Fabric self-adaptive sewing parameter recommendation method and system, electronic device and medium

Abstract

The application provides a fabric adaptive sewing parameter recommendation method and system, an electronic device and a medium. The method comprises the following steps: before the start of a sewing operation, performing non-destructive pre-detection on a fabric to be sewn to obtain an original sensing signal; preprocessing and feature extraction are performed on the original sensing signal to generate a multi-dimensional feature vector; the multi-dimensional feature vector is calculated with each standard fabric sample stored in a database to obtain a similarity; when the similarity meets a preset threshold, a corresponding verified sewing parameter is called, otherwise the multi-dimensional feature vector is input into a parameter recommendation model to obtain a sewing parameter; and the obtained sewing parameter set is sent to a target sewing device to complete sewing parameter configuration and perform a sewing operation. The application solves the problems that sewing parameter configuration relies on manual repeated trial sewing when fabric is changed, it is difficult to complete a complete parameter recommendation according to the fabric properties before the sewing operation, and verified fabric process data is difficult to continuously deposit and reuse in the production site.

Description

Technical Field

[0001] This application belongs to the field of intelligent technology of industrial sewing equipment, and relates to a method and system for recommending sewing parameters based on fabric adaptation, electronic equipment and media. Background Technology

[0002] As garment and home textile sewing production gradually moves towards multi-variety, small-batch, and rapid changeover, production lines need to frequently switch between fabrics of different materials, weave structures, and thicknesses. For different fabrics, the appropriate sewing parameters, such as needle thread tension, looper thread tension, presser foot pressure, differential ratio, feed speed, and feed dog height, often vary. Therefore, quickly configuring these sewing parameters during changeovers has a significant impact on production efficiency and first-piece quality.

[0003] Currently, after fabric changes, sewing parameter configuration still largely relies on experienced operators adjusting the settings and gradually determining the appropriate parameters through multiple trial sewings. This method consumes a significant amount of changeover time when fabric types change frequently, and the differences in experience among different operators can easily lead to inconsistent parameter configuration results. If the parameter configuration is inappropriate, sewing quality problems such as skipped stitches, loose threads, uneven stitches, fabric wrinkling, stretching deformation, or fabric damage may occur.

[0004] Furthermore, existing fabric sensing methods in sewing equipment often struggle to obtain multidimensional information that characterizes the fabric's intrinsic properties. For example, single thickness information is insufficient to fully reflect the fabric's fiber composition, blend ratio, and mechanical properties such as stretching, bending, and friction. Consequently, it is difficult to establish a stable and reusable correspondence between fabric identification results and the entire set of sewing process parameters. In addition, long-term fabric process experience accumulated on the production floor is usually scattered in personnel experience or isolated process records. Verified and effective fabric characteristics and sewing parameters are difficult to continuously accumulate with actual production and be directly reused in subsequent production changes.

[0005] Therefore, there is still room for improvement in how to obtain the multidimensional ontological attributes of the fabric to be sewn before the sewing operation, automatically recommend and configure a complete set of sewing parameters, and further utilize the sewing quality results to continuously accumulate reusable fabric process data. Summary of the Invention

[0006] This application provides a method and system for recommending adaptive sewing parameters for fabrics, an electronic device, and a medium to solve the problems of relying on repeated manual trial sewing when changing fabrics, difficulty in recommending a complete set of parameters based on the fabric's inherent properties before sewing operations, and difficulty in continuously accumulating and reusing verified fabric process data on the production site.

[0007] Firstly, this application provides a method for recommending adaptive sewing parameters for fabrics, comprising: performing non-destructive pre-detection on the fabric to be sewn before the sewing operation begins to obtain the original sensing signal of the fabric; preprocessing and extracting features from the original sensing signal to generate a multidimensional feature vector characterizing the ontological attributes of the fabric; calculating the similarity between the multidimensional feature vector and each standard fabric sample stored in a pre-constructed fabric-process knowledge graph database, wherein each standard fabric sample records at least a multidimensional feature vector of the fabric and its corresponding verified set of effective sewing parameters; when there exists When at least one standard fabric sample has a similarity exceeding a preset threshold, a target standard fabric sample is determined from the at least one standard fabric sample, and a verified and valid sewing parameter set corresponding to the target standard fabric sample is called; when no standard fabric sample has a similarity exceeding the preset threshold, the multidimensional feature vector is input into a trained parameter recommendation model, and an initial sewing parameter set is output; the verified and valid sewing parameter set or the initial sewing parameter set is sent as the sewing parameter set to be executed to each mechanism controller of the target sewing equipment to complete the sewing parameter configuration of the target sewing equipment and execute the sewing operation.

[0008] In one implementation of the first aspect, before the sewing operation begins, non-destructive pre-inspection is performed on the fabric to be sewn to obtain the original sensing signals of the fabric, including: non-destructively detecting the fiber composition and blending ratio of the fabric to be sewn using a near-infrared spectroscopy sensor to obtain the chemical composition information of the fabric to be sewn; acquiring the tensile-recovery curve, bending stiffness, and surface friction coefficient of the fabric to be sewn using a micro-force sensor and a displacement encoder in a mechanical sensing unit; and performing non-contact detection on the fabric to be sewn using a thickness sensor to obtain the thickness and thickness uniformity of the fabric to be sewn; wherein the thickness sensor is a laser displacement sensor or a capacitive sensor.

[0009] In one implementation of the first aspect, preprocessing and feature extraction of the original sensing signal to generate a multidimensional feature vector characterizing the properties of the fabric to be sewn includes: performing filtering, noise reduction, and data synchronization preprocessing on the original sensing signal; extracting spectral feature peaks, elastic modulus, plastic deformation, bending stiffness, surface friction coefficient, thickness, and thickness uniformity from the preprocessed original sensing signal; and combining the extracted features to generate a multidimensional feature vector corresponding to the fabric to be sewn.

[0010] In one implementation of the first aspect, the standard fabric samples stored in the fabric-process knowledge graph database include standard fabric samples obtained from expert calibration experiments and new fabric samples added and verified on-site based on online quality inspection results; the verified effective sewing parameter set is a set of sewing parameters verified by expert calibration experiments or on-site online quality inspection, including needle thread tension, looper thread tension, presser foot pressure, differential ratio, feed speed, and feed dog height; the parameter recommendation model is trained using at least one machine learning model among support vector machine, random forest, and deep neural network, and the parameter recommendation model takes the multidimensional feature vector as input and the initial sewing parameter set as output.

[0011] In one implementation of the first aspect, the method further includes: performing online detection of sewing quality during the sewing process to obtain a quality detection result; when the quality detection result indicates that the sewing quality is qualified, storing the multidimensional feature vector and the sewing parameter set actually used in this sewing operation and verified as valid by the online detection as a new standard fabric sample in the fabric-process knowledge graph database; when the quality detection result indicates the existence of sewing defects and the sewing defects meet the parameter fine-tuning conditions according to preset rules, fine-tuning at least one sewing parameter in the sewing parameter set used in this sewing operation according to the preset rules, and performing the sewing operation and the online detection again based on the fine-tuned sewing parameter set; when the fine-tuned quality detection result indicates that the sewing quality is qualified, storing the multidimensional feature vector and the fine-tuned sewing parameter set as a new standard fabric sample in the fabric-process knowledge graph database; when the fine-tuned quality detection result still indicates that the sewing quality is unqualified, outputting a prompt message to remind the operator to perform manual intervention and adjustment.

[0012] In one implementation of the first aspect, online detection of sewing quality during the sewing process includes: acquiring stitch images during the sewing process using a miniature industrial camera, and using an image recognition algorithm to identify whether sewing defects exist in the stitch images; the sewing defects include skipped stitches, loose threads, and uneven stitches; acquiring the spindle current signal and power fluctuation signal of the target sewing equipment, and using the spindle current signal and power fluctuation signal to assist in evaluating the sewing quality.

[0013] Secondly, this application provides a fabric adaptive sewing parameter recommendation system, comprising: a fabric adaptive sewing parameter recommendation device configured to perform the fabric adaptive sewing parameter recommendation method described in any of the above claims; and a sewing device communicatively connected to the fabric adaptive sewing parameter recommendation device, configured to receive a sewing parameter set output by the fabric adaptive sewing parameter recommendation device, and to perform sewing parameter configuration and sewing operations based on the sewing parameter set.

[0014] In one implementation of the second aspect, the fabric adaptive sewing parameter recommendation device includes: a fabric information acquisition module, located at the feeding position of the sewing equipment or integrated into an independent handheld scanning terminal, and including a near-infrared spectral sensor, a mechanical sensing unit, and a thickness sensor, configured to perform non-destructive pre-detection on the fabric to be sewn before the sewing operation begins, to obtain the original sensing signal of the fabric to be sewn; a feature vector generation module, configured to preprocess and extract features from the original sensing signal to generate a multi-dimensional feature vector characterizing the ontological attributes of the fabric to be sewn; and a sewing parameter acquisition module, configured to calculate the similarity between the multi-dimensional feature vector and each standard fabric sample stored in a pre-constructed fabric-process knowledge graph database, wherein each A standard fabric sample records a multi-dimensional feature vector of the fabric and its corresponding verified and valid sewing parameter set. When there is at least one standard fabric sample with a similarity exceeding a preset threshold, a target standard fabric sample is determined from the at least one standard fabric sample, and the verified and valid sewing parameter set corresponding to the target standard fabric sample is called. When there is no standard fabric sample with a similarity exceeding the preset threshold, the multi-dimensional feature vector is input into a trained parameter recommendation model, and an initial sewing parameter set is output. The sewing parameter execution module is configured to send the verified and valid sewing parameter set or the initial sewing parameter set as the sewing parameter set to be executed to each mechanism controller of the target sewing equipment to complete the sewing parameter configuration of the target sewing equipment and execute the sewing operation.

[0015] Thirdly, this application provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program stored in the memory to cause the electronic device to perform the method described in any of the preceding claims.

[0016] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any of the preceding claims.

[0017] As described above, the fabric adaptive sewing parameter recommendation method and system, electronic device, and medium described in this application have the following beneficial effects:

[0018] (1) By performing multi-dimensional non-destructive pre-inspection on the fabric to be sewn before the sewing operation, and obtaining information on the fiber composition, blending ratio, mechanical properties and thickness of the fabric, the system can form a more complete representation of the fabric's characteristics before the fabric actually enters the sewing process, providing a data basis for the automatic configuration of the entire set of sewing parameters.

[0019] (2) By adopting a dual-path decision-making approach that combines fabric-process knowledge graph database similarity matching with parameter recommendation model reasoning, the process parameters that have been verified in historical standard fabric samples can be reused first, and the initial sewing parameters can be output by parameter recommendation model when there are no highly similar historical samples, thus taking into account both the efficiency of historical process reuse and the adaptability to new fabrics.

[0020] (3) By automatically sending the obtained sewing parameter set to the controllers of each mechanism of the target sewing equipment, the process of manually adjusting parameters one by one and making multiple trial sewings during the production change process can be reduced, thereby improving the efficiency of fabric production change and first piece production.

[0021] (4) By detecting the stitch quality and spindle running status online during the sewing process, and storing the verified "fabric multidimensional feature vector-sewing parameter set" as a new standard fabric sample into the fabric-process knowledge graph database, we can continuously accumulate fabric process data from the production site and realize the digital accumulation and subsequent reuse of process knowledge.

[0022] (5) When a sewing defect that can be handled by parameter fine-tuning is detected, a closed-loop optimization process of "perception-recommendation-execution-evaluation-update" can be formed by fine-tuning at least one sewing parameter and performing sewing quality verification again, thereby improving the system's adaptability to flexible production scenarios with multiple varieties and small batches. Attached Figure Description

[0023] Figure 1 The diagram shown is a structural schematic of a fabric adaptive sewing parameter recommendation system according to an embodiment of this application.

[0024] Figure 2 The flowchart shown is a method for recommending adaptive sewing parameters for fabrics according to an embodiment of this application.

[0025] Figure 3 The diagram shown is a schematic representation of a non-destructive pre-inspection process performed on a sewing fabric according to an embodiment of this application.

[0026] Figure 4 The flowchart shown is a process for multimodal sensing data preprocessing and fabric multidimensional feature vector generation in one embodiment of this application.

[0027] Figure 5 The diagram shown is a schematic representation of the data organization structure of the fabric-process knowledge graph database in one embodiment of this application.

[0028] Figure 6 The diagram shown is a flowchart of the sewing parameter set acquisition process used for this sewing operation in one embodiment of this application.

[0029] Figure 7 The flowchart shown is a process for online sewing quality detection in one embodiment of this application.

[0030] Figure 8 The diagram shown is an architectural block diagram of a fabric adaptive sewing parameter recommendation system according to another embodiment of this application.

[0031] Figure 9 The diagram shown is a structural schematic of an electronic device according to an embodiment of this application. Detailed Implementation

[0032] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0033] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0034] The following embodiments of this application provide a method and system for recommending adaptive sewing parameters for fabrics, as well as electronic devices and media, primarily targeting fabric changeover processes in flexible production scenarios involving multiple varieties and small batches, such as apparel and home textiles. This application can complete the detection of the fabric's inherent properties and the recommendation of sewing parameters before the formal sewing operation begins, and verify the adopted parameters using quality inspection results after actual sewing, enabling the continuous accumulation and reuse of effective fabric process data generated on-site.

[0035] Compared to methods that rely on operators repeatedly performing trial sewings to determine parameters, this application connects pre-sewing multimodal fabric ontology detection, historical standard fabric sample retrieval, machine learning parameter recommendation, automatic execution of overall machine parameters, and production quality feedback into a continuous processing chain. Therefore, when similar historical fabrics exist, validated process parameters can be reused preferentially. When highly similar historical samples are lacking, the parameter recommendation model can provide initial parameters, which are then validated and updated in a closed-loop manner using actual sewing quality results.

[0036] The method described in this application can be deployed in sewing equipment, stand-alone handheld scanning terminals, workshop edge computing servers, or combinations thereof. The principles and implementation methods of a fabric adaptive sewing parameter recommendation method, system, electronic device, and medium according to this embodiment will be described in detail below with reference to the accompanying drawings, enabling those skilled in the art to understand the embodiments of this application without creative effort.

[0037] Please see Figure 1 The image shown is a schematic diagram of the structure of a fabric adaptive sewing parameter recommendation system in one embodiment of this application.

[0038] like Figure 1 As shown, this embodiment provides a fabric adaptive sewing parameter recommendation system, including a fabric adaptive sewing parameter recommendation device and a sewing machine.

[0039] The fabric adaptive sewing parameter recommendation device is configured to perform the fabric adaptive sewing parameter recommendation method described in this application.

[0040] The sewing equipment is communicatively connected to the fabric adaptive sewing parameter recommendation device, and is configured to receive the sewing parameter set output by the fabric adaptive sewing parameter recommendation device, and to perform sewing parameter configuration and sewing operations according to the sewing parameter set.

[0041] Please see Figure 2 The above is a flowchart of a fabric adaptive sewing parameter recommendation method in one embodiment of this application.

[0042] like Figure 2 As shown, this embodiment provides a method for recommending adaptive sewing parameters for fabrics, including the following steps S100 to S400.

[0043] In step S100, before the sewing operation begins, a non-destructive pre-inspection is performed on the fabric to be sewn to obtain the original sensor signal of the fabric to be sewn.

[0044] The term "before the start of sewing operation" in this application refers to the stage before the target sewing equipment performs formal stitch formation on the fabric to be sewn. During this stage, the fabric's intrinsic properties, such as fiber composition, mechanical properties, and thickness, are acquired, rather than instantaneous sewing condition data that can only be obtained after needle insertion, stitch formation, or sewing load has occurred. Therefore, the fabric to be sewn can generate input data for subsequent sample matching and model inference before it is formally sewn.

[0045] The original sensing signal refers to the detection data collected by the fabric adaptive sewing parameter recommendation device that has not yet completed feature extraction; the fabric adaptive sewing parameter recommendation device includes a fabric information acquisition module, which includes a near-infrared spectral sensor, a mechanical sensing unit, and a thickness sensor.

[0046] Please see Figure 3 The diagram shows a schematic representation of a non-destructive pre-inspection of a sewing fabric in one embodiment of this application.

[0047] like Figure 3 As shown, in one deployment method, the fabric information acquisition module can be set at the feeding position of the sewing equipment. Before entering the sewing area, the fabric to be sewn passes sequentially through a near-infrared spectral sensor (such as a spectral probe), a thickness sensor (such as a laser sensor), and a mechanical sensing unit (such as a mechanical roller). In another deployment method, the fabric information acquisition module can be integrated into a standalone handheld scanning terminal to scan the fabric edge or a preset sampling area. This application does not limit the specific order of the sensor arrangement.

[0048] In one embodiment of this application, the near-infrared spectral sensor is used to perform non-destructive spectral detection on the fabric to be sewn to obtain spectral information related to the fabric's fiber composition and blending ratio, thereby obtaining the chemical composition information of the fabric to be sewn. This detection does not require actual sewing to obtain fabric information, and therefore can be completed before the formal sewing begins.

[0049] The near-infrared spectral sensor can perform multi-point or continuous scanning of the fabric to be sewn within a preset near-infrared band. The processing unit can perform at least one of the following processing on the acquired spectrum: smoothing, baseline correction, or normalization, and extract spectral characteristic peaks that can characterize the differences in chemical composition of different fibers. These spectral characteristic peaks can be further used to obtain fiber category information and / or blend ratio characterization values.

[0050] In one embodiment of this application, the mechanical sensing unit includes a micro-force sensor and a displacement encoder, used to acquire the stretch-recovery curve, bending stiffness, and surface friction coefficient of the fabric to be sewn.

[0051] A micro-force sensor acquires the force acting on the fabric during the detection process, and a displacement encoder simultaneously acquires the displacement or deformation corresponding to the force, thereby forming a tension-recovery curve. The processing unit can obtain the elastic modulus or equivalent elastic characterization based on the force-displacement response during the tensioning stage, and obtain the plastic deformation based on the residual deformation after unloading; it can obtain the bending stiffness through the bending detection response, and obtain the surface friction coefficient through the relative motion response between the fabric and the preset contact component.

[0052] In one embodiment of this application, the thickness sensor is used to perform non-contact detection on the fabric to be sewn to obtain the thickness and thickness uniformity of the fabric. The thickness sensor can be a laser displacement sensor or a capacitive sensor. When obtaining thickness uniformity, multiple thickness sampling points can be obtained along a preset detection direction of the fabric, and thickness uniformity features can be generated based on the dispersion of each thickness sampling value relative to the average thickness.

[0053] Through the above-mentioned non-destructive pre-inspection before sewing, chemical composition, mechanical properties and thickness information can be obtained simultaneously on the same fabric to be sewn, avoiding the use of only a single thickness or instantaneous working condition signals during the formal sewing process to characterize the fabric, thus providing a multi-dimensional fabric ontology data basis for the pre-recommendation of the entire set of sewing parameters.

[0054] In step S200, the original sensing signal is preprocessed and features are extracted to generate a multi-dimensional feature vector characterizing the ontological properties of the fabric to be sewn.

[0055] Please see Figure 4 The flowchart shown is a process flow chart of multimodal sensing data preprocessing and fabric multidimensional feature vector generation in one embodiment of this application.

[0056] like Figure 4 As shown, the processing unit receives raw sensor signals from the fabric information acquisition module, performs filtering, noise reduction, and data synchronization on data from different sensing modes, and extracts features such as spectral characteristic peaks, elastic modulus, plastic deformation, bending stiffness, surface friction coefficient, thickness, and thickness uniformity from the pre-processed raw sensor signals. Data output from different sensors can be synchronized according to acquisition time, sampling location, or the same fabric sample identifier, ensuring that all features correspond to the current fabric to be sewn.

[0057] In one embodiment of this application, spectral characteristic peaks can be combined to form a spectral sub-vector Vs, elastic modulus, plastic deformation, bending stiffness, and surface friction coefficient can be combined to form a mechanical sub-vector Vm, and thickness and thickness uniformity can be combined to form a thickness sub-vector Vh. After scale normalization, each sub-vector is combined in a preset order to form a multidimensional feature vector V. To balance the influence of different modes on subsequent matching, modal weights can be set for Vs, Vm, and Vh respectively.

[0058] It should be noted that the multidimensional feature vector is used to characterize the intrinsic properties of the fabric to be sewn, rather than to characterize the instantaneous needle-thread interaction state after sewing has occurred. The multidimensional feature vector corresponding to the same fabric to be sewn can remain associated with the fabric sample identifier before the parameter recommendation, parameter execution, and subsequent quality verification are completed; after the quality verification is passed, the multidimensional feature vector obtained before sewing and the set of sewing parameters actually used and verified in this case will still be used to form a new standard fabric sample.

[0059] In this way, the chemical composition information provided by near-infrared spectroscopy, the deformation and friction response information provided by mechanical testing, and the geometric information provided by thickness testing are combined into a unified fabric ontology feature representation, so that the same multidimensional feature vector can be used simultaneously for historical standard fabric sample matching and parameter recommendation model inference.

[0060] In step S300, a set of sewing parameters for this sewing operation is obtained based on the multidimensional feature vector.

[0061] Please see Figure 5 The diagram shows the data organization structure of the fabric-process knowledge graph database in one embodiment of this application.

[0062] like Figure 5 As shown, the fabric-process knowledge graph database can store multiple standard fabric samples. A standard fabric sample refers to a fabric process data record in the database that can be retrieved and reused. Each standard fabric sample records at least a multi-dimensional feature vector of the fabric and its corresponding verified set of effective sewing parameters. In this embodiment, the verified set of effective sewing parameters can be understood as a set of effective sewing parameters that has been verified through expert calibration experiments or actual sewing quality testing on the production floor and can be used for the corresponding standard fabric sample; it is not required to be mathematically optimal. For example, it may include needle thread tension, looper thread tension, presser foot pressure, differential ratio, feed speed, and feed dog height.

[0063] In one implementation, the fabric-process knowledge graph database is centered on a standard fabric sample entity Si. Through relationships such as "having features", "associated parameters", "quality verified", and "source / version", it is associated with fabric ontology feature entity Fi, sewing parameter entity Pi, quality verification record Qi, and sample source and version information, respectively. Among them, fabric ontology feature entity Fi records the fabric multidimensional feature vector Vi, and sewing parameter entity Pi records the corresponding set of verified and valid sewing parameters.

[0064] Please see Figure 6 The diagram shows a flowchart of the sewing parameter set acquisition process used for this sewing operation in one embodiment of this application.

[0065] like Figure 6 As shown, the sewing parameter set for this sewing operation is obtained based on the multi-dimensional feature vector, which includes a dual-path decision-making method that prioritizes historical standard fabric sample similarity matching and is supplemented by parameter recommendation model reasoning.

[0066] Specifically, the multidimensional feature vector corresponding to the fabric to be sewn is compared with the similarity of each standard fabric sample stored in the pre-constructed fabric-process knowledge graph database. When at least one standard fabric sample has a similarity exceeding a preset threshold, a target standard fabric sample is determined from the at least one standard fabric sample, and the verified and valid sewing parameter set corresponding to the target standard fabric sample is called. When no standard fabric sample has a similarity exceeding the preset threshold, the multidimensional feature vector is input into the trained parameter recommendation model, and an initial sewing parameter set is output.

[0067] In one embodiment of this application, the similarity can be calculated separately according to different modalities and then fused. For example, the sub-similarity between the current fabric and the standard fabric sample on the spectral sub-vector, mechanical sub-vector, and thickness sub-vector can be calculated separately, and a comprehensive similarity can be obtained according to a preset modal weight. Each sub-similarity can be obtained using cosine similarity or based on normalized distance mapping. This application does not limit the specific similarity algorithm, as long as the similarity used can reflect the closeness between the current fabric to be sewn and the standard fabric sample in multidimensional ontological attributes.

[0068] When multiple standard fabric samples with similarity exceeding a preset threshold exist, the standard fabric sample with the highest similarity can be determined as the target standard fabric sample; alternatively, the sample with the highest similarity can be selected from candidate samples with consistent auxiliary conditions such as equipment model, fabric application, or quality verification level. The preset threshold can be calibrated using historical verification samples.

[0069] For example, for the current multidimensional feature vector V, the system can obtain its similarity Si with the multidimensional feature vectors Vi of each standard fabric sample, and determine the maximum similarity Smax. When Smax reaches or exceeds a preset threshold, the standard fabric sample with the highest similarity is selected as the target standard fabric sample and its validated sewing parameter set is called; when Smax does not reach the preset threshold, the current multidimensional feature vector V is input into the trained parameter recommendation model, and the initial sewing parameter set is output.

[0070] In one embodiment of this application, the parameter recommendation model can be trained using at least one machine learning model selected from Support Vector Machine (SVM), Random Forest, and Deep Neural Network. The parameter recommendation model takes the multidimensional feature vector as input and the initial sewing parameter set as output.

[0071] The model training data may include standard fabric samples obtained from expert calibration experiments and new standard fabric samples added in the closed loop of the production site and verified for quality. Each training sample takes a multi-dimensional feature vector as input and a set of sewing parameters associated with the multi-dimensional feature vector and verified to be effective as the supervision target.

[0072] When the number of newly added standard fabric samples that have passed quality verification in the production site reaches a preset quantity, a preset time period, or a preset update condition is triggered, the new samples can be added to the training dataset to perform incremental training or periodic retraining on the parameter recommendation model, and the updated model version can be saved. Thus, the sample growth of the knowledge graph database and the capability updates of the parameter recommendation model work in tandem.

[0073] In step S400, the sewing parameter set obtained in step S300 is sent as the sewing parameter set to be executed to each mechanism controller of the target sewing equipment to complete the sewing parameter configuration of the target sewing equipment and execute the sewing operation.

[0074] Specifically, the system can automatically send the corresponding parameters to the actuators of the target sewing equipment, such as the tension motor, pressure solenoid valve, spindle servo motor, and feed mechanism controller, to complete the automatic configuration of parameters such as needle thread tension, looper thread tension, presser foot pressure, differential ratio, feed speed, and feed dog height.

[0075] In one embodiment of this application, the needle thread tension and looper thread tension can be configured by a corresponding tension motor or tension adjustment mechanism; the presser foot pressure can be configured by a pressure solenoid valve or presser foot pressure actuator; the spindle servo motor is used to control the sewing speed or related operating parameters; the differential ratio, feed speed, and feed dog height can be configured by the feed mechanism controller and corresponding actuators. The parameter execution unit converts a unified set of sewing parameters to be executed into control set values ​​that can be recognized by each mechanism controller.

[0076] In some embodiments, before the parameters are issued, boundary checks can be performed on the sewing parameters to be executed based on the parameter range allowed by the target sewing equipment; when a parameter exceeds the executable range of the target equipment, it can be restricted to the allowed range or a configuration prompt can be output. After the configuration is completed, the operator can directly load the material and start sewing, or the sewing operation can be triggered by the automated production unit.

[0077] In one embodiment of this application, the fabric adaptive sewing parameter recommendation method further includes the following steps S500 and S800.

[0078] In step S500, the sewing quality is detected online during the sewing process to obtain the quality detection result.

[0079] Please see Figure 7 The above is a flowchart of online sewing quality detection in one embodiment of this application.

[0080] like Figure 7As shown, a miniature industrial camera can be used to capture stitch images during the sewing process, and an image recognition algorithm can be used to identify sewing defects such as skipped stitches, loose threads, and uneven stitches. Simultaneously, the spindle current signal and power fluctuation signal of the target sewing equipment are acquired, which are used to help characterize the operational stability during the sewing process. Based on the spindle current signal and the power fluctuation signal, the sewing quality can be assessed.

[0081] In one embodiment of this application, if the image recognition does not detect a preset sewing defect and the spindle current and power fluctuations are within the normal operating range of the corresponding equipment, the current sewing quality can be determined to be qualified; if a minor defect is detected that can be improved by adjusting one or more sewing parameters according to historical rules, it is determined to be a fine-tunable defect; if the defect is serious or cannot be eliminated after continuous fine-tuning, manual intervention is required.

[0082] In step S600, when the quality inspection result indicates that the sewing quality is qualified, the multi-dimensional feature vector and the sewing parameter set actually used in this sewing operation and verified by the online detection are stored as a new standard fabric sample in the fabric-process knowledge graph database. Thus, fabric process data verified through actual sewing in the production site can be transformed into a standard fabric sample that can be retrieved and reused subsequently.

[0083] When creating new standard fabric samples, the "current fabric multidimensional feature vector + current verified valid sewing parameter set" can be packaged together, and quality inspection results, sample source, and optional equipment identification or time information can be attached simultaneously. The sample data is then encrypted or written to the fabric-process knowledge graph database using a controlled storage method. In addition to being used for subsequent similarity matching, newly added standard fabric samples can also be incorporated into the incremental training data for the parameter recommendation model.

[0084] In step S700, when the quality inspection result indicates the presence of a sewing defect and the sewing defect meets the parameter fine-tuning conditions according to preset rules, at least one sewing parameter in the sewing parameter set used in this sewing operation is fine-tuned according to the preset rules, and the sewing operation and the online inspection are performed again based on the fine-tuned sewing parameter set; when the fine-tuned quality inspection result indicates that the sewing quality is qualified, the multidimensional feature vector and the fine-tuned sewing parameter set are stored as a new standard fabric sample in the fabric-process knowledge graph database.

[0085] As an example of preset rules, a correspondence can be established between sewing defect types and adjustable parameters. For instance, for minor wrinkling defects, the differential ratio and / or thread tension can be adjusted first; for floating thread defects, the corresponding thread tension can be adjusted first; and for uneven stitches, adjustments can be made in conjunction with feed speed, presser foot pressure, or other feed-related parameters. The direction and magnitude of adjustment for each parameter can be determined by expert calibration experiments or historical production records, and are limited to a preset fine-tuning range.

[0086] In one optional embodiment of this application, a reinforcement learning model can also be used to perform parameter fine-tuning.

[0087] Specifically, the current fabric multidimensional feature vector, current sewing parameter set, defect type and defect degree, and spindle current or power fluctuation can be used as state information; the limited adjustment amount of increasing, decreasing or maintaining at least one of the parameters of needle thread tension, looper thread tension, presser foot pressure, differential ratio, feed speed and feed dog height can be used as action; and the degree of improvement in sewing quality can be used as a positive evaluation factor.

[0088] In this embodiment, regardless of whether a preset rule or a reinforcement learning model is used, the fine-tuned parameter set will only be written into the standard fabric sample as valid verification parameters after the fine-tuned parameters are actually issued, sewing is performed again, and the online quality inspection verifies that the parameters are qualified.

[0089] In step S800, if the quality inspection result after fine-tuning still indicates that the sewing quality is unqualified, a prompt message is output to remind the operator to make manual intervention adjustments.

[0090] If the sewing parameters are verified and valid after manual adjustment, new standard fabric samples can be formed by adding new samples in a closed-loop manner on the production site.

[0091] The above method will be explained below with a specific production case.

[0092] The operator acquires a new blended fabric that has not been processed on the current production line before and scans the fabric using a handheld scanning terminal before sewing begins. The handheld scanning terminal integrates a near-infrared spectral sensor, a mechanical sensing unit, and a thickness sensor, and generates a multi-dimensional feature vector of the fabric after completing the corresponding detection.

[0093] The multidimensional feature vector is transmitted to the workshop edge computing server. The system queries the local fabric-process knowledge graph database. In a specific example, no historical standard fabric sample with a similarity greater than 95% is found. Therefore, the multidimensional feature vector is input into the trained deep neural network model, which outputs a set of recommended sewing parameters for the current fabric. In this example, the recommended sewing parameters may include: needle thread tension 1N, looper thread tension 0.1N, presser foot pressure 28N, differential ratio 1.2:1, and feed dog height 1.2mm.

[0094] The recommended sewing parameters are automatically sent to the corresponding intelligent sewing machine, and the operator starts sewing after loading the material. During the sewing process, a miniature industrial camera placed near the needle plate continuously collects stitch images; when the image recognition results indicate that the stitches are uniform, without skipped stitches or loose threads, and the spindle current runs smoothly, the sewing quality can be determined to be qualified. The system then packages and encrypts the multidimensional feature vector of the blended fabric with the valid sewing parameters, storing it as a new standard fabric sample that has passed the quality verification on the production site in the fabric-process knowledge graph database, which can be used for subsequent incremental model training.

[0095] When using the same or highly similar fabric as the standard fabric sample again, the system can directly call up the verified historical sewing parameters through similarity matching, without having to re-enter the parameter recommendation model inference path. In another scenario, if a slight wrinkling defect is detected in the fabric after sewing, the system can fine-tune the differential ratio and needle tension according to preset rules; for example, the differential ratio can be adjusted from 1.2:1 to 1.3:1, and the needle tension can be adjusted to 0.9N. After re-sewing to verify that the defect has been eliminated, the adjusted fabric feature-parameter data pair is added to the database as a new standard fabric sample.

[0096] In another specific embodiment, the workshop acquires another batch of fabrics highly similar to historical standard samples. Before sewing begins, the fabric information acquisition module obtains the fiber composition, blending ratio, mechanical properties, and thickness information of this batch of fabrics, forming a multi-dimensional feature vector. The system calculates the similarity between this multi-dimensional feature vector and each standard fabric sample in the database. When the maximum similarity reaches or exceeds a preset threshold, the target standard fabric sample corresponding to the maximum similarity is directly selected, and its complete set of sewing parameters, verified as effective in actual production, is invoked. Both decision paths share the same pre-sewing fabric multi-dimensional feature vector as input, and ultimately both enter the parameter distribution, actual sewing, online quality inspection, and standard sample closed-loop update stages.

[0097] Please see Figure 8 The diagram shown is an architectural block diagram of a fabric adaptive sewing parameter recommendation system in another embodiment of this application.

[0098] like Figure 8 As shown in the figure, this application embodiment also provides a fabric adaptive sewing parameter recommendation system, the system including a fabric adaptive sewing parameter recommendation device and a sewing machine.

[0099] In one embodiment of this application, the fabric adaptive sewing parameter recommendation device may include a fabric information acquisition module, a feature vector generation module, a sewing parameter acquisition module, and a sewing parameter execution module.

[0100] The fabric information acquisition module is located at the feeding position of the sewing equipment or integrated into an independent handheld scanning terminal, and includes a near-infrared spectral sensor, a mechanical sensing unit, and a thickness sensor. It is configured to perform non-destructive pre-inspection on the fabric to be sewn before the sewing operation begins, so as to obtain the original sensing signal of the fabric to be sewn.

[0101] The feature vector generation module is configured to preprocess and extract features from the original sensing signal to generate a multi-dimensional feature vector characterizing the inherent properties of the fabric to be sewn.

[0102] The sewing parameter acquisition module is configured to calculate the similarity between the multidimensional feature vector and each standard fabric sample stored in a pre-constructed fabric-process knowledge graph database. Each standard fabric sample records a multidimensional feature vector of the fabric and its corresponding verified and valid sewing parameter set. When there is at least one standard fabric sample with a similarity exceeding a preset threshold, a target standard fabric sample is determined from the at least one standard fabric sample, and the verified and valid sewing parameter set corresponding to the target standard fabric sample is called. When there is no standard fabric sample with a similarity exceeding the preset threshold, the multidimensional feature vector is input into a trained parameter recommendation model, and an initial sewing parameter set is output.

[0103] The sewing parameter execution module is configured to send the verified and valid sewing parameter set or the initial sewing parameter set as the sewing parameter set to be executed to each mechanism controller of the target sewing device, so as to complete the sewing parameter configuration of the target sewing device and execute the sewing operation.

[0104] It should be noted that the working principle of each module of the system in this embodiment corresponds one-to-one with the corresponding steps in the aforementioned fabric adaptive sewing parameter recommendation method, so the same content will not be repeated here.

[0105] Please see Figure 9 The image shown is a schematic diagram of the structure of an electronic device in one embodiment of this application.

[0106] like Figure 9 As shown, this application provides an electronic device, including:

[0107] A memory for storing computer programs.

[0108] A processor, the processor being configured to execute a computer program stored in the memory, so as to cause the electronic device to perform the method described in any of the preceding descriptions.

[0109] Preferably, the processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0110] The memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0111] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, or methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, and the division of modules or units is only a logical functional division; other division methods may be used in actual implementation. Multiple modules or units may be combined or integrated into another system, or some functions may be executed by the same processing node according to actual needs.

[0112] Those skilled in the art will also recognize that the example units and method steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functionality in the foregoing description; whether these functions are implemented in hardware or software can be determined based on the specific application and design constraints.

[0113] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in any of the above embodiments. Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing a processor. The program can be stored in a non-transitory computer-readable storage medium, such as random access memory, read-only memory, flash memory, hard disk, solid-state drive, magnetic tape, optical disk, and any combination thereof.

[0114] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.

[0115] This application can be applied to flexible production scenarios with multiple varieties and small batches, such as clothing and home textiles. It can also be used as a fabric process sensing and automatic configuration module for intelligent sewing units to support continuous production with reduced reliance on manual labor and intelligent manufacturing applications such as "lights-out factories".

[0116] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A method for recommending adaptive sewing parameters for fabrics, characterized in that, include: Before the sewing operation begins, a non-destructive pre-inspection is performed on the fabric to be sewn to obtain the original sensor signals of the fabric. The original sensing signal is preprocessed and features are extracted to generate a multi-dimensional feature vector characterizing the ontological properties of the fabric to be sewn. The multidimensional feature vectors are compared with the similarity of each standard fabric sample stored in the pre-constructed fabric-process knowledge graph database. Each standard fabric sample records at least one multidimensional feature vector and its corresponding verified set of effective sewing parameters. When there is at least one standard fabric sample with a similarity exceeding a preset threshold, a target standard fabric sample is determined from the at least one standard fabric sample, and the verified set of effective sewing parameters corresponding to the target standard fabric sample is called. When there is no standard fabric sample with a similarity exceeding the preset threshold, the multidimensional feature vectors are input into the trained parameter recommendation model, and an initial set of sewing parameters is output. The verified and valid sewing parameter set or the initial sewing parameter set is sent as the sewing parameter set to be executed to each mechanism controller of the target sewing equipment, so as to complete the sewing parameter configuration of the target sewing equipment and execute the sewing operation.

2. The method according to claim 1, characterized in that, Before the sewing operation begins, a non-destructive pre-inspection is performed on the fabric to be sewn to obtain the original sensor signals of the fabric, including: The fiber composition and blending ratio of the fabric to be sewn are non-destructively detected using a near-infrared spectroscopy sensor to obtain information on the chemical composition of the fabric. The tensile-recovery curve, bending stiffness, and surface friction coefficient of the fabric to be sewn are acquired through the micro-force sensor and displacement encoder in the mechanical sensing unit. The thickness of the fabric to be sewn is obtained by non-contact detection using a thickness sensor, which is a laser displacement sensor or a capacitive sensor.

3. The method according to claim 1, characterized in that, Preprocessing and feature extraction of the original sensor signal to generate a multidimensional feature vector characterizing the ontological properties of the fabric to be sewn includes: The original sensor signal is preprocessed by filtering, noise reduction, and data synchronization. Spectral characteristic peaks, elastic modulus, plastic deformation, bending stiffness, surface friction coefficient, thickness, and thickness uniformity are extracted from the preprocessed original sensor signal. The extracted features are combined to generate a multidimensional feature vector corresponding to the fabric to be sewn.

4. The method according to claim 1, characterized in that: The standard fabric samples stored in the fabric-process knowledge graph database include standard fabric samples obtained from expert calibration experiments and new fabric samples added in a closed loop based on online quality inspection results at the production site and verified. The verified and valid sewing parameter set is a set of sewing parameters that has been verified and qualified through expert calibration experiments or online quality testing on the production site, including needle thread tension, looper thread tension, presser foot pressure, differential ratio, feed speed, and feed dog height. The parameter recommendation model is trained using at least one machine learning model among support vector machine, random forest and deep neural network. The parameter recommendation model takes the multidimensional feature vector as input and the initial sewing parameter set as output.

5. The method according to claim 1, characterized in that, Also includes: The sewing quality is monitored online during the sewing process to obtain the quality inspection results; When the quality inspection result indicates that the sewing quality is qualified, the multidimensional feature vector and the sewing parameter set actually used in this sewing operation and verified by the online detection are stored as a new standard fabric sample in the fabric-process knowledge graph database. When the quality inspection result indicates the presence of a sewing defect and the sewing defect meets the parameter fine-tuning conditions according to the preset rules, at least one sewing parameter in the sewing parameter set used in this sewing operation is fine-tuned according to the preset rules, and the sewing operation and the online inspection are performed again based on the fine-tuned sewing parameter set; when the fine-tuned quality inspection result indicates that the sewing quality is qualified, the multidimensional feature vector and the fine-tuned sewing parameter set are stored as a new standard fabric sample in the fabric-process knowledge graph database; If the quality inspection results after fine-tuning still indicate that the sewing quality is unqualified, a prompt message will be output to remind the operator to make manual intervention adjustments.

6. The method according to claim 5, characterized in that, Online quality control during the sewing process includes: Images of the stitches during the sewing process are captured using a miniature industrial camera, and image recognition algorithms are used to identify whether there are sewing defects in the stitch images; the sewing defects include skipped stitches, loose threads, and uneven stitches; The spindle current signal and power fluctuation signal of the target sewing equipment are collected, and the sewing quality is evaluated based on the spindle current signal and the power fluctuation signal.

7. A fabric adaptive sewing parameter recommendation system, characterized in that, include: The fabric adaptive sewing parameter recommendation device is configured to perform the fabric adaptive sewing parameter recommendation method as described in any one of claims 1 to 6; The sewing equipment is communicatively connected to the fabric adaptive sewing parameter recommendation device, and is configured to receive the sewing parameter set output by the fabric adaptive sewing parameter recommendation device, and to perform sewing parameter configuration and sewing operations according to the sewing parameter set.

8. The system according to claim 7, characterized in that, The fabric adaptive sewing parameter recommendation device includes: The fabric information acquisition module is set at the feeding position of the sewing equipment or integrated into an independent handheld scanning terminal, and includes a near-infrared spectral sensor, a mechanical sensing unit and a thickness sensor. It is configured to perform non-destructive pre-inspection on the fabric to be sewn before the sewing operation begins, so as to obtain the original sensing signal of the fabric to be sewn. The feature vector generation module is configured to preprocess and extract features from the original sensing signal to generate a multi-dimensional feature vector characterizing the ontological properties of the fabric to be sewn. The sewing parameter acquisition module is configured to calculate the similarity between the multidimensional feature vector and each standard fabric sample stored in a pre-constructed fabric-process knowledge graph database. Each standard fabric sample records a multidimensional feature vector of the fabric and its corresponding verified and valid sewing parameter set. When there is at least one standard fabric sample with a similarity exceeding a preset threshold, a target standard fabric sample is determined from the at least one standard fabric sample, and the verified and valid sewing parameter set corresponding to the target standard fabric sample is called. When there is no standard fabric sample with a similarity exceeding the preset threshold, the multidimensional feature vector is input into a trained parameter recommendation model, and an initial sewing parameter set is output. The sewing parameter execution module is configured to send the verified and valid sewing parameter set or the initial sewing parameter set as the sewing parameter set to be executed to each mechanism controller of the target sewing device, so as to complete the sewing parameter configuration of the target sewing device and execute the sewing operation.

9. An electronic device, characterized in that, include: The memory is used to store computer programs; A processor for executing a computer program stored in the memory to cause the electronic device to perform the method of any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.

Images