Layered grinding and polishing process method and system for repairing damaged surface of aircraft radome

Abstract

The invention belongs to the technical field of intelligent grinding and polishing control, and particularly discloses a layered grinding and polishing process method and system for repairing the damaged surface of an aircraft radome. Comprising the following steps: determining repair boundaries and thickness changes of coatings with different functions in a damaged area of the aircraft radome; a layered grinding and polishing track is planned, and initial process parameters are set for each coating; layered grinding and polishing operation is executed, mechanical parameters, surface parameters and energy parameters are collected in real time in the grinding and polishing process, and an improved layered grinding and polishing contact force model is constructed to conduct local compensation control on grinding and polishing contact force; evaluating the surface quality, the repair integrity and the interlayer consistency of the current grinding and polishing layer in real time; and constructing a multi-target quality evaluation and process parameter collaborative optimization mechanism, and dynamically adjusting subsequent grinding and polishing process parameters according to the deviation between a real-time evaluation result and a preset quality target. According to the method, the material removal accuracy is remarkably improved, the problem of excessive grinding or non-uniform removal is avoided, and it is ensured that the quality of the repaired coating meets the standard.

Description

A layered grinding and polishing process and system for repairing damaged surfaces of aircraft radomes Technical Field

[0001] This invention belongs to the field of intelligent control technology for grinding and polishing, and more specifically, relates to a layered grinding and polishing process and system for repairing damaged surfaces of aircraft radomes. Background Technology

[0002] As the most critical and vulnerable structural component in the nose of an aircraft, the radome endures complex aerodynamic, thermal, humid, and electromagnetic environments over long periods of service. Aircraft operate under conditions of rain erosion, hail, ultraviolet radiation, and cyclical temperature and humidity. The multi-layered functional coatings on the radome surface (antistatic layer, anti-corrosion layer, electromagnetic wave-transmitting layer, and outer protective paint layer) are prone to complex damage such as cracking, peeling, delamination, and corrosion. Statistics show that the annual average surface erosion and lightning strike damage rate of typical aircraft radomes exceeds 10%. Damaged areas are often distributed at composite material interfaces and coating layer transitions. The quality of repair directly affects the wave transmission performance and structural safety of the aircraft's radar system.

[0003] However, the repair of radomes faces several technical bottlenecks. First, the damaged areas on the radome surface are complex in morphology and have large defects, including cracks, peeling, and corrosion spots. The uneven distribution and varying sizes of the damage mean that the repair requirements for each damaged area differ. This variability makes the repair process highly unpredictable, increasing the complexity and difficulty of the repair process. Second, the thickness of each coating layer on the radome surface varies; for example, the antistatic layer is thinner, while the electromagnetic wave-transmitting layer is thicker. The significant differences in physical properties (such as hardness and toughness) between different layers result in strong nonlinear characteristics in material removal and mechanical response during the repair process. Due to the large differences in the bonding force between coating layers, improper control during the repair process may lead to coating peeling, delamination, or surface unevenness, thereby affecting the electromagnetic performance and appearance quality of the radome. Currently, the repair of radomes mainly involves first grinding away the original damaged surface coating, and then re-spraying a protective coating with different functions. The grinding process is mainly done manually and must be done in layers—because the material composition and process requirements of each coating layer are different, it is impossible to grind the entire layer at once. The maintenance and repair of radomes in domestic and international aviation maintenance systems still primarily rely on manual inspection and experience-driven hand polishing. Traditional processes suffer from poor consistency and quality assurance when dealing with the complex multi-layered structure of radomes, and cannot effectively control the uniformity of material removal during the repair process. Specific technical difficulties include: 1) Insufficient coating removal precision: Due to the differences in physical properties of different functional coatings, traditional polishing methods struggle to ensure uniform removal of each layer. Over- or under-removal will affect the quality of the repaired radome. Especially when the protective layer is thin, improper removal control can easily expose the underlying material or cause damage to the substrate. 2) Non-uniformity of the repair area: The complex curvature of the radome surface and significant variations in interlayer hardness and thickness make it difficult for traditional manual polishing methods to adapt to minute changes in surface shape. During the repair process, the contact force and feed speed of the polishing disc cannot be precisely controlled, easily leading to localized over- or under-grinding, thus affecting surface quality. 3) Lack of quality verification mechanisms: Traditional repair processes mainly rely on manual sampling and visual inspection, lacking quantitative assessment methods for coating removal depth, surface roughness, and interlayer bonding strength. The lack of an effective quality assessment and closed-loop feedback mechanism leads to unstable surface quality after repair, and the repair effect is difficult to trace and quantify. 4) High repair difficulty and difficulty in ensuring quality: Due to the differences in physical properties of the radome coating and the non-uniformity of damage, traditional manual repair methods cannot perform targeted layer-by-layer polishing, resulting in the repair quality often failing to meet standards. During recoating, uneven interlayer bonding and excessively high surface roughness prevent the new coating from adhering evenly, affecting the electromagnetic wave transmission performance and mechanical strength of the radome. Summary of the Invention

[0004] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a layered grinding and polishing process and system for repairing damaged surfaces of aircraft radomes. Taking into account the characteristics of aircraft radomes—large differences in physical properties (hardness, toughness, thickness), complex curvature, and the tendency for localized over- or under-grinding and interlayer bonding damage in traditional repair methods—a dynamically adaptable layered grinding and polishing process verification and closed-loop control system is designed. Key components (modules), such as the layered equivalent stiffness modulation mechanism based on the Kelvin-Voigt model and the three-layer closed-loop feedback mechanism based on a heterogeneous graph neural network (HGNN) quality dynamic evaluation model, are studied and designed, along with their specific configurations. This effectively solves the technical bottlenecks of traditional grinding and polishing methods, such as difficulty in ensuring uniform removal of each coating layer, poor process consistency, and a lack of quantitative evaluation and verification mechanisms. Furthermore, it possesses high reliability features including multi-source data fusion for real-time perception, adaptive reverse dynamic parameter adjustment, and ensuring no layer mixing during multi-layer coating repair. Therefore, it is particularly suitable for applications involving the repair of multi-layer composite material damage in aircraft radomes, intelligent maintenance, and high-quality recoating.

[0005] To achieve the above objectives, according to one aspect of the present invention, a layered grinding and polishing process for repairing damaged surfaces of aircraft radomes is proposed, comprising the following steps: Step 1, identifying the spatial distribution characteristics of the damaged area of ​​the aircraft radome and the layered structure information of the multilayer functional coatings, and determining the repair boundaries and thickness variations of different functional coatings; Step 2, establishing a layered model based on the layered structure of the coatings, planning the layered grinding and polishing trajectory, and setting initial process parameters for each layer of coating.

[0006] Step 3: Perform layered grinding and polishing operations. During the grinding and polishing process, mechanical parameters, surface parameters, and energy parameters are collected in real time. Based on the Kelvin-Voyt model, a layered equivalent stiffness modulation mechanism is introduced to construct an improved layered grinding and polishing contact force model for local compensation control of the grinding and polishing contact force. Step 4: A process, state, and quality correlation network is established using a heterogeneous graph neural network. The collected multi-source parameters are integrated to evaluate the surface quality, repair integrity, and interlayer consistency of the current grinding and polishing layer in real time. Step 5: Construct a multi-objective quality assessment and process parameter collaborative optimization mechanism. Based on the deviation between the real-time assessment results and the preset quality target, the subsequent grinding and polishing process parameters are dynamically adjusted until the quality of the current layer meets the verification requirements before proceeding to the next grinding and polishing layer, thus achieving closed-loop control.

[0007] As a further preferred embodiment, in step three, the improved layered grinding and polishing contact force model includes: In the formula, The force applied by the grinding head, , The respective The equivalent stiffness and damping coefficient of the coating layer change dynamically with the layer thickness and material properties. This represents the contact compression of the layer.

[0008] As a further preferred embodiment, in step four, the multi-source parameters constitute a multi-source feature set of the heterogeneous graph neural network: In the formula, These represent mechanical, surface, and energy characteristics, respectively. This represents the number of sampling time points; after the multi-source feature set is aggregated through a time sliding window, it forms a node state vector. , which serves as the input to the heterogeneous graph neural network.

[0009] As a further preferred option, in step four, a heterogeneous graph structure is constructed for the heterogeneous graph neural network. , where the node set The edge set represents the process parameters, quality indicators, and sensor signal variables during the grinding and polishing process. Type set represents the relationship between different variables. Defining the heterogeneous edge type and its weight function, the update rule for the heterogeneous graph neural network layer is as follows: In the formula, For the first Layer node representation, For the first +1 level node representation, These are the weight matrices for different types of edges. For activation function, The normalization coefficient is... For nodes In type The set of neighboring nodes of the edge, For the first Layer neighbor nodes The feature representation vector, The self-connection weight matrix represents the node's own features; the node embedding matrix is ​​obtained through multi-layer propagation in a heterogeneous graph neural network. Embed the node into the matrix Mapping to a multi-quality index space: ,in, That is, the surface roughness index; That is, remove integrity indicators; This refers to the inter-layer consistency index.

[0010] As a further preferred embodiment, step five, the construction of the multi-objective quality assessment and process parameter collaborative optimization mechanism, includes: targeting the first... Establish process parameter vectors for layer coating ,in, It is the first Layer grinding and polishing feed rate; It is the first Layer grinding and polishing speed; It is the first Layer grinding and polishing contact pressure; using layer quality evaluation function To constrain the process parameters, joint optimization is performed: in, It is the first The optimal combination of process parameters obtained by the layer under the current quality constraints; These are the weighting coefficients for inter-layer continuity constraints. It is a penalty term for changes in interlayer parameters.

[0011] As a further preferred embodiment, in step five, the closed-loop control is based on the overall quality score function. The overall quality score function is used to make a determination. The computational models include: In the formula, As the indicator weight, For the variance of the fluctuation, The penalty coefficient is... The dynamic rate of change of each indicator; when When the polishing quality deviates from the target threshold, adaptive adjustment is triggered. The quality function threshold set for the system.

[0012] As a further preferred embodiment, the adaptive adjustment includes adjusting the polishing parameters using predicted deviations: In the formula, This is the Jacobian matrix of process parameters versus quality indicators. To adjust the step size This is the Jacobian matrix of process parameters versus quality indicators. To adjust the step size, The output of the heterogeneous graph neural network is the first... Predicted values ​​for each quality indicator The first obtained through multi-source sensors or online detection systems The actual measured values ​​of each quality indicator.

[0013] As a further preferred embodiment, the closed-loop control employs a three-layer closed-loop feedback mechanism: outer loop: the first layer of feedback based on the HGNN output... The target is the quality index of the coating layer. When the quality deviation exceeds the threshold, the target process parameters of the layer are corrected, and layer switching is only allowed after the current layer quality meets the standard. The middle ring: According to the outer ring command, combined with the interlayer continuity constraint, the grinding and polishing speed, feed rate and contact pressure within the layer are optimized in a coordinated manner to maintain a smooth transition of interlayer parameters. The inner ring: Based on the layered equivalent compliant force control model, the grinding and polishing contact force is automatically adjusted according to the detected local stiffness changes to suppress transient stress concentration.

[0014] According to another aspect of the present invention, a layered grinding and polishing process system for repairing damaged surfaces of aircraft radomes is also provided, comprising: a multi-source data acquisition and layered processing module for identifying the spatial distribution characteristics of the damaged area of ​​the aircraft radome and the layered structure information of the multi-layered functional coatings, and determining the repair boundaries and thickness variations of different functional coatings; a layered process planning module for establishing a layered model based on the layered structure, planning the layered grinding and polishing trajectory, and setting initial process parameters for each coating layer; an execution unit for executing the layered grinding and polishing operation; and a control modeling module for acquiring mechanical parameters, surface parameters, and energy parameters in real time during the grinding and polishing process, and based on... The Kelvin-Voyt model introduces a layered equivalent stiffness modulation mechanism to construct an improved layered grinding and polishing contact force model for local compensation control of the grinding and polishing contact force. The multi-objective quality assessment and feedback module is used to establish a process, state, and quality correlation network using a heterogeneous graph neural network, integrate collected multi-source parameters, and evaluate the surface quality, repair integrity, and interlayer consistency of the current grinding and polishing layer in real time. It also constructs a multi-objective quality assessment and process parameter collaborative optimization mechanism, and dynamically adjusts the subsequent grinding and polishing process parameters according to the deviation between the real-time assessment results and the preset quality target until the quality of the current layer meets the verification requirements before proceeding to the next grinding and polishing layer, thus achieving closed-loop control.

[0015] As a further preferred embodiment, the multi-objective quality assessment and process parameter collaborative optimization mechanism includes: constructing a heterogeneous graph neural network model, which is used to integrate process parameters, morphological signals and mechanical response data, establish a correlation network between process, state and quality, and output a comprehensive quality score in real time. When the score is lower than a set threshold, the system is triggered to update the grinding and polishing process parameters. The system operates based on a three-layer closed-loop feedback mechanism of outer loop, middle loop and inner loop collaboration. Among them, the outer loop is responsible for layer switching judgment and parameter correction based on quality indicators, the middle loop is responsible for collaborative optimization of process parameters within the layer and smooth transition between layers, and the inner loop is responsible for real-time compensation of contact force based on changes in material stiffness.

[0016] In summary, compared with existing technologies, the technical solutions conceived in this invention have the following main advantages: 1. Achieving precise control of multi-layer coating grinding and polishing based on dynamic adaptation: This invention, based on the Kelvin-Voigt mechanical control model, introduces a layered equivalent stiffness modulation mechanism to achieve real-time adaptive adjustment of grinding and polishing process parameters to coating properties. The system can dynamically adjust the contact force and feed speed according to changes in the curvature of the radome surface and the hardness of the coating, maintaining the stability and uniformity of the grinding and polishing process. This system significantly improves the accuracy of material removal, avoids the problems of over-grinding or uneven removal, and ensures that the quality of the repaired coating meets the standards.

[0017] 2. A dynamic evaluation and feedback control of grinding and polishing quality based on heterogeneous graph neural networks (HGNN) is realized: This invention realizes dynamic evaluation of the grinding and polishing process through an HGNN model, which can perceive the nonlinear relationship between process parameters and surface quality in real time and output a comprehensive quality score. When the quality score Reaching the set threshold During this process, the system automatically adjusts process parameters to ensure stable quality. This model allows the system to precisely monitor the removal effect of each coating layer and automatically adjust the process, ensuring the polishing process remains at its optimal state.

[0018] 3. The three-layer closed-loop control system of the invention consists of an outer-loop quality feedback adjustment layer, a middle-loop process parameter collaborative optimization layer, and an inner-loop compliant force control layer. These layers work together to ensure precise repair of different coatings. The outer loop provides real-time feedback on the polishing quality through an HGNN model, and the control system automatically adjusts process parameters based on this feedback. The middle loop optimizes various process parameters (such as polishing speed and pressure) to ensure uniform removal of each coating layer. The inner loop maintains stable contact between the grinding head and the workpiece through compliant force control, preventing stress concentration and uneven grinding, and ensuring surface quality. Attached Figure Description

[0019] Figure 1 is a flowchart of a layered grinding and polishing process for repairing damaged surfaces of an aircraft radome according to an embodiment of the present invention; Figure 2 is a schematic diagram of the radome grinding and polishing system according to an embodiment of the present invention; Figure 3 is a flowchart of dynamic evaluation of grinding and polishing quality according to an embodiment of the present invention; Figure 4 is a schematic diagram of layered process verification and closed-loop control system according to an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0021] As shown in Figure 1, this invention provides a layered grinding and polishing process for repairing damaged surfaces of aircraft radomes. Addressing the scenario of repairing damage to multi-layer functional coatings on aircraft radomes, this method achieves high consistency and high reliability repair of damaged radome surfaces through layered modeling, layered trajectory planning, multi-target quality assessment, and collaborative optimization of process parameters. The overall process, as shown in Figure 1, mainly includes: collecting surface morphology, defect distribution, and grinding and polishing process status data of the damaged area of ​​the radome; processing and extracting layered information from the collected data to obtain damage characteristics and identify the coating layer structure; planning layered grinding and polishing trajectories based on the coating layer structure information to generate corresponding grinding and polishing trajectories; and controlling a robot to perform grinding and polishing operations according to the planned trajectories. During the grinding and polishing process, the quality status of the current grinding and polishing layer is assessed and feedback results are output. Based on the quality assessment results, the grinding and polishing process parameters are collaboratively optimized and adjusted. Simultaneously, layered grinding and polishing verification is completed, and quality assessment results are output.

[0022] In this invention, the radar dome polishing system is shown in Figure 2, wherein... This represents the vertical coordinate of each measurement point, which is the vertical depth from the workpiece surface to the polishing head. This indicates the position of each measurement point in the plane coordinate system.

[0023] More specifically, the layered polishing process for repairing damaged surfaces of aircraft radomes involved in this embodiment includes the following steps: Step 1, identifying the spatial distribution characteristics of the damaged area of ​​the aircraft radome and the layered structure information of the multilayer functional coatings, and determining the repair boundaries and thickness variations of different functional coatings.

[0024] Step 2: Establish a layered model based on the layered structure, plan the layered grinding and polishing trajectory, and set initial process parameters for each coating layer.

[0025] Step 3: Perform layered grinding and polishing operations. During the grinding and polishing process, mechanical parameters, surface parameters, and energy parameters are collected in real time. Based on the Kelvin-Voyt model, a layered equivalent stiffness modulation mechanism is introduced to construct an improved layered grinding and polishing contact force model for local compensation control of the grinding and polishing contact force.

[0026] Step four: Establish a process, state, and quality correlation network using a heterogeneous graph neural network, integrate collected multi-source parameters, and evaluate the surface quality, repair integrity, and interlayer consistency of the current polishing layer in real time. Step five: Construct a multi-objective quality assessment and process parameter collaborative optimization mechanism. Based on the deviation between the real-time assessment results and the preset quality targets, dynamically adjust the subsequent polishing process parameters until the quality of the current layer meets the verification requirements before proceeding to the next polishing layer, thus achieving closed-loop control.

[0027] In the above steps, based on the multi-layer functional coating structure information of the identified radome damage area, a layered polishing model is established. The specific process includes: (1) Layered structure parameter modeling: Based on the damage area identification results, the spatial distribution and thickness variation of each functional coating are determined, and a set of layered structure parameters is constructed. in, It is the first Layer coating number, It is the first Coating thickness, It is the first Layer material property parameters.

[0028] A layered structural model of the radome damage area is established using this set of structural parameters.

[0029] (2) Layered polishing trajectory planning: Based on the layered structure model, corresponding layered polishing trajectories are generated for the damaged area: in, It is the first The set of polishing trajectories of the layers It is the first Layer A trajectory point.

[0030] By using trajectory planning, a progressive polishing path is generated from the outer layer to the inner layer, avoiding the removal of mixed layers of different coatings.

[0031] (3) Setting initial process parameters for each layer: Based on the layer structure parameters and material removal model, set the corresponding initial process parameters for each layer: in, It is the first Polishing speed of layer grinding It is the first Layer feed rate, It is the first Layer contact pressure.

[0032] This parameter serves as the initial process condition for subsequent dynamic optimization and closed-loop control.

[0033] The key to this embodiment lies in achieving precise control of multi-layer coating repair through real-time optimization and dynamic adjustment of process parameters.

[0034] (1) Construction of multi-objective layered grinding and polishing quality control targets: In the layered grinding and polishing process, there are significant differences in the quality targets corresponding to different coatings. To avoid the interlayer performance mismatch problem caused by traditional single quality index control, this invention targets the first layer... Multi-objective quality constraint vectors are constructed using layer coatings: The three categories of indicators respectively constrain surface quality, repair integrity, and interlayer consistency.

[0035] At the same time, the system sets hierarchical weight coefficients. Construct a hierarchical quality evaluation function: This allows for differentiated constraints on the quality targets of different coatings.

[0036] (2) Mechanism for collaborative optimization of layered process parameters: For each coating, the system establishes a process parameter vector: ,in, It is the first Layer grinding and polishing feed rate; It is the first Layer grinding and polishing speed; It is the first Layer grinding and polishing contact pressure. The system uses a layer quality evaluation function. To constrain the process parameters, joint optimization is performed: in, It is the first The optimal combination of process parameters obtained by the layer under the current quality constraints; λ is the weighting coefficient of the interlayer continuity constraint, used to balance the two objectives of "optimal quality" and "smooth transition of interlayer parameters"; It is a penalty term for changes in interlayer parameters, used to limit abrupt changes in process parameters between adjacent coatings and reduce the risk of over-wearing, under-wearing and quality fluctuations caused by interlayer switching.

[0037] (3) Local compensation mechanism: Based on the Kelvin–Voigt model, this invention introduces a layered equivalent stiffness modulation mechanism to construct an improved layered grinding and polishing contact force model. Under the condition of uneven contact force distribution, the constant contact force between the grinding head and the workpiece surface is ensured by compliance force control, avoiding over- or under-grinding caused by local curvature changes or uneven hardness, as shown below: in, The force applied to the grinding head; , The respective The equivalent stiffness and damping coefficient of the coating layer change dynamically with the layer thickness and material properties. This represents the contact compression of the layer. This is achieved by introducing layering parameters. The model can reflect the differences in mechanical response of different coatings during the polishing process, ensuring that the polishing process remains stable in the areas of coating switching and stiffness abrupt change.

[0038] Furthermore, as shown in Figure 3, the quality feedback and closed-loop control involved in this embodiment include: the system collects grinding and polishing process data in real time through multi-source sensors such as force, displacement, temperature, vibration, and surface roughness. After normalization and feature embedding, a multi-dimensional feature vector is formed to provide input for subsequent neural network modeling; a heterogeneous graph structure is constructed based on the collected multi-source features, and process parameters, signal nodes, and quality index nodes are associated and mapped. The HGNN network is used to learn the nonlinear coupling relationship between different variables and output quality prediction values ​​such as roughness, removal integrity, and interlayer consistency; when the deviation between the predicted value and the actual measurement result exceeds a set threshold, the system starts the time-series weighted evolution algorithm to automatically correct key process parameters (such as contact pressure and feed rate) according to the fluctuation trend of each quality index, realizing quality-driven reverse regulation; the comprehensive quality score is calculated by combining the rate of change and variance of multiple quality indicators output by the HGNN. ,like Once the stable threshold is reached, the system determines that the grinding and polishing process is qualified and outputs an evaluation report; if the standard is not met, the self-learning and adjustment loop continues until the process converges. The specific process is as follows: (1) Dynamic quality evaluation: In order to realize online quality evaluation and closed-loop correction of the grinding and polishing process of the radome surface coating, this invention proposes a dynamic quality evaluation method for grinding and polishing based on heterogeneous graph neural network (HGNN). This method establishes a "process-state-quality" correlation network by integrating multi-source process parameters, morphological signals and mechanical response data, so as to realize real-time prediction, anomaly detection and reverse control of grinding and polishing quality.

[0039] First, the following multi-dimensional signals are collected in real time during the grinding and polishing process: 1) Mechanical parameters: grinding and polishing contact force, feed rate, tangential speed, etc.; 2) Surface parameters: removal depth, roughness. Interlayer thickness variation 3) Energy parameters: power, temperature, and grinding head vibration signals. These signals, after normalization and feature embedding, constitute a multi-source feature set: in These represent mechanical, surface, and energy characteristics, respectively. This represents the number of sampling time points. After aggregation through a time sliding window, a node state vector is formed. , as input to the graph neural network.

[0040] Secondly, to characterize the nonlinear coupling relationship between process parameters and multiple quality indicators, this invention constructs a heterogeneous graph structure. , where the node set Represents various variables in the grinding and polishing process (process parameter nodes, quality index nodes, sensor signal nodes); edge set Indicates the relationship between different variables (such as pressure-roughness, speed-temperature rise, removal amount-interlayer stress, etc.); type set Define the heterogeneous edge types and their weight functions. The update rules for the HGNN layers are as follows: in, For the first Layer node representation, These are the weight matrices for different types of edges. For activation function, The normalization coefficient is denoted as . This network can simultaneously learn the dynamic dependencies between different types of nodes, achieving an adaptive representation of changes in grinding and polishing quality.

[0041] Through multi-layer propagation of HGNN, this invention obtains the node embedding matrix. And map it to a multi-quality index space: ,in, That is, the surface roughness index; That is, remove integrity indicators; This refers to the inter-layer consistency index.

[0042] The system is based on the dynamic change rate of each indicator. Calculate the overall quality score function: in As the indicator weight, For the variance of the fluctuation, This is the penalty coefficient. When When the system determines that the polishing quality deviates from the target threshold, it triggers adaptive adjustment.

[0043] When the deviation between the quality indicators predicted by HGNN and the real-time sensing data exceeds a set threshold, the system automatically updates the grinding and polishing process parameters through the controller, realizing quality-driven reverse adjustment. Specifically, this includes: 1) Node adaptive update: Based on time-series weights, automatically adjusting the influence weights of key nodes (such as pressure and feed rate); 2) Time-series weight evolution algorithm: Optimizing short-term dynamic response through a time-series attention mechanism, making the model robust to sudden vibrations or material mutations; 3) Process parameter correction: Utilizing prediction deviations... Adjust grinding and polishing parameters: .in This is the Jacobian matrix of process parameters versus quality indicators. To adjust the step size.

[0044] After multiple updates, the system automatically converges to the optimal parameter set, achieving self-learning and self-stabilization of the grinding and polishing process.

[0045] Finally, the system generates a dynamic quality assessment report, which includes: layer removal rate distribution and residual cloud map; roughness time series curve and fluctuation trend; interlayer thickness consistency distribution map; comprehensive quality score and control record. These results can be used for subsequent layering process optimization and recoating quality prediction, realizing an intelligent quality management closed loop of "process monitoring - dynamic assessment - feedback optimization".

[0046] According to another aspect of the present invention, in order to realize the layered grinding and polishing verification and quality closed-loop control in the multi-layer coating damage repair process of radar dome, the present invention constructs a process verification and closed-loop control system for layered grinding and polishing. The system framework is shown in Figure 4. The multi-source data acquisition and layered processing module acquires multi-source information in the grinding and polishing process and analyzes the layered structure of the damaged area. The results are input to the layered process planning and control modeling module. The layered process planning and control modeling module generates the layered grinding and polishing trajectory and process parameter instructions, and controls the execution unit to complete the layered grinding and polishing operation. The grinding and polishing status data and surface quality information of the execution unit are fed back to the multi-target quality assessment and feedback module in real time. The quality assessment module assesses the grinding and polishing quality of the current layer and feeds back the assessment results to the control modeling module to realize the dynamic adjustment of process parameters. Finally, the human-computer interaction and process verification module displays and verifies the layered grinding and polishing process and results.

[0047] The specific introduction of each module is as follows: (1) Data collection and processing module: collect the surface morphology and defect distribution information of the radar dome damage area, as well as the contact force, displacement and other state data during the polishing process, and analyze the collected data to identify the spatial distribution characteristics and coating layer structure of the damage area, clarify the thickness change and repair boundary of different functional coatings, and provide structural basis for layer polishing. (2) Layer process planning and control modeling module: based on the layer modeling results, according to the polishing principle of progressive polishing from the outer layer to the inner layer, polishing trajectories are generated for different coatings, and different trajectory densities and initial process parameters are set for each layer to avoid the mixed removal of multiple layers of materials; different contact force is adjusted for different coatings to suppress over-polishing or under-polishing problems caused by changes in interlayer stiffness; based on the layer quality assessment results, the process parameters of the current layer and subsequent layers are coordinated and optimized to achieve unified constraints and dynamic balance of multiple quality targets in the layer polishing process. (3) Execution unit: executes robot polishing operation according to the layer trajectory and process parameter instructions, and provides real-time feedback of execution status data during the polishing process. (4) Multi-objective quality assessment and feedback module: For the current polishing layer, comprehensively assess the removal integrity, surface roughness and interlayer consistency; integrate multiple quality indicators through the HGNN model to output the comprehensive quality status; determine whether the quality requirements of the current layer are met, and feed back the adjustment signal to the control module. (5) Human-computer interaction and process verification module: Display the layered polishing process, quality assessment results and process adjustment status, and output the layered polishing process verification report.

[0048] The core control logic of the system is based on a three-layer closed-loop feedback mechanism of “layered process parameters - layered material response - multi-objective quality index”, and its principle is as follows: (1) The outer loop outputs the first layer of HGNN model. The quality indicators of the coating layer (roughness, uniformity, and removal integrity) are used as control targets. The quality deviation is calculated by comparing the predicted values ​​with the actual measured values. .when At that time, the system only applies to the current number. The layer triggers an outer loop control command to correct the target process parameters for that layer; when the first... When the layer quality meets the standard, the outer ring determines that the polishing of the layer is complete and sends a layer switching permission signal to the system to enter the polishing control of the next layer. (2) The middle ring uses the current polishing layer as the polishing stage. The layer is the controlled object. Based on the quality control commands output from the outer loop, and combined with the material properties and layer structure information of this layer, intra-layer collaborative optimization is performed on process parameters such as grinding and polishing speed, feed rate, and contact pressure. In the first... Within a layer, the intermediate ring coordinates the joint changes of multiple process parameters to achieve the simultaneous satisfaction of multiple target quality indicators for that layer; between layers, the intermediate ring introduces inter-layer continuity constraints, ensuring that the first layer... Layer and First The process parameters between layers remain smooth to avoid over- or under-grinding caused by sudden parameter changes during layer switching, ensuring uniform removal of multiple coating layers throughout the repair process. (3) The inner ring is for the current grinding and polishing. The system employs a layered equivalent compliant force control model to perform real-time compensation control of the contact force between the grinding head and the coating layer. When the system detects the first layer... When there are local stiffness changes, thickness fluctuations, or interlayer transition areas within the layer, the inner ring automatically adjusts the polishing contact force according to the equivalent stiffness and damping parameters corresponding to that layer, maintaining a stable contact state between the grinding head and the current coating, and suppressing transient stress concentration caused by differences in interlayer mechanical properties.

[0049] Through the coordinated action of the outer, middle, and inner rings, the system sequentially performs polishing control on each functional coating of the radome. Each coating undergoes a complete closed-loop process of "intra-layer polishing—intra-layer evaluation—intra-layer optimization—layer completion judgment." Only after the quality verification of the current layer passes will the system allow the polishing operation of the next coating to proceed. This mechanism ensures that no mixed-layer removal or quality accumulation errors occur during the repair process of multi-layer coatings, achieving layer-by-layer verification and highly consistent repair of the multi-layer structure of the radome.

[0050] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A layered grinding and polishing process for repairing damaged surfaces of aircraft radomes, characterized in that, Includes the following steps: Step 1: Identify the spatial distribution characteristics of the damaged area of ​​the aircraft radome and the layered structure information of the multi-layer functional coatings, and determine the repair boundaries and thickness variations of different functional coatings. Step 2: Establish a layered model based on the coating layered structure, plan the layered polishing trajectory, and set initial process parameters for each coating layer. Step 3: Perform layered polishing operations, collecting mechanical, surface, and energy parameters in real time during the polishing process. Based on the Kelvin-Voyt model, introduce a layered equivalent stiffness modulation mechanism to construct an improved layered polishing contact force model for local compensation control of the polishing contact force. Step 4: Utilize a heterogeneous graph neural network to establish a process, state, and quality correlation network, fusing the collected multi-source parameters to evaluate the surface quality, repair integrity, and interlayer consistency of the current polished layer in real time. Step 5: Construct a multi-objective quality assessment and process parameter collaborative optimization mechanism. Based on the deviation between the real-time assessment results and the preset quality target, dynamically adjust subsequent polishing process parameters until the current layer quality meets the verification requirements before proceeding to the next polishing layer, achieving closed-loop control.

2. The layered grinding and polishing process for repairing damaged surfaces of aircraft radomes according to claim 1, characterized in that, In step three, the improved layered grinding and polishing contact force model includes: In the formula, The force applied by the grinding head, 、 The respective The equivalent stiffness and damping coefficient of the coating layer change dynamically with the layer thickness and material properties. This represents the contact compression of the layer.

3. The layered grinding and polishing process for repairing damaged surfaces of aircraft radomes according to claim 1, characterized in that, In step four, the multi-source parameters constitute the multi-source feature set of the heterogeneous graph neural network: In the formula, These represent mechanical, surface, and energy characteristics, respectively. This represents the number of sampling time points; after the multi-source feature set is aggregated through a time sliding window, it forms a node state vector. , which serves as the input to the heterogeneous graph neural network.

4. The layered grinding and polishing process for repairing damaged surfaces of aircraft radomes according to claim 3, characterized in that, In step four, the heterogeneous graph structure of the heterogeneous graph neural network is constructed. , where the node set This represents the process parameters, quality indicators, and sensor signal variables during the grinding and polishing process, and the edge set. Type set represents the relationship between different variables. Defining the heterogeneous edge type and its weight function, the update rule for the heterogeneous graph neural network layer is as follows: In the formula, For the first Layer node representation, For the first +1 level node representation, These are the weight matrices for different types of edges. For activation function, The normalization coefficient is... For nodes In type The set of neighboring nodes of the edge, For the first Layer neighbor nodes The feature representation vector, The self-connection weight matrix represents the node's own features; the node embedding matrix is ​​obtained through multi-layer propagation in a heterogeneous graph neural network. Embed the node into the matrix Mapping to a multi-quality index space: ,in, That is, the surface roughness index; That is, remove integrity indicators; This refers to the inter-layer consistency index.

5. The layered grinding and polishing process for repairing damaged surfaces of aircraft radomes according to claim 1, characterized in that, Step five, the construction of the multi-objective quality assessment and process parameter collaborative optimization mechanism includes: targeting the first... Establish process parameter vectors for layer coating ,in, It is the first Layer grinding and polishing feed rate; It is the first Layer grinding and polishing speed; It is the first Layer grinding and polishing contact pressure; using layer quality evaluation function To constrain the process parameters, joint optimization is performed: in, It is the first The optimal combination of process parameters obtained by the layer under the current quality constraints; These are the weighting coefficients for inter-layer continuity constraints. It is a penalty term for changes in interlayer parameters.

6. The layered grinding and polishing process for repairing damaged surfaces of aircraft radomes according to claim 1, characterized in that, In step five, the closed-loop control is based on the overall quality score function. The overall quality score function is used to make a determination. The computational models include: In the formula, As the indicator weight, For the variance of the fluctuation, The penalty coefficient is... The dynamic rate of change of each indicator; when When the polishing quality deviates from the target threshold, adaptive adjustment is triggered. The quality function threshold set for the system.

7. The layered grinding and polishing process for repairing damaged surfaces of aircraft radomes according to claim 6, characterized in that, The adaptive adjustment includes adjusting the polishing parameters using predicted deviations: In the formula, This is the Jacobian matrix of process parameters versus quality indicators. To adjust the step size This is the Jacobian matrix of process parameters versus quality indicators. To adjust the step size, The output of the heterogeneous graph neural network is the first... Predicted values ​​for each quality indicator The first obtained through multi-source sensors or online detection systems The actual measured values ​​of each quality indicator.

8. The layered grinding and polishing process for repairing damaged surfaces of aircraft radomes according to claim 1, characterized in that, The closed-loop control employs a three-layer closed-loop feedback mechanism: outer loop: based on the output of the HGNN... The target is the quality index of the coating layer. When the quality deviation exceeds the threshold, the target process parameters of the layer are corrected, and layer switching is only allowed after the quality of the current layer meets the standard. Middle ring: Based on the outer ring command and combined with interlayer continuity constraints, the grinding and polishing speed, feed rate and contact pressure within the layer are optimized in a coordinated manner to maintain a smooth transition of parameters between layers; Inner ring: Based on the layered equivalent compliant force control model, the grinding and polishing contact force is automatically adjusted according to the detected local stiffness changes to suppress transient stress concentration.

9. A layered grinding and polishing process system for repairing damaged surfaces of aircraft radomes, characterized in that, include: The multi-source data acquisition and layered processing module is used to identify the spatial distribution characteristics of the damaged area of ​​the aircraft radome and the layered structure information of the multi-layer functional coating, and to determine the repair boundary and thickness variation of different functional coatings; the layered process planning module is used to establish a layered model based on the layered structure, plan the layered polishing trajectory, and set the initial process parameters for each layer. The execution unit is used to perform layered grinding and polishing operations; The control modeling module is used to collect mechanical, surface, and energy parameters in real time during the grinding and polishing process. Based on the Kelvin-Voyt model, it introduces a layered equivalent stiffness modulation mechanism to construct an improved layered grinding and polishing contact force model for local compensation control of the grinding and polishing contact force. The multi-objective quality assessment and feedback module is used to establish a process, state, and quality correlation network using a heterogeneous graph neural network. It integrates the collected multi-source parameters to evaluate the surface quality, repair integrity, and interlayer consistency of the current grinding and polishing layer in real time. It also constructs a multi-objective quality assessment and process parameter collaborative optimization mechanism. Based on the deviation between the real-time assessment results and the preset quality target, it dynamically adjusts the subsequent grinding and polishing process parameters until the quality of the current layer meets the verification requirements before proceeding to the next grinding and polishing layer, thus achieving closed-loop control.

10. A layered grinding and polishing process system for repairing damaged surfaces of aircraft radomes according to claim 9, characterized in that, The multi-objective quality assessment and process parameter collaborative optimization mechanism includes: constructing a heterogeneous graph neural network model, which is used to integrate process parameters, morphological signals and mechanical response data, establish a correlation network between process, state and quality, and output a comprehensive quality score in real time. When the score is lower than a set threshold, the system is triggered to update the grinding and polishing process parameters. The system operates based on a three-layer closed-loop feedback mechanism of outer loop, middle loop and inner loop collaboration. Among them, the outer loop is responsible for layer switching judgment and parameter correction based on quality indicators, the middle loop is responsible for collaborative optimization of process parameters within the layer and smooth transition between layers, and the inner loop is responsible for real-time compensation of contact force based on changes in material stiffness.

Images