An on-board radar aperture antenna cover multi-layer curved surface frequency selection process alignment design method

Abstract

This invention discloses a method for aligning and designing a multi-layer curved surface frequency-selective process for airborne radar aperture radomes, relating to the field of radome technology. First, a multi-layer curved surface model with double curvature, including upper and lower frequency-selective surfaces, is constructed, defining the X-axis as the lateral arc direction and the Y-axis as the longitudinal arc direction. Based on prior data of the frequency-selective machinable dimensions, array-type positioning holes are generated on the curved surface. A partitioning and unfolding algorithm with area loss compensation is used to unfold the surface into a two-dimensional plane, with an area loss rate of ≤1.5‰. The biaxial spacing difference between the positioning holes of the upper and lower curved surfaces with the largest curvature is extracted. Within a safe threshold, the period of the frequency-selective unit is adjusted by coupling an adaptive biaxial period adjustment strategy. The optimal parameters are obtained through simulation verification and automatic correction using a pre-trained deep learning model, and a frequency-selective array pattern is generated. Finally, the planar frequency-selective film processing pattern is mapped to obtain the pattern. This invention can eliminate alignment gaps in the multi-layer curved surface frequency-selective structure, balancing stealth and radiation performance, and significantly improving process adaptability and processing quality.

Description

Technical Field

[0001] This invention relates to the field of radome technology, and in particular to a method for aligning and designing a multi-layer curved surface frequency-selective process for an airborne radar aperture radome. Background Technology

[0002] Frequency selective surfaces (FSS) are the core structure of airborne radar aperture radomes, enabling electromagnetic wave filtering and electromagnetic stealth. To meet the requirements of airborne applications, the radome body is usually designed as a hyperbolic curved surface structure. The multi-layered frequency selective structure loaded on its surface needs to be bonded to it and cured. Therefore, the alignment accuracy of the multi-layered frequency selective structure directly determines the electromagnetic performance of the radome.

[0003] In existing technologies, the alignment design of multi-layer curved surface frequency selective structures generally adopts the traditional process of "first unfolding the curved surface - then designing the frequency selective film - finally drilling positioning holes": first, the curved surface is unfolded into a plane, a planar frequency selective film with the same period unit size is designed, and after preparation, holes are drilled in the frequency selective film for curved surface fitting and laying. However, this process has two major drawbacks: 1. Inherent error in unfolding a double-curvature surface: a double-curvature surface cannot be unfolded into a plane without deformation. The traditional method completely ignores the stretching, compression, and area loss during the unfolding process, resulting in a deviation of more than 0.5 mm between the theoretical spacing difference and the actual spacing difference after processing; 2. Reference conversion error: the curvature line of the curved surface is the theoretical geometric reference, while the actual laying alignment uses the physical reference of the positioning holes. The two reference conversions introduce additional systematic errors.

[0004] For large-arc radomes, the difference in the distance between the arc lines of the upper and lower frequency-selective surfaces can reach more than 5 mm. The superposition of these errors will lead to obvious alignment gaps between the upper and lower frequency-selective units, compromising the integrity of the frequency-selective array. If these gaps are not addressed, the electromagnetic stealth performance of the radome will be significantly reduced; if metallization or other methods are used to fill the gaps, the local electromagnetic characteristics will be altered, worsening the radiation performance of the radar antenna. Current technology lacks an effective solution that can eliminate alignment gaps in multi-layer curved surface frequency-selective structures while ensuring fabrication feasibility and simultaneously maintaining both the stealth and radiation performance of the radome. Summary of the Invention

[0005] In view of this, this application provides a multi-layer curved surface frequency selective process alignment design method for airborne radar aperture radomes to overcome the shortcomings of the existing technology.

[0006] The first aspect of this application provides a method for aligning and designing a multi-layer curved surface frequency-selective process for an airborne radar aperture radome, comprising: Based on the digital model of the frequency-selective structure product of the radome, a multi-curvature surface model of the airborne radar aperture radome containing upper and lower frequency-selective surfaces is constructed. The X-axis is defined as the lateral curvature direction of the radome and the Y-axis is defined as the longitudinal curvature direction of the radome. Based on prior data of frequency-selectable machinable dimensions, including machinable parameters of positioning holes and matching parameters of frequency-selective units, an array of positioning holes is generated on a multi-layer frequency-selective surface. A partitioning unfolding algorithm with area loss compensation is used to unfold the hypercurvature multi-layer surface model into a two-dimensional plane, and the area loss rate of the upper and lower frequency-selected surfaces after unfolding is verified to be less than or equal to a set value. In the hypercurvature multi-layer surface, select the upper and lower corresponding frequency-selected surfaces with the largest curvature, and extract the first spacing difference in the X-axis direction and the second spacing difference in the Y-axis direction of the corresponding positioning holes at the same position respectively. A dual-axis period adaptive adjustment strategy is adopted. Within the safety threshold range determined by electromagnetic simulation, the period of the frequency-selective unit in the X-axis and Y-axis directions is coupled and adjusted based on the first spacing difference and the second spacing difference to generate the target period parameter. The target period parameters are input into a pre-trained deep learning model for geometric alignment and electromagnetic performance simulation verification and automatic correction, outputting the optimal target period parameters, and generating the optimal frequency-selective array pattern based on the optimal target period parameters. The optimal frequency-selective array pattern is mapped onto the two-dimensional plane to obtain a planar frequency-selective film processing pattern.

[0007] In one possible implementation of the first aspect, generating an array of positioning holes on a multi-layer frequency-selective surface includes: The opening positions of the positioning holes on the upper and lower frequency-selective curved surfaces are calibrated, the center of the positioning hole is aligned with the reference point of the frequency-selective unit array, and the distance between the center of the positioning hole and the center of the adjacent frequency-selective unit is controlled to be an integer multiple of the unit period.

[0008] In one possible implementation of the first aspect, the partitioning expansion algorithm with area loss compensation includes: The hypercurvature multi-layer surface model is divided into multiple sub-surface regions with uniform curvature changes; Each of the sub-surface regions is unfolded in plane, and the area loss rate during the unfolding process is calculated and compensated. At the seam between adjacent sub-surface regions, a size-gradient transition unit matching the frequency-selective unit topology is set to ensure the continuity of the frequency-selective stealth performance and radiation performance of the hyperbolic surface.

[0009] In one possible implementation of the first aspect, the dual-axis periodic adaptive adjustment strategy includes: The airborne radar radome is divided into a simple hyperbolic radome and a complex hyperbolic radome. The simple hyperbolic radome is a ruled surface with no geometric curvature difference in the Y-axis direction, while the complex hyperbolic radome is a non-ruled surface with geometric curvature difference in the Y-axis direction. For the simple hyperbolic radome, only the frequency-selective cell period adjustment is performed in the X-axis direction, while rigid alignment is maintained in the Y-axis direction; For the complex hyperbolic radome, frequency-selective element periodic coupling adjustment is performed in the X-axis and Y-axis directions.

[0010] In one possible implementation of the first aspect, performing frequency-selective unit period coupling adjustment in the X-axis and Y-axis directions includes: The period adjustment amount in the X-axis direction is calculated as the ratio of the first spacing difference in the X-axis direction to the number of frequency-selective units in that direction, and is recorded as the first period adjustment amount. Within the safety threshold range, the period of the X-axis frequency-selective units is proportionally corrected based on the first period adjustment amount. When the second spacing difference in the Y-axis direction is greater than or equal to 0.3 frequency selection unit cycles, the second cycle adjustment amount is calculated according to the ratio of the second spacing difference in the Y-axis direction to the number of frequency selection units in that direction. Within the safety threshold range, the cycle of the Y-axis frequency selection unit is proportionally corrected based on the second cycle adjustment amount. When the second spacing difference in the Y-axis direction is less than 0.3 frequency-selective unit cycles, the Y-axis unit cycle remains unchanged. Compensation is achieved through flexible tiling to ensure the physical alignment of the upper and lower frequency-selective surfaces in the Y-axis direction.

[0011] In one possible implementation of the first aspect, for the hypercurvature surface, when the X-axis and Y-axis directions are adjusted simultaneously, the safety threshold range is adaptively and dynamically adjusted according to the coupling relationship of the dual-axis adjustment amount.

[0012] In one possible implementation of the first aspect, the pre-trained deep learning model is obtained through the following steps: Simulation data of frequency-selective unit period adjustment of hypercurvature surfaces under different curvatures are collected as training sets to supervise the training of a deep learning model based on convolutional neural network and long short-term memory network until the parameter correction accuracy and electromagnetic performance prediction deviation of the deep learning model are both less than or equal to preset values.

[0013] In one possible implementation of the first aspect, the simulation verification and automatic correction of the target periodic parameters using a pre-trained deep learning model includes: When adjusting the X-axis and Y-axis periods simultaneously, the deep learning model is used to perform fusion compensation calculations on the period adjustment amounts of the two axes to eliminate alignment gaps between the multi-layer frequency-selective structures.

[0014] In one possible implementation of the first aspect, a planar frequency-selective film is prepared based on the pattern of the planar frequency-selective film. The planar frequency-selective film is then aligned and laid onto the upper and lower frequency-selective curved surfaces using positioning holes and pins. After integral curing and molding, defect identification and repair steps are performed. The surface image of the radome after it has been laid and cured is acquired using machine vision. The surface image is used to extract features using a machine learning model to identify and determine the type and level of defects in the surface image. Based on the defect type and the defect level, perform the corresponding repair operation.

[0015] Its beneficial effects are as follows: This invention discloses a method for aligning and designing a multi-layer curved surface frequency-selective process for an airborne radar aperture radome. First, a multi-layer curved surface model with double curvature, including upper and lower frequency-selective surfaces, is constructed, defining the X-axis as the lateral arc direction and the Y-axis as the longitudinal arc direction. Based on prior data of the frequency-selective machinable dimensions, array-type positioning holes are generated on the curved surface. A partitioning and unfolding algorithm with area loss compensation is used to unfold the surface into a two-dimensional plane, with a verified area loss rate ≤1.5‰. The biaxial spacing difference between the positioning holes of the upper and lower curved surfaces with the largest curvature is extracted. Within a safe threshold, the frequency-selective unit period is adjusted by coupling an adaptive biaxial period adjustment strategy. The optimal parameters are obtained through simulation verification and automatic correction using a pre-trained deep learning model, and a frequency-selective array pattern is generated. Finally, the planar frequency-selective film processing pattern is mapped to obtain the pattern. This invention can eliminate alignment gaps in the multi-layer curved surface frequency-selective structure, balancing stealth and radiation performance, and significantly improving process adaptability and processing quality. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the alignment design method for a multi-layer curved surface frequency selective process of an airborne radar aperture radome provided in an embodiment of this application; Figure 2 This is a multi-layer curved surface model of an antenna radome provided in an embodiment of this application; Figure 3 This is a schematic diagram of the alignment of upper and lower layer frequency-selective array layouts provided in an embodiment of this application; Figure 4 This is a partial schematic diagram of the alignment of upper and lower frequency-selective curved surfaces provided in an embodiment of this application; Figure 5This is a partial alignment diagram of the unfolded upper and lower frequency-selective curved surfaces provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0020] To better understand this application, the technical names involved in this application are explained below: Frequency-selective structure product digital model of radome: refers to the three-dimensional design digital model of airborne radar aperture radome. It is the only reference data for constructing a double curvature multi-layer surface model. It includes core geometric and performance parameters such as the radome's external dimensions, surface curvature parameters, skin thickness, number of frequency-selective structure design layers, frequency-selective surface placement, full-size design tolerances, and electromagnetic performance design requirements.

[0021] Machinable parameters for positioning holes: These refer to the geometric parameters and tolerance requirements that can be stably machined based on the existing mature machining capabilities of airborne radomes. They are a core component of the prior data for frequency-selective machinable dimensions, including but not limited to the hole diameter design range, hole shape requirements, hole opening accuracy, hole spacing tolerance, hole wall roughness, and through-hole coaxiality requirements, ensuring that the design scheme for positioning holes can be stably implemented through existing processes.

[0022] Frequency selective unit matching parameters: These are the associated design parameters that ensure the matching and accuracy of the positioning hole with the frequency selective unit array reference. They are a core component of the prior data of the machinable dimensions of the frequency selective unit, including but not limited to the overlap requirements of the positioning hole and the reference point of the frequency selective unit array, the distance matching requirements between the center of the positioning hole and the center of the adjacent frequency selective unit, the matching relationship between the positioning hole spacing and the integer multiple of the frequency selective unit period, and the matching tolerance between the machining accuracy of the positioning hole and the machining accuracy of the frequency selective unit.

[0023] Prior data corresponding to frequency-selectable machinable dimensions: refers to the standardized process database accumulated through a large number of airborne radome frequency-selective structure process tests and electromagnetic simulation verifications. It includes prior constraint data that can be directly used for process design, such as positioning hole machinability parameters, frequency-selective unit matching parameters, and multi-layer curved surface alignment tolerance requirements.

[0024] Example In existing technologies, the alignment design of multi-layer curved frequency-selective structures generally employs planar frequency-selective films of the same periodic unit size for curved surface fitting and mounting. However, due to the curved spatial structure characteristics of the radome, the arc length of the upper frequency-selective surface is greater than that of the lower frequency-selective surface; for radomes with large arcs, this spacing difference can reach more than 5 mm. This dimensional difference leads to alignment gaps between the upper and lower frequency-selective units, compromising the integrity of the frequency-selective array. If these gaps are not addressed, the electromagnetic stealth performance of the radome will be significantly reduced; if metallization or other methods are used to fill the gaps, the local electromagnetic characteristics will be altered, worsening the radiation performance of the radar antenna.

[0025] Therefore, this application provides a multi-layer curved surface frequency selective process alignment design method for airborne radar aperture radomes, such as... Figure 1 As shown, it includes: Based on the digital model of the frequency-selective structure product of the radome, a multi-curvature surface model of the airborne radar aperture radome containing upper and lower frequency-selective surfaces is constructed. The X-axis is defined as the lateral curvature direction of the radome and the Y-axis is defined as the longitudinal curvature direction of the radome. Based on prior data of frequency-selectable machinable dimensions, including machinable parameters of positioning holes and matching parameters of frequency-selective units, an array of positioning holes is generated on a multi-layer frequency-selective surface. A partitioning unfolding algorithm with area loss compensation is used to unfold the hypercurvature multi-layer surface model into a two-dimensional plane, and the area loss rate of the upper and lower frequency-selected surfaces after unfolding is verified to be less than or equal to a set value. In the hypercurvature multi-layer surface, select the upper and lower corresponding frequency-selected surfaces with the largest curvature, and extract the first spacing difference in the X-axis direction and the second spacing difference in the Y-axis direction of the corresponding positioning holes at the same position respectively. A dual-axis period adaptive adjustment strategy is adopted. Within the safety threshold range determined by electromagnetic simulation, the period of the frequency-selective unit in the X-axis and Y-axis directions is coupled and adjusted based on the first spacing difference and the second spacing difference to generate the target period parameter. The target period parameters are input into a pre-trained deep learning model for geometric alignment and electromagnetic performance simulation verification and automatic correction, outputting the optimal target period parameters, and generating the optimal frequency-selective array pattern based on the optimal target period parameters. The optimal frequency-selective array pattern is mapped onto the two-dimensional plane to obtain a planar frequency-selective film processing pattern.

[0026] This embodiment provides a method for aligning and designing a multi-layer curved surface frequency-selective process for an airborne radar aperture radome. This method addresses the technical problems in existing technologies where alignment gaps arise due to differences in the curvature of the multi-layer curved surface frequency-selective structure during installation, making it difficult to balance stealth and radiation performance. This method utilizes biaxial periodic adaptive adjustment, deep learning model optimization, partitioned planar unfolding compensation, and array-based high-precision positioning techniques to achieve precise alignment design of the multi-layer curved surface frequency-selective structure, improve process alignment accuracy and adaptability, and ensure the integrity of the frequency-selective array.

[0027] To achieve the above objectives, the technical solution adopted in this embodiment is as follows: 1. Construct a hyperbolic surface model of the airborne radar aperture radome, and determine the upper and lower frequency selectivity surfaces on the hyperbolic surface model, such as... Figure 2 As shown.

[0028] Specifically, based on the design drawings and processing accuracy requirements of the airborne radar aperture radome, a full-size hyperbolic surface solid model of the radome was constructed using 3D digital modeling software. The deviation between the geometric parameters of the model and the actual processing dimensions of the radome was controlled within 0.01mm. According to the design number of frequency selection structures and their placement positions, parallel upper and lower frequency selection surfaces were marked on the hyperbolic surface model. The spacing between the upper and lower frequency selection surfaces matched the layer thickness of the radome skin and was consistent with the normal direction of the radome surface.

[0029] 2. Based on prior data corresponding to frequency-selectable machinable dimensions, an array of positioning holes is generated on a multi-layer frequency-selective curved surface, such as... Figure 5 As shown.

[0030] Specifically, the prior data consists of a standardized process database accumulated through extensive process experiments and simulation verification. This database includes the machinability parameters of the positioning holes themselves, the matching parameters between the frequency-selective units and the positioning holes, and the alignment tolerance data for multi-layer curved surfaces. The principle for generating array-type positioning holes is as follows: The opening positions of the positioning holes on the upper and lower layers of frequency-selective curved surfaces are calibrated, and the alignment of the center of the positioning holes with the reference points of the frequency-selective unit array is controlled. The reference point of the positioning hole must completely coincide with the reference point of the frequency-selective array to prevent the transmission of reference deviation from the source. The spacing between the positioning holes must be an integer multiple of the frequency-selective unit period to prevent the machining error of the positioning holes from being transmitted to the alignment accuracy of the frequency-selective units. Single-point / multi-point positioning cannot guarantee the overall alignment accuracy of large-sized curved surfaces; array-type distribution can disperse the error to various local areas. Furthermore, the limits of machining capabilities are used as constraints during the design phase to prevent the design of positioning holes that are impossible to machine or whose machining accuracy does not meet standards.

[0031] The traditional process involves "making the frequency-selective film first and then drilling the positioning holes," which can easily lead to discrepancies between the positioning holes and the frequency-selective array. In this embodiment, the positioning holes are first generated on the curved surface model, and then the frequency-selective unit period is adjusted based on the spacing difference of the positioning holes. This achieves a unified reference throughout the entire process of "positioning reference - period adjustment - pattern mapping - installation alignment," which is also the key prerequisite for this invention to improve the alignment accuracy to ≤0.03mm.

[0032] 3. Design a partitioned plane unfolding algorithm with area loss compensation to unfold the hypercurvature surface model into a two-dimensional plane, and verify that the area loss rate of the upper and lower frequency-selected surfaces after unfolding is less than a set value.

[0033] Specifically, a partitioned planar unfolding algorithm with area loss compensation is designed. Based on the differential geometric properties of the surface, the algorithm adopts the principle of equal area unfolding to divide the hypercurvature surface model into multiple regular sub-surface regions. The boundary of each sub-surface region is the curvature feature line of the surface, avoiding excessive stretching or compression after unfolding. The planar unfolding of each sub-surface region is performed separately, and the unfolded area loss rate of each region is calculated in real time. The algorithm compensates and corrects the loss amount to ensure that the area loss rate of a single sub-surface region and the entire surface after unfolding is less than a set value (the set value is determined according to the processing accuracy requirements of the radome, preferably ≤1.5‰).

[0034] 4. Select a set of upper and lower frequency-selected surfaces after unfolding from the hypercurvature multi-layered surface, such as... Figure 3 As shown, the first spacing difference in the X-axis direction and the second spacing difference in the Y-axis direction of the positioning holes of the upper and lower frequency-selective surface unfolded planes are extracted respectively. Figure 4 As shown.

[0035] This step is the core calibration step of the innovative "positioning-unfolding-cycle adjustment" process in this embodiment. It completely solves the problem of inaccurate cycle adjustment caused by ignoring the inherent deformation error of the double-curvature surface unfolding and the inconsistent benchmark when directly extracting the spacing difference of the theoretical surface curvature lines in the traditional method. The principle is to first complete the unfolding of the partitioned surface with area loss compensation, and take the positioning hole that runs through the entire process of design, processing and laying as the only physical benchmark. Pair the upper and lower frequency-selected surfaces at the same position, and extract the X and Y axis spacing difference of the corresponding positioning hole on the unfolded plane. This difference has included all process deformation errors and is completely consistent with the actual offset during actual laying. At the same time, it matches the anisotropic characteristics of the double-curvature surface, providing accurate input for the subsequent biaxial cycle adaptive adjustment, eliminating benchmark conversion and unfolding errors from the source, and greatly improving the accuracy of cycle adjustment and the final alignment effect.

[0036] Calculate the distance difference between the upper and lower frequency selective surfaces in the X-axis direction, and record it as the first distance difference. Calculate the distance difference between the upper and lower frequency selective surfaces in the Y-axis direction, and record it as the second distance difference. The accuracy of length extraction and calculation is controlled within 0.0001mm.

[0037] 5. Design a dual-axis period adaptive adjustment strategy. Within a preset safety threshold range, based on the first spacing difference and the second spacing difference, couple and adjust the period of the frequency-selective unit in the X-axis direction and the Y-axis direction respectively to generate the target period parameter.

[0038] Specifically, a safe threshold range for adjusting the frequency-selective unit period in the X and Y axes is preset. This range is the adjustable interval of the frequency-selective unit period, and the electromagnetic performance of the frequency-selective structure after adjustment meets the design requirements of the airborne radar. Based on the first spacing difference and the second spacing difference, the theoretical adjustment amount of the frequency-selective unit period in the X and Y axes is calculated respectively. If the theoretical adjustment amount is within the safe threshold range, the frequency-selective unit period in the X and Y axes is adjusted proportionally. If the theoretical adjustment amount exceeds the safe threshold range, the adjustment amount is corrected to be within the safe threshold range. Finally, the target period parameters of the frequency-selective unit in the X and Y axes are generated. The target period parameters include the unit period size and the number of array units.

[0039] 6. Input the target period parameters into a pre-trained deep learning model for geometric alignment and electromagnetic performance simulation verification and automatic correction, output the optimal target period parameters, and generate the optimal frequency-selective array pattern based on the optimal target period parameters.

[0040] Specifically, a deep learning model based on a convolutional neural network and a long short-term memory network is constructed. The input layer of the model consists of multi-dimensional data such as the radius of curvature of the surface, the difference in the spacing between positioning holes, the period of the frequency selection unit, the period adjustment amount, the alignment accuracy of the adjusted frequency selection array, and electromagnetic performance indicators. The hidden layer is a multi-dimensional feature extraction and coupled calculation module, and the output layer is the optimized period parameters and the predicted value of the frequency selection performance. Simulation data (including electromagnetic simulation data and geometric alignment simulation data) of frequency selection unit period adjustment under different surface curvatures and different spacing differences are collected as training sets to supervise the training of the deep learning model. During the training process, the model weights are continuously adjusted. The bias is adjusted until the parameter correction accuracy and performance prediction accuracy of the model meet the preset requirements (prediction deviation ≤ 1%). The generated target periodic parameters are then input into the trained deep learning model. The model is aligned and subjected to geometric alignment simulation and electromagnetic performance simulation verification. If the verification results do not meet the design requirements, the target periodic parameters are automatically fused and compensated to correct them, and finally the optimal target periodic parameters are output. Based on the optimal target periodic parameters, combined with the basic pattern of the frequency selective unit (such as square ring, cross, or circular ring), the optimal frequency selective array pattern that matches the upper and lower frequency selective surfaces is generated. The resolution of the array pattern matches the fabrication accuracy of the planar frequency selective film.

[0041] 7. Map the optimal frequency-selective array pattern onto a two-dimensional plane, prepare a planar frequency-selective film, and align and lay it onto the upper and lower frequency-selective surfaces.

[0042] Specifically, the optimal frequency-selective array pattern generated above is precisely mapped onto the unfolded two-dimensional plane according to the topological relationship of the surface unfolding, forming a two-dimensional pattern that matches the fabrication process of the planar frequency-selective film. Using processes such as laser etching and photoresist development, the planar frequency-selective film is fabricated based on the mapped two-dimensional pattern. The substrate of the frequency-selective film matches the composite material of the radome skin, and the film thickness is controlled within 0.05-0.2 mm. On the upper and lower frequency-selective surfaces of the hyperbolic surface model, as well as the fabricated planar frequency-selective film, using the generated array-type positioning holes as a reference, the positioning holes are through holes with an opening accuracy controlled within 0.03 mm. The positioning holes of the planar frequency-selective film are matched with the positioning holes of the radome curved surface frequency-selective surface using positioning pins to achieve precise alignment and placement of the planar frequency-selective film on the upper and lower frequency-selective surfaces.

[0043] This embodiment uses a complex hyperbolic airborne radar aperture radome as an example. The radome has an ellipsoidal structure, and the frequency selection structure adopts a double-layer square patch array with a unit period of 5.0 mm. The radome skin is made of quartz cyanate composite material. The specific implementation steps of this embodiment are as follows: 1. Construct a hypercurvature multi-layer surface model A dual-curvature surface solid model was constructed using 3D modeling software based on the design drawings of the radome. The geometric dimensional deviation of the model was controlled within 0.01 mm. According to the design requirements of the dual-layer frequency-selective structure, the upper and lower frequency-selective surfaces were marked on the model, with a spacing of 2 mm between the upper and lower layers, both consistent with the normal of the ellipsoid. The X-axis was defined as the lateral curvature direction of the radome, and the Y-axis as the longitudinal curvature direction.

[0044] 2. Generate an array of positioning holes Based on the standardized process data (including the machinable parameters of the positioning holes and the matching parameters of the frequency-selective units) accumulated through a large number of process experiments and simulation verifications in the early stage, the opening positions of the positioning holes of the upper and lower layers of frequency-selective curved surfaces are calibrated, the center of the positioning hole is aligned with the reference point of the frequency-selective unit array, and the distance between the center of the positioning hole and the center of the adjacent frequency-selective unit is controlled to be an integer multiple of the unit period.

[0045] 3. Partitioned Plane Expansion The design incorporates a partitioned planar unfolding algorithm with area loss compensation. The ellipsoidal hypercurvature surface model is divided into four rectangular sub-surface regions with uniform curvature changes. These regions are unfolded using an equal-area unfolding method, and the algorithm compensates for area loss to ensure that the area loss rate of each region after unfolding is ≤1.5‰. Gradual transition units matching the topology of the square patch are set at the seams of adjacent sub-surface regions.

[0046] 4. Extract the difference in biaxial spacing Taking the pair of corresponding frequency-selected surfaces with the largest curvature along the major axis of the ellipsoid, and using UG software, the first spacing difference of the corresponding positioning holes at the same location in the X-axis direction is 4.3700 mm, and the second spacing difference in the Y-axis direction is 0.0300 mm, with a length calculation accuracy of 0.0001 mm. A three-dimensional coordinate system is established with the geometric center of the surface model as the origin. Using software such as UG, the X-axis arc length of the upper frequency-selected surface is extracted to be 170.0000 mm, and that of the lower surface is 165.6300 mm, with a first arc length difference of 4.3700 mm; the longest arc length on the Y-axis is 365.0000 mm for the upper surface and 364.9700 mm for the lower surface, with a second arc length difference of 0.0300 mm, and a length calculation accuracy of 0.0001 mm. The first spacing difference and the first arc length difference are equal, and the second spacing difference and the second arc length difference are equal, indicating that the extracted biaxial spacing difference data are accurate.

[0047] 5. Dual-axis periodic adaptive adjustment to generate target period parameters The safe threshold range for frequency-selective element period adjustment was pre-determined to be 4.85mm to 5.00mm using electromagnetic simulation software such as CST and HFSS. Since the radome is a complex hypercurvature surface, the theoretical adjustment amount in the X-axis direction, calculated using the ratio of the spacing difference to the number of elements in that direction, is approximately 4.37 / 34 ≈ 0.1285mm. The second spacing difference in the Y-axis direction, 0.0300mm, is less than 0.3 element periods. Compensation is achieved through flexible tiling, ultimately generating target period parameters: the upper-layer frequency-selective surface has a target period of 5.0mm on both the X and Y axes, while the lower-layer frequency-selective surface has a target period of 4.8715mm on the X-axis and 5mm on the Y-axis. Figure 5 In the diagram, 0 represents the alignment reference for the upper and lower frequency selection surfaces. The lower frequency selection surface is on the left, and the upper frequency selection surface is on the right. The target period of the lower frequency selection surface on the X-axis is 4.8715 mm, and the target period on the Y-axis is 5 mm. The target periods of the upper frequency selection surface on both the X-axis and Y-axis are 5.0 mm.

[0048] 6. Deep learning model optimization The target period parameters are input into a pre-trained deep learning model (based on convolutional neural network and long short-term memory network, with parameter correction accuracy and electromagnetic performance prediction deviation both ≤1%). After simulation verification, the model is fused and compensated to output the optimal target period parameters. Based on these parameters, the optimal frequency-selective array pattern is generated, with 34 X-axis units and 73 Y-axis units in the upper layer and 34 X-axis units and 73 Y-axis units in the lower layer.

[0049] 7. Pattern mapping and frequency-selective film preparation The optimal frequency-selective array pattern is mapped onto the unfolded two-dimensional plane to obtain the planar frequency-selective film processing pattern; the planar frequency-selective film is prepared on a polyimide substrate (film thickness 0.025 mm) using a laser etching process.

[0050] 8. Precise installation and curing By using positioning pins and positioning holes, the planar frequency selective film is precisely laid onto the upper and lower frequency selective curved surfaces and then placed in a thermostatic precipitator for integrated curing and molding.

[0051] 9. Defect Identification and Repair Images of the radome surface were captured using a 20-megapixel industrial camera. Defect features were extracted using a machine learning model, identifying one general wrinkle defect and two micro-bubbles. The wrinkle defect was repaired using hot-pressing leveling, and the bubbles were repaired using needle puncture and degassing. After repair, the alignment accuracy of the frequency-selective structure was ≤0.03mm, the radar cross-section (RCS) attenuation was ≥20dB, and the electromagnetic wave transmittance was ≥90%, meeting the design requirements of this type of airborne radar radome.

[0052] Furthermore, the generation of the array of positioning holes includes: The center of the positioning hole is aligned with the reference point of the frequency selective unit array. The reference point of the frequency selective unit array is the geometric center or corner feature point of the array. The coincidence deviation between the center of the positioning hole and the array reference point is ≤0.01mm. The distance between the center of the positioning hole and the center of the adjacent frequency selective unit is controlled to be an integer multiple of the unit period. This prevents the machining deviation of the positioning hole from being transmitted to the alignment accuracy of the frequency selective unit. In other words, the opening position of the positioning hole matches the period of the frequency selective unit array, so that the machining error of the positioning hole will not affect the alignment of the frequency selective unit and ensure the laying accuracy of the frequency selective array.

[0053] Furthermore, the partitioned planar unfolding algorithm with area loss compensation includes: The hypercurvature multi-layer surface model is divided into multiple sub-surface regions. The division principle is that the curvature of each sub-surface region is uniform and there is no obvious curvature abrupt change. The shape of the sub-surface region is a regular rectangle or fan shape, which is convenient for planar unfolding and pattern mapping. Each sub-surface region is unfolded into a plane. The equal area unfolding method is used to unfold each sub-surface region. The surface area before unfolding and the plane area after unfolding of each sub-surface region are calculated in real time to obtain the area loss rate of each region. The plane size after unfolding is compensated and corrected by an algorithm to ensure that the area loss rate of each region is controlled within ≤1.5‰. At the seam between adjacent sub-surface regions, a size-gradient transition unit matching the frequency-selective unit topology is set. The size of the transition unit smoothly transitions from the frequency-selective unit period of one sub-surface region to the frequency-selective unit period of the other sub-surface region. The topology of the transition unit is consistent with that of the frequency-selective unit to avoid frequency-selective performance breakpoints caused by abrupt changes in unit size at the seam, thus ensuring the continuity of the frequency-selective stealth performance and radiation performance of the hyperbolic surface.

[0054] Furthermore, the dual-axis periodic adaptive adjustment strategy includes: Airborne radar radomes are classified into simple hyperbolic radomes and complex hyperbolic radomes. Simple hyperbolic radomes are ruled surfaces with no geometric curvature difference in the Y-axis direction, such as cylindrical or conical surface radomes. Their Y-axis is the generatrix of the surface, which is straight, and the theoretical lengths of the arc lines in the Y-axis direction of the upper and lower frequency-selective surfaces are equal. Complex hyperbolic radomes are non-ruled surfaces with geometric curvature difference in the Y-axis direction, such as spherical caps, ellipsoids, and freeform surface radomes. Their X and Y axes are both arc directions, and there is a natural spacing difference between the upper and lower frequency-selective surfaces in both axes. For a simple hyperbolic radome, only the frequency-selective element period adjustment in the X-axis direction is performed, while the Y-axis direction remains rigidly aligned. That is, the frequency-selective element period in the Y-axis direction keeps the basic size unchanged, and the Y-axis direction is achieved by the cooperation of the positioning hole and the positioning pin. For complex hyperbolic radomes, the frequency-selective element period coupling adjustment is performed in the X-axis and Y-axis directions. That is, the frequency-selective element period in both directions is adjusted proportionally according to the difference in the positioning hole spacing in the X-axis and Y-axis directions to ensure that there are no alignment gaps in both directions.

[0055] Furthermore, performing frequency-selective cell period coupling adjustment in the X-axis and Y-axis directions includes: When the second spacing difference in the Y-axis direction is greater than or equal to 0.3 unit periods, the Y-axis unit period is adjusted proportionally according to the X-axis adjustment logic. That is, according to the correspondence between the "spacing difference and the number of units in the X-axis direction", the period adjustment ratio in the Y-axis direction is calculated, and the Y-axis frequency-selective unit period is adjusted proportionally so that the total length of the frequency-selective array in the Y-axis direction after adjustment matches the length of the arc line of the frequency-selective surface. When the second spacing difference in the Y-axis direction is less than 0.3 unit periods, compensation is made through flexible laying to ensure the physical alignment of the upper and lower frequency-selective curved surfaces in the Y-axis direction. That is, without adjusting the Y-axis unit period, the physical alignment after laying is achieved by slightly stretching or compressing the flexible characteristics of the planar frequency-selective film substrate. This flexible compensation amount is within the elastic deformation range of the substrate and will not have a significant impact on the electromagnetic performance of the frequency-selective structure.

[0056] Furthermore, the preset safety threshold range is determined by simulating the frequency-selective stealth performance and radiation performance of hypercurvature surfaces under different curvatures; Electromagnetic simulation software (such as HFSS and CST) was used to establish simulation models of frequency-selective structures with different curvature surfaces. The radar cross section (RCS), electromagnetic wave transmittance, out-of-band suppression ratio, and other stealth performance indicators of the frequency-selective structure were simulated under different frequency-selective unit period adjustment values. The gain, radiation pattern, and VSWR of the radar antenna were also simulated. The maximum adjustable value of the frequency-selective unit period in the X and Y axes was determined based on the criterion that "the stealth performance indicators meet the stealth design requirements of airborne radar and the radiation performance indicators do not deteriorate significantly". The preset safety threshold range for adjusting the frequency-selective unit period was set with the maximum adjustable value as the boundary.

[0057] Furthermore, for hypercurvature surfaces, when the X-axis and Y-axis directions are adjusted simultaneously, the safety threshold range is adaptively and dynamically adjusted according to the coupling relationship of the adjustment amount.

[0058] When periodic adjustments are made only in the X-axis or Y-axis direction, a preset fixed safety threshold range is used. When periodic adjustments are made simultaneously in both the X-axis and Y-axis directions, the impact on the frequency-selective electromagnetic performance is nonlinearly superimposed due to the coupling relationship between the two-axis adjustments. Therefore, based on the coupling relationship between the two-axis adjustments, the frequency-selective performance index under the joint adjustment of the two axes is recalculated through electromagnetic simulation, and the original safety threshold range is adaptively and dynamically corrected. The corrected safety threshold range can achieve the optimal allocation of the two-axis adjustments, eliminating alignment gaps while maximizing the electromagnetic performance of the frequency-selective structure.

[0059] Furthermore, building and training a deep learning model includes: Simulation data on the periodic adjustment of frequency-selective elements (FSEs) on hypercurvature surfaces with different curvatures were collected as a training set. This training set included multi-dimensional data such as surface curvature radius, positioning hole spacing difference, FSE period, period adjustment amount, alignment accuracy of the adjusted FSE array, and electromagnetic performance indicators. The sample size of the training set covered the common curvature range of airborne radar radomes. The training set was divided into training, validation, and test sets in an 8:1:1 ratio for supervised training of the deep learning model. During the training process, the model parameters were continuously optimized with the combined optimal alignment accuracy and electromagnetic performance indicators as the objective function. The optimization continued until the parameter correction accuracy of the deep learning model met the preset requirements, i.e., the deviation between the optimized periodic parameters output by the model and the actual optimal parameters was ≤1%, and the deviation between the predicted electromagnetic performance values ​​and the actual simulated values ​​was ≤1%.

[0060] Furthermore, the simulation verification and automatic correction of the target periodic parameters using the trained deep learning model include: When adjusting the X-axis and Y-axis periods simultaneously, the target period parameters are input into the trained deep learning model. The model first analyzes the coupling relationship between the two-axis adjustment amounts, verifies the alignment accuracy through the built-in geometric simulation module, and verifies the stealth and radiation performance through the electromagnetic simulation module. If the verification results do not meet the design requirements, the deep learning model performs a fusion compensation calculation on the period adjustment amounts of the two axes. That is, based on the influence weight of the two-axis adjustment amounts on the electromagnetic performance, the adjustment amounts of the X-axis and Y-axis are mutually compensated and corrected to achieve optimal electromagnetic performance while ensuring alignment accuracy, and finally eliminate the alignment gaps between the multi-layer frequency-selective structures.

[0061] Furthermore, the defect identification and repair steps are as follows: The surface image of the radome after curing is acquired by machine vision. An industrial camera and a telecentric lens are used to form a machine vision acquisition system. The acquisition range covers the entire frequency-selective curved surface of the radome, ensuring that the image can clearly show the alignment status of the frequency-selective unit and the bubbles and wrinkles on the film surface. The machine learning model is used to extract features from the surface image. The image preprocessing (denoising, enhancement, registration) is used to extract the defect features in the image, such as the offset of the alignment deviation, the size and position of the bubble, and the outline and area of ​​the wrinkle. The trained machine learning model is used to identify the defect features and to determine the defect type and defect level according to the severity of the defect. The defect level is divided into minor defects, general defects and severe defects. Based on the defect type and defect level, corresponding repair operations are performed: for minor alignment deviations, micro-bubbles, and shallow wrinkles, hot pressing and leveling are used for repair; for general defects, laser redrawing is used to correct the misaligned frequency-selective units, or needle punching is used to treat bubbles; for severe defects, local film repair is used to remove the frequency-selective film in the defective area, re-prepare and lay a matching local frequency-selective film, and the alignment accuracy and electromagnetic performance of the repaired frequency-selective structure must meet the design requirements.

[0062] This example also provides a comparative example, specifically: Using the existing technology of planar frequency-selective film fitting and laying process of the same period unit size, a multi-layer frequency-selective structure is laid on the above-mentioned complex double curvature airborne radar aperture radome. The upper and lower frequency-selective films both adopt a square patch array with a period of 5.0 mm.

[0063] After installation, inspection revealed a significant alignment gap in the X-axis direction due to a 4.37mm spacing difference, and a localized gap in the Y-axis direction due to a 0.03mm spacing difference, which compromised the integrity of the frequency-selective array. Electromagnetic performance testing showed that the radar cross section (RCS) attenuation was only 8dB, and the electromagnetic wave transmittance was 72%. Neither stealth nor radiation performance met the design requirements, and the gaps could not be restored to their original performance through simple repairs.

[0064] By comparing the embodiments and comparative examples, it can be seen that the multi-layer curved surface frequency-selective process alignment design method of the airborne radar aperture radome in this embodiment can effectively eliminate the alignment gaps of the multi-layer curved surface frequency-selective structure, ensure the integrity of the frequency-selective array, take into account the electromagnetic stealth performance and radiation performance of the radome, and greatly improve the laying alignment accuracy and overall processing quality. However, the existing technology cannot solve the alignment problem caused by the curvature difference of the curved surface, and the performance indicators are far below the design requirements.

[0065] This embodiment constructs a multi-curvature curved surface model and accurately extracts the difference in the spacing between the positioning holes in both the X and Y axes. Combined with a dual-axis period adaptive adjustment strategy, it achieves precise coupling adjustment of the frequency selection unit period, effectively eliminating the alignment gaps caused by the difference in curvature of the multi-curvature curved surface frequency selection structure, ensuring the integrity of the frequency selection array, and solving the problem of reduced stealth performance caused by alignment gaps in the prior art.

[0066] This embodiment introduces a pre-trained deep learning model, which is trained with a large amount of simulation data to realize the simulation verification and automatic correction of the target periodic parameters. In particular, when adjusting the dual axes simultaneously, it can perform fusion compensation calculations, eliminating alignment gaps while maximizing the electromagnetic stealth performance and radar radiation performance of the frequency-selective structure, thus achieving a balance between the two performances.

[0067] The partitioned planar unfolding algorithm with area loss compensation designed in this embodiment divides the hypercurvature surface into sub-regions for unfolding and performs area loss compensation, ensuring that the area loss rate after unfolding is controlled at ≤1.5‰. At the same time, size gradient transition units are set at the seams to avoid frequency selection performance breakpoints and improve the accuracy of surface unfolding and pattern mapping.

[0068] This embodiment achieves precise alignment and installation of planar frequency-selective film and curved frequency-selective surface by designing an array of positioning holes. The positioning holes match the period of the frequency-selective unit array, avoiding the transmission of positioning deviations, greatly improving the installation and alignment accuracy of the multi-layer frequency-selective structure, and reducing the risk of installation deviations.

[0069] This embodiment adds a defect intelligent identification and repair step after the tarpaulin is laid and cured. Through machine vision and machine learning, defects are quickly identified and classified. Combined with targeted repair operations, the overall processing quality of the tarpaulin is effectively improved, and the yield rate is guaranteed.

[0070] This embodiment is applicable to various types of airborne radar aperture radomes, including simple and complex hypercurvature radomes. It can adaptively adjust the periodic adjustment strategy according to the surface characteristics of the radome, exhibiting good adaptability and versatility. Furthermore, the technical means for each process step are clear, highly feasible, and easy to promote and apply in industrialization, significantly improving the automation and precision level of the multi-layer curved surface frequency selection process for airborne radar aperture radomes.

[0071] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computing software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0072] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0073] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for aligning and designing a multi-layer curved surface frequency-selective process for an airborne radar aperture radome, characterized in that, include: Based on the digital model of the frequency-selective structure product of the radome, a multi-curvature surface model of the airborne radar aperture radome containing upper and lower frequency-selective surfaces is constructed. The X-axis is defined as the lateral curvature direction of the radome and the Y-axis is defined as the longitudinal curvature direction of the radome. Based on prior data of frequency-selectable machinable dimensions, including machinable parameters of positioning holes and matching parameters of frequency-selective units, an array of positioning holes is generated on a multi-layer frequency-selective surface. A partitioning unfolding algorithm with area loss compensation is used to unfold the hypercurvature multi-layer surface model into a two-dimensional plane, and the area loss rate of the upper and lower frequency-selected surfaces after unfolding is verified to be less than or equal to a set value. In the hypercurvature multi-layer surface, select the upper and lower corresponding frequency-selected surfaces with the largest curvature, and extract the first spacing difference in the X-axis direction and the second spacing difference in the Y-axis direction of the corresponding positioning holes at the same position respectively. A dual-axis period adaptive adjustment strategy is adopted. Within the safety threshold range determined by electromagnetic simulation, the period of the frequency-selective unit in the X-axis and Y-axis directions is coupled and adjusted based on the first spacing difference and the second spacing difference to generate the target period parameter. The target period parameters are input into a pre-trained deep learning model for geometric alignment and electromagnetic performance simulation verification and automatic correction, outputting the optimal target period parameters, and generating the optimal frequency-selective array pattern based on the optimal target period parameters. The optimal frequency-selective array pattern is mapped onto the two-dimensional plane to obtain a planar frequency-selective film processing pattern.

2. The method for aligning multiple curved surfaces of an airborne radar aperture radome according to claim 1, characterized in that, Generating an array of positioning holes on a multi-layer frequency-selective curved surface includes: The opening positions of the positioning holes on the upper and lower frequency-selective curved surfaces are calibrated, the center of the positioning hole is aligned with the reference point of the frequency-selective unit array, and the distance between the center of the positioning hole and the center of the adjacent frequency-selective unit is controlled to be an integer multiple of the unit period.

3. The method for aligning multiple curved surfaces of an airborne radar aperture radome according to claim 1, characterized in that, The partitioning algorithm with area loss compensation includes: The hypercurvature multi-layer surface model is divided into multiple sub-surface regions with uniform curvature changes; Each of the sub-surface regions is unfolded in plane, and the area loss rate during the unfolding process is calculated and compensated. At the seam between adjacent sub-surface regions, a size-gradient transition unit matching the frequency-selective unit topology is set to ensure the continuity of the frequency-selective stealth performance and radiation performance of the hyperbolic surface.

4. The method for aligning multiple curved surfaces of an airborne radar aperture radome according to claim 1, characterized in that, The dual-axis periodic adaptive adjustment strategy includes: The airborne radar radome is divided into a simple hyperbolic radome and a complex hyperbolic radome. The simple hyperbolic radome is a ruled surface with no geometric curvature difference in the Y-axis direction, while the complex hyperbolic radome is a non-ruled surface with geometric curvature difference in the Y-axis direction. For the simple hyperbolic radome, only the frequency-selective cell period adjustment is performed in the X-axis direction, while rigid alignment is maintained in the Y-axis direction; For the complex hyperbolic radome, frequency-selective element periodic coupling adjustment is performed in the X-axis and Y-axis directions.

5. The method for aligning the multi-layer curved surface frequency selective process of an airborne radar aperture radome according to claim 4, characterized in that, Performing frequency-selective cell period coupling adjustment in the X-axis and Y-axis directions includes: The period adjustment amount in the X-axis direction is calculated as the ratio of the first spacing difference in the X-axis direction to the number of frequency-selective units in that direction, and is recorded as the first period adjustment amount. Within the safety threshold range, the period of the X-axis frequency-selective units is proportionally corrected based on the first period adjustment amount. When the second spacing difference in the Y-axis direction is greater than or equal to 0.3 frequency selection unit cycles, the second cycle adjustment amount is calculated according to the ratio of the second spacing difference in the Y-axis direction to the number of frequency selection units in that direction. Within the safety threshold range, the cycle of the Y-axis frequency selection unit is proportionally corrected based on the second cycle adjustment amount. When the second spacing difference in the Y-axis direction is less than 0.3 frequency-selective unit cycles, the Y-axis unit cycle remains unchanged. Compensation is achieved through flexible tiling to ensure the physical alignment of the upper and lower frequency-selective surfaces in the Y-axis direction.

6. The method for aligning the multi-layer curved surface frequency selective process of an airborne radar aperture radome according to claim 1, characterized in that, The safety threshold range is determined by simulating the frequency-selective stealth performance and radiation performance of hypercurvature surfaces under different curvatures.

7. The method for aligning the multi-layer curved surface frequency selective process of an airborne radar aperture radome according to claim 6, characterized in that, For the hypercurvature surface, when the X-axis and Y-axis directions are adjusted simultaneously, the safety threshold range is adaptively and dynamically adjusted according to the coupling relationship of the dual-axis adjustment amount.

8. The method for aligning the multi-layer curved surface frequency selective process of an airborne radar aperture radome according to claim 1, characterized in that, The pre-trained deep learning model is obtained through the following steps: Simulation data of frequency-selective unit period adjustment of hypercurvature surfaces under different curvatures are collected as training sets to supervise the training of a deep learning model based on convolutional neural network and long short-term memory network until the parameter correction accuracy and electromagnetic performance prediction deviation of the deep learning model are both less than or equal to preset values.

9. The method for aligning the multi-layer curved surface frequency selective process of an airborne radar aperture radome according to claim 8, characterized in that, The simulation verification and automatic correction of the target periodic parameters using a pre-trained deep learning model include: When adjusting the X-axis and Y-axis periods simultaneously, the deep learning model is used to perform fusion compensation calculations on the period adjustment amounts of the two axes to eliminate alignment gaps between the multi-layer frequency-selective structures.

10. The method for aligning the multi-layer curved surface frequency-selective process of an airborne radar aperture radome according to claim 1, characterized in that, Based on the aforementioned planar frequency-selective film processing pattern, a planar frequency-selective film is prepared. Using positioning holes and positioning pins, the planar frequency-selective film is aligned and laid onto the upper and lower frequency-selective curved surfaces. After integral curing and molding, defect identification and repair steps are performed. The surface image of the radome after it has been laid and cured is acquired using machine vision. The surface image is used to extract features using a machine learning model to identify and determine the type and level of defects in the surface image. Based on the defect type and the defect level, perform the corresponding repair operation.

Images