Intelligent execution system and method for small single fast response-oriented clothing customization order

Abstract

The present application relates to the technical field of intelligent manufacturing and garment digital production, and particularly discloses a garment customization order intelligent execution system and method for small single fast response, which comprises an order analysis unit, a fabric physical property sensing unit, a virtual physical layout unit, a cutting control unit and a feedback optimization unit. The system constructs a fabric virtual physical field by fusing computational fluid dynamics and generative adversarial networks, simulates the deformation behavior of the fabric in the cutting bed environment, generates a layout scheme containing a non-uniform physical deformation compensation area, and dynamically adjusts and controls the cutting parameters to adapt to the fabric properties. Meanwhile, the system continuously optimizes the model based on a closed-loop feedback. The present application improves the first-order qualified rate, material utilization rate and production response speed of small-batch orders.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing and digital production technology of clothing, specifically relating to an intelligent execution system and method for small-batch, fast-response clothing customization orders. Background Technology

[0002] With the rapid evolution of the fashion consumer market, the small-batch, rapid-response model has become a core driving force for the transformation and upgrading of the apparel manufacturing industry. This model requires production systems with extremely high responsiveness and flexible production capabilities to adapt to the market demands for diverse varieties, small batches, and personalized customization. In the wave of digital transformation, intelligent execution systems, by integrating the Industrial Internet of Things (IIoT) and advanced algorithms, have achieved end-to-end collaboration from design to production, improving the operational efficiency and market flexibility of the apparel supply chain.

[0003] Intelligent pattern making and cutting, as key execution steps in the garment customization process, directly determine the utilization rate of raw materials and the quality of the finished garment. These key execution steps aim to optimally arrange irregular garment pieces within a limited fabric width using complex geometric topology algorithms, combined with automated control of the cutting equipment to achieve precise cutting. With the increasing diversity of fabrics, modern execution systems not only need to handle complex geometric splicing logic but also need to adapt to the processing characteristics of different materials in real time to ensure production standardization and precision stability in mass customization scenarios.

[0004] Traditional nesting systems primarily rely on static geometric calculations for optimization, often neglecting the differences in physical properties between different batches of fabric in small-batch, high-response production. This approach, focusing solely on shape fitting while neglecting dynamic parameters such as material elasticity, drape, and heat shrinkage, makes the cut pieces highly susceptible to physical deformation during actual production. Due to the lack of precise simulation of the fabric's microscopic stress state, the system cannot predict tension changes during fabric laying and cutting, leading to increased dimensional deviations in the finished garment and a high rework rate. Furthermore, nesting schemes relying on linear logic struggle to handle the nonlinear coupling relationships between multidimensional physical variables, failing to compensate for deformation while ensuring material utilization. This results in a low first-order pass rate in applications with frequent fabric changes, becoming a technical challenge hindering the rapid execution of customized orders. Therefore, an intelligent execution system and method for small-batch, high-response garment customization orders are needed. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent execution system for small-batch, fast-response garment customization orders, which can solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: an intelligent execution system for small-batch, fast-response garment customization orders, comprising: The order parsing unit is configured to receive and parse customized order information from the client, extract garment style, size specifications, cut piece outline data and fabric type identifier, and distribute the structured information to the subsequent processing unit. The fabric physical property sensing unit is configured to call a pre-stored fabric physical parameter database according to the fabric type identifier, obtain the elasticity coefficient, drape index, heat shrinkage rate and surface friction characteristics of the corresponding batch of fabric, and dynamically calibrate the parameters in combination with real-time sensing data. The virtual physical layout unit is configured to integrate computational fluid dynamics algorithms and generative adversarial network models on the basis of geometric layout to construct a virtual physical field of the fabric in the cutting bed environment, simulate its deformation behavior under the action of gravity, tension and external force of fabric laying, and generate a layout scheme including a physical deformation compensation zone accordingly. The cutting control unit is configured to receive the fabric layout scheme, convert it into motion trajectory instructions that can be executed by the cutting bed equipment, and simultaneously adjust the cutting blade pressure, travel speed and adsorption strength to adapt to the physical properties of the current fabric and ensure the edge accuracy of the cut piece and the uniform distribution of internal stress. The feedback optimization unit is configured to collect actual cut piece size deviation data, sewn product size qualification rate, and rework records after cutting, update the training sample set of the generative adversarial network through a closed-loop learning mechanism, and optimize the boundary condition parameters in the computational fluid dynamics model.

[0007] Preferably, the virtual physical fabric laying unit models the fabric as a continuous medium field with viscous and tensile properties, and calculates the displacement and strain field distributions of the fabric during static laying and dynamic stretching based on a simplified form of the Navier-Stokes equations.

[0008] Preferably, the generative adversarial network in the virtual physical nesting unit consists of a generator and a discriminator. The generator takes the geometric cut piece outline and fabric physical parameters as input and outputs an initial nesting topology that satisfies the process constraints. The discriminator judges the authenticity of the initial nesting topology based on the operational experience data of historical excellent nesters and drives the generator to iteratively optimize until a highly feasible nesting scheme is generated.

[0009] Preferably, the physical deformation compensation area is a non-uniformly expanded area added to the outer edge of the original cut piece outline. Its expansion amount is dynamically adjusted according to the local strain gradient. A larger compensation amount is set in the high sag area and a smaller compensation amount is set in the low tension area to offset the size shrinkage of the cut piece caused by physical relaxation during subsequent sewing.

[0010] Preferably, the fabric physical property sensing unit integrates a near-infrared spectroscopy analysis module and a micro-force sensing array, which can detect the fiber density, moisture content and surface tension of the fabric in real time during the fabric feeding stage, and compare the detection results with the pre-stored parameters. If the deviation exceeds the preset threshold, the parameter recalibration process is triggered.

[0011] Preferably, the cutting control unit establishes a real-time communication link with the servo drive system of the cutting bed equipment, and dynamically adjusts the downward pressure of the cutting blade in different areas according to the local stress distribution map output by the virtual physical nesting unit, so as to ensure that an appropriate clamping force is applied in the high elasticity fabric area to suppress the rebound deformation.

[0012] Preferably, the feedback optimization unit uses the first order pass rate after each order execution as the core evaluation indicator. When the pass rate is lower than a predetermined threshold, the model retraining program is automatically started, and the newly added failure case data is used to enhance the generative adversarial network's generalization ability to abnormal physical scenarios.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. The intelligent execution system for small-batch, high-response garment customization orders provided by this invention breaks through the limitations of traditional pattern making systems that rely solely on geometric fitting, and for the first time deeply integrates computational fluid dynamics and generative artificial intelligence. By constructing a virtual physical field of the fabric, the system accurately predicts its deformation behavior in a real cutting environment and automatically generates a pattern making scheme with physical compensation, solving the problem of uncontrolled cutting precision caused by differences in the physical properties of the fabric.

[0014] 2. Relying on the efficient learning ability of generative adversarial networks to learn from human expert experience, the system can quickly generate highly feasible layouts that meet complex process constraints while ensuring material utilization, thereby improving the first-order pass rate and production response speed in small-batch, multi-variety order scenarios.

[0015] 3. The closed-loop feedback mechanism ensures the system has continuous evolution capabilities, and the physical model and material layout strategy can be continuously optimized as production data accumulates, providing a new generation of execution architecture with precision, flexibility and intelligence for intelligent garment manufacturing. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall technical solution architecture according to the present invention; Figure 2 This is a schematic diagram of the core principle framework of virtual physical material discharge that integrates computational fluid dynamics and generative adversarial networks according to the present invention. Figure 3 This is a flowchart illustrating the logical flow of fabric physical property sensing and dynamic parameter calibration according to the present invention. Figure 4This is a schematic diagram illustrating the multi-level interaction relationship and data flow between the virtual physical nesting scheme and the cutting control instructions according to the present invention; Figure 5 This is a schematic diagram illustrating the closed-loop optimization principle of model parameters based on deviation feedback according to the present invention. Detailed Implementation

[0017] Example 1: Please refer to the appendix Figure 1 To be continued Figure 5 To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0018] The intelligent execution system for small-batch, fast-response garment customization orders includes an order parsing unit, a fabric physical property sensing unit, a virtual physical nesting unit, a cutting control unit, and a feedback optimization unit. The order parsing unit receives and parses customized order information from the client, extracts garment style, size specifications, cut piece outline data, and fabric type identifiers, and distributes the structured information to subsequent processing units. The order parsing unit is deployed within an order management server with high-speed computing capabilities, and its input establishes a data communication link with the remote client via an encrypted hypertext transfer protocol. The order parsing unit internally includes an order message interception subunit, a semantic recognition subunit, and a geometric data extraction subunit. The order message interception subunit is configured to monitor asynchronous requests from the client in real time. When it captures a structured object containing customized requirements, it uses a cyclic redundancy check algorithm to verify data integrity, ensuring that no bits are lost or tampered with during order information transmission. The semantic recognition subunit is equipped with a dedicated ontology knowledge base for the apparel industry. It is configured to use natural language processing technology to perform feature word matching on the unstructured text in the order, identifying garment style categories (such as shirts, suits, dresses, etc.) and descriptive terms related to fabric materials. The geometric data extraction subunit is specifically responsible for parsing the vector graphics file associated with the order, converting the sequence of outline coordinate points of the two-dimensional cut pieces into a standardized polygon vertex matrix, and calculating the area of ​​the initial envelope rectangle of each cut piece, providing basic geometric parameters for subsequent layout calculations.

[0019] The fabric physical property sensing unit is used to retrieve the elasticity coefficient, drape index, heat shrinkage rate, and surface friction characteristics of the corresponding batch of fabric from a pre-stored fabric physical parameter database based on the fabric type identifier, and dynamically calibrate the parameters in conjunction with real-time sensing data. The fabric physical property sensing unit is physically deployed at the inlet end of the fully automatic fabric spreading machine. It not only contains a static relational database storing tens of thousands of fabric attributes, but also integrates a high-precision multimodal sensor array. The sensor array includes a near-infrared spectroscopy analysis module, a micro-force sensor array, and an environmental temperature and humidity compensation module. The near-infrared spectroscopy analysis module is configured to irradiate the fabric surface with infrared light within a specific wavelength range during fabric spreading, and analyze the fiber composition ratio, yarn twist, and moisture content percentage of the fabric by collecting the absorption band characteristics of the reflected spectrum. The micro-force sensor array is embedded in the support position of the spreading roller shaft to sense the minute tangential tension generated by the fabric during traction in real time, thereby calculating the elastic modulus change curve of the fabric at different elongation rates. The control logic inside the fabric physical property sensing unit is configured such that when the absolute value of the deviation between the real-time detected physical parameters (such as moisture content or surface tension) and the pre-stored standard values ​​in the database exceeds a preset 5% threshold, a dynamic calibration program is automatically triggered to correct the physical parameters using a first-order Taylor expansion and output a set of real-time physical description vectors that can truly reflect the fabric state under the current production environment.

[0020] The virtual physical nesting unit is used to construct a virtual physical field for the fabric in the cutting bed environment by integrating computational fluid dynamics algorithms and generative adversarial network models on the basis of geometric nesting. It simulates the deformation behavior of the fabric under the action of gravity, tension and external forces during fabric laying, and generates a nesting layout scheme that includes a physical deformation compensation zone. This is the core computing power module of this system, which realizes complex physical simulation through a high-performance graphics processing computing power cluster.

[0021] The virtual physical nesting unit internally constructs a physical field simulator based on continuum mechanics, which models the fabric as a two-dimensional fluid medium field with viscous and tensile properties. The virtual physical nesting unit is configured based on a simplified textual form of the Navier-Stokes equations to calculate the displacement field distribution of the fabric during static laying and dynamic stretching. The virtual physical nesting unit divides the fabric area to be laid into tens of thousands of triangular micro-element meshes, each of which is assigned mass, momentum, and energy conservation properties. During the simulation, the unit first calculates the vertical pressure generated by the gravitational acceleration on the fabric, and then, combined with the downward pressure generated by the cutting bed adsorption system, calculates the local normal pressure distribution between the fabric and the cutting bed surface. Furthermore, the virtual physical nesting unit simulates the lateral compressive stress generated by the cutting blade on the surrounding fibers during the cutting process; this stress is converted into a gradient change in the velocity vector within the physical field.

[0022] The generative adversarial network in the virtual physical nesting unit consists of a generator and a discriminator. The generator is configured to receive a geometric pattern matrix from the order parsing unit and a fabric physical description vector from the perception unit, mapping them to a high-dimensional feature space. Internally, the generator contains a multi-layer deconvolutional neural network configured to output an initial nesting topology while satisfying fabric width constraints and geometric constraints ensuring non-overlapping patterns. Simultaneously, the discriminator is configured to load a dataset of operational experience from historically successful nesters, containing layout patterns for different fabrics that have proven to have high utilization and high accuracy in actual production. The discriminator performs a probability score on the nesting layout output by the generator, determining whether the layout conforms to real-world process fluid dynamics and expert operating habits. If the score given by the discriminator is lower than a preset threshold of 0.8, the gradient error signal is backpropagated to the generator, driving it to adjust the rotation angle, horizontal displacement, and gap distance of the pattern pieces until both reach a Nash equilibrium, generating the globally optimal nesting scheme.

[0023] When outputting the layout plan, the virtual physical nesting unit automatically generates a physical deformation compensation zone on the outer edge of the original cut piece outline. This physical deformation compensation zone is a non-uniform geometric expansion region, the specific shape of which is determined by the local strain gradient. In the virtual physical field simulation results, if a region is identified as a high-sag region or a high-tension accumulation region, the unit is configured to apply a positive normal displacement along the outline of the high-sag region or high-tension accumulation region, i.e., increase the compensation amount; while in regions with stable physical properties and uniform stress distribution, a smaller or zero compensation amount is set. This dynamic compensation mechanism effectively counteracts the physical relaxation of the fabric after cutting and loss of tension restraint, ensuring that the actual size of the cut piece before sewing is completely consistent with the design size.

[0024] The cutting control unit receives the fabric layout scheme, converts it into motion trajectory commands executable by the cutting machine, and synchronously adjusts the cutting blade pressure, travel speed, and adsorption strength to adapt to the physical properties of the current fabric, ensuring edge precision and uniform internal stress distribution. The cutting control unit connects to the industrial control computer of the cutting machine via a real-time industrial Ethernet protocol. The cutting control unit includes a trajectory interpolation subunit, a servo drive control subunit, and a multi-source actuator synchronization subunit. The trajectory interpolation subunit is configured to decompose the complex curved contours in the fabric layout into micron-level straight line segments or arc segments and calculate the continuous motion coordinates of the cutting blade head in three-dimensional space. The servo drive control subunit establishes a real-time communication link with the servo motor driver of the cutting machine and dynamically adjusts the downward pressure of the cutting blade as it passes through different areas based on the local stress distribution map output by the virtual physical fabric layout unit. For example, in areas with high elasticity and high pressure stress in high-elasticity fabrics, the unit is configured to increase the vertical clamping force of the cutting blade and reduce its horizontal travel speed to suppress the rebound deformation of the fabric during the cutting process. Meanwhile, the multi-source actuator synchronization subunit will also control the output power of the vacuum pump under the cutting bed in real time according to the air permeability and thickness of the fabric, and dynamically adjust the adsorption intensity to ensure that the fabric remains in a flat physical state throughout the cutting process.

[0025] The feedback optimization unit is used to collect actual cut piece size deviation data, finished product size pass rate, and rework records after cutting. It updates the training sample set of the generative adversarial network through a closed-loop learning mechanism and optimizes the boundary condition parameters in the computational fluid dynamics model. The feedback optimization unit integrates a machine vision-based size detection device deployed at the end of the cutting line. The size detection device is configured to use a high-resolution industrial camera to capture images of each cut piece and perform pixel-level alignment comparison with the original design template, calculating the average offset of the edge contour and the size error values ​​of key parts. The feedback optimization unit associates and stores these measured deviation data with the corresponding order's physical parameters and nesting strategy, forming closed-loop data pairs. After the system completes the current batch of orders, the size detection unit calculates the first-order pass rate of the current batch. If the pass rate is lower than a predetermined 95%, the model retraining program is automatically started. During retraining, the system uses failed cases as negative samples and fine-tunes the weight parameters of the generative adversarial network using the gradient descent algorithm. At the same time, based on the deviation distribution characteristics, it automatically corrects the boundary condition parameters such as viscosity coefficient and elastic modulus in the computational fluid dynamics model, enabling the system to have adaptive evolution capabilities for specific fabrics and specific process scenarios.

[0026] Example 2: To further improve the execution accuracy in scenarios with extremely complex fabrics (such as high-count silk and multi-layer composite fabrics), this example provides an intelligent execution system for small-batch, fast-response garment customization orders based on a cloud-edge collaborative architecture.

[0027] The intelligent execution system for small-batch, fast-response garment customization orders consists of an edge-aware execution unit deployed on the production site and a high-performance simulation optimization cluster deployed in the cloud. The edge-aware execution unit includes an embedded control system and a sensor array integrated into the intelligent cutting bed. The edge-aware execution unit is not only responsible for receiving initial order data, but its core task is to perform microscopic sampling of the fabric's physical properties. The sensor array incorporates an acoustic emission sensor to monitor the ultrasonic signals generated when the cutting blade interacts with the fabric fibers. The edge-aware execution unit is internally configured with lightweight signal processing logic to convert the frequency distribution characteristics of the acoustic emission signal into internal fiber breaking energy parameters of the fabric, and then uploads these parameters, along with spectral analysis data, to the cloud in real time via a high-speed fiber optic interface.

[0028] The cloud-based high-performance simulation optimization cluster, serving as the core carrier of virtual physical nesting in this embodiment, possesses a heterogeneous computing pool composed of thousands of central processing unit cores and tens of thousands of stream processors. Within the cloud-based high-performance simulation optimization cluster, the operation of virtual physical nesting is further refined. The cloud system is configured to construct a fully parameterized digital twin model, which not only simulates the macroscopic deformation of the fabric but also delves into the microscopic mechanical behavior at the yarn level. During computational fluid dynamics simulation, the cloud cluster is configured to employ the large eddy simulation algorithm to accurately capture the subtle flutter behavior of the fabric under the adsorption of high-speed airflow on the cutting bed. This flutter behavior is identified as a key physical variable affecting edge quality, and the physical safety gap between the cut pieces in the nesting diagram is adjusted accordingly.

[0029] In this embodiment, the generative adversarial network is configured as a conditional generative adversarial network. Its input includes not only geometric data but also a vector of real-time environmental influence factors (such as workshop humidity fluctuations and tool wear status) fed back from the edge. The generator, through a multi-level residual connection structure, learns how to generate the most robust nesting layout in this variable environment. The discriminator introduces a reinforcement learning-based evaluation mechanism, configured to provide a comprehensive performance index based on the estimated material utilization rate, expected cutting time, and estimated physical deformation risk value. Through its interaction with the discriminator, the generator continuously optimizes the nesting scheme, ensuring that the final generated layout is not only geometrically compact but also the most stable during physical processing.

[0030] In this embodiment, the cutting control unit is a smart controller with feedforward compensation. The smart controller receives hyperdimensional trajectory commands from the cloud, which include not only coordinate information but also the predicted compensation vector corresponding to each trajectory segment. The cutting control unit is configured to pre-calculate the mechanical inertia during motor acceleration and deceleration before executing the trajectory movement, and, combined with the transient strain response of the fabric, adjust the motor's current loop and speed loop parameters in real time through a predictive control algorithm to achieve sub-millimeter-level path tracking accuracy.

[0031] In this embodiment, the feedback optimization unit uses distributed ledger technology to store quality feedback data. The size data of each cut piece and the quality inspection results of each garment are recorded in an immutable, encrypted chain. The feedback optimization unit is configured to use clustering analysis algorithms to deeply mine historical data and identify hidden factors affecting the production pass rate. For example, when the system identifies that the heat shrinkage rate of a certain fabric exhibits a non-linear transition within a specific humidity range, the feedback optimization unit automatically updates the fluid dynamics boundary condition settings of the cloud simulation cluster, enabling automatic knowledge transfer and model iteration within the system.

[0032] Example 3: This example focuses on describing a highly flexible hardware implementation method for ultra-large-scale customization scenarios, aiming to improve the adaptability of intelligent execution systems through modular hardware combinations.

[0033] The intelligent execution system for small-batch, fast-response garment customization orders has a hardware architecture that includes a modular sensing front-end, a heterogeneous computing power nesting back-end, a multi-axis collaborative cutting terminal, and a multi-source quality monitoring closed-loop unit. The modular sensing front end is configured as a replaceable slider structure and mounted on the crossbeam of the fabric spreading machine. Depending on the fabric to be processed, the user can attach different sensing modules. For example, for dark or coated fabrics, the sensing front end is configured to mount a laser ultrasonic detection module, which uses laser pulses to excite elastic waves on the fabric surface, thereby non-contactly measuring the fabric's thickness consistency and internal damage. The signal output from the laser ultrasonic detection module is converted into a digital sequence by a high-speed analog-to-digital converter and transmitted to the material feeding backend in real time.

[0034] The heterogeneous computing power nesting backend consists of multiple Field-Programmable Gate Array (FPGA) boards and a dedicated AI acceleration chip. With this hardware configuration, the virtual physical nesting unit is configured to achieve true parallel simulation. The FPGA boards are specifically configured to handle iterative hydrodynamic calculations of large-scale triangular meshes, accelerating the discretization solution of the Navier-Stokes equations to the millisecond level through a hardware-based pipeline structure. The dedicated AI acceleration chip is configured to run a generative adversarial network inference engine, achieving nesting scheme generation in seconds.

[0035] In this embodiment, the generation of the physical deformation compensation zone employs a logic based on a shape memory algorithm. The unit is configured to first simulate the process of fabric being stretched to its limit in a virtual physical field, and then record the recovery trajectory of each node after stress release. The boundary of the compensation zone is configured as the envelope of these trajectories, geometrically pre-leaving sufficient margin so that the cut fabric reaches the target size precisely after natural shrinkage.

[0036] In addition to a conventional XY-axis motion platform, the multi-axis collaborative cutting terminal also includes a controlled Z-axis tool deflection mechanism and an independently controlled adsorption matrix control system. The adsorption matrix control system comprises hundreds of independently operable solenoid valves, each corresponding to a small area on the cutting bed surface. The cutting control unit is configured to precisely activate the vacuum adsorption force in the corresponding area based on the position of the cut pieces in the nesting plan, while deactivating adsorption in non-processing areas, thus reducing energy consumption while maintaining adsorption strength. The Z-axis tool deflection mechanism is configured to adjust the cutting angle of the cutter in real time based on the predicted local stress direction in the nesting plan, ensuring that the blade always travels along the path of least internal stress in the material, further reducing dimensional deviations caused by fiber tension.

[0037] The multi-source quality monitoring closed-loop unit employs semantic segmentation technology based on deep learning. After cutting, the unit acquires real-time images of the entire fabric using a top-mounted wide-angle vision sensor. Utilizing a segmentation model trained on a massive dataset of labeled samples, the unit automatically identifies the edge boundaries of each cut piece and marks pieces with defects or dimensional deviations. This information is fed back to the front-end order parsing unit in real time. For defective pieces, the system automatically triggers a replacement order logic, prioritizing replacement cutting tasks in the next fabric arrangement cycle, thus achieving complete automation and self-healing of order execution.

[0038] In the system logic of this embodiment, the fabric physical property sensing unit also integrates a real-time environmental monitoring submodule, which is configured to monitor the ambient temperature, air humidity, and atmospheric pressure in the production workshop in real time. The virtual physical nesting unit is configured to use these environmental variables as auxiliary boundary conditions for the computational fluid dynamics model. For example, when increased air humidity causes the fabric fibers to absorb moisture and expand, the system automatically increases the heat shrinkage rate parameter and correspondingly reduces the reserved value of the compensation zone when generating the nesting scheme. This sensitivity to the external environment enables the system to maintain a high degree of consistent output in different seasons and geographical locations.

[0039] The cutting control unit is also equipped with a tool condition monitoring algorithm, which is configured to analyze the tool wear level by collecting the current fluctuation signal of the spindle motor and using wavelet transform technology. When tool wear is detected to increase cutting resistance, the cutting control unit is configured to automatically compensate the compensation value output by the virtual physical nesting unit, that is, to increase the compensation area to counteract the extrusion deformation caused by tool dullness, thereby extending the tool's service life and ensuring machining accuracy.

[0040] Example 4: This example describes a fast-response system configuration for extremely small batch customized orders, with particular emphasis on completing the entire process from order parsing to cutting and execution in a very short time.

[0041] The core feature of the intelligent execution system for small-batch, fast-response garment customization orders lies in its use of a rapid material layout logic based on prior physical feature map prediction.

[0042] The order parsing unit is configured to support direct parsing of 3D human body scan data. Internally, it includes an automated flattening subunit for 3D-to-2D conversion. This subunit is configured to flatten the 3D garment model into 2D pieces in virtual space based on the principle of minimum energy consumption, and simultaneously calculate the local strain distribution during the flattening process. This distribution data is directly input as the initial physical feature to the virtual physical nesting unit.

[0043] In this embodiment, the fabric physical property sensing unit is simplified to a handheld multi-functional fabric feature meter. Before processing small-batch orders, the operator only needs to use the handheld multi-functional fabric feature meter to perform three random samples of the fabric currently in use. The handheld multi-functional fabric feature meter integrates a miniature spectrometer and a piezoelectric hardness tester, and is configured to quickly obtain the basic physical vector of the fabric. The basic physical vector is transmitted to the system's core controller via Bluetooth protocol.

[0044] In this embodiment, the virtual physical nesting unit employs a strategy combining pre-computation and real-time correction. The system pre-trains a series of standard nesting template libraries for different fabric categories and style classifications using a generative adversarial network in the background. When a new order arrives, the virtual physical nesting unit is configured to first retrieve the base template from the template library that has the closest physical properties and the highest geometric similarity. Subsequently, the computational fluid dynamics model no longer performs global physics field iteration, but only performs local perturbation analysis on the differences between the current order and the base template (such as size scaling and local detail modifications). This local simulation strategy reduces the computational load, enabling the system to output a final nesting scheme including physical deformation compensation within seconds.

[0045] In this embodiment, the cutting control unit is configured to support dynamic path optimization. Its internal logic is configured to predict the physical state changes of the next cut piece while performing cutting. If the cutting of the current cut piece causes displacement of the surrounding remaining fabric, the unit is configured to update the coordinate offset of subsequent cut pieces in real time, ensuring that extremely high cutting accuracy can still be maintained even if the fabric drifts as a whole.

[0046] The feedback optimization unit focuses on achieving "first-piece qualification." It is configured to perform a flash image comparison immediately after the first piece is cut. If the error is within acceptable limits, the system maintains the current parameters; if the error exceeds the limit, the system uses a backpropagation algorithm to correct the core parameters in the physical model within milliseconds and applies the corrected compensation strategy to the remaining pieces. This ultra-fast feedback mechanism perfectly matches the extreme pursuit of response speed in the small-piece, fast-response mode.

[0047] In this embodiment, the system also incorporates an augmented reality (AR)-based auxiliary guidance unit. This AR-based unit is configured to project the physical field distribution map generated by the virtual physical nesting unit and the predicted compensation zone range directly onto the fabric surface of the actual cutting bed using a projection device. Operators can visually observe the expected stress on the fabric in different areas, improving the transparency of system operation and providing an intuitive reference for manual intervention in complex processes.

[0048] When constructing the physical field, the virtual physical nesting unit also takes into account the fabric's texture directionality (such as warp and weft differences). The virtual physical nesting unit is configured to assign different orthogonal anisotropic properties to the grid elements, i.e., setting different elastic moduli and damping coefficients in the warp and weft directions. This refined modeling approach enables the system to accurately predict pattern shifts caused by physical shrinkage when handling matching requirements for fabrics such as checks and stripes. It automatically adjusts the cut piece positions when generating the nesting scheme, ensuring that the final product achieves uniformity in both aesthetic layout and physical dimensions.

[0049] In this embodiment, the cutting control unit also includes an emergency suspension subroutine. The emergency suspension subroutine is configured to monitor the pressure feedback signal of the adsorption pump in real time. If the adsorption pressure suddenly drops below the safety threshold due to air leakage or other reasons, the cutting control unit will immediately issue an emergency stop command to the servo drive system, and automatically resume the breakpoint cutting based on the data from the position encoder after the pressure is restored, thus avoiding fabric waste.

[0050] The intelligent execution system for small-batch, high-response garment customization orders provided in this embodiment reduces the processing cycle of a single order from several hours to minutes through end-to-end optimization from front-end perception to back-end execution. This efficient, precise system architecture with high physical perception capabilities provides strong technical support for garment manufacturers to cope with rapidly changing market demands.

[0051] In summary, the intelligent execution system for small-batch, high-response garment customization orders provided by this invention achieves a profound leap from geometric nesting to physical nesting by integrating computational fluid dynamics and generative adversarial networks. The system can not only accurately predict and compensate for the physical deformation of fabrics during processing, but also continuously evolve through a closed-loop feedback mechanism, solving the technical bottlenecks caused by material diversity and process instability in garment customization, thereby improving production efficiency and product quality.

[0052] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent execution system for small-batch, fast-response garment customization orders, characterized in that: include: The order parsing unit is used to receive and parse customized order information from the client, extract garment style, size specifications, cut piece outline data and fabric type identifier, and distribute the information to the subsequent processing unit after structuring it. The fabric physical property sensing unit is used to call the pre-stored fabric physical parameter database according to the fabric type identifier, obtain the physical parameters of the corresponding batch of fabric, and dynamically calibrate the physical parameters in combination with real-time sensing data. The virtual physical nesting unit is used to construct a virtual physical field of fabric in the cutting bed environment by integrating computational fluid dynamics algorithms and generative adversarial network models on the basis of geometric nesting, simulating its deformation behavior under external forces, and generating a nesting layout scheme that includes a physical deformation compensation zone. The cutting control unit is used to receive the layout scheme, convert it into motion trajectory instructions that can be executed by the cutting bed equipment, and synchronously adjust the cutting bed execution parameters to adapt to the physical characteristics of the current fabric, so as to ensure the edge accuracy of the cut piece and the uniform distribution of internal stress. The feedback optimization unit is used to collect actual cut piece size deviation data and sewn product size qualification rate after cutting, update the training sample set of the generative adversarial network through a closed-loop learning mechanism, and optimize the boundary condition parameters in the computational fluid dynamics model.

2. The intelligent execution system for small-batch, fast-response garment customization orders according to claim 1, characterized in that, The order parsing unit includes: The order message interception subunit is used to listen for asynchronous requests from the client in real time. When a structured object containing customized requirements is captured, the cyclic redundancy check algorithm is used to verify the integrity of the data. The semantic recognition subunit has a dedicated ontology knowledge base for the apparel industry. It uses natural language processing technology to match feature words in unstructured text in orders, and identifies the style classification of garments and descriptive words related to fabric materials. The geometric data extraction subunit is used to parse the vector graphics file associated with the order, convert the sequence of outline coordinate points of the two-dimensional pattern pieces into a polygon vertex matrix, and calculate the area of ​​the initial envelope rectangle of each pattern piece.

3. The intelligent execution system for small-batch, fast-response garment customization orders according to claim 2, characterized in that, The fabric physical property sensing unit includes a sensor array integrated at the feed inlet of the fully automatic fabric spreading machine, the sensor array comprising: The near-infrared spectroscopy analysis module is used to irradiate the fabric surface with infrared light during the fabric unfolding process and analyze the fiber composition ratio, yarn twist and moisture content percentage of the fabric by collecting the absorption band characteristics of the reflection spectrum; the micro-force sensor array is embedded in the support position of the fabric laying roller to sense the small tangential tension generated by the fabric during the traction process in real time and calculate the elastic modulus change curve of the fabric at different elongation rates. An environmental temperature and humidity compensation module is used to monitor the temperature and humidity changes of the production environment in real time. The fabric physical property sensing unit is used to trigger a dynamic calibration program when the absolute value of the deviation between the real-time detected physical parameters and the pre-stored standard values ​​in the database exceeds a preset 5% threshold. The physical parameters are corrected using a first-order Taylor expansion, and a real-time physical description vector reflecting the fabric state under the current production environment is output.

4. The intelligent execution system for small-batch, fast-response garment customization orders according to claim 3, characterized in that, The virtual physical material laying unit is equipped with a physical field simulator based on continuum mechanics. The simulator models the material as a two-dimensional fluid medium field with viscous and tensile properties. The virtual physical nesting unit is based on the textual description of the Navier-Stokes equations to calculate the displacement field distribution of the fabric during static laying and dynamic stretching processes. The virtual physical material laying unit divides the material laying area to be laid into a triangular micro-element mesh, and assigns mass attribute, momentum attribute and energy conservation attribute to each micro-element. The virtual physical layout unit calculates the vertical pressure generated by the gravitational acceleration of the fabric and, in combination with the downward pressure generated by the cutting bed adsorption system, calculates the local normal pressure distribution between the fabric and the cutting bed surface. At the same time, it simulates the lateral compressive stress generated by the cutting blade on the surrounding fibers during the cutting process and converts the lateral compressive stress into the gradient change of the velocity vector.

5. The intelligent execution system for small-batch, fast-response garment customization orders according to claim 4, characterized in that, The generative adversarial network in the virtual physical nesting unit includes a generator and a discriminator: the generator receives the geometric piece matrix and the fabric physical description vector, and outputs the initial nesting topology map through the internal multi-layer deconvolutional neural network, under the premise of satisfying the fabric width constraint and the geometric constraint that the pieces do not overlap. The discriminator loads a dataset of operational experience from historical outstanding material layout engineers, performs probability scoring on the material layout output by the generator, and determines whether the material layout conforms to the laws of process fluid dynamics and the operating habits of experts. The generative adversarial network works as follows: if the score given by the discriminator is lower than a preset score threshold, the gradient error signal is backpropagated to the generator, driving the generator to adjust the rotation angle, horizontal displacement and gap distance of the cut pieces until the generator and the discriminator reach a Nash equilibrium state, and a material layout scheme is generated.

6. The intelligent execution system for small-batch, fast-response garment customization orders according to claim 5, characterized in that, The virtual physical nesting unit generates a physical deformation compensation area on the outer edge of the original cut piece outline. The physical deformation compensation area is a non-uniform geometric expansion area. The virtual physical material feeding unit dynamically adjusts the expansion of the physical deformation compensation zone based on the local strain gradient output by the virtual physical field simulator. Apply positive normal displacement to the contour lines identified as high sag areas or high tension accumulation areas to increase the compensation amount, while set 0 compensation amount in areas with uniform stress distribution to counteract the physical relaxation of the fabric after it loses its tension restraint.

7. The intelligent execution system for small-batch, fast-response garment customization orders according to claim 6, characterized in that, The trimming control unit includes a trajectory interpolation subunit, a servo drive control subunit, and a multi-source actuator synchronization subunit: The trajectory interpolation subunit decomposes the curved contour in the material layout into straight line segments or arc segments, and calculates the continuous motion coordinates of the cutting head in three-dimensional space. The servo drive control subunit establishes a real-time communication link with the servo motor driver of the cutting bed, so as to dynamically adjust the downward pressure of the cutting blade when passing through different areas according to the local stress distribution map output by the virtual physical nesting unit. The multi-source actuator synchronization subunit controls the output power of the vacuum pump under the cutting bed in real time according to the air permeability and thickness of the fabric, and dynamically adjusts the adsorption intensity; the cutting control unit increases the vertical clamping force of the cutting blade and reduces the horizontal travel speed of the cutting blade in areas with high elasticity and high pressure stress.

8. The intelligent execution system for small-batch, fast-response garment customization orders according to claim 7, characterized in that, The feedback optimization unit integrates a machine vision-based size detection device. The size detection device uses an industrial camera to capture images of the cut pieces and performs pixel-level alignment comparison with the original design template to calculate the average offset of the edge contour and the size error value of key parts. The feedback optimization unit associates and stores the measured deviation data with the physical parameters and material arrangement strategy of the corresponding order to form a closed-loop data pair. When the first order qualification rate of a batch is lower than the predetermined 95% threshold, the model retraining program is automatically started, and the gradient descent algorithm is used to fine-tune the weight parameters of the generative adversarial network and correct the viscosity coefficient and elastic modulus parameters in the computational fluid dynamics model.

9. The intelligent execution system for small-batch, fast-response garment customization orders according to claim 8, characterized in that, The system adopts a cloud-edge collaborative architecture, including edge-aware execution units deployed on the production site and a high-performance simulation optimization cluster deployed in the cloud: The edge-sensing execution unit includes a sensor array, in which an acoustic emission sensor is introduced to monitor the ultrasonic signals generated when the cutting blade interacts with the fabric fibers. The edge-aware execution unit converts the frequency distribution characteristics of the ultrasonic signal into the internal fiber fracture energy parameters of the fabric and uploads them to the cloud-based high-performance simulation optimization cluster. The cloud-based high-performance simulation optimization cluster is used to construct a fully parameterized digital twin model. It employs the large eddy simulation algorithm to capture the flutter behavior of fabric under the airflow adsorption of the cutting bed, and adjusts the physical safety gap between the cut pieces in the layout accordingly. The generative adversarial network is a conditional generative adversarial network, and its input also includes a real-time environmental influence factor vector fed back by the edge-aware execution unit. The real-time environmental influence factor vector includes workshop humidity fluctuation data and tool wear status data.

10. A smart execution method for small-batch, fast-response garment customization orders, characterized in that: The intelligent execution system for small-batch, fast-response garment customization orders described in any one of claims 1-9 is used to realize the intelligent execution of garment customization orders.

Images