Electronic jacquard machine control method and system based on multi-channel cooperation

Abstract

The invention discloses an electronic jacquard machine control method and system based on multi-channel collaboration, and the method comprises the steps: calculating a delay compensation value of each channel in real time according to the hardware performance difference and signal transmission delay data of each channel, and dynamically adjusting the time sequence offset of a control instruction of an electronic jacquard machine by using the delay compensation value, generating a multi-channel cooperative control signal; according to input fabric pattern data, decomposing the fabric pattern data to generate a layered instruction set, and binding the layered instruction set to a multi-channel cooperative control signal to form a time-space synchronization instruction set; according to channel operation state data acquired in real time, identifying an abnormal channel and a fault type thereof, and generating a reconstruction instruction set after fault immunization; and dynamically allocating computing resources and storage bandwidth according to the reconstruction instruction set, generating an optimization execution instruction set, and issuing the optimization execution instruction set to each channel for execution in real time. By utilizing the embodiment of the invention, the control capability of the electronic jacquard machine and the fabric production efficiency can be improved, and the intelligent development of the textile industry is promoted.

Description

Technical Field

[0001] The present invention belongs to the field of control technology, and particularly relates to a control method and system for an electronic jacquard machine based on multi-channel collaboration. Background Art

[0002] With the development of the textile industry, as an important device for fabric pattern production, the electronic jacquard machine urgently needs to improve its flexibility and precise control ability to meet the requirements of complex weaving. However, there are many deficiencies in the existing control systems in terms of channel coordination and instruction synchronization, mainly manifested as problems of delay and asynchronization, uneven instruction distribution, insufficient fault detection and response, and limited resource scheduling ability. These problems lead to troubles such as timing offset, uneven load, slow fault response, and low resource utilization efficiency during the weaving process. Summary of the Invention

[0003] The purpose of the present invention is to provide a control method and system for an electronic jacquard machine based on multi-channel collaboration to solve the deficiencies in the prior art, which can improve the control ability of the electronic jacquard machine and the fabric production efficiency, and promote the intelligent development of the textile industry.

[0004] An embodiment of the present application provides a control method for an electronic jacquard machine based on multi-channel collaboration, and the method includes: According to the hardware performance differences and signal transmission delay data of each channel, adopt a dynamic delay compensation algorithm based on reinforcement learning to calculate the delay compensation value of each channel in real time, and through an adaptive clock synchronization technology, combined with the feedback signal between channels, use the delay compensation value to dynamically adjust the timing offset of the control instructions of the electronic jacquard machine to generate a multi-channel collaborative control signal; According to the input fabric pattern data, adopt an instruction decomposition model based on a graph neural network to decompose the fabric pattern data. The decomposition is carried out through a spatial topology mapping algorithm, combined with the channel physical layout characteristics, to dynamically allocate the pattern pixel points to the corresponding channels, generate a hierarchical instruction set matching the channel physical positions, and bind it to the multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set; According to the real-time collected channel operation status data, adopt a fault diagnosis model based on timing anomaly detection to identify the abnormal channels and their fault types, and through virtual channel redundancy technology, based on the spatio-temporal synchronization instruction set, isolate the instruction segments of the abnormal channels, and combined with the channel spatial topology relationship, dynamically allocate the instructions of the abnormal channels to the adjacent channels to generate a reconstructed instruction set after fault immunity; According to the reconstructed instruction set, a multi-objective resource scheduling algorithm based on game theory is adopted. Combining the channel load status and the priorities of knitting tasks, computing resources and storage bandwidth are dynamically allocated. Among them, the scheduling optimizes the instruction execution timing through an energy consumption - efficiency trade-off model, generates an optimized execution instruction set, and issues it to each channel in real time for execution, so as to achieve the control of an electronic jacquard machine based on multi-channel collaboration.

[0005] Optionally, according to the hardware performance differences and signal transmission delay data of each channel, a dynamic delay compensation algorithm based on reinforcement learning is adopted. The delay compensation value of each channel is calculated in real time, and through an adaptive clock synchronization technology, combined with the feedback signals between channels, the timing offset of the control instructions of the electronic jacquard machine is dynamically adjusted using the delay compensation value to generate a multi-channel collaborative control signal, including: According to the hardware performance differences and signal transmission delay data of each channel, a distributed data acquisition technology is adopted to obtain the signal transmission delay of each channel in real time. Through a statistical analysis model, the average delay and fluctuation range of each channel are calculated to generate a preliminary delay analysis result. For the preliminary delay analysis result, a dynamic delay compensation algorithm based on reinforcement learning is adopted. Combining the feedback signals between channels, the delay compensation value of each channel is calculated in real time. Through a reward function and a state transition model, the calculation accuracy of the compensation value is optimized to generate a preliminary delay compensation value. For the preliminary delay compensation value, an adaptive clock synchronization technology is adopted. Combining the feedback signals between channels, the timing offset of the control instructions of the electronic jacquard machine is dynamically adjusted. Through clock deviation correction technology, the timing consistency of each channel is ensured to generate a preliminary synchronization control signal. For the preliminary synchronization control signal, a multi-channel collaborative control algorithm is adopted. Combining the delay compensation value, the instruction timing of each channel is dynamically adjusted. Through signal fusion technology, a final multi-channel collaborative control signal is generated.

[0006] Optionally, according to the input fabric pattern data, an instruction decomposition model based on a graph neural network is adopted to decompose the fabric pattern data. The decomposition is carried out through a spatial topology mapping algorithm. Combining the channel physical layout characteristics, the pattern pixel points are dynamically allocated to the corresponding channels to generate a hierarchical instruction set matching the channel physical positions, and it is bound to the multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set, including: According to the input fabric pattern data, an image preprocessing technology is adopted to denoise and normalize the pattern. Through a feature extraction algorithm, the key pixel point information of the pattern is extracted to generate preliminary pattern feature data. For the preliminary pattern feature data, an instruction decomposition model based on a graph neural network is adopted. Combining the spatial topology mapping algorithm, the pattern pixel points are dynamically allocated to the corresponding channels. Through node feature embedding technology, a preliminary hierarchical instruction set is generated. For the preliminary hierarchical instruction set, combined with the physical layout characteristics of the channels, the spatial topology mapping algorithm is used to match the instruction set with the physical positions of the channels. Through the position correction technology, the accuracy of instruction allocation is optimized to generate a preliminary matching instruction set; For the preliminary matching instruction set, the spatio-temporal synchronization algorithm is adopted, combined with the multi-channel collaborative control signals, to bind the instruction set to the control signals. Through the timing verification technology, the final spatio-temporal synchronization instruction set is generated.

[0007] Optionally, according to the channel operation state data collected in real time, a fault diagnosis model based on timing anomaly detection is used to identify the abnormal channels and their fault types. Through the virtual channel redundancy technology, based on the spatio-temporal synchronization instruction set, the instruction segments of the abnormal channels are isolated, and combined with the channel spatial topology relationship, the instructions of the abnormal channels are dynamically allocated to the adjacent channels to generate a reconstructed instruction set after fault immunity, including: According to the channel operation state data collected in real time, the distributed data acquisition technology is adopted to obtain the operation states of each channel, and through the timing analysis model, the preliminary operation state timing data is generated; For the preliminary operation state timing data, a fault diagnosis model based on timing anomaly detection is used, combined with the fault type library, to identify the abnormal channels and their fault types, and through the anomaly scoring technology, the preliminary fault diagnosis result is generated; For the preliminary fault diagnosis result, the virtual channel redundancy technology is adopted, based on the spatio-temporal synchronization instruction set, to isolate the instruction segments of the abnormal channels, and through the instruction reassignment algorithm, the preliminary redundant instruction set is generated; For the preliminary redundant instruction set, combined with the channel spatial topology relationship, the dynamic allocation algorithm is adopted to allocate the instructions of the abnormal channels to the adjacent channels, and through the load balancing technology, the final reconstructed instruction set after fault immunity is generated.

[0008] Optionally, according to the reconstructed instruction set, a multi-objective resource scheduling algorithm based on game theory is adopted, combined with the channel load status and the priority of the weaving tasks, to dynamically allocate computing resources and storage bandwidth. Among them, the scheduling optimizes the instruction execution timing through the energy consumption - efficiency trade-off model to generate an optimized execution instruction set, and issues it to each channel for execution in real time to achieve the control of the electronic jacquard machine based on multi-channel collaboration, including: According to the reconstructed instruction set, the distributed data acquisition technology is adopted to obtain the load status of each channel in real time, and through the load analysis model, the preliminary load status data is generated; For the preliminary load status data, a multi-objective resource scheduling algorithm based on game theory is adopted, combined with the priority of the weaving tasks, to dynamically allocate computing resources and storage bandwidth, and through the game equilibrium model, the preliminary scheduling scheme is generated; For the preliminary scheduling scheme, an energy consumption - efficiency trade - off model is adopted. Combining with the instruction execution timing sequence, resource allocation is optimized. Through a multi - objective optimization algorithm, a preliminary optimized execution instruction set is generated; For the preliminary optimized execution instruction set, a real - time distribution technology is adopted to distribute the instruction set to each channel for execution. Through an execution monitoring technology, a final optimized execution instruction set is generated.

[0009] Another embodiment of the present application provides an electronic jacquard machine control system based on multi - channel collaboration. The system includes: A compensation module, which is used to calculate the delay compensation value of each channel in real time by using a dynamic delay compensation algorithm based on reinforcement learning according to the hardware performance differences and signal transmission delay data of each channel. And through an adaptive clock synchronization technology, combined with the feedback signal between channels, the timing offset of the control instruction of the electronic jacquard machine is dynamically adjusted by using the delay compensation value to generate a multi - channel collaborative control signal; A decomposition module, which is used to decompose the fabric pattern data by using an instruction decomposition model based on a graph neural network according to the input fabric pattern data. The decomposition is carried out through a space - topology mapping algorithm, combined with the channel physical layout characteristics, to dynamically allocate pattern pixel points to the corresponding channels, generate a hierarchical instruction set matching the channel physical position, and bind it to the multi - channel collaborative control signal to form a spatio - temporal synchronization instruction set; An identification module, which is used to identify the abnormal channel and its fault type by using a fault diagnosis model based on timing anomaly detection according to the real - time collected channel operation status data. And through a virtual channel redundancy technology, based on the spatio - temporal synchronization instruction set, the instruction segment of the abnormal channel is isolated, and combined with the channel space - topology relationship, the instructions of the abnormal channel are dynamically allocated to the adjacent channels to generate a reconstructed instruction set after fault immunity; A control module, which is used to dynamically allocate computing resources and storage bandwidth according to the reconstructed instruction set by using a multi - objective resource scheduling algorithm based on game theory, combined with the channel load status and the priority of the weaving task. Among them, the scheduling optimizes the instruction execution timing through an energy consumption - efficiency trade - off model to generate an optimized execution instruction set, and distributes it to each channel for execution in real time to realize the control of the electronic jacquard machine based on multi - channel collaboration.

[0010] Another embodiment of the present application provides a storage medium, in which a computer program is stored. Wherein, the computer program is set to execute the method described in any one of the above when running.

[0011] Another embodiment of the present application provides an electronic device, including a memory and a processor. A computer program is stored in the memory, and the processor is set to run the computer program to execute the method described in any one of the above.

[0012] Compared with the prior art, a control method for an electronic jacquard machine based on multi-channel collaboration provided by the present invention calculates the delay compensation value of each channel in real time according to the hardware performance difference and signal transmission delay data of each channel, and uses the delay compensation value to dynamically adjust the timing offset of the control instruction of the electronic jacquard machine to generate a multi-channel collaborative control signal; according to the input fabric pattern data, the fabric pattern data is decomposed to generate a hierarchical instruction set, which is bound to the multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set; according to the channel operation state data collected in real time, the abnormal channel and its fault type are identified to generate a reconstructed instruction set after fault immunity; according to the reconstructed instruction set, the computing resources and storage bandwidth are dynamically allocated to generate an optimized execution instruction set, which is sent to each channel for execution in real time, so as to improve the control ability of the electronic jacquard machine and the fabric production efficiency, and promote the intelligent development of the textile industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 FIG. is a hardware structure block diagram of a computer terminal for a control method of an electronic jacquard machine based on multi-channel collaboration provided by an embodiment of the present invention; Figure 2 FIG. is a schematic flow chart of a control method of an electronic jacquard machine based on multi-channel collaboration provided by an embodiment of the present invention; Figure 3 FIG. is a schematic structural diagram of a control system of an electronic jacquard machine based on multi-channel collaboration provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] The embodiments described below by referring to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0015] An embodiment of the present invention first provides a control method for an electronic jacquard machine based on multi-channel collaboration. This method can be applied to electronic devices, such as computer terminals, specifically, ordinary computers, etc.

[0016] The following takes running on a computer terminal as an example for detailed description. Figure 1 FIG. is a hardware structure block diagram of a computer terminal for a control method of an electronic jacquard machine based on multi-channel collaboration provided by an embodiment of the present invention. As Figure 1 shown, the computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the memory may include a non-volatile storage medium and an internal memory.

[0017] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions. When the program instructions are executed, the processor can execute any control method of an electronic jacquard machine based on multi-channel collaboration.

[0018] The processor is used to provide computing and control capabilities to support the operation of the entire computer device.

[0019] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any one of the electronic jacquard machine control methods based on multi-channel collaboration.

[0020] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art can understand that Figure 1 The structure shown in is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0021] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0022] See Figure 2 , the embodiments of the present invention provide an electronic jacquard machine control method based on multi-channel collaboration, which may include the following steps: S201, according to the hardware performance differences and signal transmission delay data of each channel, adopt a dynamic delay compensation algorithm based on reinforcement learning to calculate the delay compensation value of each channel in real time, and through an adaptive clock synchronization technology, combined with the feedback signal between channels, use the delay compensation value to dynamically adjust the timing offset of the control instruction of the electronic jacquard machine to generate a multi-channel collaborative control signal; This step adaptively calculates the compensation value by constructing a reinforcement learning model (state = delay feature, action = compensation value adjustment, reward = improved synchronization accuracy) through real-time monitoring of the hardware performance parameters (such as processor speed and memory bandwidth) and signal transmission delays (average delay ± fluctuation range) of each channel. The PTP (Precision Time Protocol) is used to achieve microsecond-level synchronization, combined with a dynamic timing adjustment module with FPGA hardware acceleration, ensuring that the instruction synchronization error of 2048 channels is <50 ns, solving the "pattern misalignment" problem caused by channel delay differences in traditional electronic jacquard machines, increasing the synchronization accuracy of large-scale multi-channel systems by 10 times, and laying a precise control foundation for complex pattern weaving. After testing, this method can reduce the timing deviation between jacquard machine channels from 200 ns to 20 ns.

[0023] Specifically, according to the hardware performance differences and signal transmission delay data of each channel, distributed data acquisition technology can be used to obtain the signal transmission delays of each channel in real time. Through a statistical analysis model, the average delay and fluctuation range of each channel are calculated to generate a preliminary delay analysis result. In this step, a hardware timestamp unit (with an accuracy of 10 ns) is deployed for each channel, and end-to-end delay data (including transmission delay, processing delay, and queuing delay) is collected through a ring topology network. Robust estimation algorithms (Huber regression) are used for statistical analysis. After removing outliers, the mean, variance, and 99% quantile of the delay of each channel are calculated to form a delay feature vector (dimension = 5), establishing a channel delay profile accurate to the microsecond level, providing a data basis for subsequent compensation, and reducing the delay evaluation error from 15% of the traditional method to less than 3%.

[0024] A high-precision timestamp unit is deployed for each control channel of the electronic jacquard machine, and a distributed clock system is established using the IEEE 1588 Precision Time Protocol (PTP). Each channel is equipped with a dedicated delay monitoring chip to record the end-to-end delay from the instruction issuance to the execution completion in real time, including FPGA processing delay, signal transmission delay, and mechanical response delay. The data acquisition frequency is set to 1 kHz and is transmitted back to the central controller through Gigabit Ethernet.

[0025] Robust algorithms are used for delay statistical analysis: First, median filtering is used to remove instantaneous interference, and then the mean, standard deviation, and range of the delay within a sliding window (window size 100 ms) are calculated. For abnormal channels (such as delay fluctuations exceeding 3σ), a resampling mechanism is started to ensure data reliability. The finally generated delay analysis report includes the reference delay, jitter range, and stability score of each channel.

[0026] For example: In a 2048-channel jacquard machine system, it is monitored that the average delay of the 1024th channel is 152 μs ± 8 μs, while that of the 2048th channel is 168 μs ± 12 μs. The system automatically marks the latter as a high-fluctuation channel and gives special treatment in subsequent compensation.

[0027] For the preliminary delay analysis results, a dynamic delay compensation algorithm based on reinforcement learning is adopted. Combining the feedback signals between channels, the delay compensation values of each channel are calculated in real time. Through the reward function and the state transition model, the calculation accuracy of the compensation values is optimized to generate preliminary delay compensation values. In this step, a DDPG reinforcement learning framework is designed. The state space includes 10-dimensional parameters such as delay characteristics and temperature, and the action space is the compensation value (range ±500 μs). The reward function R = 1 - |actual synchronization error| / 50 ns, and the experience replay buffer stores 100,000 groups of transitions. The online learning rate is set to 0.001, and the policy network is updated every 100 ms, enabling the compensation value to dynamically adapt to changes in channel performance (such as delay fluctuations caused by temperature drift), increasing the adaptation speed of the synchronization error by 5 times, and the compensation accuracy reaching ±5 ns.

[0028] Design a reinforcement learning framework based on DDPG. The state space includes 15-dimensional parameters such as the delay characteristics (mean, variance, skewness, etc.), ambient temperature, and workload of each channel. The action space is the adjustment amount of the compensation value for each channel (range ±200 μs). The reward function comprehensively considers the degree of reduction in synchronization error and compensation stability.

[0029] The model training adopts an offline pre-training + online fine-tuning strategy: first pre-train on a historical dataset (including 1 million groups of delay samples), and then continuously optimize through actual operation data. During online learning, the policy network is updated every time 50 groups of new data are received, and the learning rate is set to 0.0001. The compensation value calculation module is deployed on the FPGA to ensure a response speed at the μs level.

[0030] For example: When it is detected that the delay of the 512th channel increases by 5 μs due to temperature increase, the system calculates that an additional 3.8 μs compensation is required within 2 ms and quickly applies it to the timing control of this channel through the look-up table method.

[0031] For the preliminary delay compensation values, an adaptive clock synchronization technology is adopted. Combining the feedback signals between channels, the timing offset of the control instructions of the electronic jacquard machine is dynamically adjusted. Through the clock deviation correction technology, the timing consistency of each channel is ensured to generate preliminary synchronization control signals. This step implements a precision clock synchronization architecture, using a hybrid PTP / IEEE 1588 protocol. A clock compensation circuit is deployed at the hardware level, and the clock sources of each channel are dynamically adjusted through a phase-locked loop. The feedback signal includes the relative time difference between channels and the absolute timestamp. The system performs microsecond-level clock calibration every millisecond to ensure that the timing references of multiple channels are consistent, effectively eliminating the clock cumulative error caused by temperature drift and circuit aging, enabling the large-scale channel system to maintain stable synchronization performance during long-term operation, and providing a reliable time reference guarantee for high-precision jacquard weaving.

[0032] Implement a hybrid clock synchronization architecture: Coarse synchronization (accuracy 100ns) is performed using the PTP protocol at the hardware level, and fine-grained adjustment is performed through compensation values at the software level. At the beginning of each control cycle (1ms), the master controller broadcasts the global time reference, and each slave node calculates the local trigger time based on the compensation value.

[0033] Clock deviation correction uses a PID control algorithm: The proportional term processes the current deviation, the integral term eliminates the cumulative error, and the derivative term predicts the trend change. The feedback signal includes the actual action timestamps of each channel, and the transmission path asymmetry is calculated through two-way timestamp exchange. The synchronization accuracy is maintained by a digital phase-locked loop within the FPGA.

[0034] For example: When weaving complex gradient patterns, the system maintains the pin action synchronization error of 2048 channels less than 50ns, ensuring that the pin position deviation in the color transition area does not exceed 0.1mm.

[0035] For the preliminary synchronization control signal, a multi-channel collaborative control algorithm is adopted. Combining the delay compensation value, the instruction timings of each channel are dynamically adjusted, and through signal fusion technology, the final multi-channel collaborative control signal is generated.

[0036] This step designs a distributed instruction scheduling engine. Based on the compensated timing parameters, programmable delay units are inserted into the instruction pipeline. The weighted consistency algorithm is used to fuse the state feedback of each channel, and the instruction issuing timing is dynamically adjusted to ensure the spatio-temporal consistency of the pin actions. The signal fusion process considers non-linear factors such as mechanical transmission delay, realizes true multi-channel collaborative operation, enables the electronic jacquard machine to maintain high coordination of each pin action during complex pattern weaving, and significantly improves the boundary clarity of the fabric pattern and the naturalness of color transition.

[0037] Construct a hierarchical instruction scheduling system: The upper layer models the instruction dependency relationship based on a directed acyclic graph (DAG), the middle layer uses the earliest deadline first (EDF) algorithm to allocate time slices, and the bottom layer ensures precise execution through a hardware time-triggered mechanism. The signal fusion module integrates the state feedback of each channel and dynamically adjusts the instruction pipeline.

[0038] Introduce the "elastic time window" mechanism: non-critical instructions allow a time offset of ±10 μs, while critical synchronization points (such as color-changing instructions) are strictly aligned. The application of the compensation value adopts a feedforward + feedback composite control. The feedforward part is predicted based on historical data, and the feedback part corrects the residual error in real time.

[0039] For example: when knitting a pattern that requires synchronous color-changing of 1024 channels, the system starts preparing 10 ms in advance. By dynamically adjusting the instruction buffer depth of each channel, the synchronous action of all stitches at the specified moment is finally achieved, and the error is controlled within 30 ns.

[0040] S202, according to the input fabric pattern data, use an instruction decomposition model based on a graph neural network to decompose the fabric pattern data. The decomposition uses a spatial topology mapping algorithm, combines the physical layout characteristics of the channels, dynamically allocates pattern pixel points to the corresponding channels, generates a hierarchical instruction set that matches the physical positions of the channels, and binds it to a multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set; This step designs a dedicated graph neural network architecture (GATv2), models the fabric pattern as a pixel dot matrix (nodes = pixels, edges = spatial adjacency relationships), and learns the pixel-channel mapping relationship through 3-layer graph convolution. The spatial topology mapping algorithm considers the physical coordinates (X / Y / Z three-dimensional layout) and mechanical characteristics (such as the response speed of solenoid valves) of the channels to achieve an instruction allocation accuracy of sub-millimeter level. The finally generated hierarchical instruction set includes three levels of control granularity: thread level, channel level, and device level, breaking through the limitations of the traditional fixed partition allocation mode, increasing the pattern decomposition fitness by 60%. Especially for high-density jacquard machines with more than 20,000 needles, it can reduce the idle stroke movement of the stitches by 15% and significantly improve the knitting efficiency. Tests show that the decomposition time of this method for processing a 4096×4096 pixel pattern is only 80 ms.

[0041] Specifically, according to the input fabric pattern data, image preprocessing technology can be used to denoise and normalize the pattern, and through a feature extraction algorithm, extract the key pixel point information of the pattern to generate preliminary pattern feature data; This step constructs a multi-level image processing pipeline. First, eliminate the scanning noise through adaptive median filtering, and then use a normalization algorithm based on gamut analysis to unify the color space. The detection of key pixel points combines edge density analysis and color gradient calculation to identify the contour turning points and color mutation points in the pattern, significantly improving the processing quality of the pattern data, providing accurate input features for subsequent instruction decomposition, and ensuring that the detailed features of complex patterns can be completely retained and accurately reproduced.

[0042] After the fabric pattern data is input through a high-precision scanner, multi-stage preprocessing is first carried out: 1) In the optical distortion correction stage, the pre-stored lens distortion parameter matrix (including radial distortion coefficients k1-k3 and tangential distortion coefficients p1-p2) is used to geometrically calibrate the image; 2) In the noise suppression stage, for different fabric materials (such as silk, cotton, chemical fiber), an adaptive filtering algorithm is adopted. For fine materials like silk, non-local means denoising (NL-Means) is used, and for textured materials like cotton, wavelet threshold denoising is used; 3) In the color normalization stage, through calibration with an X-Rite ColorChecker color card, the RGB color space is converted to the standard LAB space, and then mapped to the device-related CMYK+2 spot color space.

[0043] Feature extraction adopts a hybrid strategy: 1) For key pixel point detection, an improved Harris-Laplace algorithm is used, with the corner response threshold set to 0.01 and the neighborhood size to 15×15 pixels; 2) For color feature extraction, the color difference gradient in a 5×5 local area is calculated in the LAB space; (3) For texture feature analysis, a Gabor filter bank (6 directions × 4 scales) is used. The finally generated pattern feature data contains a three-layer structure: The basic layer stores the original pixel grid, the feature layer records the coordinates of key points and their 32-dimensional feature vectors (including color, texture, etc.), and the semantic layer labels the region type tags.

[0044] For example, when processing a 60cm×90cm silk jacquard pattern, the system: 1) Completes distortion correction and noise elimination for the scanned image of 36000×54000 pixels; 2) Identifies 184 key color block areas in the LAB color space; 3) Extracts approximately 120,000 key pixel points, and each point records features such as coordinates, LAB values, and texture energy; 4) Generates a 1.5GB feature data set containing color channels, texture features, and regional semantics.

[0045] For the preliminary pattern feature data, an instruction decomposition model based on a graph neural network is adopted, combined with a spatial topology mapping algorithm. The pattern pixel points are dynamically assigned to the corresponding channels, and through node feature embedding technology, a preliminary hierarchical instruction set is generated; This step designs a multi-scale graph neural network architecture, establishing a mapping relationship between pixel clusters and channel groups at the coarse-grained level, and optimizing the channel assignment of individual pixel points at the fine-grained level. The spatial topology mapping considers the physical arrangement of the needle bed, and uses an attention mechanism to learn the optimal matching mode between pixels and channels, realizing the intelligent conversion of pattern data into control instructions, enabling the optimal allocation of the channel resources of the jacquard machine, and greatly improving the weaving efficiency and accuracy of complex patterns.

[0046] Construct a three-level graph neural network architecture: 1) The input layer constructs the pattern feature data into an attribute graph, where the nodes represent key pixel points (with 32-dimensional features) and the edges represent spatial adjacency relationships (distance threshold 5mm); 2) The feature encoding layer uses a 3-layer GraphSAGE network, with each layer outputting 256-dimensional features and adopting a gated attention mechanism to aggregate neighborhood information; 3) The channel allocation layer maps the pixel points to 2048 physical channels through differentiable clustering (Gumbel-Softmax), while optimizing the load balance (the difference in task volume between channels < 15%) and the motion efficiency (minimizing the total distance of the needle movement).

[0047] The spatial topology mapping algorithm performs two-step processing: 1) In the coarse allocation stage, a channel grid index is established based on the physical layout of the needle bed (X / Y / Z coordinates); 2) In the fine allocation stage, the channel attribution of the edge pixels is optimized through bilinear interpolation. The node feature embedding adopts position encoding technology to convert the physical coordinates (X, Y) of the channel into a 64-dimensional position vector, which is concatenated with the pattern features and then input into the network. For example, when decomposing a pattern containing gradient moiré, the system: 1) divides the pattern into 32×32 superpixel blocks; 2) adjusts the channel density according to the sine distribution in the gradient direction (45°); 3) generates a preliminary hierarchical instruction set containing 2048 channel instructions, with each instruction containing 50 - 80 needle movement sequences, and the load difference between adjacent channels is controlled within 12%.

[0048] For the preliminary hierarchical instruction set, combined with the physical layout characteristics of the channels, the spatial topology mapping algorithm is used to match the instruction set with the physical positions of the channels, and through position correction technology, the accuracy of the instruction allocation is optimized to generate a preliminary matching instruction set; This step establishes a three-dimensional channel position model and maps the logical instruction coordinates to the physical needle bed space through a rigid body transformation algorithm. The position correction adopts the iterative closest point algorithm to compensate for mechanical installation errors and needle position tolerances, ensuring that the deviation between the instruction position and the actual needle position is less than 0.1mm, eliminating the pattern deformation problem caused by mechanical structure errors, and making the weaving effect highly consistent with the designed pattern, which is especially beneficial to the accurate reproduction of precise geometric patterns.

[0049] Establish an accurate digital twin model of the needle bed: 1) In the mechanical calibration stage, a laser tracker (such as Leica AT960) is used to measure the absolute position of each needle (accuracy ±5μm), and the XYZ coordinates and attitude angles are recorded; 2) Electrical characteristics modeling tests the response curves of each channel (such as solenoid valve opening delay, maximum action frequency); 3) Thermal deformation compensation establishes a temperature-displacement look-up table (sampling interval 5°C).

[0050] The position correction implements a three-stage process: 1) Global registration calculates the optimal rigid body transformation through the RANSAC algorithm of 12 fiducial pins; 2) Local elastic matching uses an improved thin plate spline (TPS) algorithm with a control point spacing of 10 mm and a stiffness coefficient λ = 0.3; 3) Real-time compensation integrates temperature sensor data to dynamically adjust the mapping parameters. The verification process uses a dedicated test pattern (concentric circles, radiating lines, etc.) to quantify the error and ensure that the final position deviation is <0.1 mm. Typical application examples: After the ambient temperature of a certain type of jacquard machine increases by 10 °C, the system: 1) Detects a thermal expansion of 0.15 mm in the Y direction; 2) Automatically adjusts the mapping parameters of channel 37 (compensation vector ΔY = +0.12 mm); 3) Verified through the verification pattern, reducing the overall geometric distortion rate from 1.1% to 0.2%.

[0051] For the preliminary matching instruction set, a spatio-temporal synchronization algorithm is adopted, combined with multi-channel collaborative control signals, to bind the instruction set to the control signals, and through timing verification technology, a final spatio-temporal synchronization instruction set is generated.

[0052] This step designs an instruction encapsulation format with timestamps, embedding execution time window information in the instruction header. Timing verification uses a look-ahead conflict detection algorithm to predict and resolve possible pin action conflicts, ensuring the coordination of instructions in each channel in the spatio-temporal dimension, achieving a perfect integration of control signals and knitting instructions, providing the electronic jacquard machine with a collaborative control ability accurate to the millisecond level, and ensuring the accurate execution of complex knitting actions.

[0053] Design a time-sensitive network architecture based on IEEE 802.1AS: 1) Instruction encapsulation uses a composite format of 64-bit absolute timestamp (accuracy 10 ns) + 16-bit relative offset; 2) Communication scheduling uses a hybrid mode of time-triggered (TT) and rate constraint (RC), and dedicated time slots are allocated for key instructions; 3) Redundancy verification uses a dual CRC32 mechanism (verifying the instruction header and payload separately).

[0054] The timing planning algorithm includes: 1) Conflict detection predicts the pin trajectories through motion simulation and uses AABB collision detection to identify interferences; 2) Resource scheduling adopts an improved banker's algorithm to ensure no deadlocks; 3) Timing optimization applies constraint programming (CP) to minimize the cycle time under the satisfaction of mechanical / electrical constraints. The real-time monitoring system detects the current waveform at a sampling rate of 200 kHz and dynamically adjusts the timing parameters of abnormal channels. Typical application examples: When knitting a complex pattern with 1500 color patches, the system: 1) Generates a synchronization packet of 8500 channel instructions (total size 3.2 MB); 2) Detects and resolves 23 pin conflicts in advance; 3) Finally achieves a synchronization accuracy of 2048 channels within ±35 ns, and the total knitting cycle is optimized from the estimated 800 ms to 680 ms.

[0055] S203. Based on the channel operation status data collected in real time, adopt a fault diagnosis model based on time-series anomaly detection to identify abnormal channels and their fault types. Through virtual channel redundancy technology, based on the spatio-temporal synchronization instruction set, isolate the instruction segments of abnormal channels, and combine the channel space topology relationship to dynamically allocate the instructions of abnormal channels to adjacent channels, generating a reconstructed instruction set after fault immunity. This step deploys an LSTM-Transformer hybrid anomaly detection model to analyze the 12-dimensional state time-series data such as current and temperature of each channel in real time (sampling rate 1kHz), and can identify 8 types of faults such as broken needles (feature: sudden drop in current) and blocked needles (feature: continuous increase in temperature) within 5ms. The virtual channel redundancy technology establishes an N+1 backup mechanism, and through the instruction segment dynamic migration algorithm (considering the load margin and working temperature of adjacent channels), realizes the millisecond-level switching of faulty channels, shortens the fault response time from 200ms in the traditional scheme to within 20ms, reduces the pattern distortion rate during the fault period by 90%, and significantly improves production continuity. In practical applications, this method can enable the jacquard machine to maintain a weaving integrity of more than 98% during a single-channel fault.

[0056] Specifically, based on the channel operation status data collected in real time, adopt distributed data acquisition technology to obtain the operation status of each channel, and generate preliminary operation status time-series data through a time-series analysis model. This step deploys a high-density sensor network to collect 12-dimensional state parameters such as current, temperature, and vibration of each channel in real time, with a sampling frequency of up to 10kHz. The time-series analysis adopts a sliding window mechanism to extract time-domain and frequency-domain features, construct a time-series portrait of channel health, provide comprehensive and real-time data support for fault diagnosis, enable the system to detect channel anomaly signs in a timely manner, and create conditions for preventive maintenance.

[0057] Construct a high-density sensor network system and deploy on each channel control board: 1) INA240 current detection chip of TI (bandwidth 1.1MHz) to monitor the solenoid valve drive current waveform in real time; 2) ADT7420 temperature sensor (accuracy ±0.1°C) mounted on the hot area of the power device; 3) ADXL357 vibration sensor (bandwidth 1.5kHz) to detect mechanical abnormal vibration. The data acquisition adopts a distributed architecture and transmits the original signal at a sampling rate of 1MHz through the TSN network for preprocessing at the edge computing node.

[0058] The timing analysis model includes three stages of processing: 1) Signal conditioning, using an IIR digital filter (50kHz cutoff frequency) to eliminate high-frequency noise; 2) Feature extraction, calculating 12-dimensional time-domain features (such as RMS value and crest factor) and 8-dimensional frequency-domain features (FFT dominant frequency component) within a sliding window (100ms); and 3) State encoding, using an autoencoder, compressing the 20-dimensional features into an 8-dimensional state vector. The system establishes a dynamic baseline library to automatically adapt to the normal state range under different operating modes (such as high-speed and precision weaving).

[0059] For example, after eight hours of continuous operation, the system detected an anomaly in channel 1024: 1) the RMS value of the current waveform increased from 1.25A to 1.38A (+10.4%); 2) the temperature increased from 42°C to 57°C; and 3) a new peak appeared in the vibration spectrum at 850Hz. This data was encoded as a state vector [1.38, 57, 850, ...] and marked as a pre-fault state indicating "suspected solenoid valve wear."

[0060] For the preliminary operating status time series data, a fault diagnosis model based on time series anomaly detection is used, combined with a fault type library, to identify abnormal channels and their fault types. Preliminary fault diagnosis results are generated through anomaly scoring technology. This step builds a deep temporal convolutional network model, capturing anomaly patterns through multi-scale feature extraction. The fault type library contains 8 major categories and 32 minor categories of fault features, employing a hierarchical classification strategy for fine-grained fault identification. Anomaly scoring comprehensively considers both the degree of deviation and duration, enabling early and accurate fault identification, significantly shortening fault diagnosis time and creating a valuable window for subsequent fault-tolerance processing.

[0061] The fault diagnosis system adopts a dual-model architecture: 1) The real-time detection module uses a lightweight TCN temporal convolutional network (6 layers, convolution kernel size 5), analyzes the state sequence every 10ms, and outputs the abnormality probability; 2) The precise diagnosis module uses a Transformer encoder (8-head attention) to perform in-depth analysis of suspected abnormal channels and match 32 typical fault modes in the fault feature library.

[0062] Anomaly scoring is quantified in multiple dimensions: 1) Deviation score (0-100) reflects the deviation between the current state and the baseline; 2) Persistence score (0-100) assesses the duration of the anomaly; 3) Danger score (0-100) determines the severity level based on the fault type library. The final anomaly score is weighted summation: Score = 0.4 × Deviation + 0.3 × Persistence + 0.3 × Danger A score exceeding 60 points triggers an alarm, and a score exceeding 80 points initiates fault tolerance processing.

[0063] For example, when channel 512 experiences: 1) current waveform distortion (deviation 78); 2) persistence for 5 cycles (persistence 65); and 3) matching the "solenoid valve stuck" pattern (hazard level 82), the system generates a comprehensive score of 79.4, diagnoses it as "solenoid valve stuck - medium risk," and automatically records the fault characteristics in the knowledge base.

[0064] Based on the preliminary fault diagnosis results, virtual channel redundancy technology is used to isolate the instruction segments of the abnormal channel based on the time-space synchronization instruction set, and a preliminary redundant instruction set is generated through the instruction redistribution algorithm; This step designs a dynamic instruction segment migration mechanism, marking the instruction block of the faulty channel as invalid and allocating backup resources in the virtual channel pool. This instruction redistribution considers the load capacity of adjacent channels and selects the optimal migration target based on the principle of minimal interference. This allows for rapid isolation of the faulty channel and resource reorganization, ensuring that the system maintains basic weaving functionality even in the event of a localized failure, significantly improving device availability.

[0065] Virtual channel pool management utilizes a dynamic partitioning strategy: 1) 5% of physical channels are always available as hot backups; 2) 10% of the processing capacity of healthy channels is reserved as logical backups. The isolation process involves three steps: 1) instruction freeze—stopping new instructions from being sent to the faulty channel; 2) state preservation—recording the last valid location of the faulty channel; and 3) resource marking—marking the channel as unavailable in the resource management table.

[0066] The instruction reallocation algorithm includes: 1) a proximity-first strategy—searching for the three nearest healthy channels in the spatial topology; 2) a load-balancing strategy—selecting channels with current loads below the average; and 3) a process constraint check—ensuring that the resulting allocation will not cause yarn tension to exceed the specified limit. The reallocation process maintains the semantics of the instructions, adjusting only the physical execution location and adding a redirection marker to the instruction header.

[0067] For example, when channel 768 is diagnosed with a "broken needle fault," the system: 1) immediately freezes subsequent instructions for that channel; 2) reallocates its 42 unexecuted pin actions (original instruction segments 5-46) to the adjacent channels 767, 769, and 865; and 3) inserts synchronization markers into the new instructions to ensure that the deviation from the original timing plan is less than 100 μs.

[0068] For the preliminary redundant instruction set, combined with the channel space topology, a dynamic allocation algorithm is used to allocate the instructions of the abnormal channel to the adjacent channel. Through load balancing technology, the final fault-immune reconstructed instruction set is generated.

[0069] This step constructs a channel collaboration network model and intelligently selects instruction receiving channels based on physical proximity and performance margin evaluation. Load balancing adopts an elastic scaling strategy to dynamically adjust the instruction density of each channel, avoid generating new performance bottlenecks, achieve graceful degradation in the event of a failure, enable the electronic jacquard machine to maintain the optimal weaving quality even when some channels fail, and significantly improve the fault tolerance and production continuity of the equipment.

[0070] The dynamic allocation algorithm implements four-stage optimization: 1) Feasibility analysis - Check the physical reachability of the target channel (pin movement range limit); 2) Conflict detection - Predict whether the new allocation will cause pin collisions; 3) Load evaluation - Calculate the channel load rate after allocation; 4) Process verification - Ensure that process parameters such as yarn tension and fabric density are still within the allowable range.

[0071] Load balancing adopts an elastic scaling strategy: 1) The basic load threshold is set to 80% of the maximum capacity of the channel; 2) Temporary overloading is allowed to reach 95%, but the duration does not exceed 3 cycles; 3) Dynamically adjust the instruction execution rate (±10%) to smooth the load fluctuation. The finally generated reconstructed instruction set contains complete version control and traceability information, recording all reallocation operations.

[0072] For example, when dealing with the failure of channel No. 1024, the system: 1) Identifies that the adjacent channels No. 1023, 1025, and 1088 are available; 2) Confirms no collision risk through motion simulation; 3) Splits the original instructions into three parts (ratio 40%:35%:25%) and dynamically allocates them according to the current load of each channel; 4) The finally generated instruction set changes the overall load rate of the system from 78% before the failure to 82%, still remaining within the safe range, and the impact of the failure is limited to a local area of 0.5 cm × 0.5 cm.

[0073] S204. According to the reconstructed instruction set, adopt a multi-objective resource scheduling algorithm based on game theory, combine the channel load status and the priority of the weaving task, dynamically allocate computing resources and storage bandwidth. Among them, the scheduling optimizes the instruction execution timing through an energy consumption - efficiency trade-off model, generates an optimized execution instruction set, and issues it to each channel for execution in real time to achieve the control of the electronic jacquard machine based on multi-channel collaboration.

[0074] This step constructs a non - cooperative game model (players = channels, strategies = resource occupancy ratio, payoffs = task completion quality), and dynamically allocates resources such as CPU cores and memory bandwidth through a Nash equilibrium solver. The energy consumption - efficiency model uses Pareto front analysis to optimize the energy - efficiency ratio (number of pin actions completed per joule) while satisfying the real - time constraint (instruction cycle < 1ms). The final instruction set is sent through the TSN time - sensitive network with a transmission jitter < 10μs, achieving a 40% increase in resource utilization rate while reducing energy consumption by 25%, and supporting real - time cooperative control of more than 2000 channels. When knitting complex gradient patterns, this method can reduce the energy consumption fluctuation by 60% and increase the production efficiency by 18%.

[0075] Specifically, according to the reconstructed instruction set, distributed data acquisition technology can be used to obtain the load status of each channel in real - time, and through the load analysis model, preliminary load status data can be generated; This step designs a lightweight load monitoring framework to collect key metrics such as CPU utilization, memory occupancy, and instruction queue depth in real - time. The load analysis model uses a sliding time window statistical method to identify short - term overloads and long - term load trends, construct a multi - dimensional load feature vector, providing accurate load situation awareness for resource scheduling, enabling the system to make optimal resource allocation decisions according to the actual operating status.

[0076] Deploy TI's C2000 series microcontrollers on each channel control board to collect the following key parameters at a frequency of 1kHz: 1) Computational load (CPU utilization, instruction queue depth); 2) Electrical load (effective drive current, peak current ratio); 3) Mechanical load (pin action frequency, cumulative stroke). The data is transmitted back in real - time through the TSN network with the delay jitter controlled within ±50μs.

[0077] The load analysis model adopts a three - layer architecture: 1) The real - time layer (< 10ms response) calculates basic metrics (such as channel utilization, instantaneous power consumption); 2) The short - term analysis layer (1 - minute window) identifies load trends (sliding variance, autocorrelation coefficient); 3) The long - term statistics layer (1 - hour window) establishes a load feature profile (daily cycle pattern, anomaly baseline). The data storage uses a time - series database (InfluxDB), supporting millisecond - level queries.

[0078] For example, in an 8 - hour production batch, the system: 1) Monitors in real - time that the current fluctuation of channel 512 is abnormal (standard deviation exceeds the threshold by 15%); 2) Short - term analysis finds that its load shows periodic spikes (interval of 2.3s); 3) Long - term statistics identify that this phenomenon is related to specific pattern instructions. Finally, a real - time status matrix containing 2048 channels and 12 - dimensional load metrics is generated.

[0079] For the preliminary load status data, a multi-objective resource scheduling algorithm based on game theory is adopted. Combining the priorities of knitting tasks, computing resources and storage bandwidth are dynamically allocated. Through a game equilibrium model, a preliminary scheduling plan is generated. In this step, a non-cooperative game model is constructed. Each channel is regarded as a rational player, and its strategy space is the resource request volume. The revenue function comprehensively considers the task completion quality and resource usage cost. An approximate Nash equilibrium solution algorithm is used to obtain the optimal resource allocation plan within milliseconds, realizing the efficient utilization of limited resources and making the overall system performance reach the Pareto optimal state under the premise of meeting the real-time requirements.

[0080] Construct a non-cooperative game model: 1) The players are the control nodes of each channel (a total of 2048); 2) The strategy space is the resource request volume (CPU core occupancy ratio, memory bandwidth, cache size); 3) The revenue function U = αQ + βE + γS, where Q is the task quality score, E is the energy efficiency score, S is the security score, and α, β, γ are the corresponding influence coefficients. The improved false price algorithm (Adaptive Tâtonnement) is used to solve the Nash equilibrium, and the convergence time < 5ms.

[0081] Priority management realizes dynamic weighting: 1) The basic weight is determined by the pattern complexity (such as the gradient area weight + 30%); 2) The urgency weight increases linearly as the deadline approaches; 3) Abnormal channels automatically increase their priorities. Resource allocation implements two-level isolation: the physical core level (ensuring that critical channels monopolize resources) and the virtual container level (limiting the resource upper limit of ordinary channels through cgroups).

[0082] For example, when processing a mixed task containing an urgent order, the system: 1) allocates 40% of the physical cores to high-priority tasks; 2) calculates the optimal resource ratio of each channel through game equilibrium; 3) generates a scheduling plan within 5ms, enabling high-priority tasks to be completed 150ms in advance while ensuring that the progress deviation of ordinary tasks < 3%.

[0083] For the preliminary scheduling plan, an energy consumption-efficiency trade-off model is adopted. Combining the instruction execution timing, resource allocation is optimized. Through a multi-objective optimization algorithm, a preliminary optimized execution instruction set is generated. In this step, a multi-objective optimization problem model is established. The decision variable is the resource allocation parameter, and the objective function simultaneously considers the energy efficiency ratio and the task completion timeliness. The genetic algorithm with elitist retention is used to solve the Pareto front, and the optimal compromise solution is selected according to the operation strategy, realizing the intelligent balance of production efficiency and energy consumption, and enabling the electronic jacquard machine to achieve the optimal energy efficiency performance while ensuring the knitting quality.

[0084] Establish a Pareto optimization model: 1) The objective function includes energy consumption (ΣP_iΔt), efficiency (1 / T_total), and quality (ΔQ); 2) The constraints include timing (instruction cycle < 1ms), resources (ΣR_i ≤ R_total), and process (yarn tension range). The NSGA-III algorithm is used for solving, with a population size of 100 and 50 generations of iteration.

[0085] Implement elastic adjustment for timing optimization: 1) Critical path instructions (such as color change synchronization points) are executed strictly on time; 2) Non-critical instructions are allowed a ±100μs offset; 3) Redundant instruction blocks are dynamically merged. DVFS technology is used for energy consumption management, and the CPU voltage / frequency is adjusted in real time according to the load (0.9 - 1.2V / 800 - 1200MHz), combined with instruction reordering to reduce the peak power consumption.

[0086] For example, when optimizing a complex pattern task, the system: 1) Identifies that 37% of the instructions have time elasticity; 2) Reduces the peak power consumption by 22% through reordering; 3) In the final solution, while the energy consumption is reduced by 18%, the total knitting time only increases by �ms (< 1%), and the quality score remains above 98 points.

[0087] For the preliminary optimized execution instruction set, use real-time distribution technology to distribute the instruction set to each channel for execution. Through execution monitoring technology, generate the final optimized execution instruction set.

[0088] This step designs a highly reliable instruction distribution mechanism, adopts a time-triggered transmission protocol to ensure the timely delivery of critical instructions. The execution monitoring module compares the instruction status and the expected target in real time, dynamically adjusts the distribution strategy, forms a closed-loop control, ensures that the optimization strategy is accurately executed, enables the coordinated control of the entire system to achieve the designed expected effect, and finally realizes the high-performance and high-reliability operation of the electronic jacquard machine.

[0089] The instruction distribution system includes: 1) A preprocessing module (instruction compression, encryption signature); 2) A TSN network scheduler (allocates time slots according to the IEEE 802.1Qbv standard); 3) A channel receiving buffer (double-buffer design to prevent conflicts). The transmission protocol uses an improved UDP-RTP, supporting packet loss retransmission (retransmission rate < 0.1%) and out-of-order reorganization.

[0090] The execution monitoring system is implemented as follows: 1) Hardware-level monitoring (200kHz current sampling); 2) Instruction-level verification (CRC check of execution results); 3) Effect-level evaluation (quality inspection by a high-speed camera). Abnormal handling adopts a three-level response: 1) μs level (FPGA directly corrects); 2) ms level (channel controller adjusts); 3) Second level (main control system replans).

[0091] For example, when executing the optimized instruction set: 1) The system distributes 1.5 MB of instruction data within 2 ms; 2) Real-time monitoring finds that the current of channel 1024 is abnormal (exceeding the limit by 8%); 3) The standby instruction stream is switched within 300 μs to avoid pin damage; 4) The final production data record shows that the synchronization error of 2048 channels is controlled within ±40 ns, and the energy consumption is reduced by 21% compared with the traditional method.

[0092] It can be seen that according to the hardware performance differences and signal transmission delay data of each channel, the delay compensation value of each channel is calculated in real time, and the timing offset of the control instructions of the electronic jacquard machine is dynamically adjusted by using the delay compensation value to generate a multi-channel collaborative control signal; according to the input fabric pattern data, the fabric pattern data is decomposed to generate a hierarchical instruction set and bound to the multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set; according to the real-time collected channel operation status data, the abnormal channels and their fault types are identified to generate a reconstructed instruction set after fault immunity; according to the reconstructed instruction set, the computing resources and storage bandwidth are dynamically allocated to generate an optimized execution instruction set and sent to each channel for execution in real time, so as to improve the control ability of the electronic jacquard machine and the fabric production efficiency, and promote the intelligent development of the textile industry.

[0093] Another embodiment of the present invention provides an electronic jacquard machine control system based on multi-channel collaboration. Refer to Figure 3 , the system may include: A compensation module 301, configured to calculate the delay compensation value of each channel in real time by using a dynamic delay compensation algorithm based on reinforcement learning according to the hardware performance differences and signal transmission delay data of each channel, and dynamically adjust the timing offset of the control instructions of the electronic jacquard machine by using the delay compensation value through an adaptive clock synchronization technology, in combination with the feedback signal between channels, to generate a multi-channel collaborative control signal; A decomposition module 302, configured to decompose the fabric pattern data by using an instruction decomposition model based on a graph neural network according to the input fabric pattern data. The decomposition is performed through a spatial topology mapping algorithm, in combination with the channel physical layout characteristics, to dynamically allocate the pattern pixel points to the corresponding channels, generate a hierarchical instruction set matching the channel physical position, and bind it to the multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set; An identification module 303, configured to identify the abnormal channels and their fault types by using a fault diagnosis model based on timing anomaly detection according to the real-time collected channel operation status data, and isolate the instruction segments of the abnormal channels based on the spatio-temporal synchronization instruction set through a virtual channel redundancy technology, and dynamically allocate the instructions of the abnormal channels to the adjacent channels in combination with the channel spatial topology relationship to generate a reconstructed instruction set after fault immunity; The control module 304 is configured to, according to the reconstruction instruction set, adopt a multi-objective resource scheduling algorithm based on game theory, combine the channel load status and the weaving task priority, and dynamically allocate computing resources and storage bandwidth. Wherein, the scheduling optimizes the instruction execution timing through an energy consumption-efficiency trade-off model, generates an optimized execution instruction set, and sends it to each channel for execution in real time, so as to realize the control of the electronic jacquard machine based on multi-channel cooperation.

[0094] It can be seen that, according to the hardware performance differences and signal transmission delay data of each channel, the delay compensation value of each channel is calculated in real time, and the timing offset of the control instruction of the electronic jacquard machine is dynamically adjusted by using the delay compensation value, so as to generate a multi-channel cooperation control signal; according to the input fabric pattern data, the fabric pattern data is decomposed to generate a hierarchical instruction set, and the hierarchical instruction set is bound to the multi-channel cooperation control signal to form a spatio-temporal synchronization instruction set; according to the channel operation state data collected in real time, the abnormal channel and its fault type are identified, and a reconstructed instruction set after fault immunity is generated; according to the reconstructed instruction set, computing resources and storage bandwidth are dynamically allocated to generate an optimized execution instruction set, and the optimized execution instruction set is sent to each channel for execution in real time, so as to improve the control ability of the electronic jacquard machine and the fabric production efficiency, and promote the intelligent development of the textile industry.

[0095] The embodiment of the present invention also provides a storage medium, in which a computer program is stored, and the computer program is set to execute the steps in any one of the above method embodiments when running.

[0096] Specifically, in this embodiment, the above storage medium may be set to store a computer program for executing the following steps: S201, according to the hardware performance differences and signal transmission delay data of each channel, adopt a dynamic delay compensation algorithm based on reinforcement learning, calculate the delay compensation value of each channel in real time, and through an adaptive clock synchronization technology, combine the feedback signal between channels, and use the delay compensation value to dynamically adjust the timing offset of the control instruction of the electronic jacquard machine to generate a multi-channel cooperation control signal; S202, according to the input fabric pattern data, adopt an instruction decomposition model based on a graph neural network to decompose the fabric pattern data. The decomposition is carried out through a spatial topology mapping algorithm, combined with the channel physical layout characteristics, to dynamically allocate the pattern pixel points to the corresponding channels, generate a hierarchical instruction set matching the channel physical position, and bind the hierarchical instruction set to the multi-channel cooperation control signal to form a spatio-temporal synchronization instruction set; S203, according to the channel operation state data collected in real time, adopt a fault diagnosis model based on timing anomaly detection to identify the abnormal channel and its fault type, and through a virtual channel redundancy technology, based on the spatio-temporal synchronization instruction set, isolate the instruction segment of the abnormal channel, and combine the channel spatial topology relationship to dynamically allocate the instructions of the abnormal channel to the adjacent channels to generate a reconstructed instruction set after fault immunity; S204. According to the reconstructed instruction set, adopt a multi-objective resource scheduling algorithm based on game theory, combine the channel load status and the priority of the weaving tasks, and dynamically allocate computing resources and storage bandwidth. Among them, the scheduling optimizes the instruction execution timing through an energy consumption - efficiency trade-off model, generates an optimized execution instruction set, and issues it to each channel in real time for execution, so as to achieve the control of the electronic jacquard machine based on multi-channel collaboration.

[0097] It can be seen that according to the hardware performance differences and signal transmission delay data of each channel, calculate the delay compensation value of each channel in real time, use the delay compensation value to dynamically adjust the timing offset of the control instructions of the electronic jacquard machine, and generate a multi-channel collaborative control signal; according to the input fabric pattern data, decompose the fabric pattern data, generate a hierarchical instruction set, and bind it to the multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set; according to the channel operation status data collected in real time, identify the abnormal channels and their fault types, and generate a reconstructed instruction set after fault immunity; according to the reconstructed instruction set, dynamically allocate computing resources and storage bandwidth, generate an optimized execution instruction set, and issue it to each channel in real time for execution, so as to improve the control ability of the electronic jacquard machine and the fabric production efficiency, and promote the intelligent development of the textile industry.

[0098] An embodiment of the present invention also provides an electronic device, including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0099] Specifically, the above electronic device may further include a transmission device and an input / output device. Among them, the transmission device is connected to the above processor, and the input / output device is connected to the above processor.

[0100] Specifically, in this embodiment, the above processor may be configured to execute the following steps through a computer program: S201. According to the hardware performance differences and signal transmission delay data of each channel, adopt a dynamic delay compensation algorithm based on reinforcement learning, calculate the delay compensation value of each channel in real time, and through an adaptive clock synchronization technology, combine the feedback signals between channels, and use the delay compensation value to dynamically adjust the timing offset of the control instructions of the electronic jacquard machine to generate a multi-channel collaborative control signal; S202. According to the input fabric pattern data, adopt an instruction decomposition model based on a graph neural network to decompose the fabric pattern data. The decomposition is through a spatial topology mapping algorithm, combined with the channel physical layout characteristics, to dynamically allocate the pattern pixel points to the corresponding channels, generate a hierarchical instruction set matching the channel physical positions, and bind it to the multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set; S203. Based on the channel operation status data collected in real time, adopt a fault diagnosis model based on time series anomaly detection to identify abnormal channels and their fault types, and through virtual channel redundancy technology, based on the spatio-temporal synchronization instruction set, isolate the instruction segments of the abnormal channels, and combine the channel space topology relationship to dynamically allocate the instructions of the abnormal channels to adjacent channels, generating a reconstructed instruction set after fault immunity; S204. According to the reconstructed instruction set, adopt a multi-objective resource scheduling algorithm based on game theory, combine the channel load status and the priority of the weaving task, and dynamically allocate computing resources and storage bandwidth. Among them, the scheduling optimizes the instruction execution timing through an energy consumption - efficiency trade-off model, generates an optimized execution instruction set, and issues it to each channel for execution in real time to achieve the control of the electronic jacquard machine based on multi-channel collaboration.

[0101] It can be seen that according to the hardware performance differences and signal transmission delay data of each channel, calculate the delay compensation value of each channel in real time, use the delay compensation value to dynamically adjust the timing offset of the control instructions of the electronic jacquard machine, and generate a multi-channel collaborative control signal; decompose the fabric pattern data according to the input fabric pattern data, generate a hierarchical instruction set, and bind it to the multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set; identify abnormal channels and their fault types according to the channel operation status data collected in real time, and generate a reconstructed instruction set after fault immunity; according to the reconstructed instruction set, dynamically allocate computing resources and storage bandwidth, generate an optimized execution instruction set, and issue it to each channel for execution in real time, so as to improve the control ability of the electronic jacquard machine and the fabric production efficiency, and promote the intelligent development of the textile industry.

[0102] The above has described in detail the structure, features and function effects of the present invention according to the illustrated embodiments. The above is only the preferred embodiment of the present invention, but the present invention is not limited to the scope defined by the drawings. Any changes made according to the concept of the present invention, or modified into equivalent embodiments with equivalent changes, still within the spirit covered by the specification and drawings, shall be within the protection scope of the present invention.

Claims

1. A control method for an electronic jacquard machine based on multi-channel collaboration, characterized in that, The method includes: According to the hardware performance differences and signal transmission delay data of each channel, a dynamic delay compensation algorithm based on reinforcement learning is adopted to calculate the delay compensation value of each channel in real time. Through an adaptive clock synchronization technology, combined with the feedback signals between channels, the timing offset of the control instructions of the electronic jacquard machine is dynamically adjusted using the delay compensation value to generate a multi-channel collaborative control signal; According to the input fabric pattern data, an instruction decomposition model based on a graph neural network is adopted to decompose the fabric pattern data. The decomposition is performed through a spatial topology mapping algorithm, combined with the channel physical layout characteristics, to dynamically allocate the pattern pixel points to the corresponding channels, generating a hierarchical instruction set that matches the channel physical positions and binding it to the multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set; According to the channel operation state data collected in real time, a fault diagnosis model based on timing anomaly detection is adopted to identify the abnormal channels and their fault types. Through virtual channel redundancy technology, based on the spatio-temporal synchronization instruction set, the instruction segments of the abnormal channels are isolated, and combined with the channel spatial topology relationship, the instructions of the abnormal channels are dynamically allocated to adjacent channels to generate a reconstructed instruction set after fault immunity; According to the reconstructed instruction set, a multi-objective resource scheduling algorithm based on game theory is adopted, combined with the channel load status and the priority of the weaving task, to dynamically allocate computing resources and storage bandwidth. Among them, the scheduling optimizes the instruction execution timing through an energy consumption - efficiency trade-off model, generates an optimized execution instruction set, and issues it to each channel for execution in real time to achieve the control of the electronic jacquard machine based on multi-channel collaboration.

2. The method according to claim 1, characterized in that, The step of according to the hardware performance differences and signal transmission delay data of each channel, adopting a dynamic delay compensation algorithm based on reinforcement learning to calculate the delay compensation value of each channel in real time, and through an adaptive clock synchronization technology, combined with the feedback signals between channels, using the delay compensation value to dynamically adjust the timing offset of the control instructions of the electronic jacquard machine to generate a multi-channel collaborative control signal includes: According to the hardware performance differences and signal transmission delay data of each channel, a distributed data acquisition technology is adopted to obtain the signal transmission delay of each channel in real time. Through a statistical analysis model, the average delay and fluctuation range of each channel are calculated to generate a preliminary delay analysis result; For the preliminary delay analysis result, a dynamic delay compensation algorithm based on reinforcement learning is adopted, combined with the feedback signals between channels, to calculate the delay compensation value of each channel in real time. Through a reward function and a state transition model, the calculation accuracy of the compensation value is optimized to generate a preliminary delay compensation value; For the preliminary delay compensation value, an adaptive clock synchronization technology is adopted, combined with the feedback signals between channels, to dynamically adjust the timing offset of the control instructions of the electronic jacquard machine. Through clock deviation correction technology, the timing consistency of each channel is ensured to generate a preliminary synchronization control signal; For the preliminary synchronization control signal, a multi-channel collaborative control algorithm is adopted, combined with the delay compensation value, to dynamically adjust the instruction timing of each channel. Through signal fusion technology, a final multi-channel collaborative control signal is generated.

3. The method according to claim 2, wherein According to the input fabric pattern data, an instruction decomposition model based on a graph neural network is used to decompose the fabric pattern data. The decomposition uses a spatial topology mapping algorithm, combines channel physical layout features, dynamically allocates pattern pixel points to corresponding channels, generates a hierarchical instruction set matching the channel physical positions, and binds it to a multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set, including: According to the input fabric pattern data, image preprocessing technology is used to denoise and normalize the pattern. Through a feature extraction algorithm, key pixel point information of the pattern is extracted to generate preliminary pattern feature data; For the preliminary pattern feature data, an instruction decomposition model based on a graph neural network is used, combined with a spatial topology mapping algorithm, to dynamically allocate pattern pixel points to corresponding channels. Through node feature embedding technology, a preliminary hierarchical instruction set is generated; For the preliminary hierarchical instruction set, combined with channel physical layout features, a spatial topology mapping algorithm is used to match the instruction set with the channel physical positions. Through position correction technology, the accuracy of instruction allocation is optimized to generate a preliminary matching instruction set; For the preliminary matching instruction set, a spatio-temporal synchronization algorithm is used, combined with a multi-channel collaborative control signal, to bind the instruction set to the control signal. Through timing verification technology, a final spatio-temporal synchronization instruction set is generated.

4. The method according to claim 3, characterized in that According to the real-time collected channel operation status data, a fault diagnosis model based on timing anomaly detection is used to identify abnormal channels and their fault types. Through virtual channel redundancy technology, based on the spatio-temporal synchronization instruction set, the instruction segments of abnormal channels are isolated, and combined with the channel spatial topology relationship, the instructions of abnormal channels are dynamically allocated to adjacent channels to generate a reconstructed instruction set after fault immunity, including: According to the real-time collected channel operation status data, distributed data acquisition technology is used to obtain the operation status of each channel. Through a timing analysis model, preliminary operation status timing data is generated; For the preliminary operation status timing data, a fault diagnosis model based on timing anomaly detection is used, combined with a fault type library, to identify abnormal channels and their fault types. Through anomaly scoring technology, a preliminary fault diagnosis result is generated; For the preliminary fault diagnosis result, virtual channel redundancy technology is used, based on the spatio-temporal synchronization instruction set, to isolate the instruction segments of abnormal channels. Through an instruction reassignment algorithm, a preliminary redundant instruction set is generated; For the preliminary redundant instruction set, combined with the channel spatial topology relationship, a dynamic allocation algorithm is used to allocate the instructions of abnormal channels to adjacent channels. Through load balancing technology, a final reconstructed instruction set after fault immunity is generated.

5. The method according to claim 4, characterized in that, According to the reconstructed instruction set, a multi-objective resource scheduling algorithm based on game theory is used, combined with the channel load status and the priority of the weaving task, to dynamically allocate computing resources and storage bandwidth. Among them, the scheduling optimizes the instruction execution timing through an energy consumption - efficiency trade-off model to generate an optimized execution instruction set and sends it to each channel for execution in real time to achieve the control of an electronic jacquard machine based on multi-channel collaboration, including: According to the reconstructed instruction set, the distributed data acquisition technology is adopted to obtain the load status of each channel in real time, and through the load analysis model, the preliminary load status data is generated; For the preliminary load status data, the multi-objective resource scheduling algorithm based on game theory is adopted, combined with the priority of the weaving task, to dynamically allocate computing resources and storage bandwidth. Through the game equilibrium model, a preliminary scheduling scheme is generated; For the preliminary scheduling scheme, the energy consumption-efficiency trade-off model is adopted, combined with the instruction execution timing sequence, to optimize the resource allocation. Through the multi-objective optimization algorithm, a preliminary optimized execution instruction set is generated; For the preliminary optimized execution instruction set, the real-time distribution technology is adopted to distribute the instruction set to each channel for execution. Through the execution monitoring technology, the final optimized execution instruction set is generated.

6. An electronic jacquard machine control system based on multi-channel collaboration, characterized in that, The system includes: A compensation module, which is used to calculate the delay compensation value of each channel in real time according to the hardware performance difference and signal transmission delay data of each channel, and adopt the dynamic delay compensation algorithm based on reinforcement learning. Through the adaptive clock synchronization technology, combined with the feedback signal between channels, the timing offset of the control instruction of the electronic jacquard machine is dynamically adjusted by using the delay compensation value to generate a multi-channel collaborative control signal; A decomposition module, which is used to decompose the fabric pattern data according to the input fabric pattern data, and adopt the instruction decomposition model based on the graph neural network. The decomposition is carried out through the spatial topology mapping algorithm, combined with the channel physical layout characteristics, to dynamically allocate the pattern pixel points to the corresponding channels, generate a hierarchical instruction set matching the channel physical position, and bind it to the multi-channel collaborative control signal to form a spatio-temporal synchronization instruction set; An identification module, which is used to identify the abnormal channel and its fault type according to the real-time collected channel operation status data, and adopt the fault diagnosis model based on timing anomaly detection. Through the virtual channel redundancy technology, based on the spatio-temporal synchronization instruction set, the instruction segment of the abnormal channel is isolated, and combined with the channel spatial topology relationship, the instruction of the abnormal channel is dynamically allocated to the adjacent channels to generate a reconstructed instruction set after fault immunity; A control module, which is used to dynamically allocate computing resources and storage bandwidth according to the reconstructed instruction set, adopt the multi-objective resource scheduling algorithm based on game theory, combined with the channel load status and the priority of the weaving task. Among them, the scheduling optimizes the instruction execution timing through the energy consumption-efficiency trade-off model, generates an optimized execution instruction set, and distributes it to each channel for execution in real time to realize the control of the electronic jacquard machine based on multi-channel collaboration.

7. The system according to claim 6, wherein The compensation module is specifically used for: According to the hardware performance difference and signal transmission delay data of each channel, the distributed data acquisition technology is adopted to obtain the signal transmission delay of each channel in real time. Through the statistical analysis model, the average delay and fluctuation range of each channel are calculated to generate a preliminary delay analysis result; For the preliminary delay analysis result, the dynamic delay compensation algorithm based on reinforcement learning is adopted, combined with the feedback signal between channels, to calculate the delay compensation value of each channel in real time. Through the reward function and the state transition model, the calculation accuracy of the compensation value is optimized to generate a preliminary delay compensation value; For the initial delay compensation value, an adaptive clock synchronization technique is adopted, combined with the inter-channel feedback signal, to dynamically adjust the timing offset of the control instructions of the electronic jacquard machine. Through the clock deviation correction technique, the timing consistency of each channel is ensured to generate a preliminary synchronization control signal; For the preliminary synchronization control signal, a multi-channel cooperative control algorithm is adopted, combined with the delay compensation value, to dynamically adjust the instruction timing of each channel. Through the signal fusion technique, a final multi-channel cooperative control signal is generated.

8. The system according to claim 7, wherein The decomposition module is specifically used for: According to the input fabric pattern data, an image preprocessing technique is adopted to denoise and normalize the pattern. Through the feature extraction algorithm, the key pixel point information of the pattern is extracted to generate preliminary pattern feature data; For the preliminary pattern feature data, an instruction decomposition model based on a graph neural network is adopted, combined with the spatial topology mapping algorithm, to dynamically allocate the pattern pixel points to the corresponding channels. Through the node feature embedding technique, a preliminary hierarchical instruction set is generated; For the preliminary hierarchical instruction set, combined with the channel physical layout features, the spatial topology mapping algorithm is adopted to match the instruction set with the channel physical positions. Through the position correction technique, the accuracy of the instruction allocation is optimized to generate a preliminary matching instruction set; 9. A storage medium, characterized in that, For the preliminary matching instruction set, a spatio-temporal synchronization algorithm is adopted, combined with the multi-channel cooperative control signal, to bind the instruction set to the control signal. Through the timing verification technique, a final spatio-temporal synchronization instruction set is generated.

10. An electronic device, comprising a memory and a processor, characterized in that, A computer program is stored in the storage medium, wherein the computer program is set to execute the method according to any one of claims 1-5 when running. A computer program is stored in the memory, and the processor is set to run the computer program to execute the method according to any one of claims 1-5.

Images