a risk state or fail within a preset time window, the system executes a closed-loop control process of prediction-judgment-control, the overall logical relationship of which is as follows: Figure 9 As shown. The control system constructs an attitude feedforward control problem:

[0096] Under the conditions of satisfying flight dynamics and control constraints, by adjusting the angle of attack of the aircraft and roll angle This alters the aerodynamic heat flow distribution on the outer surface to reduce the thermal load on the target insulation tile.

[0097] The control process can be implemented using model predictive control (MPC), rule-driven control, or learning-based policy mapping methods, forming a prediction-control closed loop.

[0098] After completing attitude feedforward control, this embodiment further includes system state visualization and human-computer interaction steps. In this embodiment, the visualization interface formed during system operation is as follows: Figure 10 As shown, the visualization interface integrates a state transition diagram (top left), a thermal insulation tile state heat map (bottom left), a time series curve of temperature and attitude parameters (top right), and a control log display module (bottom right), which is used to realize an intuitive presentation of the system's operating status and post-task evaluation.

[0099] Although the embodiments of this application have been described above in conjunction with the accompanying drawings, this application is not limited to the specific embodiments and application fields described above. The specific embodiments described above are merely illustrative and instructive, not restrictive. Those skilled in the art can make many other forms based on the guidance of this specification and without departing from the scope of protection of the claims of this application, and these are all within the scope of protection of this application.

Claims

1. A method for predicting and controlling the surface temperature of an aircraft based on a time-series graph neural network, characterized in that, Includes the following steps: S1. A high-temperature resistant fiber optic grating sensor array is installed inside the modular heat insulation tile to perform real-time distributed monitoring of the temperature of the upper and lower layers of the heat insulation tile and obtain temperature data. S2. Model all modular thermal insulation tiles on the outer surface of the aircraft as a graph structure. , where the set of nodes Each node in Corresponding to a heat insulation tile; edge set The edge in Represents a node and The corresponding heat insulation tiles are adjacent in physical space; S3. Based on the temperature data and the aircraft attitude parameters received in real time from the flight control system, within the time interval... Each node Built on continuous Temporal feature sequences at each time step: Among them, time step Node feature vectors Defined as: in, Temperature of the upper layer of the heat insulation tile; Temperature of the lower layer of the insulation tile; The temperature difference between the upper and lower shelves; This represents the rate of change of temperature difference. The current angle of attack of the aircraft; This is the aircraft's current roll angle; S4. The time-series feature sequence The input is fed into a time-series graph neural network model to predict the future time window for each thermal insulation tile. Internal temperature or thermal risk indicators ; S5. Based on the prediction results of step S4 With the preset threshold set Compare and determine the health status of each insulation tile. ,in, ; S6. When at least one heat insulation tile reaches a preset risk or failure level, construct and solve the attitude feedforward control optimization problem, and change the aerodynamic heat flow distribution on the outer surface by adjusting the aircraft's angle of attack and / or roll angle to reduce the thermal load on the target heat insulation tile.

2. The method for predicting and controlling the surface temperature of an aircraft according to claim 1, characterized in that, In step S1, the modular heat insulation tile includes a ceramic substrate shell designed as a hollow regular hexagonal prism. The ceramic substrate shell includes an upper plate and a lower plate. The upper plate and the lower plate have three optical fiber embedding channels distributed along the center lines of opposite sides. The high-temperature resistant fiber grating sensing array includes high-temperature resistant sapphire optical fibers arranged in the optical fiber embedding channels. The sapphire optical fibers have multiple equally spaced fiber grating nodes engraved on them.

3. The method for predicting and controlling aircraft surface temperature according to claim 2, characterized in that, The diameter of the fiber optic embedded channel is 0.4-0.6 mm; the spacing between the fiber grating nodes is 10-20 mm.

4. The method for predicting and controlling the surface temperature of an aircraft according to claim 1, characterized in that, During the assembly of modular thermal insulation tiles, structural bonding and composite filling processes are used between adjacent tiles; specifically including: A fast-curing ceramic adhesive film is applied to the surface of the aircraft skin to bond the heat insulation tiles; A flexible ceramic fiber sealing strip is installed at the joint between adjacent heat insulation tiles, and the outside of the flexible ceramic fiber sealing strip is filled with fast-curing high-temperature resistant ceramic adhesive; Apply a quick-drying, high-temperature erosion-resistant coating to the outer surface of the joint.

5. The method for predicting and controlling the surface temperature of an aircraft according to claim 1, characterized in that, In step S4, the time-series graph neural network model includes: The graph neural network module is used to aggregate the features of nodes at each time step to obtain spatial embedding features that reflect the spatial thermal coupling relationship between the node and its neighboring nodes; the graph neural network module is a graph convolutional network, a graph attention network, or a message passing neural network. The temporal neural network module is used to model the temporal evolution of spatially embedded feature sequences at multiple consecutive time steps and output the prediction results within a preset future time window; the temporal neural network module is a long short-term memory network, a gated recurrent unit, or a temporal graph neural network.

6. The method for predicting and controlling the surface temperature of an aircraft according to claim 1, characterized in that, Step S5 is as follows: Set multiple thresholds corresponding to the predicted temperature or thermal risk index of the insulation tile; The predicted result for each heat insulation tile is compared with the threshold, and its health status is divided into four levels: normal, warning, risk, and failure.

7. The method for predicting and controlling the surface temperature of an aircraft according to claim 1, characterized in that, In step S6, the objective function of the attitude feedforward control optimization problem is to minimize the predicted temperature or thermal risk index of the target heat insulation tile, the optimization variables are the aircraft angle of attack and roll angle, and the constraints are flight dynamics and control constraints; the attitude feedforward control optimization problem is solved using model predictive control, rule-driven control, or learning-based policy mapping methods.

8. The method for predicting and controlling the surface temperature of an aircraft according to claim 1, characterized in that, The method also includes state visualization and interaction: It can receive and display the temperature data, health status, and aircraft attitude parameters of each heat insulation tile in real time. The layout of the heat insulation tiles on the surface of the aircraft is displayed graphically, and different colors are used to map their health status. Dynamically plot the curves of temperature, temperature difference, and attitude parameters of each heat insulation tile over time; Record and display control commands, posture adjustment actions, and health status switching events.

9. A modular heat-insulating tile applicable to the aircraft surface temperature prediction and control method according to any one of claims 1 to 8, characterized in that, include: The ceramic-based shell is constructed as a hollow regular hexagonal prism shell, consisting of an upper plate, a lower plate, and a sidewall connecting the upper plate and the lower plate; optical fiber embedding channels are provided in the upper plate and the lower plate. A lattice support layer and a filling layer are disposed within the internal space of the ceramic substrate shell; the lattice support layer comprises a plurality of closely arranged permeable lattice units; the filling layer is made of a lightweight material with low thermal conductivity and fills the spaces between adjacent permeable lattice units and the internal cavities of each permeable lattice unit. The real-time temperature monitoring module comprises a high-temperature resistant fiber Bragg grating sensor array embedded in the optical fiber channel, and an optical signal processing unit that is communicatively connected to the fiber Bragg grating sensor array.

Images