Control system and method for an automatic toe-puff device

Abstract

The application provides a control system and method of an automatic sock moving and toe sewing device, which realizes intelligent closed-loop control of the whole process of sock moving, turning, alignment and sewing by introducing multi-sensor fusion perception, model predictive control, machine learning algorithm and real-time optimization strategy. The toe position and grabbing state are collected in real time by a visual sensor and a force sensor, and edge recognition and alignment detection are performed in combination with a convolutional neural network; a model predictive control algorithm is used to dynamically optimize the turning tube motion trajectory and the sewing seat rotation speed, effectively avoiding turning wrinkles and toe misalignment; a reinforcement learning model is used to adaptively adjust the grabbing force and sewing parameters according to historical data, improving the adaptability to different sock materials; the application significantly improves the intelligent level and process consistency of the sock moving and toe sewing device, overcomes the problems of relying on manual experience, poor adaptability and insufficient precision in the traditional control mode, and has good industrial application prospect.

Description

Technical Field

[0001] This invention relates to the field of automatic sock machines, and in particular to a control system and method for an automatic sock-sewing device. Background Technology

[0002] As an everyday knitted item, socks' production process mainly includes three major steps: sock body knitting, turning, and toe sewing. With the improvement of automation levels in textile machinery, fully automatic sock machines are widely used in production lines. Among these, the degree of automation in the sock transfer and toe sewing stages directly affects production efficiency and product quality. Traditional sock machines typically use separate sock transfer and toe sewing mechanisms. After the sock body is knitted, it needs to be transferred to the toe sewing station using devices such as transfer trays and suction tubes. Then, it is turned inside out manually or mechanically, the toe edges are aligned, and then sewn.

[0003] In existing technologies, such as the automatic sock-sewing device disclosed in Chinese patent CN113215731A, the sock body is gripped by a toothed needle and a gripping needle, the turning tube automatically turns the sock inside out, the toothed needle disc flips and aligns with the edge of the sock toe, and then the toe is sewn by a rotating sewing mechanism. Although this device achieves automation to a certain extent, its control method is still relatively basic, relying heavily on preset programs and simple servo drives. It lacks the ability to adapt to variables such as sock material, thickness, and elasticity, and cannot dynamically adjust according to real-time working conditions.

[0004] In practical applications, existing technologies have the following problems:

[0005] 1. Insufficient control precision: Traditional control systems have difficulty adjusting the gripping force, turning speed or sewing parameters in real time according to the physical characteristics of socks such as thickness and elasticity, which can easily lead to problems such as unstable gripping of the sock toe, wrinkles at the edge after turning, and misalignment of the seam.

[0006] Second, lack of intelligent sensing and feedback mechanisms: Existing devices rely heavily on position sensors and timing control, lacking intelligent sensing methods such as visual recognition and force feedback, and cannot accurately determine whether the toe edge of the sock is aligned, whether the inside is turned out completely, and whether the stitching is flat.

[0007] 3. Poor adaptability: When dealing with socks of different specifications and materials, frequent manual adjustments to equipment parameters are required, resulting in low production efficiency and difficulty in ensuring consistency.

[0008] Fourth, the lack of algorithm models: The existing control system does not introduce predictive models, adaptive algorithms or machine learning methods, which makes it impossible to achieve self-optimization and long-term learning of process parameters, thus restricting the improvement of the intelligence level of the equipment.

[0009] Therefore, there is an urgent need in this field for an intelligent control system with sensing, decision-making and adaptive capabilities to achieve precise control of the entire process of the automatic sock-sewing device, thereby improving the applicability of the equipment, production efficiency and product quality. Summary of the Invention

[0010] The purpose of this invention is to provide a control system and method for an automatic sock-sewing device to solve the problems existing in the prior art.

[0011] To achieve the above objectives, the present invention provides the following solution:

[0012] This invention provides a control system for an automatic sock-sewing device, comprising:

[0013] The sensor module is used to collect real-time data on sock body position, toe edge image, gripping force, turn-up tube position, and sewing resistance.

[0014] The main control module, connected to the sensor module, is used to run control algorithms, process sensor data, and generate control commands.

[0015] An actuator drive module, connected to the main control module, is used to drive the actuators of the sock-shifting mechanism, the turning mechanism, and the sewing mechanism.

[0016] The data storage and processing module is used to store operation data, model parameters, and control logs;

[0017] The human-computer interaction module is used for parameter setting, status monitoring, and alarm handling.

[0018] The communication module is used for data interaction with the host system or cloud platform.

[0019] Preferably, the sensor module includes:

[0020] A visual sensor is used to capture images of the toe edge of a sock;

[0021] Position sensors are used to detect the positions of the toothed needle disc, flipping tube, and suture seat;

[0022] Force sensor used to measure the gripping force of the needle and the suturing resistance of the needle;

[0023] A proximity sensor is used to detect whether the sock is in place.

[0024] Preferably, the main control module has the following built-in features:

[0025] A convolutional neural network model for sock toe edge recognition and alignment detection;

[0026] Model predictive control algorithm is used to optimize the lifting trajectory of the flipping tube and the rotation speed of the sewing seat;

[0027] A reinforcement learning model is used to adaptively adjust the gripping force and stitching parameters;

[0028] Kalman filtering algorithm is used for multi-sensor data fusion and state estimation.

[0029] Preferably, the cost function of the model predictive control algorithm is:

[0030] ;

[0031] in, To predict the time domain, The height of the flip tube. For the target height, For speed weighting coefficients, For speed, For acceleration weighting coefficients, It is acceleration.

[0032] Preferably, the reward function of the reinforcement learning model is:

[0033] ;

[0034] in, To score the quality of sutures, Total operation time. This is the time weighting coefficient.

[0035] Preferably, the actuator drive module includes:

[0036] The film gripper needle driver is used to control the opening / closing and raising / lowering of the film gripper needle;

[0037] A toothed needle disc driver is used to control the flipping of the semi-circular disc and the sock-removing structure;

[0038] Flipping tube driver, used to control the lifting and lowering movement of the flipping tube;

[0039] The suture seat driver is used to control the rotation of the suture seat and the movement of the automatic suture needle.

[0040] The present invention also provides a control method for an automatic sock-sewing device, comprising the following steps:

[0041] S1. System initialization, loading pre-trained model, performing sensor calibration and actuator zeroing;

[0042] S2. Locate the edge of the sock toe using a vision sensor, and control the gripping needle to grab the sock body and transfer it to the toothed needle;

[0043] S3. Control the rising of the turning tube to complete the turning of the sock body, and detect the turning quality in real time;

[0044] S4. Control the toothed needle disc to rotate so that the edge of the sock toe is aligned, detect the alignment and adjust it;

[0045] S5. Control the rotation of the sewing seat and the movement of the automatic sewing needle to complete the toe sewing;

[0046] S6. Visually inspect the suture quality, record the data, and update the control model.

[0047] Preferably, in step S2, a convolutional neural network model is used for sock toe edge recognition, and its loss function is cross-entropy loss, expressed as:

[0048] ;

[0049] in, For real labels, For predicted values, This represents the number of samples.

[0050] Preferably, in step S3, the height control of the flipping tube adopts model predictive control, and its state-space model is as follows:

[0051] ;

[0052] in, Let be the derivative of the height of the flipping tube with respect to time. Let be the derivative of the flipping tube speed with respect to time. It is acceleration.

[0053] Preferably, in step S6, an LSTM network is used to predict the suture quality. If the quality score is lower than the threshold, the suture parameters are adjusted and the suture is re-sutured.

[0054] The present invention achieves the following beneficial technical effects compared to the prior art:

[0055] This invention provides a control system and method for an automatic sock-shifting and sewing device. By introducing multi-sensor fusion sensing, model predictive control, machine learning algorithms, and real-time optimization strategies, it achieves intelligent closed-loop control of the entire process of sock shifting, turning, alignment, and sewing. Visual and force sensors are used to collect the sock toe position and gripping status in real time, combined with convolutional neural networks for edge recognition and alignment detection. A model predictive control algorithm dynamically optimizes the turning tube's trajectory and the sewing seat's rotation speed, effectively avoiding turning wrinkles and sewing misalignment. A reinforcement learning model adaptively adjusts the gripping force and sewing parameters based on historical data, improving adaptability to different sock materials. This invention significantly improves the intelligence level and process consistency of the sock-shifting and sewing device, overcoming the problems of reliance on human experience, poor adaptability, and insufficient precision in traditional control methods, and has promising prospects for industrial applications. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 Control relationship diagram of the control system of the automatic sock-sewing device provided by the present invention;

[0058] Figure 2 A flowchart illustrating the control method for the automatic sock-sewing device provided by this invention. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] The purpose of this invention is to provide a control system and method for an automatic sock-sewing device to solve the problems existing in the prior art.

[0061] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0062] Example 1:

[0063] like Figure 1 As shown, the control system of the automatic sock-sewing device provided by this invention consists of a sensor module, a main control module, an actuator drive module, a data storage and processing module, a human-machine interaction module, and a communication module. These modules are interconnected via an industrial bus (such as EtherCAT or CAN) to form a closed-loop control system.

[0064] The sensor module is responsible for collecting production process data in real time. This includes a vision sensor that uses a 5-megapixel industrial camera to capture images of the sock toe edge; a position sensor that uses an absolute encoder to monitor the precise positions of the toothed needle disc, turning tube, and seam stop; a force sensor that uses a miniature strain gauge to measure the gripping force of the gripping needle and the sewing resistance of the sewing needle; and a proximity sensor that uses a photoelectric sensor to detect the sock's positioning. This sensor data is transmitted to the main control module in real time via a bus.

[0065] Furthermore, the main control module employs an embedded industrial computer, equipped with an Intel i7 processor and an NVIDIA Jetson GPU, running a real-time operating system. This module integrates multiple intelligent algorithms: a convolutional neural network model using the U-Net architecture is used for sock toe edge recognition and alignment detection, with its training data containing over 100,000 labeled images of socks of different materials; a model predictive control algorithm is used to optimize the trajectory of the turning tube, with a prediction time domain set to 2 seconds; a reinforcement learning model employs a proximal policy optimization algorithm, pre-trained in a simulation environment and then transferred to the actual system; and a Kalman filter algorithm is used to fuse multi-sensor data to improve state estimation accuracy. The main control module processes real-time data, generates control commands, and sends them to the actuator driver module.

[0066] Furthermore, the actuator drive module includes four dedicated drivers: a gripper needle driver that controls the opening / closing of the gripper needle, using a servo motor to achieve a positioning accuracy of ±0.1mm; a geared needle disc driver that controls the rotation of the semi-circular disc, achieving a precise 180° rotation via a geared motor; a flipping tube driver that controls the lifting motion, using a linear motor to achieve a repeatability of 0.01mm; and a suture seat driver that controls the rotational motion and the movement of the automatic suture needle, using a servo system to achieve 0.1° angle control. All these drivers communicate with the main control module via the EtherCAT protocol, achieving microsecond-level synchronous control.

[0067] Furthermore, the data storage and processing module uses a 1TB solid-state drive to store historical process parameters, quality data, and model versions, while simultaneously performing data preprocessing and feature extraction via Python scripts. This module automatically performs model retraining weekly, using newly acquired production data to optimize control parameters.

[0068] Furthermore, the human-machine interaction module is equipped with a 10-inch touchscreen. The operation interface, developed based on the Qt framework, can display the toe alignment, stitch quality score, and equipment status in real time, and supports online modification of process parameters and fault diagnosis.

[0069] Furthermore, the communication module interfaces with the factory's MES system via the OPC UA protocol, uploading production statistics and equipment utilization data every 5 minutes, and receiving production order information.

[0070] Example 2:

[0071] like Figure 2 As shown, the control method of the present invention is implemented according to the following process: After the system starts, an initialization program is first executed, including visual sensor calibration, force sensor zero-position calibration, and zero-return operation of each actuator. The pre-trained CNN model and reinforcement learning strategy are loaded from the storage module into memory, and the initialization time is usually no more than 30 seconds.

[0072] During the sock-grabbing phase, a vision sensor acquires images of the needle area at a rate of 30 frames per second, and a CNN model performs real-time inference to identify the position of the sock toe edge. Once the edge coordinates are detected, the main control module calls a trajectory planning algorithm to calculate the optimal movement path of the needle based on a reinforcement learning-optimized objective function.

[0073] ;

[0074] in, For the real-time position of the film-grabbing needle, For the target location, To control the input, Weighting coefficient (default value 0.1). During the movement of the gripping pin, the force sensor monitors the gripping force in real time, and the PID controller dynamically adjusts the output:

[0075] ;

[0076] in, For a moment The force error, , , The values ​​are 1.2, 0.5, and 0.1 respectively.

[0077] During the turning stage, the transfer arm moves the sock body directly above the turning tube, and the sock body enters the tube cavity through an opening in the tube wall. As the turning tube rises, the MPC controller operates according to the state-space model:

[0078] ;

[0079] Calculate the optimal control quantity in real time, where... Let be the derivative of the height of the flipping tube with respect to time. Let be the derivative of the flipping tube speed with respect to time. For acceleration. The cost function is defined as:

[0080] ;

[0081] in, To predict the time domain, The height of the flip tube. For the target height, For speed weighting coefficients, For speed, For acceleration weighting coefficients, For acceleration, weighting coefficient , , The values ​​were set to 1.0, 0.5, and 0.2 respectively. During the turning process, the vision system continuously monitored the degree of wrinkles on the sock body. When the wrinkle area exceeded the threshold (set to 5%), the system automatically reduced the rising speed by 20%.

[0082] During the sock toe edge alignment stage, the active semi-circular disc of the toothed needle plate rotates downwards by 180°, causing the sock toe edges to align in a semi-circular shape. The vision system acquires the alignment image, fits the edge curve using a Hough transform, and calculates the alignment error.

[0083] ;

[0084] in,( , ) represents the coordinates of the edge point, ( , ( ) is the center of the fitted circle. The radius is used. When the error exceeds 1.5mm, the system uses the gradient descent method to adjust the flip angle.

[0085] ;

[0086] Learning rate Setting it to 0.01 usually achieves the required accuracy after 2-3 iterations.

[0087] During the sewing stage, the toothed needle disc transfers the edge of the sock toe onto the fixed needles of the sewing station. The sewing station operates at a reference speed. =120rpm rotation, while dynamically adjusting the rotation speed based on the stitching resistance measured by the force sensor:

[0088] ;

[0089] Adaptive gain coefficient A value of 10 is used to ensure that the resistance is controlled within the range of 0.5-1.5N. The LSTM network predicts the suture quality in real time. The input sequence contains speed and resistance data from the most recent 50 cycles, and the output is a quality score Q. When Q < 0.8, the system automatically adjusts the suture needle trajectory and re-sutures the current segment.

[0090] During the quality inspection phase, the vision system acquires images of the stitching area and performs defect detection using a CNN classifier. Simultaneously, the system applies a reward function:

[0091] ;

[0092] Update reinforcement learning strategies, among which A value of 0.1 is used to encourage increased efficiency while ensuring quality. All process data is stored in a database for model iterative optimization.

[0093] In practical applications, for 1.5mm thick stockings, the system automatically sets the closing interval to 28mm, the turning tube rising speed to 40mm / s, and the entire processing cycle to approximately 3.2 seconds, with a consistently high stitching quality score above 0.92. Through continuous learning, the system can adapt to stockings of different materials and specifications, significantly improving production adaptability and quality consistency.

[0094] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0095] It should be noted that the components mentioned in the above embodiments are all general standard parts or components known to those skilled in the art. Their structures and principles can be learned by those skilled in the art through technical manuals or conventional experimental methods.

[0096] This invention has illustrated its principles and implementation methods using specific examples. The descriptions of these embodiments are merely illustrative of the method and its core ideas; furthermore, those skilled in the art will recognize that modifications may be made to the specific implementation methods and application scope based on the principles of this invention. Therefore, the content of this specification should not be construed as limiting the invention.

Claims

1. A control system for an automatic sock-sewing device, characterized in that, include: The sensor module is used to collect real-time data on sock body position, toe edge image, gripping force, turn-up tube position, and sewing resistance. The main control module, connected to the sensor module, is used to run control algorithms, process sensor data, and generate control commands. An actuator drive module, connected to the main control module, is used to drive the actuators of the sock-shifting mechanism, the turning mechanism, and the sewing mechanism. The data storage and processing module is used to store operation data, model parameters, and control logs; The human-computer interaction module is used for parameter setting, status monitoring, and alarm handling. The communication module is used for data interaction with the host system or cloud platform.

2. The control system of the automatic sock-sewing device according to claim 1, characterized in that, The sensor module includes: A visual sensor is used to capture images of the toe edge of a sock; Position sensors are used to detect the positions of the toothed needle disc, flipping tube, and suture seat; Force sensor used to measure the gripping force of the needle and the suturing resistance of the needle; A proximity sensor is used to detect whether the sock is in place.

3. The control system of the automatic sock-sewing device according to claim 1, characterized in that, The main control module has the following built-in features: A convolutional neural network model for sock toe edge recognition and alignment detection; Model predictive control algorithm is used to optimize the lifting trajectory of the flipping tube and the rotation speed of the sewing seat; A reinforcement learning model is used to adaptively adjust the gripping force and stitching parameters; Kalman filtering algorithm is used for multi-sensor data fusion and state estimation.

4. The control system of the automatic sock-sewing device according to claim 3, characterized in that, The cost function of the model predictive control algorithm is: ； in, To predict the time domain, The height of the flip tube. For the target height, For speed weighting coefficients, For speed, For acceleration weighting coefficients, It is acceleration.

5. The control system of the automatic sock-sewing device according to claim 3, characterized in that, The reward function of the reinforcement learning model is: ； in, To score the quality of sutures, Total operation time. This is the time weighting coefficient.

6. The control system of the automatic sock-sewing device according to claim 1, characterized in that, The actuator drive module includes: The film gripper needle driver is used to control the opening / closing and raising / lowering of the film gripper needle; A toothed needle disc driver is used to control the flipping of the semi-circular disc and the sock-removing structure; Flipping tube driver, used to control the lifting and lowering movement of the flipping tube; The suture seat driver is used to control the rotation of the suture seat and the movement of the automatic suture needle.

7. A control method for an automatic sock-sewing device, characterized in that, Includes the following steps: S1. System initialization, loading pre-trained model, performing sensor calibration and actuator zeroing; S2. Locate the edge of the sock toe using a vision sensor, and control the gripping needle to grab the sock body and transfer it to the toothed needle; S3. Control the rising of the turning tube to complete the turning of the sock body, and detect the turning quality in real time; S4. Control the toothed needle disc to rotate so that the edge of the sock toe is aligned, detect the alignment and adjust it; S5. Control the rotation of the sewing seat and the movement of the automatic sewing needle to complete the toe sewing; S6. Visually inspect the suture quality, record the data, and update the control model.

8. The control method for the automatic sock-sewing device according to claim 7, characterized in that, In step S2, a convolutional neural network model is used for sock toe edge recognition, and its loss function is cross-entropy loss, expressed as: ； in, For real labels, For predicted values, This represents the number of samples.

9. The control method for the automatic sock-sewing device according to claim 7, characterized in that, In step S3, the height control of the flipping tube adopts model predictive control, and its state-space model is as follows: ； in, Let be the derivative of the height of the flipping tube with respect to time. Let be the derivative of the flipping tube speed with respect to time. It is acceleration.

10. The control method for the automatic sock-sewing device according to claim 7, characterized in that, In step S6, an LSTM network is used to predict the suture quality. If the quality score is lower than the threshold, the suture parameters are adjusted and the suture is re-sutured.

Images