Three-dimensional digital model generation method and system based on glass fiber reinforced plastic streamline headstock

Abstract

The invention relates to the technical field of model generation, in particular to a three-dimensional digital model generation method and system based on a glass fiber reinforced plastic streamline headstock. The method comprises the following steps: testing the whole hot-press molding process of a standard glass fiber reinforced plastic sample to obtain the thermal expansion coefficient of the glass fiber reinforced plastic material; constructing a glass fiber reinforced plastic shrinkage tensor field based on the glass fiber reinforced plastic material thermal expansion coefficient; performing paving layer path topological relation construction on the headstock fiber paving layer to obtain a headstock paving layer path diagram; performing path-process coupling deformation prediction on the headstock fiber bundle based on the headstock layering path diagram to obtain layering path deformation prediction data; and obtaining an initial vehicle head three-dimensional model. According to the method, the risks of superspeed train rendezvous pressure wave exceeding and sensor false triggering are remarkably reduced, the design precision is integrally improved, the iteration period is shortened, and the wind tunnel test cost is saved.

Description

Technical Field

[0001] The present invention relates to the technical field of model generation, and in particular to a method and system for generating a three-dimensional digital model based on a glass fiber reinforced plastic streamlined vehicle head. Background Art

[0002] In high-speed rail vehicle manufacturing, fiberglass reinforced plastic (FRP) is widely used in the production of streamlined front shells due to its lightweight and high plasticity. In the current design process, engineers often manually adjust the 3D front shell model based on wind tunnel test data. This is particularly true for the design of the micro-curved fairings for maglev trains operating at speeds above 350 km / h, where traditional parametric modeling approaches face significant challenges. A long-overlooked difficulty lies in the implicit conflict between the non-uniform material shrinkage characteristics of FRP during the hot press forming process and the ideal geometric topology of the digital model. When designers construct aerodynamically consistent surfaces with continuous curvature in CAD software, they often assume the material isotropic. However, the variability in the layup orientation of fiberglass cloth in complex hyperbolic regions (such as the transition zone between the nose cone and the wing) can lead to unpredictable deformations of 0.5-2 mm in localized areas after curing. This results in systematic deviations between the high-precision aerodynamic simulation results of the digital model and the actual wind tunnel test data. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide a method and system for generating a three-dimensional digital model based on a fiberglass streamlined front end to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, a method for generating a three-dimensional digital model based on a fiberglass streamlined front end includes the following steps:

[0005] Step S1: Conducting a full hot pressing forming process test on a standard FRP sample to obtain the thermal expansion coefficient of the FRP material; constructing a FRP contraction tensor field based on the thermal expansion coefficient of the FRP material;

[0006] Step S2: constructing a topological relationship of the fiber layup path for the headstock to obtain a headstock layup path diagram; performing path-process coupled deformation prediction on the headstock fiber bundle based on the headstock layup path diagram to obtain layup path deformation prediction data;

[0007] Step S3: Obtaining an initial three-dimensional model of the vehicle head; spatially distributing the fiberglass laminate material properties based on the initial three-dimensional model and the predicted deformation data of the ply path to obtain the material property distribution data of the steel layer; performing aerodynamic-structural bidirectional coupling iterative correction on the initial three-dimensional model of the vehicle head based on the material property distribution data of the steel layer to obtain a coupled corrected vehicle head model;

[0008] Step S4: performing inverse compensation and discrete point cloud generation on the coupled modified vehicle head model according to the FRP shrinkage tensor field to obtain vehicle head compensation point cloud data; performing radial basis function-curvature constraint fusion reconstruction on the compensation surface of the vehicle head according to the vehicle head compensation point cloud data to obtain vehicle head geometric compensation contour data;

[0009] Step S5: performing virtual manufacturing simulation on the vehicle head according to the vehicle head geometric compensation contour data to obtain the vehicle head virtual manufacturing result data; and constructing a streamlined vehicle head three-dimensional digital model based on the vehicle head virtual manufacturing result data.

[0010] The present invention obtains the nonlinear shrinkage tensor field of FRP through hot pressing test, quantifying the millimeter-level thermally induced deformation that was originally ignored in advance; through the ply path topology and path-process coupling deformation prediction, the random error caused by the manual differences of different process technicians is eliminated, making the deformation law calculable and reproducible; by assigning laminated material properties according to the real spatial distribution in the initial model and performing aerodynamic-structural bidirectional coupling iteration, the simulated pressure field is ensured to be highly consistent with the subsequent physical wind tunnel data; through inverse compensation and radial basis function-curvature constraint reconstruction, mathematical compensation is used to replace repeated manual mold repair, and the curing shrinkage amount is "preset" into the geometric contour at one time; finally, virtual manufacturing simulation verifies the compensation effect again, and outputs a streamlined three-dimensional digital model of the front end that can be directly used for production. As a result, the design-manufacturing-testing chain has been shortened from multiple rounds of trial and error to "one-time modeling, one-time forming", which not only avoids the 2.5kPa→3.9kPa aerodynamic deviation caused by unpredictable deformation of 0.5-2mm in traditional methods, but also significantly reduces the risk of excessive pressure waves and false triggering of sensors during ultra-high-speed train crossings, thereby improving overall design accuracy, shortening iteration cycles and saving wind tunnel test costs.

[0011] Preferably, step S1 includes the following steps:

[0012] Step S11: measuring the dimensions of the standard FRP sample to obtain the geometric parameters of the FRP sample;

[0013] Step S12: Designing the fiber layup angles of the standard FRP specimens according to the FRP specimen geometric parameters to obtain a multi-angle fiber layup scheme. The fiber layup angle design specifically includes designing fiber layup schemes according to seven angle gradients: 0°, 15°, 30°, 45°, 60°, 75°, and 90°. Five replicate specimens are prepared for each angle. A prepreg layup process is used, and the fiber volume fraction is controlled at 55% ± 2%.

[0014] Step S13: Arranging a strain gauge array on the standard FRP specimen based on the fiber multi-angle layup scheme to obtain strain monitoring points;

[0015] Step S14: installing a thermocouple temperature sensor on the standard FRP sample according to the strain monitoring points to obtain the temperature monitoring points;

[0016] Step S15: setting hot pressing process parameters for the standard FRP sample based on the temperature monitoring points and the strain monitoring points to obtain FRP forming process parameters;

[0017] Step S16: performing real-time data acquisition on a standard FRP sample according to the FRP molding process parameters to obtain a FRP strain-temperature parameter set;

[0018] Step S17: Identifying the strain-temperature relationship of the standard FRP specimen based on the FRP strain-temperature parameter set to obtain the thermal expansion coefficient of the FRP material;

[0019] Step S18: constructing a FRP contraction tensor field based on the thermal expansion coefficient of the FRP material.

[0020] Preferably, step S18 includes the following steps:

[0021] Step S181: Calculating the shrinkage strain of a standard FRP specimen based on the FRP material's thermal expansion coefficient and the FRP strain-temperature parameter set to obtain original shrinkage strain data;

[0022] Step S182: performing fiber orientation angle correlation identification on the shrinkage strain raw data to obtain angle-shrinkage correlation data;

[0023] Step S183: identifying the principal strain direction of the standard FRP specimen based on the angle-contraction correlation data to obtain the principal strain direction data of the FRP;

[0024] Step S184: reconstructing the material temperature gradient field of the standard FRP sample according to the FRP principal strain direction data and the temperature monitoring points to obtain the FRP temperature gradient field;

[0025] Step S185: calculating the material shrinkage tensor components of the standard FRP specimen based on the FRP temperature gradient field and the FRP principal strain direction data to obtain the FRP material shrinkage tensor components;

[0026] Step S186: performing material contraction behavior space interpolation expansion on the FRP material contraction tensor component to obtain a continuous tensor field of FRP material contraction;

[0027] Step S187: verifying the shrinkage prediction model of the standard FRP material based on the FRP material shrinkage continuous tensor field to obtain the FRP shrinkage tensor field.

[0028] Preferably, constructing a topological relationship of the laying path for the fiber layer of the headstock in step S2 includes:

[0029] Acquire the front surface geometry data; extract NURBS surface parameters from the front surface geometry data to obtain the front surface control points;

[0030] Gaussian curvature of the vehicle head surface is calculated based on the vehicle head surface control points to obtain the curvature distribution data of the vehicle head surface;

[0031] According to the curvature distribution data of the vehicle head surface, the hyperbolic area of ​​the vehicle head surface is identified to obtain the marking data of the hyperbolic area of ​​the vehicle head;

[0032] Based on the hyperbolic region marking data of the vehicle head, the vehicle head surface is divided into a geodesic grid to obtain a geodesic grid; the vehicle head surface is evaluated for developability based on the geodesic grid to obtain surface developability evaluation data;

[0033] Based on the surface developability evaluation data, the front surface is decomposed into quasi-planar units to obtain a set of quasi-planar units on the front surface;

[0034] The starting point of the fiber layer of the front of the vehicle is planned according to the quasi-planar unit set of the front surface, and the starting point data of the fiber layer of the front of the vehicle is obtained;

[0035] Based on the starting point data of the fiber layer of the head and the quasi-planar unit set of the head surface, the fiber bundle path of the head is planned to obtain the fiber bundle path of the head;

[0036] Identify the ply turning point based on the fiber bundle path of the front of the vehicle and obtain the coordinates of the ply turning point of the front of the vehicle;

[0037] The ply topology relationship is constructed based on the coordinates of the turning points of the front ply and the fiber bundle path of the front ply, and the front ply path diagram is obtained.

[0038] Preferably, in step S2, performing path-process coupled deformation prediction on the headstock fiber bundle based on the headstock layup path diagram includes:

[0039] The stress state of the fiber bundle of the head is modeled according to the head layup path diagram to obtain the stress state parameters of the fiber bundle tension field;

[0040] Based on the stress state parameters of the fiber bundle tension field, the curvature response coupling mapping of the fiber bundle of the headstock is performed to obtain the fiber bundle curvature-tension coupling response data;

[0041] The path memory factor of the fiber bundle of the head is calculated based on the curvature-tension coupling response data of the fiber bundle and the turning point coordinates of the head ply to obtain the path memory factor data;

[0042] Perform multi-angle visual and mechanical synchronous acquisition of the actual layup process of the vehicle head to obtain layup behavior time series data;

[0043] Clustering and identifying the layup habit features of layup behavior time series data to obtain the layup habit pattern of the technicians;

[0044] Based on the path memory factor data, the randomness model of the technician's layup habit pattern is built to obtain the random perturbation parameters of the layup process;

[0045] The ply deformation of the vehicle head is predicted according to the random disturbance parameters of the ply process and the path memory factor data, and the ply path deformation prediction data is obtained.

[0046] Preferably, in step S3, obtaining an initial vehicle head 3D model; and spatially allocating the properties of the fiberglass laminated material according to the ply path deformation prediction data based on the initial vehicle head 3D model comprises:

[0047] Obtain an initial three-dimensional model of the vehicle head; discretize the initial three-dimensional model of the vehicle head to obtain a CFD calculation grid for the vehicle head;

[0048] The boundary conditions of the vehicle head flow field are set based on the vehicle head CFD calculation grid to obtain the boundary conditions of the vehicle head flow field;

[0049] According to the FRP shrinkage tensor field and the boundary conditions of the vehicle head flow field, the CFD solver is set for the initial vehicle head 3D model to obtain the CFD solution parameters;

[0050] The front flow field is numerically solved based on the CFD solution parameters to obtain the initial front flow field solution data;

[0051] The pressure distribution on the vehicle head surface is extracted based on the initial flow field solution data to obtain the pressure distribution data on the vehicle head surface. The pressure distribution extraction specifically includes extracting pressure values ​​at 2156 nodes on the vehicle head surface and calculating the unit pressure using the area-weighted average method. The pressure data accuracy is ±1Pa.

[0052] Generate FEA structural mesh of the vehicle head based on the surface pressure distribution data of the vehicle head to obtain the vehicle head structural calculation mesh;

[0053] The properties of the FRP laminated materials are spatially distributed based on the deformation prediction data of the ply path and the calculation grid of the vehicle head structure to obtain the distribution data of the steel layer material properties.

[0054] Preferably, in step S3, performing aerodynamic-structural bidirectional coupling iterative correction on the initial vehicle head 3D model based on the steel layer material property distribution data includes:

[0055] Based on the steel layer material property distribution data and the head surface pressure distribution data, the head thermal-mechanical coupling solution is performed to obtain the head thermal stress distribution data;

[0056] The shrinkage deformation of the vehicle head is calculated based on the thermal stress distribution data of the vehicle head to obtain the deformation response data of the vehicle head structure; the geometry of the vehicle head is updated and transformed based on the deformation response data of the vehicle head structure to obtain the deformed geometry data of the vehicle head;

[0057] Regenerate the aerodynamic mesh based on the deformed geometric data of the vehicle head to obtain the updated CFD mesh of the vehicle head;

[0058] Based on the updated CFD mesh of the vehicle head, deformation-aerodynamic sensitivity mapping calculation is performed to obtain the vehicle head deformation sensitivity table;

[0059] According to the vehicle head deformation sensitivity table, the convergence criterion of the vehicle head aerodynamic-structural coupling iteration is set to obtain the vehicle head coupling convergence criterion parameters;

[0060] Based on the front coupling convergence criterion parameters, the front aerodynamic-structural coupling iterative solution is performed to obtain the front coupling iterative solution data;

[0061] The final geometry of the initial three-dimensional vehicle head model is determined based on the vehicle head coupling iterative solution data to obtain the coupled modified vehicle head model.

[0062] Preferably, step S4 includes the following steps:

[0063] Step S41: performing deformation statistics on the coupled modified vehicle head model to obtain vehicle head deformation statistics;

[0064] Step S42: performing reverse deformation prediction on the vehicle head based on the fiberglass shrinkage tensor field and the vehicle head deformation statistical data to obtain vehicle head reverse deformation prediction data;

[0065] Step S43: dividing the vehicle head surface into compensation areas according to the vehicle head reverse deformation prediction data to obtain vehicle head compensation area division data;

[0066] Step S44: allocating compensation weights to the compensation areas based on the vehicle head compensation area division data to obtain vehicle head compensation weight allocation data;

[0067] Step S45: performing pre-deformation compensation calculation on the vehicle head according to the vehicle head compensation weight distribution data and the vehicle head reverse deformation prediction data to obtain vehicle head pre-deformation compensation data;

[0068] Step S46: generating a discrete point cloud of the vehicle head surface based on the vehicle head pre-deformation compensation data to obtain vehicle head compensation point cloud data;

[0069] Step S47: reconstructing the compensation surface of the vehicle head by using radial basis function and curvature constraint fusion according to the vehicle head compensation point cloud data to obtain the vehicle head geometric compensation contour data.

[0070] Preferably, step S5 includes the following steps:

[0071] Step S51: performing virtual layup process simulation on the vehicle head geometry compensation profile data to obtain virtual layup simulation data;

[0072] Step S52: performing a virtual curing simulation on the vehicle head based on the virtual layup simulation data and the fiberglass shrinkage tensor field to obtain the vehicle head virtual curing simulation data;

[0073] Step S53: performing demoulding cooling simulation on the locomotive head according to the locomotive head virtual solidification simulation data to obtain locomotive head demoulding cooling simulation data;

[0074] Step S54: reconstructing the residual stress field of the lathe head based on the lathe head demoulding cooling simulation data to obtain the lathe head residual stress distribution data;

[0075] Step S55: performing final deformation calculation on the vehicle head according to the residual stress distribution data of the vehicle head and the geometric compensation profile data of the vehicle head to obtain the virtual manufacturing result data of the vehicle head;

[0076] Step S56: constructing a streamlined three-dimensional digital model of the vehicle front based on the vehicle front virtual manufacturing result data.

[0077] Preferably, the present invention further provides a three-dimensional digital model generation system based on a fiberglass streamlined front end, which is used to execute the three-dimensional digital model generation method based on a fiberglass streamlined front end as described above. The three-dimensional digital model generation system based on a fiberglass streamlined front end comprises:

[0078] The material property calibration module is used to test the entire hot pressing process of standard FRP samples to obtain the thermal expansion coefficient of FRP materials; and to construct the FRP contraction tensor field based on the thermal expansion coefficient of FRP materials;

[0079] The ply path modeling module is used to construct the ply path topology relationship of the head fiber ply to obtain the head ply path diagram; based on the head ply path diagram, the path-process coupling deformation prediction of the head fiber bundle is performed to obtain the ply path deformation prediction data;

[0080] The aerodynamic-structural coupling correction module is used to obtain an initial three-dimensional model of the vehicle head. Based on the initial three-dimensional model, the fiberglass laminate material properties are spatially distributed according to the layup path deformation prediction data to obtain the steel layer material property distribution data. Based on the steel layer material property distribution data, the initial three-dimensional model of the vehicle head is subjected to aerodynamic-structural bidirectional coupling iterative correction to obtain a coupled corrected vehicle head model.

[0081] The inverse compensation and reconstruction module is used to perform inverse compensation on the coupled modified vehicle head model based on the FRP shrinkage tensor field, generate discrete point clouds, and obtain vehicle head compensation point cloud data; based on the vehicle head compensation point cloud data, the compensation surface of the vehicle head is reconstructed using radial basis function and curvature constraint fusion to obtain vehicle head geometric compensation contour data;

[0082] The virtual manufacturing verification module is used to perform virtual manufacturing simulation on the vehicle head according to the vehicle head geometric compensation contour data to obtain the vehicle head virtual manufacturing result data; and to construct a streamlined vehicle head three-dimensional digital model based on the vehicle head virtual manufacturing result data.

[0083] The present invention encapsulates five functional modules, namely material property calibration, ply path modeling, aerodynamic-structural coupling correction, inverse compensation reconstruction and virtual manufacturing verification, into an integrated platform, thus realizing for the first time a digital closed loop of the entire process from thermal compression shrinkage tensor to manufacturable geometry. After the system is running, engineers only need to input the initial aerodynamic shape to automatically complete the work that originally required multiple rounds of wind tunnel-physical iterations on a single workstation, significantly shortening the design cycle and reducing testing costs. At the same time, since each module shares a unified data interface and shrinkage tensor field, the information distortion caused by traditional cross-software and cross-departmental collaboration is eliminated, so that the error between the final output streamlined front three-dimensional digital model and the physical contour is controlled within the millimeter level, fundamentally avoiding the aerodynamic performance deviation and sensor false triggering risk caused by non-uniform shrinkage of fiberglass, and providing a replicable engineering platform for the safe, efficient and low-cost research and development of ultra-high-speed trains with speeds above 350km / h. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Other features, objects and advantages of the present invention will become more apparent from reading the detailed description made with reference to the following drawings:

[0085] Figure 1 A schematic flow chart of the steps of a method for generating a three-dimensional digital model of a fiberglass streamlined front vehicle according to an embodiment is shown.

[0086] Figure 2 A detailed flowchart of step S4 of an embodiment is shown.

[0087] Figure 3 A detailed flowchart of step S5 of an embodiment is shown.

[0088] Figure 4 A schematic diagram of the operation of measuring the size specifications of a standard fiberglass reinforced plastic sample in step S11 of one embodiment is shown. DETAILED DESCRIPTION

[0089] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0090] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0091] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0092] To achieve this, please refer to Figures 1 to 4 The present invention provides a method for generating a three-dimensional digital model based on a fiberglass streamlined front end, comprising the following steps:

[0093] Step S1: Conducting a full hot pressing forming process test on a standard FRP sample to obtain the thermal expansion coefficient of the FRP material; constructing a FRP contraction tensor field based on the thermal expansion coefficient of the FRP material;

[0094] Step S2: constructing a topological relationship of the fiber layup path for the headstock to obtain a headstock layup path diagram; performing path-process coupled deformation prediction on the headstock fiber bundle based on the headstock layup path diagram to obtain layup path deformation prediction data;

[0095] Step S3: Obtaining an initial three-dimensional model of the vehicle head; spatially distributing the fiberglass laminate material properties based on the initial three-dimensional model and the predicted deformation data of the ply path to obtain the material property distribution data of the steel layer; performing aerodynamic-structural bidirectional coupling iterative correction on the initial three-dimensional model of the vehicle head based on the material property distribution data of the steel layer to obtain a coupled corrected vehicle head model;

[0096] Step S4: performing inverse compensation and discrete point cloud generation on the coupled modified vehicle head model according to the FRP shrinkage tensor field to obtain vehicle head compensation point cloud data; performing radial basis function-curvature constraint fusion reconstruction on the compensation surface of the vehicle head according to the vehicle head compensation point cloud data to obtain vehicle head geometric compensation contour data;

[0097] Step S5: performing virtual manufacturing simulation on the vehicle head according to the vehicle head geometric compensation contour data to obtain the vehicle head virtual manufacturing result data; and constructing a streamlined vehicle head three-dimensional digital model based on the vehicle head virtual manufacturing result data.

[0098] Preferably, step S1 includes the following steps:

[0099] Step S11: measuring the dimensions of the standard FRP sample to obtain the geometric parameters of the FRP sample;

[0100] Step S12: Designing the fiber layup angles of the standard FRP specimens according to the FRP specimen geometric parameters to obtain a multi-angle fiber layup scheme. The fiber layup angle design specifically includes designing fiber layup schemes according to seven angle gradients: 0°, 15°, 30°, 45°, 60°, 75°, and 90°. Five replicate specimens are prepared for each angle. A prepreg layup process is used, and the fiber volume fraction is controlled at 55% ± 2%.

[0101] Step S13: Arranging a strain gauge array on the standard FRP specimen based on the fiber multi-angle layup scheme to obtain strain monitoring points;

[0102] Step S14: installing a thermocouple temperature sensor on the standard FRP sample according to the strain monitoring points to obtain the temperature monitoring points;

[0103] Step S15: setting hot pressing process parameters for the standard FRP sample based on the temperature monitoring points and the strain monitoring points to obtain FRP forming process parameters;

[0104] Step S16: performing real-time data acquisition on a standard FRP sample according to the FRP molding process parameters to obtain a FRP strain-temperature parameter set;

[0105] Step S17: Identifying the strain-temperature relationship of the standard FRP specimen based on the FRP strain-temperature parameter set to obtain the thermal expansion coefficient of the FRP material;

[0106] Step S18: constructing a FRP contraction tensor field based on the thermal expansion coefficient of the FRP material.

[0107] In this embodiment, please refer to the attached Figure 4In a laboratory with a constant temperature of 20℃ and a humidity of 45%, a bridge-type three-dimensional coordinate measuring machine 101 was used to calibrate the dimensions of a 150mm×150mm×3mm epoxy fiberglass standard specimen 102 placed on a storage table 103. A ruby-trigger probe with a diameter of 2mm was selected, the measuring speed was set to 5mm / s, and the detection force was 0.3N. Five measuring points were evenly taken along the length, width, and thickness of the specimen. The system automatically fitted the least squares plane and output the geometric parameters: length 150.02±0.01mm, width 149.98±0.01mm, thickness 2.97±0.01mm. A rectangular coordinate system was established with the geometric center of the specimen as the origin and recorded. T300-grade carbon fiber-epoxy prepreg was used, with a nominal single-layer thickness of 0.2 mm and a resin mass fraction of 42%. Sheets were cut into 150 mm × 150 mm single-layer sheets at seven angles: 0°, 15°, 30°, 45°, 60°, 75°, and 90°, using an automatic cutting machine. The weight of each layer was controlled to be 12.5 ± 0.2 g. Each layer was weighed using a precision electronic balance and stacked 14 layers symmetrically in the order of [0 / 15 / 30 / 45 / 60 / 75 / 90] s on a clean bench to form a laminate with a total thickness of approximately 2.8 mm. Five replicate specimens were prepared in each direction, for a total of 35 specimens. After stacking, the laminates were sealed in vacuum bags and stored at -18°C, allowing them to be warmed for 4 h before use. A 5mm×5mm reference grid was marked on the specimen surface using a laser marking instrument. Foil strain gauges with a nominal resistance of 350Ω and a grid length of 3mm were selected and adhered point by point using quick-drying cyanoacrylate adhesive. Strain gauges were arranged at the 81 grid intersections in a 9×9 array. After attachment, the gauges were rolled three times in one direction with a 2kg rubber roller to remove bubbles. A layer of polytetrafluoroethylene film was applied and the gauges were cured at room temperature for 24 hours. The resistance of each gauge was tested using a 6.5-digit digital multimeter, and the resistance deviation was controlled within ±0.1Ω. The edges of the specimens were marked with a serial number using a marker pen. K-type thermocouple wire with a diameter of 0.5mm and an insulating layer of woven glass fiber was selected; it was arranged in three layers in the thickness direction of the specimen: 3 points were arranged in an equilateral triangle on the surface layer 0.5mm away from the upper surface, 3 points were arranged in the middle layer at the center of the thickness, and 3 points were arranged on the bottom layer 0.5mm away from the lower surface, for a total of 9 points; the thermocouple junctions were fixed with high-temperature resistant epoxy glue, the wires were tied along the edge of the specimen with polyimide tape, and connected to a 32-channel temperature patrol meter, and the cold end was placed in a 0℃ ice-water mixture to achieve cold end compensation. The laminated sample is placed in an electric heating flat vulcanizer, with 12 800W heating rods built into the upper and lower hot plates; the heating program is set: room temperature → 125℃ linearly at 2℃ / min, and constant at 125℃ for 90 minutes; the pressure is controlled at 0.6±0.05MPa through a servo hydraulic station closed loop; a PID temperature controller is used for real-time adjustment, with temperature overshoot not exceeding ±1℃ and pressure fluctuation not exceeding ±0.02MPa; the process parameters are stored in the PLC after curing to ensure batch consistency. A synchronous data acquisition system is used, and the strain channel adopts a 24-bit The temperature channel uses a 16-bit successive approximation ADC (ADC is the abbreviation of Analog-to-Digital Converter, which is called analog-to-digital converter or analog-to-digital converter in Chinese. In this embodiment, it refers to the real-time conversion of continuous analog signals (such as small voltage changes output by strain gauges and thermoelectric potential output by thermocouples) into discrete digital signals). The sampling frequency is set to 1kHz; the strain signal is connected to the strain conditioning module through a four-wire connection, and the temperature signal is directly connected to the temperature module after cold-end compensation; the data is transmitted to the workstation in real time via Gigabit Ethernet and stored in TDMS format. A single test lasts about 3 hours, with a data volume of about 2GB, and is automatically backed up locally every 60 seconds. The collected strain-temperature data is imported into the data analysis software, and after eliminating the unstable section at the initial stage of heating, stable data in the range of 50-125℃ is selected; a linear regression is performed on the strain gauge data in each direction, and the slope is the thermal expansion coefficient in that direction, for example, in the 0° direction / ℃, 90° direction / °C, with R² greater than 0.98 in all directions. The principal strain directions are calculated using the differential method, and the temperature gradient through the thickness is reconstructed based on the data from the nine thermocouples. Finally, the three-dimensional thermal expansion coefficient tensor of the FRP specimen is output. For the detailed implementation of step S18, please refer to the sub-steps of step S18.

[0108] Preferably, step S18 includes the following steps:

[0109] Step S181: Calculating the shrinkage strain of a standard FRP specimen based on the FRP material's thermal expansion coefficient and the FRP strain-temperature parameter set to obtain original shrinkage strain data;

[0110] Step S182: performing fiber orientation angle correlation identification on the shrinkage strain raw data to obtain angle-shrinkage correlation data;

[0111] Step S183: identifying the principal strain direction of the standard FRP specimen based on the angle-contraction correlation data to obtain the principal strain direction data of the FRP;

[0112] Step S184: reconstructing the material temperature gradient field of the standard FRP sample according to the FRP principal strain direction data and the temperature monitoring points to obtain the FRP temperature gradient field;

[0113] Step S185: calculating the material shrinkage tensor components of the standard FRP specimen based on the FRP temperature gradient field and the FRP principal strain direction data to obtain the FRP material shrinkage tensor components;

[0114] Step S186: performing material contraction behavior space interpolation expansion on the FRP material contraction tensor component to obtain a continuous tensor field of FRP material contraction;

[0115] Step S187: verifying the shrinkage prediction model of the standard FRP material based on the FRP material shrinkage continuous tensor field to obtain the FRP shrinkage tensor field.

[0116] In this embodiment, the thermal expansion coefficients in each direction obtained in step S17 and the real-time strain-temperature parameter set are imported into the numerical calculation software. First, the non-steady-state section of heating is eliminated, and only the data from the end of the constant temperature section at 125°C to the cooling section at 30°C are retained; for each angle ply sample, the formula Calculate the pure contraction strain, where represents the measured total strain, represents the pure curing shrinkage strain, is the thermal expansion coefficient in the corresponding direction, The temperature change is calculated point by point at 81 measuring points, and a total of 2835 shrinkage strain raw data of 35 groups of samples are obtained. The data interval is 0.1s and the accuracy is ±1με. The results are stored in matrix form, with the rows being the measuring point numbers and the columns being the time series. The shrinkage strain raw data are grouped according to the fiber orientation angles of 0°, 15°, 30°, 45°, 60°, 75°, and 90°, and a single-factor variance analysis is performed using statistical software with a significance level of 0.05. The second-order Fourier series is used to calculate the shrinkage strain raw data. Fitting angle-contraction relationship, goodness of fit R²>0.93, where The fiber orientation angle is Pure curing shrinkage strain value, It is the angle between the fiber direction and the reference axis (usually the length direction of the specimen), and can be 0°, 15°, 30°, ..., 90°. is the constant term of the Fourier series, representing the average contraction level at all angles, 、 is the first harmonic coefficient, describing 2 The amplitude of symmetric and antisymmetric contraction changes within the period, 、 is the second harmonic coefficient, describing 4 The higher frequency contraction variation within the cycle; integrate the fitting coefficients and residuals into an angle-contraction correlation data table, with an interpolation point every 5°. For each set of angle-contraction correlation data obtained in step S182, construct a 2×2 plane strain tensor , solve the eigenvalue problem by Jacobi iteration method and get the principal strain 、 and its corresponding main direction angle ; The iterative convergence tolerance is set to , with a maximum number of iterations of 100 times; the principal direction results of the 7 angle specimens were summarized and fitted with a Gaussian distribution to obtain an average principal strain direction of 10.2°±1.5°, which was output as the principal strain direction data of the FRP. The temperature values ​​measured by the 9 thermocouples were imported into the gridding software according to the spatial coordinates, and a three-dimensional linear interpolation function T(x,y,z) was established; the grid size was 1mm×1mm×0.5mm, and an eight-node hexahedral unit was used; the gradient of the interpolation result was calculated to obtain the temperature gradient field , the gradient value range is 0–4℃ / mm; align the gradient field with the principal strain direction data, and output the temperature gradient field file consistent with the specimen coordinate system. Based on the generalized Hooke's law and the thermal-humidity-elastic coupling constitutive model, substitute the principal strain direction data and the temperature gradient field into the formula ,in is the direction-dependent thermal expansion coefficient tensor, is the material shrinkage tensor component in the i-th row and j-th column, which represents the strain component caused by curing shrinkage and thermal shrinkage in unit volume. E is the elastic modulus of FRP in the current fiber direction. v is the Poisson's ratio of the material, which describes the ratio of transverse to axial strain. is the Kronecker symbol (1 when i=j, otherwise 0), used to separate the volumetric strain term, is the volume strain, which is the sum of the diagonals of the strain tensor, is the temperature change relative to the reference temperature; the tensor operation library is used to calculate point by point to obtain 6 independent contraction tensor components S 11 、S 12 、S 13 、S 22 、S 23 、S 33 , retain four decimal places in numerical precision; the results are output in CSV format, with each line corresponding to an integration point coordinate and six component values. The discrete contraction tensor components obtained in step S185 are imported into the CAE software, and the multi-quadratic radial basis function is selected. Spatial interpolation was performed, where r was the Euclidean distance between the interpolated point and the known compensation point. The shape parameter c was determined to be 3.5 mm through cross-validation, and the support radius was 15 mm. The interpolation region covered the entire specimen (150 mm × 150 mm × 3 mm), generating a continuous tensor field with a grid spacing of 0.5 mm. The interpolation results were smoothed and filtered to ensure that the rate of change of adjacent unit tensors was less than 1%. Finally, a continuous three-dimensional shrinkage tensor field was output. 20% of the data points in the continuous tensor field were randomly selected as the validation set, and the remaining 80% were used for training. Using 10-fold cross-validation, the root mean square error (RMSE) between the predicted shrinkage values ​​and the measured values ​​was calculated, with a target RMSE of ≤0.05%. A coefficient of determination (R²) of ≥0.92 was considered passed. If this was not achieved, the radial basis function shape parameter c was adjusted and re-interpolated until the error criteria were met, ultimately resulting in the solidified FRP shrinkage tensor field.

[0117] Preferably, constructing a topological relationship of the laying path for the fiber layer of the headstock in step S2 includes:

[0118] Acquire the front surface geometry data; extract NURBS surface parameters from the front surface geometry data to obtain the front surface control points;

[0119] Gaussian curvature of the vehicle head surface is calculated based on the vehicle head surface control points to obtain the curvature distribution data of the vehicle head surface;

[0120] According to the curvature distribution data of the vehicle head surface, the hyperbolic area of ​​the vehicle head surface is identified to obtain the marking data of the hyperbolic area of ​​the vehicle head;

[0121] Based on the hyperbolic region marking data of the vehicle head, the vehicle head surface is divided into a geodesic grid to obtain a geodesic grid; the vehicle head surface is evaluated for developability based on the geodesic grid to obtain surface developability evaluation data;

[0122] Based on the surface developability evaluation data, the front surface is decomposed into quasi-planar units to obtain a set of quasi-planar units on the front surface;

[0123] The starting point of the fiber layer of the front of the vehicle is planned according to the quasi-planar unit set of the front surface, and the starting point data of the fiber layer of the front of the vehicle is obtained;

[0124] Based on the starting point data of the fiber layer of the head and the quasi-planar unit set of the head surface, the fiber bundle path of the head is planned to obtain the fiber bundle path of the head;

[0125] Identify the ply turning point based on the fiber bundle path of the front of the vehicle and obtain the coordinates of the ply turning point of the front of the vehicle;

[0126] The ply topology relationship is constructed based on the coordinates of the turning points of the front ply and the fiber bundle path of the front ply, and the front ply path diagram is obtained.

[0127] In this example, a GOM ATOS Q 12 M blue light scanner was used to perform a full-range white light scan of a 1:1 fiberglass streamlined front end. The lens working distance was 550 mm, the single-frame measurement range was 300 mm × 200 mm, and the scanning pitch was 0.2 mm. A total of 420 images were captured. Automatic stitching and point cloud denoising were performed in the GOM Inspect software to obtain an STL model with approximately 12 million triangles. The STL model was imported into Rhinoceros, and the "MeshToNURB" command was executed to convert the mesh into a single NURBS surface. The "Rebuild" command was then called to set the number of nodes in the U and V directions to 128 each and the degree to 5. Finally, the "What" command was used in the property bar to read the node vectors, weights, and control point coordinates. A total of 12,348 control points were obtained with a coordinate accuracy of 0.01 mm, and the surface was exported in CSV format. Read the CSV control points from the previous step into MATLAB and load the NURBS toolbox. Use the nrbderiv and nrbdeval functions to uniformly sample the parameter domain (u, v) at a spacing of 1 mm to generate a 150×200 grid point. Use the classical differential geometry formula Gaussian curvature was calculated point by point, with E, F, G, L, M, and N all derived from the first and second fundamental forms of the surface. This yielded continuously distributed data with a curvature range of -0.052 to 0.031 mm². This data was stored as an ASCII raster file, with each row containing the values ​​of three columns: u, v, and K. The ASCII raster file was imported into ImageJ, and Image→Adjust→Threshold was used to set |K| ≥ 0.002 mm² to white and all others to black, resulting in a binary image. AnalyzeParticles was used with a minimum pixel size of 100 mm² (corresponding to 100 pixels × 1 mm² / pixel) to filter out noisy areas. This resulted in an 8-bit mask PNG file, where the white area covered 37% of the vehicle's front surface area, representing the hyperbolic region marker data. Launch Grasshopper in Rhinoceros and use the Image Sampler to read the mask PNG. Using the nose tip as a fixed source, invoke the Shortest Walk plugin to calculate the geodesic distance field with a step size of 5mm. Use the Contour component to generate 64 equidistant geodesics with a spacing of 5mm. Use the Delaunay Mesh component to triangulate the geodesic intersections, limiting the maximum side length to 8mm and the minimum angle to ≥30°. This results in a geodesic mesh containing 38,426 triangles, which is then exported and saved in OBJ format. Import the OBJ mesh into EvoluteTools T.MAP and execute Analyze → Discrete Curvature to calculate the discrete Gaussian curvature integral for each triangle, setting the threshold. The software automatically marked expandable faces green and non-expandable faces red. The green area accounted for 63% of the total area and was reported in a CSV report containing face number, integral value, and color label. The green expandable face set was imported into Geomagic Design X. Using the Region Growing command, the normal deviation threshold was set to 15° and the minimum area was set to 400 mm². Adjacent faces were merged. The software automatically segmented the face into 156 quasi-planar elements with an average area of ​​410 mm² and an edge length of ≤120 mm. Each element was exported separately in IGES format, with file names numbered sequentially from Plane_001 to Plane_156. 156 IGES quasi-planar elements were batch-opened in SolidWorks, and the MassProperties function was used to calculate the centroid coordinates of each element. If the centroid was less than 15 mm from the element boundary, it was shifted 5 mm inward along the element normal and reprojected onto the head surface. This yielded 156 starting points with a coordinate accuracy of 0.1 mm. All data was summarized in an Excel spreadsheet, listing the element number, X, Y, Z, and normal directions nX, nY, and nZ. The starting point Excel spreadsheet and the IGES elements were imported into FiberSIM, with settings for a fiber width of 12.7 mm, a minimum turning radius of 50 mm, and a maximum curvature change of 0.03 mm⁻¹. The software generated a continuous fiber path based on the Dijkstra algorithm. The path was smoothed using cubic splines with a node spacing of 2 mm. A total of 156 curves with a total length of 3.8 km were output and saved in PLY format, with each curve accompanied by the start, end, and node coordinates. The PLY fiber path was read in MATLAB, and the curvature κ = dθ / ds was calculated using cubic spline interpolation. Turning points were defined as values ​​greater than 0.05 mm⁻¹. The findpeaks function was used to automatically extract all peaks, merging adjacent points with a spacing of < 5 mm. A total of 480 valid turning points were retained, with coordinates stored to three decimal places in TXT format, with each line formatted as PathID, X, Y, and Z. The TXT turning points and PLY curves were imported into Gephi, where a directed graph was constructed with turning points as nodes and curve segments as edges. Edge weights were set to the corresponding curve segment lengths, and the "Average PathLength" function was used to calculate the average edge length of 118 mm. Finally, an adjacency matrix (CSV file size of 480 × 480) and a GEXF visualization network with coordinates and weights were exported, totaling 2.1 MB.

[0128] Preferably, in step S2, performing path-process coupled deformation prediction on the headstock fiber bundle based on the headstock layup path diagram includes:

[0129] The stress state of the fiber bundle of the head is modeled according to the head layup path diagram to obtain the stress state parameters of the fiber bundle tension field;

[0130] Based on the stress state parameters of the fiber bundle tension field, the curvature response coupling mapping of the fiber bundle of the headstock is performed to obtain the fiber bundle curvature-tension coupling response data;

[0131] The path memory factor of the fiber bundle of the head is calculated based on the curvature-tension coupling response data of the fiber bundle and the turning point coordinates of the head ply to obtain the path memory factor data;

[0132] Perform multi-angle visual and mechanical synchronous acquisition of the actual layup process of the vehicle head to obtain layup behavior time series data;

[0133] Clustering and identifying the layup habit features of layup behavior time series data to obtain the layup habit pattern of the technicians;

[0134] Based on the path memory factor data, the randomness model of the technician's layup habit pattern is built to obtain the random perturbation parameters of the layup process;

[0135] The ply deformation of the vehicle head is predicted according to the random disturbance parameters of the ply process and the path memory factor data, and the ply path deformation prediction data is obtained.

[0136] In this embodiment, the 480 turning points and 156 fiber bundle paths obtained in the previous step are imported into ANSYS Composite PrepPost to establish a shell unit model with a unit side length of 2mm, a material set to T300 / epoxy, and a single layer thickness of 0.2mm. A constant tension of 50N is applied at both ends of each path, the middle turning point is set to roller constraint, the large deformation switch is turned on, and a static analysis is run. The software outputs the axial stress, bending moment, and shear force of each fiber bundle. The results are written to CSV as the average stress value of the node. The fields include turning point ID, σ_axial, M_bending, and V_shear as tension field stress state parameters. The CSV stress results of the previous step are loaded into MATLAB together with the PLY curve, and the spline toolbox is called to calculate the path curvature. ; For each fiber bundle Draw scattered points on the horizontal axis and σ_axial on the vertical axis, and use a quadratic polynomial fit to get , fitting ;Will 、 The corresponding turning point coordinates are organized into a new table with fields including turning point ID, κ, σ, and coupling coefficient as curvature-tension coupling response data. Continue in the same MATLAB session to read the coupling response data and define the path memory factor , where 450 MPa is the fiber yield stress and L is the arc length of the current segment. Trapezoidal integration is used for segment-by-segment calculations, with a 2 mm step size for each segment. This results in 480 M values ​​ranging from 0.85 to 0.99. The results are directly appended to the previous table, with a MemoryFactor column added to the fields, retaining four decimal places. Eight 4K industrial cameras were arranged in a circular array within the cleanroom, covering the entire curved surface of the vehicle head. Six-dimensional force sensors were also installed on the layup roller handles with a sampling frequency of 1000 Hz. Three experienced technicians each completed a layup according to the standard process. The cameras recorded their hand trajectories, while the force sensors recorded pressure and torque. Timestamps were synchronized using the same pulse generator. Three sets of synchronized video and CSV mechanical data were collected, each approximately 45 minutes long and at a 30 fps frame rate. Import the video frame by frame into Kinovea for key point tracking, extract the three-dimensional coordinates of the wrist and fingertips, and combine with the force sensor CSV to form a nine-dimensional vector containing x, y, z, Fx, Fy, Fz, Tx, Ty, and Tz for each frame; select K-means clustering in SPSS, set the number of clusters to 4, and iterate 100 times. After convergence, four typical action modes are obtained, including "fast paving", "slow compaction", "edge back pressure", and "turn correction". The number of samples in each category is exported together with the center vector. The covariance matrix of each nine-dimensional vector is calculated with the cluster center as the mean; a multivariate normal distribution model is established in @Risk 8, and 1,000 random perturbation samples are generated for each type of action with a step size of 2 mm; at the same time, the path memory factor M is read as a weight multiplier to finally obtain the perturbation angle of each fiber bundle. obey , disturbance tension obey All parameters were packaged into a perturbation library. The perturbation library and memory factor data were imported into Abaqus, and an explicit dynamic step was established with a step size of 0.1s and a total duration of 480s. Δθ and ΔT perturbations were applied to each element integration point. The material model used the Hashin failure criterion and geometric nonlinearity was enabled. After the calculation, the node displacements were extracted, with the maximum out-of-plane deformation of 1.7mm located in the nose cone transition area. The displacement field was resampled to a 2mm grid to obtain the ply path deformation prediction data, with the fields containing the node coordinates.

[0137] Preferably, in step S3, obtaining an initial three-dimensional model of the vehicle head; and spatially allocating the properties of the fiberglass laminated material according to the ply path deformation prediction data based on the initial three-dimensional model of the vehicle head comprises:

[0138] Obtain an initial three-dimensional model of the vehicle head; discretize the initial three-dimensional model of the vehicle head to obtain a CFD calculation grid for the vehicle head;

[0139] The boundary conditions of the vehicle head flow field are set based on the vehicle head CFD calculation grid to obtain the boundary conditions of the vehicle head flow field;

[0140] According to the FRP shrinkage tensor field and the boundary conditions of the vehicle head flow field, the CFD solver is set for the initial vehicle head 3D model to obtain the CFD solution parameters;

[0141] The front flow field is numerically solved based on the CFD solution parameters to obtain the initial front flow field solution data;

[0142] The pressure distribution on the vehicle head surface is extracted based on the initial flow field solution data to obtain the pressure distribution data on the vehicle head surface. The pressure distribution extraction specifically includes extracting pressure values ​​at 2156 nodes on the vehicle head surface and calculating the unit pressure using the area-weighted average method. The pressure data accuracy is ±1Pa.

[0143] Generate FEA structural mesh of the vehicle head based on the surface pressure distribution data of the vehicle head to obtain the vehicle head structural calculation mesh;

[0144] The properties of the FRP laminated materials are spatially distributed based on the deformation prediction data of the ply path and the calculation grid of the vehicle head structure to obtain the distribution data of the steel layer material properties.

[0145] In this example, the original geometry of the 350 km / h streamlined front end was directly read from the vehicle CAD platform. The mesh format was Parasolid X_T, and all surfaces were G2 continuous. In Siemens NX, the single closed outer shell was generated by selecting "File→Export→STL" and selecting "Chordal tolerance 0.1mm." This outer shell was imported into ANSYS FluentMeshing, using a CutCell Cartesian mesh with a global size of 0.05m, a first boundary layer height of 0.01mm, an increase rate of 1.2, a total of 8 layers, and a y+target of 1. The total mesh size was approximately 2.8M, including 2.5M of volume mesh and 300k of boundary layer mesh. The mesh was then exported to Fluent MSH format. In Fluent, set the inlet to a velocity inlet of 97.22 m / s (350 km / h), a turbulence intensity of 5%, and a hydraulic diameter of 0.4 m; the outlet to a pressure outlet of 0 Pa; the ground moving wall velocity of 97.22 m / s; the vehicle body to a no-slip wall; and zero shear stress on the symmetry plane. All boundary conditions were set in one go through the Boundary Conditions panel. In Fluent, select the pressure-based coupled solver, spatially discretize the second-order upwind solver, use SIMPLEC for transients, and a time step of 0.001 s; select the SST k-ω turbulence model, and the residual convergence criterion of 1. The FRP shrinkage tensor field obtained in step S1 was attached to the material properties using a UDF, with a density of 1.6 g / cm³, an elastic modulus of 72 GPa, and a Poisson's ratio of 0.3, as input for the structural side. A Fluent parallel calculation was started, with 3000 iterations. The lift-drag coefficient was monitored, and the calculation was terminated when the lift coefficient changed by <0.1% over 100 consecutive steps. The calculation took approximately 8 hours, and the converged pressure, velocity, and turbulent kinetic energy fields were output as the initial flow field solution. In Fluent, the front surface was created using Surface→Plane, and then the static pressure at 2156 wall nodes was extracted using Report→Surface Integrals. The node values ​​were converted to the element center using an area-weighted average, with a pressure accuracy of ±1 Pa. The results were directly exported to a CSV file, listing the node number, X, Y, Z, and P. The CSV pressure data, along with the outer shell, was imported into ANSYS Mechanical. Tetrahedral solid elements were used, with a global size of 8mm and local refinement of the nose cone and guide grooves to 2mm. The total element size was approximately 1.5M, with a minimum Jacobian of 0.3. The mesh was checked for negative volumes via Mesh Validation. The predicted ply deformation data (2mm mesh for the displacement field) was mapped to the structural mesh using the Workbench External Data module, with a 5mm mapping radius and inverse distance weighting. Each ply was 0.25mm thick, with a total of 14 layers. The fiber orientation was aligned with a ply angle sequence of 0 / 45 / -45 / 90 degrees, with an angular tolerance of ±1°. The software automatically generated direction vectors, thickness, and volume fractions at each element center, generating complete FRP laminate material property distribution data.

[0146] Preferably, in step S3, performing aerodynamic-structural bidirectional coupling iterative correction on the initial vehicle head 3D model based on the steel layer material property distribution data includes:

[0147] Based on the steel layer material property distribution data and the head surface pressure distribution data, the head thermal-mechanical coupling solution is performed to obtain the head thermal stress distribution data;

[0148] The shrinkage deformation of the vehicle head is calculated based on the thermal stress distribution data of the vehicle head to obtain the deformation response data of the vehicle head structure; the geometry of the vehicle head is updated and transformed based on the deformation response data of the vehicle head structure to obtain the deformed geometry data of the vehicle head;

[0149] Regenerate the aerodynamic mesh based on the deformed geometric data of the vehicle head to obtain the updated CFD mesh of the vehicle head;

[0150] Based on the updated CFD mesh of the vehicle head, deformation-aerodynamic sensitivity mapping calculation is performed to obtain the vehicle head deformation sensitivity table;

[0151] According to the vehicle head deformation sensitivity table, the convergence criterion of the vehicle head aerodynamic-structural coupling iteration is set to obtain the vehicle head coupling convergence criterion parameters;

[0152] Based on the front coupling convergence criterion parameters, the front aerodynamic-structural coupling iterative solution is performed to obtain the front coupling iterative solution data;

[0153] The final geometry of the initial three-dimensional vehicle head model is determined based on the vehicle head coupling iterative solution data to obtain the coupled modified vehicle head model.

[0154] In this example, the mapped front structure mesh was loaded into ANSYS Mechanical. The mesh consisted of 1.5 million tetrahedral elements, with element sizes refined to 2 mm at the nose and 5 mm elsewhere. Material properties were set using the Engineering Data module's composite material library, using 14 layers of T300 / epoxy laminate, each 0.25 mm thick, with a ply sequence of [0 / 45 / -45 / 90]s. The principal elastic moduli for the single layer were E1 = 135 GPa, E2 = 8.5 GPa, and G12 = 4.5 GPa, with a Poisson's ratio ν12 = 0.3 and a density of 1600 kg / m³. Regarding boundary conditions, the pressure load from step S3 was applied point by point to each of the 2156 nodes on the exterior surface of the solid. The pressure values ​​were imported in tabular form with an accuracy of ±1 Pa. A uniform temperature field of 125°C was applied through the thickness of the solid, using the Thermal Condition function. The solver uses steady-state thermal-structural coupling, turns on the Large Deflection option, sets the load step to 1, the substep to 20, and the convergence tolerance to 1. , the results output the three principal stresses σ1, σ2, σ3 and equivalent stress of each node, forming a CSV file indexed by the node ID, which is the thermal stress distribution data of the headstock. The thermal stress distribution data is imported into ABAQUS together with the temperature load of 125℃, and a HeatTransfer step is created in the Step module with a time step of 0.1s and a total time of 900s to calculate the temperature field; a Static General step is created in the same model with a time step of 0.05s and a total time of 1800s, using a sequential coupling method. The material model uses the Hashin failure criterion, and the node-level thermal expansion coefficient tensor is written point by point through the user subroutine USDFLD. The tensor is provided by the fiberglass shrinkage tensor field of step S1. The node temperature is obtained through GETVRM in the subroutine to calculate the curing shrinkage strain. ,in The node-level tensor, ΔT = 125–25°C, is used. After the calculation is complete, all node displacement vectors (U, V, and W) are output. The maximum out-of-plane displacement of 1.73 mm is located at the nose cone apex. The results are saved as a CSV file in the form of node ID, U, V, and W, representing the deformation response data of the front structure. Open the original front NURBS surface in Rhinoceros, use the Apply Displacement plugin, select the CSV displacement file, select Closest Point for mapping, set the displacement units to mm, and superimpose the displacement vectors using the Lagrangian description of the control point coordinates. After the surface is reconstructed, execute the RebuildEdges command with a tolerance of 0.01 mm. Check for self-intersections and negative volumes, and confirm that the surface continuity is G2. After the coordinate update is complete, use the Export function to export the surface to STEP format with three decimal places of precision. Import the updated STEP file into ANSYS Fluent Meshing, select CutCell Cartesian mesh, global base size 0.05m, boundary layer parameter first layer height 0.01mm, growth rate 1.2, number of layers 8, total thickness 0.1m; check the RemeshDeformed Geometry option, the software automatically identifies the deformed area and refines the nose cone and guide groove to 0.02m. The final updated CFD mesh of the front of the vehicle is about 2.82M units, with a maximum slope angle of 0.8° and a minimum orthogonal quality of 0.15, which meets the CFD calculation requirements. Activate Adjoint Solver in Fluent, set the objective function to the aerodynamic drag coefficient Cd, and set the convergence residual 1 for the adjoint solution. A 0.1mm positive displacement perturbation is applied to each of the 2156 surface nodes, which is implemented by the User Defined Function. After the perturbation, Cd is recalculated using the finite difference method. , get a 2156×2156 sensitivity matrix; the matrix element range is -3.2 to 4.7 The results are output as a CSV file through Report Definition, which is the head deformation sensitivity table. In ANSYS Workbench, a System Coupling system is established, and Fluent and Mechanical are dragged into the coupling domain; in the System Coupling settings panel, the displacement convergence criterion is set to relative displacement change. <1 , the pressure convergence criterion is set to relative pressure change <1 , the maximum coupling iteration step is set to 50 steps; the initial value of the Aitken relaxation factor is 0.5, and each step is based on the formula Dynamic updates ensure stable iterative convergence. Parallel computing is started in SystemCoupling, allocating 16 CPU cores, 8 cores each for Fluent and Mechanical; the first round of Fluent calculates aerodynamic loads, the second round of Mechanical calculates structural deformation, and the third round updates the geometry and regenerates the CFD mesh, iterating repeatedly; at step 27, the dual convergence criteria are met, the maximum displacement change is reduced to 0.002mm, the maximum pressure change is reduced to 0.8Pa, and the coupled iterative solution data is automatically saved. Extract the converged node coordinates in the Workbench result module and use the Export function to export them to STEP format; open the STEP file in Rhinoceros 7, run the Check Geometry command, set the tolerance to 0.01mm, confirm that there are no broken surfaces, no self-intersections, and the overall dimensional error is less than 0.05mm; after confirmation, use the surface as the coupled modified vehicle head model.

[0155] Preferably, step S4 includes the following steps:

[0156] Step S41: performing deformation statistics on the coupled modified vehicle head model to obtain vehicle head deformation statistics;

[0157] Step S42: performing reverse deformation prediction on the vehicle head based on the fiberglass shrinkage tensor field and the vehicle head deformation statistical data to obtain vehicle head reverse deformation prediction data;

[0158] Step S43: dividing the vehicle head surface into compensation areas according to the vehicle head reverse deformation prediction data to obtain vehicle head compensation area division data;

[0159] Step S44: allocating compensation weights to the compensation areas based on the vehicle head compensation area division data to obtain vehicle head compensation weight allocation data;

[0160] Step S45: performing pre-deformation compensation calculation on the vehicle head according to the vehicle head compensation weight distribution data and the vehicle head reverse deformation prediction data to obtain vehicle head pre-deformation compensation data;

[0161] Step S46: generating a discrete point cloud of the vehicle head surface based on the vehicle head pre-deformation compensation data to obtain vehicle head compensation point cloud data;

[0162] Step S47: reconstructing the compensation surface of the vehicle head by using radial basis function and curvature constraint fusion according to the vehicle head compensation point cloud data to obtain the vehicle head geometric compensation contour data.

[0163] In this embodiment, after the coupled modified front model is completed, its STEP file is imported into Geomagic ControlX. Using the built-in 3D Compare function, the original design surface is used as a reference datum, with a sampling interval of 1mm. The software automatically matches and calculates the deviation of each point from the datum. The statistical range is set to ±5mm, and a statistical table is generated while outputting the cloud map, recording the maximum deformation of 1.73mm, the average deformation of 0.48mm, and the standard deviation of 0.32mm. The deformation of the nose cone and the guide groove area is the most significant. These values ​​are summarized as the front deformation statistics. The obtained deformation statistics are loaded into COMSOL Multiphysics together with the fiberglass shrinkage tensor field. An inverse model is established in the structural mechanics module: the deformation is used as the target displacement, and the shrinkage tensor field is used as the material constitutive input. The required prestrain is inversely calculated using the least squares method. The mesh follows the structural mesh, and the inverse prestrain value of each node is obtained after the solution. The maximum prestrain is 0.23%, corresponding to a displacement of -1.78mm. These values ​​constitute the inverse deformation prediction data of the front. The inverse deformation prediction data was imported into Rhinoceros, and the pre-strain was mapped back to the front surface using the Grasshopper plug-in Human. Thresholds of 0.5mm and 1.5mm were set to divide the surface into high compensation areas (|δ|>1.5mm), medium compensation areas (0.5mm<|δ|≤1.5mm), and low compensation areas (|δ|≤0.5mm). The software automatically colored and calculated the area proportions to 12%, 35%, and 53%, respectively. A list containing region IDs, areas, and center coordinates was generated, representing the front compensation area division data. The compensation area division data was opened in Excel, and weights were assigned to each region based on the CFD sensitivity table: 1.0 for the high compensation area of ​​the nose cone, 0.8 for the middle compensation area of ​​the guide trough, 0.6 for the low compensation area of ​​the wing, and 0.4 for the low compensation area of ​​the tail. The sum of the weights was normalized to 1, and the compensation weight of each element was recalculated using the simple proportional method. The results were written into the same table to form the front compensation weight distribution data. The compensation weight distribution data and the inverse deformation prediction data were imported into MATLAB, and the pre-deformation displacement was calculated using the weighted least squares method: the high-weight area was multiplied by 1.0, the medium-weight area by 0.8, and the low-weight area by 0.6, and then multiplied by the safety factor of 1.1 to obtain the final compensation displacement; the grid node spacing was 2mm, the maximum compensation displacement was -1.95mm, and the minimum was -0.12mm. After the calculation was completed, a complete list of node coordinates and compensation vectors was output, which was the pre-deformation compensation data of the vehicle head.In Rhino, the pre-deformation compensation data was loaded into the Point Deviation plug-in. Points were evenly distributed across the vehicle head surface at a 2mm x 2mm spacing, generating a total of 78,000 discrete points. Each point recorded its original coordinates and its new, compensated coordinates. The software automatically removed points less than 5mm from the boundary, ultimately retaining approximately 75,000 points with a coordinate accuracy of 0.01mm. These points were directly exported in XYZ format to form the vehicle head compensation point cloud data. For the detailed implementation process of step S47, please refer to the substeps of step S47.

[0164] It is particularly important that step S47 further includes the following steps:

[0165] Step S471: performing radial basis function kernel selection on the vehicle head according to the vehicle head compensation point cloud data to obtain the vehicle head RBF kernel function parameters;

[0166] Step S472: Calculating the interpolation weight of the vehicle head based on the RBF kernel function parameters and the vehicle head compensation point cloud data to obtain the vehicle head surface interpolation weight coefficient;

[0167] Step S473: reconstructing the vehicle head surface through continuous compensation curves according to the vehicle head surface interpolation weight coefficients to obtain a vehicle head continuous compensation curve; identifying pressure wave sensitive areas on the vehicle head based on the vehicle head continuous compensation curve to obtain pressure wave sensitive area marking data;

[0168] Step S474: performing local mesh encryption on the pressure wave sensitive area according to the pressure wave sensitive area marking data to obtain an encrypted mesh of the vehicle head pressure wave sensitive area;

[0169] Step S475: performing curvature constraint calculation on the pressure wave sensitive area based on the encrypted grid of the vehicle head pressure wave sensitive area and the vehicle head continuous compensation surface to obtain vehicle head surface curvature constraint data;

[0170] Step S476: performing local curvature optimization on the pressure wave sensitive area according to the curvature constraint data of the vehicle head surface to obtain a local optimized surface of the vehicle head;

[0171] Step S477: Perform global surface fusion on the vehicle head based on the local optimized surface of the vehicle head and the continuous compensation surface of the vehicle head to obtain the vehicle head fusion compensation surface; perform final geometric compensation transformation on the vehicle head based on the vehicle head fusion compensation surface to obtain the vehicle head geometric compensation contour data.

[0172] In this embodiment, the 75,000 compensated point clouds generated in step S46 were imported into Rhino, and the RBF component of the Millipede plug-in was called in Grasshopper, and a multi-quadratic radial basis function was selected. The prediction error was tested in four ranges of c = 2mm, 3mm, 4mm, and 5mm by cross-validation. Finally, c = 3.5mm and a support radius of 15mm were selected to obtain the RBF kernel function parameters for the head and save them as a structured parameter table. Based on the determined c = 3.5mm and the 75,000 point clouds, the linear equation system was constructed in MATLAB using the RBF toolbox. ,in is the distance matrix, is the weight to be sought, To compensate the displacement vector; SVD decomposition is used to solve, and the condition number is controlled at Within 10000, 75,000 interpolation weight coefficients of the front surface were obtained, with a value range of -0.82 to 1.15. The interpolation weight coefficients and RBF kernel function parameters were brought back to Rhino, and the RBF Surface component of Grasshopper was used to sample the entire front area at a spacing of 1mm to reconstruct a continuous compensation surface. Using the CFD results of ANSYS Fluent, the pressure gradient was used on the compensation surface. >500Pa / m is used as the threshold to mark the sensitive area. After connected domain analysis, 12 pressure wave sensitive areas are obtained, accounting for 23% of the front surface area, and the pressure wave sensitive area marking data is generated. The marking data and the continuous compensation surface are imported into ANSYS Fluent Meshing, and the local encryption function is enabled in the sensitive area. The mesh size is refined from the global 5mm to 0.5mm, and the transition layer uses the quadtree subdivision algorithm to maintain a smooth transition. A total of 1.2M additional units are generated in the encrypted area, with a maximum deflection angle of 0.7° and a minimum orthogonal quality of 0.18, to obtain the encrypted mesh of the front pressure wave sensitive area. The encrypted mesh and the continuous compensation surface are loaded into Geomagic Design, and the surface curvature tool is used to calculate the principal curvatures κ1 and κ2 of each node; the constraints are set. <0.01mm⁻¹, <0.01mm⁻¹, the Lagrange multiplier method is used to solve the local surface patch, output the curvature constraint value and the corresponding Lagrange multiplier, and form the curvature constraint data of the head surface. Based on the curvature constraint data, the T-Splines plug-in is used in Rhinoceros to perform local surface optimization on sensitive areas: with the goal of minimizing energy, the constraint boundary maintains G2 continuity, the iteration step size is 0.5mm, and the convergence tolerance is 1× After optimization, the maximum principal curvature of the sensitive area was reduced to 0.008mm⁻¹, resulting in a locally optimized surface for the front end. The locally optimized surface and the continuously compensated surface were imported into Catia's GSD module. Using the Blend Surface function, G2 continuity was set at the boundary, with a transition length of 10mm. After global surface fusion, Check Topology was performed to ensure gap-free surfaces. Finally, the fused and compensated surface for the front end was exported in STEP format with a coordinate accuracy of 0.01mm, representing the geometrically compensated contour data for the front end.

[0173] Preferably, step S5 includes the following steps:

[0174] Step S51: performing virtual layup process simulation on the vehicle head geometry compensation profile data to obtain virtual layup simulation data;

[0175] Step S52: performing a virtual curing simulation on the vehicle head based on the virtual layup simulation data and the fiberglass shrinkage tensor field to obtain the vehicle head virtual curing simulation data;

[0176] Step S53: performing demoulding cooling simulation on the locomotive head according to the locomotive head virtual solidification simulation data to obtain locomotive head demoulding cooling simulation data;

[0177] Step S54: reconstructing the residual stress field of the lathe head based on the lathe head demoulding cooling simulation data to obtain the lathe head residual stress distribution data;

[0178] Step S55: performing final deformation calculation on the vehicle head according to the residual stress distribution data of the vehicle head and the geometric compensation profile data of the vehicle head to obtain the virtual manufacturing result data of the vehicle head;

[0179] Step S56: constructing a streamlined three-dimensional digital model of the vehicle front based on the vehicle front virtual manufacturing result data.

[0180] In this embodiment, the compensated head STEP surface is imported into MSC Apex Iberian Lynx, the Composite module is enabled, and the model is modeled layer by layer according to the [0 / 45 / -45 / 90] 2s laying scheme. The thickness of each layer is 0.25mm, the bandwidth is 12.7mm, and the starting point is taken from the center of the fused surface in step S47. The automatic tape laying machine path algorithm is used to generate a wrinkle-free trajectory, the roller pressure is 0.2MPa, the speed is 30mm / s, the simulation time is 480s, the step size is 0.1s, and the fiber angle, tension and compaction of each layer are output to form virtual laying simulation data. The virtual laying data is imported into Autodesk MoldflowInsight together with the FRP shrinkage tensor field to establish a three-dimensional curing model; the mold temperature is set to 125℃, the heating rate is 2℃ / min, and the heat preservation is 90min. The Kamal-Sourour curing kinetic model is used, and the curing degree is 0. ,in, is the instantaneous curing rate, that is, the increment of curing degree per unit time, The curing degree currently achieved, ranging from 0 to ≤1, 0 means completely uncured, 1 means completely cured, k1 is the primary reaction rate constant, k2 is the secondary autocatalytic reaction rate constant, m describes the order of the autocatalytic reaction, n describes the reaction order of the unreacted resin, k1=8.5 s⁻¹, k2=1.2 s⁻¹, m=0.75, n=1.8; after the calculation is completed, the node temperature, curing degree and shrinkage strain are output to form the virtual curing simulation data of the front of the car. The temperature field after curing is loaded into Abaqus to establish a thermal-structural coupling demolding cooling analysis; the contact surface between the mold and the workpiece adopts hard contact, and the cooling stage is reduced from 125℃ to 25℃ at a rate of 1℃ / min; a transient heat conduction step is used, with a time increment of 10s and a total time of 6000s, to record the node temperature and thermal shrinkage displacement. After the calculation is completed, the displacement field at the moment of demolding is output to form the demolding cooling simulation data of the front of the car. The temperature and displacement field after demolding cooling are imported into COMSOL Multi physics, the viscoelastic mechanics module is activated, and the Maxwell-Wiechert model is used to describe the resin relaxation, with a relaxation time τ=3600s; the curing shrinkage and cooling shrinkage strains are input into the model to calculate the residual stress distribution; the solver is selected as steady state, and the convergence tolerance is 1 The three principal stresses and von Mises stresses at the nodes are output to form the residual stress distribution data for the front of the vehicle. The residual stress distribution data and the compensation contour surface are loaded back into Rhino. The residual stresses are converted into nodal loads using Grasshopper's FEMesh component, and the elastic deformation is recalculated. Material parameters are E1 = 135 GPa, E2 = 8.5 GPa, G12 = 4.5 GPa, and Poisson's ratio is 0.3. The residual stress deformation is superimposed with the geometric compensation displacement to obtain the final displacement of each node. The maximum total deformation of 1.98 mm is located at the nose cone. The results are output as nodal coordinates + displacement, which constitute the virtual manufacturing result data for the front of the vehicle. For the detailed implementation process of step S56, please refer to the substeps of step S56.

[0181] It is particularly important that step S56 further includes the following steps:

[0182] Step S561: arranging virtual measurement points on the vehicle head surface based on the vehicle head virtual manufacturing result data to obtain vehicle head virtual measurement point arrangement data;

[0183] Step S562: performing a virtual laser scanning simulation on the vehicle head according to the arrangement data of the virtual measurement points on the vehicle head to obtain the virtual scanning data of the vehicle head;

[0184] Step S563: performing measurement point cloud filtering based on the vehicle head virtual scanning data to obtain vehicle head purified point cloud data; performing three-dimensional reconstruction of the vehicle head surface based on the vehicle head purified point cloud data to obtain vehicle head virtual point cloud data;

[0185] Step S564: Acquire the vehicle head design target geometry data; perform registration and alignment based on the vehicle head virtual point cloud data and the vehicle head design target geometry data to obtain vehicle head registration transformation data;

[0186] Step S565: performing deviation calculation between the vehicle head virtual point cloud data and the vehicle head design target geometry data according to the vehicle head registration transformation data to obtain vehicle head geometry deviation data;

[0187] Step S566: performing a prediction model accuracy assessment based on the vehicle head geometry deviation data to obtain vehicle head model error assessment data; performing key parameter sensitivity decoupling based on the vehicle head model error assessment data to obtain vehicle head parameter sensitivity data;

[0188] Step S567: constructing a vehicle head model reinforcement learning environment based on the vehicle head parameter sensitivity data; extracting the current model parameter vector based on the vehicle head model error evaluation data and the vehicle head parameter sensitivity data to obtain the vehicle head prediction model parameters;

[0189] Step S568: performing gradient reverse update on the vehicle head prediction model parameters according to the vehicle head model reinforcement learning environment to obtain a streamlined vehicle head three-dimensional digital model.

[0190] In this example, the front-end compensation contour surface was opened in CATIA's Generative Shape Design workbench. Using the Point on Surface command, the spacing in the U and V directions was set to 8mm, with a 4mm spacing in the nose cone and air duct areas, and a 5mm buffer zone extending outward from the edge. After verifying the normal vector using the Measure Item function, the X, Y, and Z coordinates of 5,000 measurement points were batch exported, with three decimal places and units in mm. The coordinate origin was kept consistent with the vehicle coordinates, generating the virtual measurement point layout data for the front end. The measurement point list was then imported into the Virtual Laser Scanner module of PolyWorks|Inspector, using a simulated FARO Focus Premium laser with a 905nm laser wavelength, ±0.02mm ranging accuracy, 0.05° angular step, and a 1MHz scanning frequency. A simulated laser was projected at each point, and the round-trip time was recorded. Gaussian noise with a mean of 0 and a standard deviation of 0.01mm was added to simulate on-site jitter. Finally, virtual scan data containing X, Y, and Z coordinates, intensity I, and normal N were output. Open the virtual scan data in Cloud Compare and first perform Statistical Outlier Removal with 20 neighborhood points and a standard deviation multiplier of 2.0. After removing outliers, approximately 74,600 points remain. Then invoke Moving Least Squares smoothing with a search radius of 3mm and a polynomial order of 2. The smoothed RMS error is less than 0.005mm. Finally, use Poisson Surface Reconstruction with an octree depth of 10 and a sampling density of 8 to generate approximately 1.5 million triangles with an average side length of 1mm, thus obtaining the virtual point cloud data for the vehicle head. The virtual point cloud of the front of the vehicle and the target STEP surface were simultaneously imported into Geo Magic Design. N-point coarse registration was first performed, selecting three points as seeds: the nose tip, side edge, and tail end. ICP fine registration was then initiated, with a maximum of 200 iterations, a convergence threshold of 0.05mm, and a sampling rate of 20%. After registration, the software output a 4×4 homogeneous transformation matrix with a translation of 0.02mm (X), -0.01mm (Y), and 0.04mm (Z), and a rotation angle of <0.05°, generating the front of the vehicle registration transformation data. The registered point cloud and target surface were then imported into GOM Inspect, where point-to-surface distance calculation was selected with a sampling interval of 1mm. The deviation value for each point was output. Statistical results showed a maximum deviation of 2.10mm at the nose cone transition, an average deviation of 0.22mm, and a standard deviation of 0.18mm. A rainbow cloud map was then used to generate the front of the vehicle geometric deviation data, colored within ±0.5mm intervals.The SALib global sensitivity analysis package was called in MATLAB. Ten variables were input: the six principal components of the FRP shrinkage tensor field, a curing temperature of 125°C, a heating rate of 2°C / min, a holding time of 90 minutes, and a ply tension of 50 N. A Sobol sequence was used to generate 1000 Latin hypercube samples. After running the surrogate model, the first-order sensitivity index S1 and the total sensitivity index ST were calculated. The ST for heating rate and ply tension were 0.42 and 0.31, respectively, with all other sensitivity indexes less than 0.10. This yielded the sensitivity data for the headliner parameters. The model error (RMS) was also recorded, which was 0.22 mm. A reinforcement learning environment was constructed in Python using the DDPG algorithm from Stable-Baselines. The state space was 10-dimensional, corresponding to the aforementioned parameters, the action space was continuous [-0.1, +0.1], and the reward function R = -RMS deviation. The network consisted of a three-layer fully connected network with 256 neurons per layer, using the Reluctant Unified Unit (ReLU) activation function, a learning rate of 0.001, a batch size of 256, and an experience replay capacity of 1. The current vehicle head prediction model parameter vector was used as the initial state input to complete environment initialization and initiate training. Training was run for 10,000 rounds, with the parameter vector updated once per round. A discount factor of 0.99 was used, and the ε-greedy strategy was linearly reduced from 1.0 to 0.01. At round 8736, the RMS dropped below 0.05 mm for the first time and remained stable for 100 rounds. The corresponding parameter set was extracted: a heating rate of 1.8°C / min, a ply tension of 47 N, and a fine-tuning of the principal components of the shrinkage tensor by ±2%. The updated parameters were written to CATIA V5-6R2023, and the surface was updated in real time using control point coordinates. The resulting 3D digital model of the streamlined vehicle head with an RMS of 0.047 mm was obtained.

[0191] Preferably, the present invention further provides a three-dimensional digital model generation system based on a fiberglass streamlined front end, which is used to execute the three-dimensional digital model generation method based on a fiberglass streamlined front end as described above. The three-dimensional digital model generation system based on a fiberglass streamlined front end comprises:

[0192] The material property calibration module is used to test the entire hot pressing process of standard FRP samples to obtain the thermal expansion coefficient of FRP materials; and to construct the FRP contraction tensor field based on the thermal expansion coefficient of FRP materials;

[0193] The ply path modeling module is used to construct the ply path topology relationship of the head fiber ply to obtain the head ply path diagram; based on the head ply path diagram, the path-process coupling deformation prediction of the head fiber bundle is performed to obtain the ply path deformation prediction data;

[0194] The aerodynamic-structural coupling correction module is used to obtain an initial three-dimensional model of the vehicle head. Based on the initial three-dimensional model, the fiberglass laminate material properties are spatially distributed according to the layup path deformation prediction data to obtain the steel layer material property distribution data. Based on the steel layer material property distribution data, the initial three-dimensional model of the vehicle head is subjected to aerodynamic-structural bidirectional coupling iterative correction to obtain a coupled corrected vehicle head model.

[0195] The inverse compensation and reconstruction module is used to perform inverse compensation on the coupled modified vehicle head model based on the FRP shrinkage tensor field, generate discrete point clouds, and obtain vehicle head compensation point cloud data; based on the vehicle head compensation point cloud data, the compensation surface of the vehicle head is reconstructed using radial basis function and curvature constraint fusion to obtain vehicle head geometric compensation contour data;

[0196] The virtual manufacturing verification module is used to perform virtual manufacturing simulation on the vehicle head according to the vehicle head geometric compensation contour data to obtain the vehicle head virtual manufacturing result data; and to construct a streamlined vehicle head three-dimensional digital model based on the vehicle head virtual manufacturing result data.

[0197] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0198] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating a three-dimensional digital model based on a fiberglass streamlined front end, characterized in that: The following steps are involved: Step S1: Conducting a full hot pressing forming process test on a standard FRP sample to obtain the thermal expansion coefficient of the FRP material; constructing a FRP contraction tensor field based on the thermal expansion coefficient of the FRP material; Step S2: constructing a topological relationship of the fiber layup path for the headstock to obtain a headstock layup path diagram; performing path-process coupled deformation prediction on the headstock fiber bundle based on the headstock layup path diagram to obtain layup path deformation prediction data; Step S3: Obtaining an initial vehicle head 3D model; Based on the initial three-dimensional model of the front of the vehicle, the properties of the fiberglass laminate are spatially distributed according to the deformation prediction data of the ply path to obtain the distribution data of the material properties of the steel layer. Based on the distribution data of the material properties of the steel layer, the initial three-dimensional model of the front of the vehicle is iteratively corrected by aerodynamic-structural bidirectional coupling to obtain a coupled corrected front of the vehicle model. Step S4: performing inverse compensation and discrete point cloud generation on the coupled modified vehicle head model according to the FRP shrinkage tensor field to obtain vehicle head compensation point cloud data; performing radial basis function-curvature constraint fusion reconstruction on the compensation surface of the vehicle head according to the vehicle head compensation point cloud data to obtain vehicle head geometric compensation contour data; Step S5: performing virtual manufacturing simulation on the vehicle head according to the vehicle head geometric compensation contour data to obtain the vehicle head virtual manufacturing result data; and constructing a streamlined vehicle head three-dimensional digital model based on the vehicle head virtual manufacturing result data.

2. The method for generating a three-dimensional digital model based on a fiberglass streamlined front end according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: measuring the dimensions of the standard FRP sample to obtain the geometric parameters of the FRP sample; Step S12: Designing the fiber layup angles of the standard FRP specimens according to the FRP specimen geometric parameters to obtain a multi-angle fiber layup scheme. The fiber layup angle design specifically includes designing fiber layup schemes according to seven angle gradients: 0°, 15°, 30°, 45°, 60°, 75°, and 90°. Five replicate specimens are prepared for each angle. A prepreg layup process is used, and the fiber volume fraction is controlled at 55% ± 2%. Step S13: Arranging a strain gauge array on the standard FRP specimen based on the fiber multi-angle layup scheme to obtain strain monitoring points; Step S14: installing a thermocouple temperature sensor on the standard FRP sample according to the strain monitoring points to obtain the temperature monitoring points; Step S15: setting hot pressing process parameters for the standard FRP sample based on the temperature monitoring points and the strain monitoring points to obtain FRP forming process parameters; Step S16: performing real-time data acquisition on a standard FRP sample according to the FRP molding process parameters to obtain a FRP strain-temperature parameter set; Step S17: Identifying the strain-temperature relationship of the standard FRP specimen based on the FRP strain-temperature parameter set to obtain the thermal expansion coefficient of the FRP material; Step S18: constructing a FRP contraction tensor field based on the thermal expansion coefficient of the FRP material.

3. The method for generating a three-dimensional digital model based on a fiberglass streamlined front end according to claim 2, characterized in that: Step S18 includes the following steps: Step S181: Calculating the shrinkage strain of a standard FRP specimen based on the FRP material's thermal expansion coefficient and the FRP strain-temperature parameter set to obtain original shrinkage strain data; Step S182: performing fiber orientation angle correlation identification on the shrinkage strain raw data to obtain angle-shrinkage correlation data; Step S183: identifying the principal strain direction of the standard FRP specimen based on the angle-contraction correlation data to obtain the principal strain direction data of the FRP; Step S184: reconstructing the material temperature gradient field of the standard FRP sample according to the FRP principal strain direction data and the temperature monitoring points to obtain the FRP temperature gradient field; Step S185: calculating the material shrinkage tensor components of the standard FRP specimen based on the FRP temperature gradient field and the FRP principal strain direction data to obtain the FRP material shrinkage tensor components; Step S186: performing material contraction behavior space interpolation expansion on the FRP material contraction tensor component to obtain a continuous tensor field of FRP material contraction; Step S187: verifying the shrinkage prediction model of the standard FRP material based on the FRP material shrinkage continuous tensor field to obtain the FRP shrinkage tensor field.

4. The method for generating a three-dimensional digital model based on a fiberglass streamlined front end according to claim 1, characterized in that: The construction of the topological relationship of the laying path for the fiber laying of the headstock in step S2 includes: Acquire the front surface geometry data; extract NURBS surface parameters from the front surface geometry data to obtain the front surface control points; Gaussian curvature of the vehicle head surface is calculated based on the vehicle head surface control points to obtain the curvature distribution data of the vehicle head surface; According to the curvature distribution data of the vehicle head surface, the hyperbolic area of ​​the vehicle head surface is identified to obtain the marking data of the hyperbolic area of ​​the vehicle head; Based on the hyperbolic region marking data of the vehicle head, the vehicle head surface is divided into a geodesic grid to obtain a geodesic grid; the vehicle head surface is evaluated for developability based on the geodesic grid to obtain surface developability evaluation data; Based on the surface developability evaluation data, the front surface is decomposed into quasi-planar units to obtain a set of quasi-planar units on the front surface; The starting point of the fiber layer of the front of the vehicle is planned according to the quasi-planar unit set of the front surface, and the starting point data of the fiber layer of the front of the vehicle is obtained; Based on the starting point data of the fiber layer of the head and the quasi-planar unit set of the head surface, the fiber bundle path of the head is planned to obtain the fiber bundle path of the head; Identify the ply turning point based on the fiber bundle path of the front of the vehicle and obtain the coordinates of the ply turning point of the front of the vehicle; The ply topology relationship is constructed based on the coordinates of the turning points of the front ply and the fiber bundle path of the front ply, and the front ply path diagram is obtained.

5. The method for generating a three-dimensional digital model based on a fiberglass streamlined front end according to claim 1, characterized in that: In step S2, path-process coupled deformation prediction of the headstock fiber bundle based on the headstock layup path diagram includes: The stress state of the fiber bundle of the head is modeled according to the head layup path diagram to obtain the stress state parameters of the fiber bundle tension field; Based on the stress state parameters of the fiber bundle tension field, the curvature response coupling mapping of the fiber bundle of the headstock is performed to obtain the fiber bundle curvature-tension coupling response data; The path memory factor of the fiber bundle of the head is calculated based on the curvature-tension coupling response data of the fiber bundle and the turning point coordinates of the head ply to obtain the path memory factor data; Perform multi-angle visual and mechanical synchronous acquisition of the actual layup process of the vehicle head to obtain layup behavior time series data; Clustering and identifying the layup habit features of layup behavior time series data to obtain the layup habit pattern of the technicians; Based on the path memory factor data, the randomness model of the technician's layup habit pattern is built to obtain the random perturbation parameters of the layup process; The ply deformation of the vehicle head is predicted according to the random disturbance parameters of the ply process and the path memory factor data, and the ply path deformation prediction data is obtained.

6. The method for generating a three-dimensional digital model based on a fiberglass streamlined front vehicle according to claim 1, characterized in that: In step S3, an initial vehicle head 3D model is obtained; Based on the initial 3D model of the vehicle head and the predicted deformation data of the ply path, the spatial distribution of the FRP laminate material properties includes: Obtain an initial three-dimensional model of the vehicle head; discretize the initial three-dimensional model of the vehicle head to obtain a CFD calculation grid for the vehicle head; The boundary conditions of the vehicle head flow field are set based on the vehicle head CFD calculation grid to obtain the boundary conditions of the vehicle head flow field; According to the FRP shrinkage tensor field and the boundary conditions of the vehicle head flow field, the CFD solver is set for the initial vehicle head 3D model to obtain the CFD solution parameters; The front flow field is numerically solved based on the CFD solution parameters to obtain the initial front flow field solution data; The pressure distribution on the vehicle head surface is extracted based on the initial flow field solution data to obtain the pressure distribution data on the vehicle head surface. The pressure distribution extraction specifically includes extracting pressure values ​​at 2156 nodes on the vehicle head surface and calculating the unit pressure using the area-weighted average method. The pressure data accuracy is ±1Pa. Generate FEA structural mesh of the vehicle head based on the surface pressure distribution data of the vehicle head to obtain the vehicle head structural calculation mesh; The properties of the FRP laminated materials are spatially distributed based on the deformation prediction data of the ply path and the calculation grid of the vehicle head structure to obtain the distribution data of the steel layer material properties.

7. The method for generating a three-dimensional digital model based on a fiberglass streamlined front end according to claim 1, characterized in that: In step S3, the aerodynamic-structural bidirectional coupling iterative correction of the initial vehicle head 3D model based on the steel layer material property distribution data includes: Based on the steel layer material property distribution data and the head surface pressure distribution data, the head thermal-mechanical coupling solution is performed to obtain the head thermal stress distribution data; The shrinkage deformation of the vehicle head is calculated based on the thermal stress distribution data of the vehicle head to obtain the deformation response data of the vehicle head structure; the geometry of the vehicle head is updated and transformed based on the deformation response data of the vehicle head structure to obtain the deformed geometry data of the vehicle head; Regenerate the aerodynamic mesh based on the deformed geometric data of the vehicle head to obtain the updated CFD mesh of the vehicle head; Based on the updated CFD mesh of the vehicle head, deformation-aerodynamic sensitivity mapping calculation is performed to obtain the vehicle head deformation sensitivity table; According to the vehicle head deformation sensitivity table, the convergence criterion of the vehicle head aerodynamic-structural coupling iteration is set to obtain the vehicle head coupling convergence criterion parameters; Based on the front coupling convergence criterion parameters, the front aerodynamic-structural coupling iterative solution is performed to obtain the front coupling iterative solution data; The final geometry of the initial three-dimensional vehicle head model is determined based on the vehicle head coupling iterative solution data to obtain the coupled modified vehicle head model.

8. The method for generating a three-dimensional digital model based on a fiberglass streamlined front vehicle according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: performing deformation statistics on the coupled modified vehicle head model to obtain vehicle head deformation statistics; Step S42: performing reverse deformation prediction on the vehicle head based on the fiberglass shrinkage tensor field and the vehicle head deformation statistical data to obtain vehicle head reverse deformation prediction data; Step S43: dividing the vehicle head surface into compensation areas according to the vehicle head reverse deformation prediction data to obtain vehicle head compensation area division data; Step S44: allocating compensation weights to the compensation areas based on the vehicle head compensation area division data to obtain vehicle head compensation weight allocation data; Step S45: performing pre-deformation compensation calculation on the vehicle head according to the vehicle head compensation weight distribution data and the vehicle head reverse deformation prediction data to obtain vehicle head pre-deformation compensation data; Step S46: generating a discrete point cloud of the vehicle head surface based on the vehicle head pre-deformation compensation data to obtain vehicle head compensation point cloud data; Step S47: reconstructing the compensation surface of the vehicle head by using radial basis function and curvature constraint fusion according to the vehicle head compensation point cloud data to obtain the vehicle head geometric compensation contour data.

9. The method for generating a three-dimensional digital model based on a fiberglass streamlined front end according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: performing virtual layup process simulation on the vehicle head geometry compensation profile data to obtain virtual layup simulation data; Step S52: performing a virtual curing simulation on the vehicle head based on the virtual layup simulation data and the fiberglass shrinkage tensor field to obtain the vehicle head virtual curing simulation data; Step S53: performing demoulding cooling simulation on the locomotive head according to the locomotive head virtual solidification simulation data to obtain locomotive head demoulding cooling simulation data; Step S54: reconstructing the residual stress field of the lathe head based on the lathe head demoulding cooling simulation data to obtain the lathe head residual stress distribution data; Step S55: performing final deformation calculation on the vehicle head according to the residual stress distribution data of the vehicle head and the geometric compensation profile data of the vehicle head to obtain the virtual manufacturing result data of the vehicle head; Step S56: constructing a streamlined three-dimensional digital model of the vehicle front based on the vehicle front virtual manufacturing result data.

10. A three-dimensional digital model generation system based on a fiberglass streamlined front end, characterized in that: For executing the method for generating a three-dimensional digital model based on a fiberglass streamlined front end according to claim 1, the three-dimensional digital model generating system based on a fiberglass streamlined front end comprises: The material property calibration module is used to test the entire hot pressing process of standard FRP samples to obtain the thermal expansion coefficient of FRP materials; and to construct the FRP contraction tensor field based on the thermal expansion coefficient of FRP materials; The ply path modeling module is used to construct the ply path topology relationship of the head fiber ply to obtain the head ply path diagram; based on the head ply path diagram, the path-process coupling deformation prediction of the head fiber bundle is performed to obtain the ply path deformation prediction data; The aerodynamic-structural coupling correction module is used to obtain an initial three-dimensional model of the vehicle head. Based on the initial three-dimensional model, the fiberglass laminate material properties are spatially distributed according to the layup path deformation prediction data to obtain the steel layer material property distribution data. Based on the steel layer material property distribution data, the initial three-dimensional model of the vehicle head is subjected to aerodynamic-structural bidirectional coupling iterative correction to obtain a coupled corrected vehicle head model. The inverse compensation and reconstruction module is used to perform inverse compensation on the coupled modified vehicle head model based on the FRP shrinkage tensor field, generate discrete point clouds, and obtain vehicle head compensation point cloud data; based on the vehicle head compensation point cloud data, the compensation surface of the vehicle head is reconstructed using radial basis function and curvature constraint fusion to obtain vehicle head geometric compensation contour data; The virtual manufacturing verification module is used to perform virtual manufacturing simulation on the vehicle head according to the vehicle head geometric compensation contour data to obtain the vehicle head virtual manufacturing result data; and to construct a streamlined vehicle head three-dimensional digital model based on the vehicle head virtual manufacturing result data.

Images