A 3D modeling-assisted garment pattern making method and system
By performing body shape clustering and multi-objective optimization on the target population, and automatically generating sewing process parameters based on stress distribution, the problems of body shape adaptability and sewing consistency in 3D clothing modeling were solved, achieving efficient clothing pattern optimization.
Patent Information
- Application Number
- CN202611336960.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-31
- Publication Date
- 2026-11-17
AI Technical Summary
Existing 3D clothing modeling technology has poor adaptability when facing people with diverse body types. Sewing process parameters are independent of pattern design, resulting in insufficient garment fit and sewing reliability.
By clustering the target population by body type, a representative virtual human body model is constructed. The pattern parameters are adjusted by combining a multi-objective optimization algorithm, and dynamic sewing process parameters are automatically generated based on stress distribution, forming a closed-loop feedback mechanism to optimize the pattern.
It significantly improves the adaptability of ready-made garments to different body types, reduces sewing defects, and ensures consistency between virtual try-on and actual production, as well as the quality of finished products.
Smart Images

Figure CN123118993A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of digital design and manufacturing technology for clothing, specifically a 3D modeling-assisted method and system for clothing pattern making. Background Technology
[0002] Traditional garment pattern making methods primarily rely on pattern makers' empirical formulas and two-dimensional grading rules. They typically use standard human body models or a few sizes as a base, generating different sizes of patterns through proportional scaling. With the development of 3D garment modeling technology, methods have emerged that utilize 3D virtual human models for simulated garment fitting. This involves virtually sewing and fitting the initial pattern in 3D software, observing the fit and pressure distribution between the garment and the body, and then having the pattern maker manually adjust the outline of the cut pieces until a satisfactory visual effect is achieved. Furthermore, some commercial 3D garment design software integrates finite element analysis functions, capable of calculating the stress distribution of garments under static or dynamic conditions. However, these analysis results are usually only used as visual references and do not form a closed loop with subsequent sewing process parameters or automatic pattern correction.
[0003] However, the existing technologies still have significant shortcomings. First, most current 3D-assisted pattern making methods optimize for a single virtual human body model, ignoring the diversity of body types within the target consumer group. This results in poor adaptability of the generated patterns to non-standard body types, making it difficult to improve the fit of finished garments. Second, there is a significant discrepancy between the idealized sewing conditions used in virtual simulation and the actual sewing process, which involves changes in process parameters such as thread tension, stitch density, and presser foot pressure, leading to fabric elongation loss and stress concentration effects. Traditional methods rely solely on simulation results to determine the pattern without considering real sewing constraints, resulting in inconsistencies between the actual finished product and the virtual fitting effect, easily leading to defects such as seam breakage, localized over-tightness, or wrinkling. Furthermore, the determination of pattern parameters and sewing process parameters is independent in the current process, lacking a two-way feedback adjustment mechanism. This prevents the reverse optimization of the pattern outline through process constraints, limiting further improvement in garment quality. Therefore, there is an urgent need for a garment pattern making method that can integrate and couple group body type distribution, 3D simulation stress analysis, and dynamic sewing process parameters. Summary of the Invention
[0004] In order to solve the problems of the prior art, the present invention provides a 3D modeling-assisted garment pattern making method and system.
[0005] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: Firstly, a 3D modeling-assisted garment pattern making method, comprising the following steps: Obtain 3D human body scan data of the target population, extract body shape feature parameters of each person, and construct a human body feature vector set; An unsupervised clustering algorithm is used to cluster the human body feature vector set to obtain several body type clusters, and a representative virtual human body model is constructed for each body type cluster. Based on the preset initial pattern parameters and the fabric parameters in the fabric physical property database, a three-dimensional clothing simulation try-on is performed on the representative virtual human body model to obtain the pressure distribution data and gap data between the clothing and the human body. Using the pressure distribution data and void volume data as optimization objectives, a multi-objective optimization algorithm is employed to adjust the initial pattern parameters, thereby obtaining an optimized pattern parameter set corresponding to each body cluster. The optimized pattern parameter set is input into 3D clothing modeling software to generate a 3D clothing model. The 3D clothing model is then virtually sewn together, and stress distribution data near the sewing stitches is extracted during virtual try-on. Based on the stress distribution data, dynamic sewing process parameters are automatically generated, including sewing thread tension, stitch density, and presser foot pressure. Using the dynamic sewing process parameters as constraints, the three-dimensional garment simulation fitting step is re-executed, and the corrected fabric elongation rate after considering the influence of the sewing process is calculated. Based on the corrected fabric elongation rate, the optimized pattern parameters are adjusted again to obtain the final pattern parameters. The final pattern parameters are output to the garment production equipment.
[0006] In one specific implementation of the first aspect, the body shape feature parameters include: chest circumference, waist circumference, hip circumference, shoulder width, arm length, neck circumference, thigh circumference, and knee circumference; the clustering algorithm adopts a Gaussian mixture model or a K-means algorithm, and the number of clusters is 5 to 12.
[0007] In one specific embodiment of the first aspect, the fabric physical property database includes at least the following parameters: the fabric's elastic modulus E (unit: MPa), shear stiffness G (unit: MPa), and flexural stiffness B (unit: N·mm). 2 Thickness t (in mm), weight per unit area ρ (in g / m²) 2 ) and the sag coefficient D (dimensionless, with a value range of 0~1).
[0008] In one specific implementation of the first aspect, the objective function of the multi-objective optimization algorithm is: minF=w1·P_stress+w2·S_gap+w3·C_prod Wherein, P_stress is the average pressure value of the garment on the human body surface (unit: kPa), S_gap is the average gap between the garment and the human body (unit: mm), C_prod is the production compatibility constraint, which represents the difference in the shape of the cut pieces between different body type clusters when cutting the same batch of fabric; the difference in the shape of the cut pieces is obtained by calculating the average Euclidean distance between the outline curves of the cut pieces corresponding to different body type clusters; w1, w2, and w3 are weighting coefficients, satisfying w1+w2+w3=1.
[0009] In one specific embodiment of the first aspect, the process of extracting the stress distribution data includes: sampling points at equal intervals along the virtual sewing stitch on the three-dimensional clothing model, calculating the maximum principal stress value σ_max (unit MPa) and shear stress value τ (unit MPa) of each sampling point; when σ_max or τ of any sampling point exceeds a preset stress threshold, the generation of dynamic sewing process parameters is triggered.
[0010] In one specific embodiment of the first aspect, the generation rule for the dynamic sewing process parameters is as follows: If σ_max or τ exceeds the threshold, the sewing thread tension value will decrease by 5%~15%, the stitch density will increase by 10%~30%, and the presser foot pressure value will decrease by 10%~20%. If both σ_max and τ are below 50% of the threshold, then the default sewing process parameters are maintained.
[0011] In one specific embodiment of the first aspect, the formula for calculating the corrected fabric elongation is: E'_eff = E_eff × (1 - k·ΔS) Wherein, E_eff is the effective elongation of the original fabric under seamless sewing constraints (dimensionless), ΔS is the local shrinkage rate of the sewing stitch caused by dynamic sewing process parameters (unit mm / mm), and k is the fabric-sewing coupling coefficient (value range 0.1~0.8, pre-calibrated by fabric type and sewing thread specification); the further adjustment and optimization of the pattern parameters refers to: using the corrected fabric elongation E'_eff as a constraint, by adjusting the coordinates of the control points of each piece outline curve in the optimized pattern parameters, regenerating the piece outline, so that the corrected virtual fitting pressure distribution falls back into the preset comfort range.
[0012] In one specific implementation of the first aspect, when the final pattern parameters are output to the garment production equipment, a corresponding sewing process parameter file is also output. This file contains segmented markings of sewing thread tension, stitch length, and presser foot pressure values, which are used to guide automatic sewing machines or provide real-time prompts to sewing workers.
[0013] Secondly, a 3D modeling-assisted garment pattern making system is provided for executing a 3D modeling-assisted garment pattern making method, including: The human body database module is used to store 3D human body scan data and body shape feature parameters; The clustering analysis module is used to cluster the human feature vector set and output a representative virtual human model. Fabric parameter library, used to store and update the physical property parameters of fabrics; The 3D simulation module is used for virtual try-on, stress analysis, and pressure distribution calculation. The multi-objective optimization module is used to adjust pattern parameters and generate an optimized pattern parameter set; The sewing process generation module is used to generate dynamic sewing process parameters based on stress distribution data. The feedback adjustment module is used to convert dynamic sewing process parameters into correct fabric elongation and drive the multi-objective optimization module to make secondary adjustments. The output interface module is used to transmit the final pattern parameters and sewing process parameters to external production equipment.
[0014] In one specific implementation of the second aspect, the feedback adjustment module has a built-in fabric-sewing coupling coefficient database, which obtains the k value under different combinations of fabric types and different sewing thread specifications through experimental calibration; the system also includes a self-learning unit, which is used to dynamically update the k value and the weight coefficients w1, w2, and w3 in the multi-objective optimization algorithm based on the actual wearing pressure test data or rework rate data of finished garments.
[0015] The beneficial effects of this invention are as follows: 1. This invention constructs a pattern generation framework that combines group body shape clustering with multi-objective optimization. First, unsupervised clustering is performed on the 3D human body data of the target population to obtain several representative body shape clusters, and virtual human body models are established for each. Based on this, an initial 3D simulation fitting is conducted in conjunction with the physical properties of the fabric. Pressure distribution, porosity, and the difference in pattern shape between different body shape clusters are used as joint optimization objectives. The optimal pattern parameters matching each body shape cluster are automatically iteratively solved. This technique allows pattern design to no longer rely on a single standard human body or empirical grading rules, but can simultaneously consider the diversity of body shapes within a group and the compatibility of patterns in mass production. This significantly improves the garment's ability to accommodate differences in human body shape and reduces discomfort caused by partial ill-fitting areas.
[0016] 2. After obtaining the optimized pattern, this invention refines the stress distribution near the virtual stitches. When the stress exceeds a preset threshold, it automatically generates dynamic sewing process parameters adapted to the local stress level, including thread tension, stitch density, and presser foot pressure. Then, using these sewing parameters as constraints, it re-executes the 3D simulation, calculates the local shrinkage rate of the sewing stitches to obtain the corrected effective fabric elongation, and uses this as a basis to adjust the control point coordinates of the cut piece contour curve, achieving a secondary correction of the pattern parameters. This closed-loop feedback mechanism incorporates the influence of sewing technology on the actual mechanical behavior of the garment into the pattern optimization process, effectively eliminating fabric performance loss and stress concentration defects caused by the separation of process and design in traditional pattern making. The final output pattern and sewing parameter combination can simultaneously ensure the fit of the virtual fitting and the sewing reliability and durability of the finished product in actual production, forming an adaptive optimization chain from population data to production process. Attached Figure Description
[0017] Figure 1 This is the overall flowchart of the present invention.
[0018] Figure 2 This is a block diagram of the module composition of the 3D modeling-assisted garment pattern making system of the present invention.
[0019] Figure 3 This is a schematic diagram illustrating the feedback adjustment between the pattern parameters, stress distribution, and sewing process of this invention. Detailed Implementation
[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0021] like Figures 1 to 3 This paper presents a 3D modeling-assisted garment pattern making method and system.
[0022] I. Overview of the Overall Technical Solution This invention provides a 3D modeling-assisted garment pattern making method and system. The core idea is to perform cluster analysis on the body shape data of the target population to obtain several representative body shape clusters, and then optimize the pattern parameters for each cluster. Based on this, 3D simulation is used to extract the stress distribution near virtual sewing stitches, dynamically generating sewing process parameters. These parameters are then used as constraints for resimulation to calculate the corrected fabric elongation, and subsequently, the pattern piece outline is adjusted in reverse, forming a closed-loop optimization. This method can significantly improve the adaptability of garments to different body shapes while reducing sewing defects.
[0023] II. Specific Implementation Steps Step 1: Human body data collection and clustering.
[0024] Input: 3D human body scan data of the target population, containing at least 500 valid samples. Scanning equipment can be a handheld 3D scanner or a fixed human body scanning chamber.
[0025] Processing: Body shape feature parameters are automatically extracted from each scanned model, including: chest circumference (in cm), waist circumference, hip circumference, shoulder width, arm length, neck circumference, thigh circumference, and knee circumference. The feature vector is represented as: x_i=[bust_i,waist_i,hip_i,shoulder_i,arm_i,neck_i,thigh_i,knee_i].
[0026] Z-score standardization is applied to all feature vectors to eliminate the influence of dimensions.
[0027] Clustering was performed using a Gaussian Mixture Model (GMM), with the number of clusters set to 6. The Expectation-Maximization (EM) algorithm was used iteratively to determine the optimal number of clusters, and the Bayesian Information Criterion (BIC) was used to determine this optimal number. In this embodiment, the BIC reached its minimum value when the number of clusters k=6.
[0028] The centroid of each body size cluster (i.e., the mean of all eigenvectors within the cluster) is used as the representative human body size for that cluster. This size is then used to construct a representative virtual human body model using parametric human modeling software (MakeHuman or a Blender plugin). The model surface is a triangular mesh with approximately 20,000 vertices, and includes a standard pose (standing, arms slightly outstretched at 15°).
[0029] Output: Representative virtual human models of 6 body type clusters, denoted as H_c (c=1,…,6). Each model is accompanied by the corresponding feature parameter range (e.g., cluster 1: chest circumference 82-86cm, waist circumference 64-68cm, hip circumference 90-94cm).
[0030] Step 2: Setting Initial Pattern Parameters and Fabric Properties Input: A basic pattern drawing (DXF format) of the design (e.g., high-waisted straight-leg jeans), including front, back, and waistband pieces. The basic pattern is the industry-standard size M (waist 68cm, hip 92cm).
[0031] Pattern parametricization: Define the outline of the cut pieces as a parametric curve. Key control points include: Waistline length (L_waist); Hip circumference length (L_hip); Front crotch depth (D_front); Crotch depth (D_back); The curvature of the side seam (determined by the coordinates of the control points of the Bézier curve).
[0032] The initial template parameter vector is denoted as: P0=[L_waist0,L_hip0,D_front0,D_back0,…].
[0033] Fabric physical property database: Selected cotton stretch denim (98% cotton + 2% spandex), the following parameters were measured using the Kawabata Evaluation System (KES-FB): Elastic modulus E: 45MPa (warp), 32MPa (weft); Shear stiffness G: 2.8 MPa; Bending stiffness B: 0.12 N·mm 2 ; Thickness t: 0.45mm; Weight per unit area ρ: 360g / m² 2 ; Overhang coefficient D: 0.65 (dimensionless); Effective elongation E_eff: 0.18 (i.e., 18% elongation); All parameters are stored in the fabric parameter library and indexed by fabric ID.
[0034] Output: Initial pattern parameter set P0 and fabric attribute set F.
[0035] Step 3: Initial 3D Simulation and Multi-Objective Optimization Input: Representative virtual human body model H_c, initial pattern parameters P0, fabric properties F.
[0036] Simulation settings: Virtual sewing and fitting were performed using CLO 3D or Browzwear VStitcher software. The simulation engine adopted a mass-spring model with a time step of 0.01 seconds and a simulation duration of 2 seconds to ensure the stability of the garment's gravity.
[0037] Data extraction: Pressure distribution data: Calculate the normal contact pressure on the triangular mesh nodes on the human body surface, and take the average value of all nodes to obtain P_stress (unit kPa).
[0038] Gap data: Calculate the shortest distance between the inner surface of the garment and the outer surface of the human body, and take the average of all distances to obtain S_gap (unit: mm).
[0039] Multi-objective optimization: The non-dominated sorting genetic algorithm (NSGA-II) with an elitist strategy is used to optimize the pattern parameters. The population size is 50, the number of generations is 30, the crossover probability is 0.9, and the mutation probability is 0.1. The objective function is: min F=w1·P_stress+w2·S_gap+w3·C_prod Where: P_stress: average pressure, comfort range 0.5~1.5kPa, the smaller the target value, the more relaxed the comfort.
[0040] S_gap: Average gap, typically ≤2mm for tight-fitting styles and 5~10mm for straight-leg styles. The target for this embodiment is 6~8mm.
[0041] C_prod: Production compatibility constraint, calculated using the following formula: C_prod=(1 / N)Σ_{i=1}^{N}d_i, d_i is the average Euclidean distance (calculated over all sampling points) between the current body type cluster and the baseline cluster (the cluster with the most samples) on the i-th piece outline curve, and N is the total number of pieces. The smaller this value is, the smaller the difference in pattern between different clusters, which is more conducive to batch cutting.
[0042] Weighting coefficients are set as follows: w1=0.3, w2=0.4, w3=0.3 (these can be adjusted according to brand positioning; for example, w1 can be increased for sportswear). After optimization, the optimized pattern parameter set P_opt_c for each body type cluster is obtained.
[0043] Output: 6 sets of optimized pattern parameters P_opt_c, and the corresponding 3D clothing model G_c.
[0044] Step 4: Stress extraction and dynamic sewing parameter generation.
[0045] Input: 3D clothing model G_c, optimized pattern parameter P_opt_c.
[0046] Stress analysis: Identify all virtual sewing stitches (such as front and back side seams, crotch seam) on G_c.
[0047] Sampling points were set along each sewing stitch at equal arc length intervals of 2 mm.
[0048] For each sampling point, the finite element stress tensor is calculated, and the maximum principal stress σ_max (unit MPa) and shear stress τ (unit MPa) are extracted.
[0049] Threshold setting: Maximum principal stress threshold σ_thr = 0.8MPa (the seam breaking strength of cotton denim is about 1.2MPa, leaving a safety margin). The shear stress threshold τ_thr = 0.4 MPa; Dynamic sewing parameter generation rules: If any sampling point satisfies σ_max>σ_thr or τ>τ_thr, then an adjustment is triggered: Sewing thread tension: Reduced by 12% from the default 50 cN (centine Newtons) to 44 cN; Stitch density: Increased by 25% from the default 8 stitches / cm to 10 stitches / cm; Presser foot pressure: Reduced by 15% from the default 30N to 25.5N; If σ_max and τ at all sampling points are less than 50% of the corresponding threshold (i.e., σ_max < 0.4 MPa and τ < 0.2 MPa), then the default parameters are maintained.
[0050] If the value falls between these two, linear interpolation should be used for adjustment.
[0051] The adjustment range (5%~15%, 10%~30%, 10%~20%) is selected from specific values; in this embodiment, the middle value is used.
[0052] Output: A dynamic sewing process parameter file, Sew_par_c, containing segment parameters for each sewing stitch. For example, high-stress areas such as the crotch seam are marked separately; specific parameters are shown in the table below. ; Step 5: Feedback adjustment and secondary version correction.
[0053] Input: Dynamic sewing process parameter Sew_par_c, original fabric attribute F, optimized pattern parameter P_opt_c.
[0054] Correcting fabric stretch calculation: First, the fabric-sewing coupling coefficient k was calibrated experimentally. Taking cotton-elasto denim and polyester sewing thread (size 40 / 2) as examples, five sets of samples were prepared, and the stitch shrinkage rate under different sewing tensions was measured. The fitted result was k=0.35 (dimensionless).
[0055] The local shrinkage rate ΔS of the sewing stitch is determined by a combination of dynamic parameters. In this embodiment, due to a 12% reduction in tension and a 25% increase in stitch density, the empirical formula is: ΔS = α·(ΔT / T0) + β·(ΔD / D0) Where ΔT / T0 is the relative change in tension (-0.12), ΔD / D0 is the relative change in needle spacing density (+0.25), and the coefficients α=0.4 and β=0.6. The calculated ΔS=0.4*(-0.12)+0.6*0.25=-0.048+0.15=0.102 (i.e., shrinkage rate 10.2%).
[0056] The corrected effective elongation is: E'_eff=E_eff×(1-k·ΔS)=0.18×(1-0.35×0.102)=0.18×(1-0.0357)=0.1736 This represents a decrease of approximately 3.6% compared to the original E_eff=0.18.
[0057] Second version revision: Using E'_eff as the new fabric elongation constraint, the pattern parameter optimization is re-executed. The optimization objective remains unchanged, but the initial solution is set to P_opt_c, and only local searches are performed (such as using gradient descent or differential evolution) to reduce computational load. The goal is to adjust the average pressure P_stress after resimulation back to the preset comfort range (0.8~1.2 kPa in this embodiment).
[0058] The specific operation involves adjusting the coordinates of the control points on the fabric piece's outline curve, especially the curvature of the side seams and crotch seams. For example, due to reduced elongation, the outer seam of the trouser leg needs to be shifted outward by 0.5~1mm to compensate for the loss of elasticity. Control point adjustment employs parametric inverse mapping: a neural network surrogate model (3 layers, 32 neurons per layer) is established to map pattern parameters to pressure distribution, and gradient corrections are calculated through backpropagation.
[0059] After 3 to 5 iterations, the model converges, yielding the final version parameter P_final_c.
[0060] Output: Final pattern parameter P_final_c and corresponding cut piece outline data (DXF format).
[0061] Step Six: Output to production equipment.
[0062] Output content: Final pattern parameter P_final_c: Converted into a CAD / CAM readable pattern file (supports DXF, AAMA, and ASTM formats).
[0063] Dynamic sewing process parameter file: Utilizing a structured data format (such as XML or JSON), it contains segmented, annotated sewing parameters. For clarity, the table below lists the parameters for different areas in the example: ; Layout diagram: Based on the final cut piece shape, the layout is arranged within the fabric width using algorithms, and the layout instructions are output to the automatic cutting bed (such as Lectra Vector iX2).
[0064] III. System Module Description.
[0065] The 3D modeling-assisted garment pattern making system of the present invention includes the following modules: Human Body Database Module: Stores raw 3D scan data and standardized feature vectors. The database uses MySQL and supports CRUD operations as well as batch export.
[0066] Clustering analysis module: This module uses Python's scikit-learn library to implement GMM clustering and outputs size clusters and representative models.
[0067] Fabric parameter library: Stores fabric IDs, physical property parameters, and corresponding k-values in NoSQL format. Supports experimental data import and automatic calibration.
[0068] 3D simulation module: Integrates CLO 3D API or open-source simulation engine, providing post-processing interfaces such as pressure and stress.
[0069] Multi-objective optimization module: Implements the NSGA-II algorithm and can be coupled iteratively with the simulation module. The convergence criterion is that the Pareto front change is less than 1% for 10 consecutive generations.
[0070] Sewing process generation module: Generates Sew_par based on stress threshold and custom rule table.
[0071] Feedback adjustment module: includes a fabric-sewing coupling coefficient database (the k-values for different fabric / sewing thread combinations are obtained through experimental calibration) and a reverse pattern correction algorithm.
[0072] Output interface module: Converts to a common format for production equipment, supports OPC UA protocol and communication with Industrial Internet of Things (IIoT).
[0073] In addition, the system includes a self-learning unit. This unit collects data such as rework rates and customer fit ratings after actual production, and uses Bayesian optimization to dynamically adjust the weight coefficients w1, w2, w3, and the fabric-sewing coupling coefficient k in the multi-objective optimization. For example, if the rework rate of a certain body type cluster is too high, the w1 (pressure weight) of that cluster is increased to prioritize improving comfort; if the shrinkage rate of a certain fabric deviates significantly from the calibration value during actual sewing, the corresponding k value is automatically corrected.
[0074] Example Example: Designing high-waisted straight-leg jeans for women with pear-shaped figures Background: Data from an e-commerce brand shows that about 40% of women aged 25-35 have a "pear-shaped body" (larger hips and relatively smaller waist). The return rate of standard size M jeans is as high as 32%, mainly because "the waist fits but the hips are too tight" or "the hips fit but the waist is too loose".
[0075] Implementation process: Data Clustering: 3D scan data of 2000 women in this age group were collected. Using GMM clustering (k=6), a "pear-shaped body cluster" (cluster number C2) was obtained, with the following characteristic parameters: waist circumference 74±2 cm, hip circumference 96±2 cm, and waist-to-hip ratio 0.77. A representative virtual human body model H_C2 was constructed.
[0076] Initial pattern: The initial pattern P0 is based on a commercially available size M jeans (waist 74 cm, hip 96 cm). The fabric used is cotton stretch denim (E=45 MPa, E_eff=0.18).
[0077] Initial optimization: Simulation of H_C2 revealed the following pressure distribution under the initial pattern: P_stress at the back of the hips = 2.3 kPa (too high), and S_gap = 3 mm (too tight). NSGA-II optimization was performed with w1 = 0.3, w2 = 0.4, and w3 = 0.3, resulting in P_opt: waist circumference increased to 76 cm, hip circumference increased to 98 cm, back crotch depth increased by 1.2 cm, and side seam widened by 0.8 cm. After optimization, P_stress decreased to 1.1 kPa, and S_gap increased to 7 mm.
[0078] Stress extraction: Simulation revealed that σ_max at the crotch seam reached 0.92 MPa, exceeding the threshold of 0.8 MPa. Dynamic sewing parameters were triggered: tension decreased from 50 cN to 44 cN, stitch length increased from 8 stitches / cm to 10 stitches / cm, and presser foot pressure decreased from 30 N to 25.5 N.
[0079] Feedback Correction: Given k=0.35, ΔS=0.102, calculate E'_eff=0.18×(1-0.0357)=0.1736. With a 3.6% reduction in elongation as a constraint, reverse the pattern correction: shift the side seam control point outward by 0.8 mm, and reduce the curvature of the back crotch seam curve by 5%. After resimulation, P_stress=1.15 kPa (still within the comfortable range), and the back crotch seam σ_max drops to 0.75 MPa, below the threshold.
[0080] Output: Generates the final pattern DXF file and a sewing process parameter table (as shown in the previous section). The brand used this pattern to produce 500 pairs of jeans. After launching them on the market, the return rate dropped to 18%, and customer satisfaction ratings rose from 3.7 to 4.4 (out of 5).
[0081] Comparative experiment: The control group used traditional grading (M size unidirectional scaling to L size), with the same fabric and process. The results showed that the average hip pressure of the control group on pear-shaped figures was 1.9 kPa, while that of the present invention was 1.1 kPa, an improvement of 42% in comfort. At the same time, the control group showed 3% stitch breakage at the back crotch seam during actual wear, while the present invention showed 0%.
[0082] Summary of parameter value ranges (applicable to common woven fabrics): ; Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A 3D modeling-assisted garment pattern making method, characterized in that, Includes the following steps: Obtain 3D human body scan data of the target population, extract body shape feature parameters of each person, and construct a human body feature vector set; An unsupervised clustering algorithm is used to cluster the human body feature vector set to obtain several body type clusters, and a representative virtual human body model is constructed for each body type cluster. Based on the preset initial pattern parameters and the fabric parameters in the fabric physical property database, a three-dimensional clothing simulation try-on is performed on the representative virtual human body model to obtain the pressure distribution data and gap data between the clothing and the human body. Using the pressure distribution data and void volume data as optimization objectives, a multi-objective optimization algorithm is employed to adjust the initial pattern parameters, thereby obtaining an optimized pattern parameter set corresponding to each body cluster. The optimized pattern parameter set is input into 3D clothing modeling software to generate a 3D clothing model. The 3D clothing model is then virtually sewn together, and stress distribution data near the sewing stitches is extracted during virtual try-on. Based on the stress distribution data, dynamic sewing process parameters are automatically generated, including sewing thread tension, stitch density, and presser foot pressure. Using the dynamic sewing process parameters as constraints, the three-dimensional garment simulation fitting step is re-executed, and the corrected fabric elongation rate after considering the influence of the sewing process is calculated. Based on the corrected fabric elongation rate, the optimized pattern parameters are adjusted again to obtain the final pattern parameters. The final pattern parameters are output to the garment production equipment.
2. The method according to claim 1, characterized in that, The body shape features include: chest circumference, waist circumference, hip circumference, shoulder width, arm length, neck circumference, thigh circumference, and knee circumference; the clustering algorithm adopts Gaussian mixture model or K-means algorithm, and the number of clusters is 5 to 12.
3. The method according to claim 1, characterized in that, The fabric physical property database includes at least the following parameters: elastic modulus E (MPa), shear stiffness G (MPa), and flexural stiffness B (N·mm). 2 Thickness t (in mm), weight per unit area ρ (in g / m²) 2 ) and the overhang coefficient D (dimensionless, with a value range of 0~1).
4. The method according to claim 1, characterized in that, The objective function of the multi-objective optimization algorithm is: minF=w1·P_stress+w2·S_gap+w3·C_prod Wherein, P_stress is the average pressure value of the garment on the human body surface (unit: kPa), S_gap is the average gap between the garment and the human body (unit: mm), C_prod is the production compatibility constraint, which represents the difference in the shape of the cut pieces between different body type clusters when cutting the same batch of fabric; the difference in the shape of the cut pieces is obtained by calculating the average Euclidean distance between the outline curves of the cut pieces corresponding to different body type clusters; w1, w2, and w3 are weighting coefficients, satisfying w1+w2+w3=1.
5. The method according to claim 1, characterized in that, The process of extracting the stress distribution data includes: sampling points at equal intervals along the virtual sewing stitch on the three-dimensional clothing model, calculating the maximum principal stress value σ_max (unit MPa) and shear stress value τ (unit MPa) of each sampling point; when σ_max or τ of any sampling point exceeds the preset stress threshold, the generation of dynamic sewing process parameters is triggered.
6. The method according to claim 5, characterized in that, The generation rules for the dynamic sewing process parameters are as follows: If σ_max or τ exceeds the threshold, the sewing thread tension value will decrease by 5% to 15%, the stitch density will increase by 10% to 30%, and the presser foot pressure value will decrease by 10% to 20%. If both σ_max and τ are below 50% of the threshold, then the default sewing process parameters are maintained.
7. The method according to claim 1, characterized in that, The formula for calculating the elongation of the modified fabric is as follows: E'_eff = E_eff × (1 - k·ΔS) Wherein, E_eff is the effective elongation of the original fabric under seamless sewing constraints (dimensionless), ΔS is the local shrinkage rate of the sewing stitch caused by dynamic sewing process parameters (unit mm / mm), and k is the fabric-sewing coupling coefficient (value range 0.1~0.8, pre-calibrated by fabric type and sewing thread specification); the further adjustment and optimization of the pattern parameters refers to: using the corrected fabric elongation E'_eff as a constraint, by adjusting the coordinates of the control points of each piece outline curve in the optimized pattern parameters, the piece outline is regenerated, so that the corrected virtual fitting pressure distribution falls back into the preset comfort range.
8. The method according to claim 1, characterized in that, When the final pattern parameters are output to the garment production equipment, the corresponding sewing process parameter file is also output. This file contains segmented markings of sewing thread tension, stitch length, and presser foot pressure values, which are used to guide automatic sewing machines or provide real-time prompts to sewing workers.
9. A 3D modeling-assisted garment pattern making system, used to perform the method according to any one of claims 1 to 8, characterized in that, include: The human body database module is used to store 3D human body scan data and body shape feature parameters; The clustering analysis module is used to cluster the human feature vector set and output a representative virtual human model. Fabric parameter library, used to store and update the physical property parameters of fabrics; The 3D simulation module is used for virtual try-on, stress analysis, and pressure distribution calculation. The multi-objective optimization module is used to adjust pattern parameters and generate an optimized pattern parameter set; The sewing process generation module is used to generate dynamic sewing process parameters based on stress distribution data. The feedback adjustment module is used to convert dynamic sewing process parameters into correct fabric elongation and drive the multi-objective optimization module to make secondary adjustments. The output interface module is used to transmit the final pattern parameters and sewing process parameters to external production equipment.
10. The system according to claim 9, characterized in that, The feedback adjustment module has a built-in fabric-sewing coupling coefficient database, which obtains the k value under different combinations of fabric types and different sewing thread specifications through experimental calibration. The system also includes a self-learning unit, which dynamically updates the k value and the weight coefficients w1, w2, and w3 in the multi-objective optimization algorithm based on the actual wearing pressure test data or rework rate data of finished garments.
