Textile production multi-device collaborative intelligent management and control system

Abstract

The application discloses a kind of textile production multi-device collaborative intelligent management and control systems, it is related to collaborative management and control technical field, comprising: data acquisition module, for collecting textile production each process equipment working condition data, generates independent input unit;Topology reconstruction module is used to divide out equipment inherent working condition component matrix, cross-device topology coupling component matrix and random noise component matrix, and stripping coupling disturbance and random noise obtains equipment independent working condition component matrix;Subgroup division module is used to calculate the topological correlation between devices and constructs device topological correlation matrix, carries out collaborative management and control subgroup division, generates the collaborative weight matrix corresponding to each subgroup;Collaborative control module is used to construct three-level hierarchical collaborative control architecture, complete production task decomposition and global scheduling, subgroup process parameter collaborative optimization calculation and equipment process parameter real-time adjustment and feedback, effectively improve the dynamic, accuracy and stability of textile production multi-device collaborative management and control.

Description

Technical Field

[0001] This invention relates to the field of collaborative management and control technology, specifically to a collaborative intelligent management and control system for multiple textile production equipment. Background Technology

[0002] Textile production is a typical multi-process, multi-equipment, and continuous production process, encompassing multiple core processes such as spinning, warping, sizing, and weaving. Each process involves numerous pieces of equipment, complex operating conditions, and strong temporal correlations. With the deep penetration of intelligent manufacturing technology into the textile industry, most large and medium-sized textile enterprises in China have completed the automation transformation of single equipment and the deployment of basic data acquisition systems. This has enabled the monitoring of the operating status of single equipment, fault warnings, and independent process parameter control, significantly improving the production efficiency and operational stability of single equipment. However, at the level of multi-equipment collaborative management, existing systems are still in the stage of passive scheduling based on preset process routes. They have failed to fully explore the implicit correlation information contained in the temporal data of multiple equipment, making it difficult to adapt to the urgent need for dynamic collaboration of multiple equipment in the flexible production mode of small batches, multiple varieties, and fast delivery in the modern textile industry.

[0003] Existing collaborative management technologies for multiple equipment in textile production suffer from three major technical bottlenecks. First, data processing is flawed. Existing systems often employ simple univariate data preprocessing methods, failing to effectively separate the inherent operating condition components, cross-equipment topological coupling components, and random noise components in the joint time-series data of multiple equipment. This results in severe interference from coupling disturbances and noise in the operating condition analysis results, making it impossible to accurately reflect the true operating status of the equipment. Second, the construction of equipment relationships is rigid. Most existing technologies pre-define fixed relationships between equipment based on physical process links, failing to identify dynamic implicit topological couplings caused by factors such as vibration transmission, tension fluctuations, and speed synchronization during production. The division of collaborative management subgroups lacks data-driven dynamic adjustment capabilities, leading to low equipment coordination and matching within subgroups. Third, the collaborative control architecture is unreasonable. Most systems adopt a centralized global control architecture, which suffers from high control response delays, concentrated computational loads, and conflicts between local operating condition adjustments and global optimization objectives, making it difficult to achieve real-time collaborative optimization of process parameters for multiple equipment.

[0004] The aforementioned technical deficiencies directly lead to problems such as insufficient coordination accuracy, slow response speed, and poor robustness in the practical application of existing multi-equipment collaborative management and control systems. Specifically, this manifests as large fluctuations in product quality consistency, high equipment failure rates, and persistently high energy consumption per unit product during the production process, severely restricting the improvement of the overall production efficiency of textile enterprises. At the same time, the existing system cannot automatically adjust the control strategy according to real-time changes in equipment operating conditions. When production tasks change or equipment status becomes abnormal, manual intervention is required to adjust parameters and redistribute tasks, further increasing production management costs and production cycles. Summary of the Invention

[0005] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a collaborative intelligent control system for multiple devices in textile production. By constructing a joint time-series matrix of multiple devices and combining singular spectral entropy topological reconstruction to separate inherent operating conditions of devices, cross-device topological coupling, and random noise components, a density clustering algorithm based on dynamic topological correlation is used to divide collaborative control subgroups and update collaborative weights in real time. A three-level hierarchical collaborative control architecture is built to achieve task decomposition, parameter collaborative optimization, and real-time feedback. This solves the technical problems of existing technologies, such as difficulty in separating data coupling interference between multiple devices, fixed device relationships, high collaborative control response delay, and insufficient optimization accuracy. It effectively improves the dynamism, accuracy, and stability of collaborative control of multiple devices in textile production.

[0006] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: a collaborative intelligent control system for multiple textile production equipment, comprising a data acquisition module, a topology reconstruction module, a subgroup partitioning module, and a collaborative control module; The data acquisition module is used to collect equipment operating data for each process in textile production, construct a joint time series matrix of multiple devices, and use a sliding time window to extract and generate time series data segments of multiple devices as independent input units. The topology reconstruction module is used to perform singular value decomposition on each independent input unit to obtain multiple singular component matrices, calculate the normalized spectral entropy value of each singular component matrix and perform double threshold segmentation to divide the device inherent operating condition component matrix, cross-device topology coupling component matrix and random noise component matrix. Based on the cross-device topology coupling component matrix, an undirected graph of multi-device topology coupling relationship is constructed, and coupling disturbances and random noise are removed to obtain the device independent operating condition component matrix. The subgroup partitioning module is used to calculate the topological correlation degree between devices based on the undirected graph of multi-device topological coupling relationship and construct the device topological correlation degree matrix. It uses the DBSCAN density-based spatial clustering algorithm to perform collaborative management subgroup partitioning, generate the collaborative weight matrix corresponding to each subgroup and update it in real time. The collaborative control module is used to build a three-level hierarchical collaborative control architecture, complete the decomposition and global scheduling of production tasks, the collaborative optimization calculation of process parameters within subgroups, the real-time adjustment and feedback of equipment process parameters, and periodically perform global parameter optimization operations.

[0007] Furthermore, the step of using a sliding time window to extract multi-device time series data segments as independent input units specifically involves: performing Z-score normalization on each row of data in the multi-device joint time series matrix; using a sliding time window method to extract segments of the normalized multi-device joint time series matrix, ensuring that adjacent sliding time windows have a 50% data overlap rate; and sequentially extracting windows according to the sliding step size to generate continuous multi-device time series data segments, each of which is a matrix of dimension M×1000, serving as an independent input unit.

[0008] Furthermore, the calculation of the normalized spectral entropy value of each singular component matrix specifically involves: performing standard singular value decomposition on each independent input unit to obtain a left singular matrix, a singular value matrix, and a right singular matrix; sequentially taking values ​​for each integer i from 1 to M, and reconstructing the i-th singular component matrix based on the i-th column vector of the left singular matrix, the i-th diagonal element of the singular value matrix, and the i-th row vector of the right singular matrix; taking the arithmetic mean of all row vectors of each singular component matrix column by column to obtain a one-dimensional time series of length 1000 corresponding to the singular component matrix; performing a fast Fourier transform on the one-dimensional time series to obtain the power spectral density function of the one-dimensional time series, and calculating the normalized probability distribution of the power spectral density function; and normalizing the information entropy value of the normalized probability distribution to obtain the normalized spectral entropy value corresponding to the singular component matrix.

[0009] Furthermore, the dual-threshold segmentation specifically involves: arranging all normalized spectral entropy values ​​in ascending order, traversing all possible combinations of the two thresholds, calculating the inter-class variance of the three classes of data under each threshold combination, and selecting the two thresholds that maximize the inter-class variance as the final first and second spectral entropy thresholds, wherein the first spectral entropy threshold is less than the second spectral entropy threshold; classifying the singular component matrix whose normalized spectral entropy value is less than the first spectral entropy threshold into the equipment inherent operating condition component matrix, classifying the singular component matrix whose normalized spectral entropy value is greater than or equal to the first spectral entropy threshold and less than the second spectral entropy threshold into the cross-equipment topology coupling component matrix, and classifying the singular component matrix whose normalized spectral entropy value is greater than or equal to the second spectral entropy threshold into the random noise component matrix.

[0010] Furthermore, the step of constructing an undirected graph of multi-device topological coupling relationships based on the cross-device topological coupling component matrix and obtaining the device-independent operating condition component matrix by removing coupling disturbances and random noise specifically involves: for each singular component matrix that is divided into the cross-device topological coupling component matrix, calculating the projection coefficients of M sensors on the cross-device topological coupling component matrix; using the M sensors as nodes, with the node numbers corresponding one-to-one with the row numbers of the multi-device joint time series matrix, constructing an undirected graph of multi-device topological coupling relationships; adding element-wise all singular component matrices that are divided into device-inherent operating condition component matrices to obtain the device-inherent operating condition total matrix; adding element-wise all singular component matrices that are divided into cross-device topological coupling component matrices to obtain the cross-device topological coupling total matrix; adding element-wise all singular component matrices that are divided into random noise component matrices to obtain the random noise total matrix; and obtaining the device-independent operating condition component matrix based on the original multi-device joint time series matrix, the cross-device topological coupling total matrix, and the random noise total matrix within the current sliding time window.

[0011] Furthermore, the topological correlation degree between the computing devices specifically involves: extracting the set of M sensor projection coefficients corresponding to all cross-device topological coupling component matrices; based on the pre-established device-sensor mapping relationship, for any two different first devices and second devices, extracting the projection coefficients of all sensors belonging to the first device on all cross-device topological coupling component matrices, and the projection coefficients of all sensors belonging to the second device on all cross-device topological coupling component matrices, calculating the mutual information value of the two sets of projection coefficients, and defining the mutual information value as the topological correlation degree between the first device and the second device; sequentially calculating the topological correlation degree between all pairs of devices, and constructing a device topological correlation degree matrix of dimension K×K, where K is the total number of devices with deployed sensors.

[0012] Furthermore, the process of dividing the collaborative management subgroups and generating and updating the collaborative weight matrix corresponding to each subgroup in real time specifically involves: converting the device topology correlation matrix into a distance matrix between devices; grouping devices whose distance is less than the neighborhood radius and whose density is reachable from each other into the same collaborative management subgroup, and dividing devices that cannot be grouped into any collaborative management subgroup into independent devices, with each independent device forming a collaborative management subgroup containing only itself; for each collaborative management subgroup, extracting the rows and columns corresponding to all devices within the subgroup from the device topology correlation matrix to form the subtopology correlation matrix corresponding to the subgroup; performing row normalization on each row of the subtopology correlation matrix to obtain the collaborative weight matrix corresponding to the collaborative management subgroup; and immediately generating an updated collaborative weight matrix after each singular spectral entropy topology decomposition operation for a sliding time window, overwriting the collaborative weight matrix generated in the previous sliding time window.

[0013] Furthermore, the three-tiered collaborative control architecture consists of a global collaboration layer, a subgroup collaboration layer, and an equipment execution layer. The global collaboration layer receives production order information and all equipment operating condition data collected by the data acquisition module, and is responsible for the decomposition and global scheduling of the overall production task. The subgroup collaboration layer receives sub-task instructions issued by the global collaboration layer and the collaboration weight matrix of the corresponding collaborative control subgroup generated by the subgroup partitioning module, and is responsible for the collaborative optimization calculation of process parameters of each device within the subgroup. The equipment execution layer receives process parameter adjustment instructions issued by the subgroup collaboration layer, is responsible for the real-time adjustment of process parameters of specific devices, and feeds back the adjusted equipment operating condition data to the data acquisition module in real time.

[0014] Furthermore, the specific calculation of the collaborative optimization of process parameters within the subgroup is as follows: combining the collaborative weight matrix corresponding to the collaborative control subgroup and the equipment independent operating condition component matrix of all equipment within the subgroup generated by the topology reconstruction module, a sum-of-squares form of the process parameter optimization objective function is constructed using a multi-objective weighted summation method; using the adjustable process parameters of each equipment within the subgroup as decision variables, the rated value range and unit adjustable step size of each decision variable are pre-set; a set of constraints including equipment safe operation boundary constraints, process parameter coupling constraints, and production task capacity constraints are established; the three optimization objectives of maximizing the overall production efficiency of the subgroup, minimizing product quality deviation, and minimizing unit product energy consumption are transformed into deviation terms relative to the baseline operating condition; using the corresponding elements of the collaborative weight matrix as weight coefficients, the three deviation terms of each equipment are weighted and a sum-of-squares form of the overall objective function is constructed; a sequential quadratic programming algorithm is used to iteratively solve the constrained overall objective function to obtain the optimal process parameter adjustment values ​​of each equipment within the subgroup.

[0015] Furthermore, the global parameter optimization operation specifically involves: performing a global parameter optimization operation once every 24 hours; extracting product quality inspection results and equipment operating condition data from the past 24 hours, and calculating product quality deviation rate, equipment failure rate, and production efficiency indicators; setting preset thresholds for product quality deviation rate, equipment failure rate, and production efficiency as judgment benchmarks; when the real-time calculated indicators exceed the corresponding judgment benchmarks, successively fine-tuning the first spectral entropy threshold, the second spectral entropy threshold, the neighborhood radius of the DBSCAN algorithm, and the minimum number of samples in the DBSCAN algorithm according to a fixed fine-tuning step size, until the real-time calculated indicators fall back to the range of the corresponding judgment benchmarks, and updating the fixed parameters after fine-tuning in real time.

[0016] (III) Beneficial Effects This invention provides a collaborative intelligent control system for multiple textile production devices, which has the following advantages: (1) Accurate and synchronous acquisition of operating data of textile multi-process equipment is achieved. Missing value filling, outlier removal and Z-score standardization are used to eliminate dimensional interference. Then, a standard time sequence segment is generated by sliding time window to ensure data integrity and consistency, remove invalid noise, and output high-quality input units. This provides reliable data support for topology reconstruction, subgroup division and collaborative control, and improves the accuracy, stability and real-time performance of subsequent system analysis and control.

[0017] (2) By using singular value decomposition and normalized spectral entropy double threshold segmentation, the three types of components—the inherent operating conditions of the equipment, cross-equipment topological coupling, and random noise—are accurately separated, effectively removing coupling disturbances and noise interference, and restoring the true independent operating conditions of the equipment. At the same time, a multi-equipment topological coupling undirected graph is constructed to clearly depict the dynamic implicit associations between equipment, providing a reliable topological basis for subgroup division, and significantly improving the accuracy of operating condition analysis and the effectiveness of subsequent collaborative management.

[0018] (3) Based on the dynamic calculation of the correlation matrix of the equipment topology coupling relationship, the DBSCAN algorithm is used to accurately divide the collaborative control subgroups, breaking the limitations of the traditional fixed correlation; the collaborative weight matrix of each subgroup is generated and updated in real time, quantifying the degree of mutual influence between equipment, adapting to the dynamic changes of working conditions, improving the collaborative matching degree of equipment within the subgroup, providing accurate weight basis for collaborative control, and enhancing the flexibility and efficiency of control.

[0019] (4) Construct a three-level hierarchical collaborative architecture to realize production task decomposition, global scheduling, subgroup parameter collaborative optimization and equipment real-time feedback, avoiding the defects of high latency and concentrated load of centralized control; solve the optimal process parameters with production efficiency, product quality and unit energy consumption as the goal, and take into account multi-dimensional production needs; regularly carry out global parameter optimization, dynamically calibrate key parameters of the algorithm, and continuously improve the system control accuracy, operation stability and flexible adaptation capability. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the structure of the intelligent control system for collaborative management of multiple textile production equipment according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 This invention provides a collaborative intelligent control system for multiple textile production devices, comprising: The data acquisition module is used to collect equipment operating data for each process in textile production, construct a joint time series matrix of multiple devices, and use a sliding time window to extract and generate time series data segments of multiple devices as independent input units. Specifically, following the physical sequence of the textile production process, piezoelectric accelerometers are deployed on the spindle bearing seats and roving roller bearing seats of the spinning machine in the spinning process, the guide rollers of the batch warping machine in the warping process, the pressure rollers of the sizing machine in the sizing process, and the reed feet of the air-jet loom in the weaving process. Strain gauge tension sensors are deployed along the yarn running path of the above-mentioned equipment. Photoelectric speed sensors are deployed on the main shaft of the above-mentioned equipment. Hall current sensors are deployed at the main motor inlet of the above-mentioned equipment. Platinum resistance temperature sensors are deployed on the main motor windings of the above-mentioned equipment. All sensors are synchronized across the entire sensor network of the textile production workshop through the IEEE 1588 precision time protocol, and all sensors are sampled synchronously at a uniform sampling frequency of 10kHz. The collected raw multi-source heterogeneous data underwent missing value imputation and outlier removal. Linear interpolation was used to imput data segments with no more than 3 consecutive missing sampling points. Data segments with more than 3 consecutive missing sampling points were marked as invalid and discarded, and the time interval corresponding to the invalid data segment was recorded. Outliers were removed using the 3σ criterion. The normal range of a single variable was determined by the factory rated parameters of the corresponding sensor equipment and the statistical mean ± 2 standard deviations of the sensor's normal operation data over the past 30 days. Data points exceeding 3 standard deviations of the statistical mean of the normal range of a single variable were judged as outliers and removed. After preprocessing, all valid data from each sensor are aligned one-to-one according to a unified nanosecond-level timestamp, constructing a multi-device joint time series matrix of dimension M×T, where M is the total number of deployed sensors, T is the total number of valid sampling points for each sensor after alignment, each row of the matrix corresponds to the univariate time series data of one sensor, and each column corresponds to the sampling data of all sensors at the same timestamp; Z-score standardization is performed on the data in each row of the matrix, that is, for each data point in the row, the mean of all data points in the row is subtracted and then divided by the standard deviation of all data points in the row, so that the mean of each row of data after standardization is 0 and the standard deviation is 1, thus eliminating the influence of different physical dimensions; The standardized multi-device joint time series matrix is ​​segmented by segment using a sliding time window method. The length of a single sliding time window is set to 1000 sampling points, corresponding to a time length of 0.1 seconds, and the sliding step size is set to 500 sampling points, so that two adjacent sliding time windows form a 50% data overlap rate. The data is segmented window by window according to the sliding step size to generate continuous multi-device time series data segments. Each data segment is a matrix of dimension M×1000, which serves as an independent input unit.

[0023] The data acquisition module enables precise and synchronous acquisition of operating data from multiple textile equipment processes. It eliminates dimensional interference through missing value imputation, outlier removal, and Z-score standardization, and then generates standardized time-series segments using a sliding time window. This module ensures data integrity and consistency, removes invalid noise, and outputs high-quality input units, providing reliable data support for topology reconstruction, subgroup partitioning, and collaborative control. This enhances the accuracy, stability, and real-time performance of subsequent system analysis and management.

[0024] The topology reconstruction module is used to perform singular value decomposition on each independent input unit to obtain multiple singular component matrices, calculate the normalized spectral entropy value of each singular component matrix and perform double threshold segmentation to divide the device inherent operating condition component matrix, cross-device topology coupling component matrix and random noise component matrix. Based on the cross-device topology coupling component matrix, an undirected graph of multi-device topology coupling relationship is constructed, and coupling disturbances and random noise are removed to obtain the device independent operating condition component matrix. Specifically, for each independent input unit, i.e., the multi-device joint timing matrix with dimension M×1000 within each sliding time window, a standard singular value decomposition operation is performed to obtain a left singular matrix of dimension M×M, a singular value matrix of dimension M×1000, and a right singular matrix of dimension 1000×1000. Each column of the left singular matrix corresponds to a left singular vector, each row of the right singular matrix corresponds to a right singular vector, the diagonal elements of the singular value matrix are arranged in descending order of value, all off-diagonal elements of the singular value matrix are zero, each diagonal element corresponds to a unique singular value, and the number of singular values ​​is the same as the number of left and right singular vectors, which is M. For each integer i from 1 to M, the i-th column vector of the left singular matrix, the i-th diagonal element of the singular value matrix, and the i-th row vector of the right singular matrix are multiplied sequentially to reconstruct the i-th singular component matrix with dimension M×1000. The arithmetic mean of each row vector in each singular component matrix is ​​then calculated column-wise. The arithmetic mean of all M elements in each column is arranged in column order to obtain a one-dimensional time series of length 1000 corresponding to the singular component matrix. A 1024-point Fast Fourier Transform is then performed on this one-dimensional time series to obtain... To obtain the power spectral density function of the time series, sum all the values ​​of the power spectral density function to get the total power. Divide the power spectral density value of each frequency point by the total power to obtain the normalized probability distribution of the power spectrum of the time series. Calculate the information entropy value of the normalized probability distribution, and then divide the information entropy value by log2(1024) for normalization to obtain the normalized spectral entropy value corresponding to the singular component matrix. The normalized spectral entropy value ranges from 0 to 1. The larger the normalized spectral entropy value, the stronger the randomness of the information contained in the singular component matrix. The smaller the normalized spectral entropy value, the stronger the structure of the information contained in the singular component matrix. Otsu's dual-threshold adaptive segmentation method is used to segment the normalized spectral entropy values ​​corresponding to all M singular component matrices using dual-threshold segmentation. The specific execution process is as follows: All normalized spectral entropy values ​​are arranged in ascending order. All possible combinations of the two thresholds are iterated, and the inter-class variance of the three classes of data under each threshold combination is calculated. The two thresholds that maximize the inter-class variance are selected as the final first and second spectral entropy thresholds. Here, the three classes of data refer to the data after being segmented by the first and second spectral entropy thresholds, with the first spectral entropy threshold being less than the second spectral entropy threshold. Singular component matrices with normalized spectral entropy values ​​less than the first spectral entropy threshold are classified as equipment-inherent operating condition component matrices. Singular component matrices with normalized spectral entropy values ​​greater than or equal to the first spectral entropy threshold and less than the second spectral entropy threshold are classified as cross-equipment topology coupling component matrices. Singular component matrices with normalized spectral entropy values ​​greater than or equal to the second spectral entropy threshold are classified as random noise component matrices. After completing the classification of all singular component matrices, the number of equipment-inherent operating condition component matrices, cross-equipment topology coupling component matrices, and random noise component matrices are counted respectively. For each singular component matrix divided into cross-device topological coupling component matrices, the projection coefficients of M sensors on the cross-device topological coupling component matrix are calculated. The projection coefficient is defined as the L2 norm of the corresponding row vector of the sensor in the cross-device topological coupling component matrix. That is, for the j-th row of the cross-device topological coupling component matrix, j from 1 to M, the sum of squares of all 1000 elements in the row is calculated, and the arithmetic square root of the sum of squares is taken to obtain the projection coefficient of the j-th sensor on the cross-device topological coupling component matrix. With the M sensors as nodes, the node number corresponds one-to-one with the row number of the multi-device joint time series matrix. The product of the projection coefficients of any two sensors is used as the weight of the corresponding edge to construct an undirected graph of multi-device topological coupling relationship. If the product of the projection coefficients of two sensors is zero, no corresponding edge is constructed. This undirected graph is the intermediate representation carrier for topological reconstruction. The edge weight represents the degree of energy contribution of the coupling component to the sensor. Finally, the topological correlation degree between devices is obtained by mutual information calculation. The singular component matrices, which are divided into equipment inherent operating condition component matrices, are summed element-wise, i.e., elements with the same row number and column number are summed separately, to obtain a total equipment inherent operating condition matrix with dimension M×1000. The singular component matrices, which are divided into cross-device topology coupling component matrices, are summed element-wise, to obtain a total cross-device topology coupling matrix with dimension M×1000. The singular component matrices, which are divided into random noise component matrices, are summed element-wise, to obtain a total random noise matrix with dimension M×1000. The total cross-device topology coupling matrix is ​​subtracted element-wise from the original multi-device joint time series matrix within the current sliding time window, and then the total random noise matrix is ​​subtracted element-wise, to obtain an equipment independent operating condition component matrix with dimension M×1000 after removing coupling disturbances. This matrix retains only the inherent operating characteristics of the equipment to which each sensor belongs, eliminating the influence of cross-device implicit coupling disturbances and random noise.

[0025] The topology reconstruction module accurately separates three components—inherent equipment operating conditions, cross-equipment topological coupling, and random noise—through singular value decomposition and normalized spectral entropy dual-threshold segmentation. This effectively removes coupling disturbances and noise interference, restoring the true independent operating conditions of the equipment. Simultaneously, it constructs a multi-equipment topological coupling undirected graph, clearly depicting the dynamic implicit relationships between equipment, providing a reliable topological basis for subgroup partitioning, and significantly improving the accuracy of operating condition analysis and the effectiveness of subsequent collaborative management.

[0026] The subgroup partitioning module is used to calculate the topological correlation degree between devices based on the undirected graph of multi-device topological coupling relationship and construct the device topological correlation degree matrix. It uses the DBSCAN density-based spatial clustering algorithm to perform collaborative management subgroup partitioning, generate the collaborative weight matrix corresponding to each subgroup and update it in real time. Specifically, based on the undirected graph of multi-device topological coupling relationships, a set of M sensor projection coefficients corresponding to all cross-device topological coupling component matrices is extracted. According to the pre-established device-sensor mapping relationship, for any two different first and second devices, the projection coefficients of all sensors belonging to the first device on all cross-device topological coupling component matrices, and the projection coefficients of all sensors belonging to the second device on all cross-device topological coupling component matrices, are extracted respectively. The mutual information value of the two sets of projection coefficients is calculated, and this mutual information value is defined as the topological correlation degree between the first and second devices. The topological correlation degree between all pairs of devices is calculated sequentially, constructing a K×K device topological correlation degree matrix, where K is the total number of devices with deployed sensors, the element in the p-th row and q-th column of the matrix represents the topological correlation degree between the p-th device and the q-th device, the matrix is ​​symmetric, and the diagonal elements represent the topological correlation degree between a device and itself, with a value of 1. The DBSCAN density-based spatial clustering algorithm is used to divide all devices into collaborative control subgroups. The device topology correlation matrix is ​​converted into a distance matrix between devices. The element in the p-th row and q-th column of the distance matrix is ​​1 minus the element in the p-th row and q-th column of the device topology correlation matrix. The neighborhood radius of the DBSCAN algorithm is set to 0.3 and the minimum number of samples is 2. Devices whose distance in the distance matrix is ​​less than the neighborhood radius and are density-reachable are grouped into the same collaborative control subgroup. Devices that cannot be grouped into any collaborative control subgroup are classified as independent devices. Mutual density reachability means that the first device is within the neighborhood radius of the second device, and the second device is also within the neighborhood radius of the first device. After clustering, several collaborative control subgroups are generated. Each collaborative control subgroup contains at least 2 devices with strong topological correlation. Each independent device constitutes a collaborative control subgroup containing only itself. For each collaborative control subgroup, extract the rows and columns corresponding to all devices within the subgroup from the device topology correlation matrix to form the subtopology correlation matrix corresponding to the subgroup. Perform row normalization processing on each row of the subtopology correlation matrix, that is, divide each element of the row by the arithmetic sum of all elements in the row to obtain the collaborative weight matrix corresponding to the collaborative control subgroup. The dimension of the collaborative weight matrix is ​​consistent with the dimension of the subtopology correlation matrix. The element in the p-th row and q-th column of the matrix represents the collaborative influence weight of the q-th device on the p-th device. The larger the weight value, the greater the impact of the change in the working condition of the q-th device on the operating state of the p-th device. The arithmetic sum of all elements in each row is 1. After each singular spectral entropy topological decomposition operation for a sliding time window is completed, all the above operations are immediately executed to generate an updated equipment topology correlation matrix, collaborative control subgroup partitioning results, and collaborative weight matrices corresponding to each subgroup. The updated collaborative weight matrix is ​​then overwritten with the collaborative weight matrix generated in the previous sliding time window and used as the input basis for the process parameter optimization calculation of the subgroup collaborative layer in the next time step, ensuring that the collaborative weights can dynamically adapt to changes in equipment operating conditions, process parameter adjustments, and minor changes in equipment topology in real time.

[0027] The subgroup partitioning module dynamically calculates the correlation matrix based on the topological coupling relationship of the equipment and uses the DBSCAN algorithm to accurately partition the collaborative control subgroups, breaking the limitations of traditional fixed correlations. It generates and updates the collaborative weight matrix of each subgroup in real time, quantifies the degree of mutual influence between equipment, adapts to dynamic changes in operating conditions, improves the collaborative matching degree of equipment within the subgroup, provides accurate weight basis for collaborative control, and enhances the flexibility and efficiency of control.

[0028] The collaborative control module is used to build a three-level hierarchical collaborative control architecture, complete the decomposition and global scheduling of production tasks, the collaborative optimization calculation of process parameters within subgroups, the real-time adjustment and feedback of equipment process parameters, and periodically perform global parameter optimization operations.

[0029] Specifically, a three-tiered collaborative control architecture is constructed, consisting of a global collaboration layer, a subgroup collaboration layer, and an equipment execution layer from top to bottom. The global collaboration layer receives production order information from the textile production management system and all equipment operating condition data collected by the data acquisition module, and is responsible for the decomposition and global scheduling of the overall production tasks. The subgroup collaboration layer receives subtask instructions from the global collaboration layer and the collaboration weight matrix of the corresponding collaborative control subgroup generated by the subgroup partitioning module, and is responsible for the collaborative optimization calculation of process parameters of each device within the subgroup. The equipment execution layer receives process parameter adjustment instructions from the subgroup collaboration layer, is responsible for the real-time adjustment of process parameters of specific devices, and feeds back the adjusted equipment operating condition data to the data acquisition module in real time. The global collaboration layer, based on the product specifications, planned production quantity, and agreed delivery deadline specified in the production order, and combining the total number of equipment, real-time equipment status data, and real-time production load of each collaborative control subgroup, breaks down the overall production task into several sub-tasks corresponding to each process unit, according to the sequence of textile processes and product process adaptation rules. Based on the equipment configuration specifications, real-time capacity output values, and the degree of adaptation to the order's process requirements of each collaborative control subgroup, and following the principle of process priority matching, each sub-task is precisely assigned to the appropriate collaborative control subgroup, generating standardized sub-task instructions and issuing them to the corresponding subgroup's collaboration layer. The sub-task instructions clearly include the planned production quantity, process quality threshold, and task completion time requirements for the corresponding sub-task. After receiving the subtask instructions from the global coordination layer, the subgroup coordination layer combines the coordination weight matrix corresponding to the coordinated control subgroup generated by the subgroup partitioning module and the equipment independent operating condition component matrix of all devices in the subgroup generated by the topology reconstruction module. It then constructs a sum-of-squares form objective function for process parameter optimization using a multi-objective weighted summation method. The specific execution process is as follows: The system uses the adjustable process parameters of each piece of equipment within the subgroup as decision variables, covering six core adjustable parameters: spindle speed and draft ratio of spinning equipment, pressure of sizing rollers and temperature of sizing tank of sizing equipment, and warp tension and weft speed of weaving equipment. The system pre-sets the value range of each decision variable based on the equipment's factory rated parameters. For example, the spindle speed range for the spinning frame is 12000 r / min to 18000 r / min, the sizing tank temperature range is 85℃ to 95℃, and the warp tension range is 2000 N to 3500 N. Simultaneously, the adjustable step size for each decision variable is set: 100 r / min for speed parameters, 0.5℃ for temperature parameters, and 10 N for tension parameters. Three types of constraints are set, including equipment safety operation boundary constraints, process parameter coupling constraints, and production task capacity constraints, which together constitute the feasible domain of the optimization problem: Equipment safety operation boundary constraints: The values ​​of all decision variables must not exceed the above-mentioned rated value range, and the real-time calculated values ​​of equipment motor current and winding temperature must not exceed the rated safety threshold of the equipment; Process parameter coupling constraints: The coupling relationship between parameters is set according to the textile process mechanism. For example, the warp tension is positively correlated with the loom speed, and the tension increase is not more than 50N for every 100r / min increase in speed; the sizing speed is positively correlated with the pressure of the sizing roller, and the pressure increase is not more than 0.02MPa for every 5m / min increase in speed; Production task capacity constraints: The overall unit time capacity output of the subgroup multiplied by the available production time must be greater than or equal to the planned production quantity limited by the subtask instruction to meet the task completion time limit requirements; The three optimization objectives—maximizing the overall production efficiency of the subgroup, minimizing product quality deviation, and minimizing energy consumption per unit product—are transformed into deviation terms relative to the current baseline operating conditions: the production efficiency deviation term is the normalized form of the difference between the target efficiency and the current efficiency; the quality deviation term is the absolute value of the difference between the product quality inspection parameters and the standard values; and the energy consumption deviation term is the difference between the energy consumption per unit product and the baseline energy consumption. Using the weight elements of the corresponding equipment in the collaborative weight matrix as coefficients, the three deviation terms of all equipment in the subgroup are weighted separately to construct a total objective function in the form of a sum of squares. This total objective function is a continuously differentiable nonlinear function with respect to the decision variables, which meets the solution premise of numerical optimization algorithms. A sequential quadratic programming algorithm is used to iteratively solve the constrained overall objective function: In each iteration, the original nonlinear constrained optimization problem is expanded into a quadratic programming subproblem at the current iteration point. The gradient vector of the objective function and the Jacobian matrix of the constraints are calculated. The iteration step size is obtained by solving the quadratic programming subproblem, and the values ​​of the decision variables are updated. The iteration terminates when the iteration step size is less than the preset convergence threshold 1e-6. The current values ​​of the decision variables are output as the optimal process parameter adjustment values ​​for each device in the subgroup. Process parameter adjustment instructions are generated and sent to the device execution layer of the corresponding device. After receiving the process parameter adjustment command issued by the subgroup coordination layer, the equipment execution layer completes the real-time adjustment of the corresponding equipment process parameters. After the equipment process parameters are adjusted, the corresponding sensors deployed in the data acquisition module immediately collect the adjusted equipment vibration, tension, speed, current, and temperature operating condition data. The collected adjusted equipment operating condition data is fed back to the data acquisition module in real time as input data for the next sliding time window singular spectral entropy topological decomposition. A global parameter optimization operation is performed every 24 hours; product quality inspection results and equipment operating data from the past 24 hours are extracted to calculate product quality deviation rate, equipment failure rate, and production efficiency indicators; preset product quality deviation rate threshold, equipment failure rate threshold, and production efficiency benchmark threshold are used as judgment benchmarks. When the real-time calculated indicators exceed the corresponding judgment benchmarks, the first spectral entropy threshold and the second spectral entropy threshold in step two, as well as the neighborhood radius and minimum sample number of the DBSCAN algorithm in step three, are finely adjusted step by step according to a fixed fine-tuning step size until the real-time calculated indicators fall back to the corresponding judgment benchmark range; the fixed parameters after fine-tuning are updated to the corresponding algorithm module for subsequent singular spectral entropy topological decomposition and collaborative control subgroup partitioning in the sliding time window, continuously improving the accuracy and stability of multi-device collaborative control.

[0030] The collaborative control module constructs a three-level hierarchical collaborative architecture to achieve production task decomposition, global scheduling, subgroup parameter collaborative optimization, and real-time equipment feedback, avoiding the defects of high latency and concentrated load in centralized control. It solves for optimal process parameters with production efficiency, product quality, and unit energy consumption as objectives, taking into account multi-dimensional production needs. It regularly conducts global parameter optimization, dynamically calibrates key algorithm parameters, and continuously improves the system's control accuracy, operational stability, and flexible adaptability.

[0031] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, and combinations thereof. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0032] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0033] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A collaborative intelligent control system for multiple textile production equipment, characterized in that: It includes a data acquisition module, a topology reconstruction module, a subgroup partitioning module, and a collaborative control module; The data acquisition module is used to collect equipment operating data for each process in textile production, construct a joint time series matrix of multiple devices, and use a sliding time window to extract and generate time series data segments of multiple devices as independent input units. The topology reconstruction module is used to perform singular value decomposition on each independent input unit to obtain multiple singular component matrices, calculate the normalized spectral entropy value of each singular component matrix and perform double threshold segmentation to divide the device inherent operating condition component matrix, cross-device topology coupling component matrix and random noise component matrix. Based on the cross-device topology coupling component matrix, an undirected graph of multi-device topology coupling relationship is constructed, and coupling disturbances and random noise are removed to obtain the device independent operating condition component matrix. The subgroup partitioning module is used to calculate the topological correlation degree between devices based on the undirected graph of multi-device topological coupling relationship and construct the device topological correlation degree matrix. It uses the DBSCAN density-based spatial clustering algorithm to perform collaborative management subgroup partitioning, generate the collaborative weight matrix corresponding to each subgroup and update it in real time. The collaborative control module is used to build a three-level hierarchical collaborative control architecture, complete the decomposition and global scheduling of production tasks, the collaborative optimization calculation of process parameters within subgroups, the real-time adjustment and feedback of equipment process parameters, and periodically perform global parameter optimization operations.

2. The intelligent control system for collaborative management of multiple textile production equipment according to claim 1, characterized in that: The specific steps of using a sliding time window to extract multi-device time series data segments as independent input units are as follows: Z-score standardization is performed on each row of data in the multi-device joint time series matrix; a sliding time window method is used to extract segments of the standardized multi-device joint time series matrix, ensuring a 50% data overlap between adjacent sliding time windows; and continuous multi-device time series data segments are generated by sequentially extracting each window according to the sliding step size. Each data segment is a matrix of dimension M×1000, serving as an independent input unit.

3. The intelligent control system for collaborative management of multiple textile production equipment according to claim 1, characterized in that: The calculation of the normalized spectral entropy of each singular component matrix is ​​specifically as follows: Perform standard singular value decomposition on each independent input unit to obtain a left singular matrix, a singular value matrix, and a right singular matrix; sequentially take values ​​for each integer i from 1 to M, and reconstruct the i-th singular component matrix based on the i-th column vector of the left singular matrix, the i-th diagonal element of the singular value matrix, and the i-th row vector of the right singular matrix; take the arithmetic mean of all row vectors of each singular component matrix column by column to obtain a one-dimensional time series of length 1000 corresponding to that singular component matrix; perform a fast Fourier transform on the one-dimensional time series to obtain the power spectral density function of the one-dimensional time series, and calculate the normalized probability distribution of the power spectral density function; normalize the information entropy value of the normalized probability distribution to obtain the normalized spectral entropy value corresponding to that singular component matrix.

4. The intelligent control system for collaborative management of multiple textile production equipment according to claim 3, characterized in that: The specific steps of performing dual-threshold segmentation are as follows: Arrange all normalized spectral entropy values ​​in ascending order, traverse all possible combinations of the two thresholds, calculate the inter-class variance of the three classes of data under each threshold combination, and select the two thresholds that maximize the inter-class variance as the final first and second spectral entropy thresholds, wherein the first spectral entropy threshold is less than the second spectral entropy threshold; classify the singular component matrix whose normalized spectral entropy value is less than the first spectral entropy threshold as the equipment inherent operating condition component matrix, classify the singular component matrix whose normalized spectral entropy value is greater than or equal to the first spectral entropy threshold and less than the second spectral entropy threshold as the cross-equipment topology coupling component matrix, and classify the singular component matrix whose normalized spectral entropy value is greater than or equal to the second spectral entropy threshold as the random noise component matrix.

5. The intelligent control system for collaborative management of multiple textile production equipment according to claim 4, characterized in that: The process of constructing an undirected graph of multi-device topological coupling relationships based on the cross-device topological coupling component matrix and removing coupling disturbances and random noise to obtain the device-independent operating condition component matrix is ​​as follows: For each singular component matrix that is divided into the cross-device topological coupling component matrix, the projection coefficients of M sensors on the cross-device topological coupling component matrix are calculated; using the M sensors as nodes, with the node numbers corresponding one-to-one with the row numbers of the multi-device joint time series matrix, an undirected graph of multi-device topological coupling relationships is constructed; all singular component matrices that are divided into device-inherent operating condition component matrices are summed element-wise to obtain the device-inherent operating condition total matrix; all singular component matrices that are divided into cross-device topological coupling component matrices are summed element-wise to obtain the cross-device topological coupling total matrix; all singular component matrices that are divided into random noise component matrices are summed element-wise to obtain the random noise total matrix; based on the original multi-device joint time series matrix, the cross-device topological coupling total matrix, and the random noise total matrix within the current sliding time window, the device-independent operating condition component matrix is ​​obtained.

6. The intelligent control system for collaborative management of multiple textile production equipment according to claim 1, characterized in that: The topological correlation degree between computing devices is specifically as follows: extract the set of M sensor projection coefficients corresponding to all cross-device topological coupling component matrices; based on the pre-established device sensor mapping relationship, for any two different first devices and second devices, extract the projection coefficients of all sensors belonging to the first device on all cross-device topological coupling component matrices, and the projection coefficients of all sensors belonging to the second device on all cross-device topological coupling component matrices, respectively, calculate the mutual information value of the two sets of projection coefficients, and define the mutual information value as the topological correlation degree between the first device and the second device; calculate the topological correlation degree between all pairs of devices in turn, and construct a device topological correlation degree matrix of dimension K×K, where K is the total number of devices with deployed sensors.

7. The intelligent control system for collaborative management of multiple textile production equipment according to claim 6, characterized in that: The process of dividing the collaborative management subgroups and generating and updating the collaborative weight matrix for each subgroup in real time involves: converting the device topology correlation matrix into a distance matrix between devices; grouping devices whose distance to each other in the distance matrix is ​​less than the neighborhood radius and whose density allows them to be mutually reached into the same collaborative management subgroup; dividing devices that cannot be assigned to any collaborative management subgroup into independent devices, with each independent device forming a collaborative management subgroup containing only itself; for each collaborative management subgroup, extracting the rows and columns corresponding to all devices within the subgroup from the device topology correlation matrix to form the subtopology correlation matrix corresponding to the subgroup; and performing row normalization on each row of the subtopology correlation matrix to obtain the collaborative weight matrix corresponding to the collaborative management subgroup. After each singular spectral entropy topological decomposition operation for a sliding time window is completed, an updated collaborative weight matrix is ​​immediately generated, overwriting the collaborative weight matrix generated in the previous sliding time window.

8. The intelligent control system for collaborative management of multiple textile production equipment according to claim 7, characterized in that: The three-tiered collaborative control architecture consists of a global collaboration layer, a subgroup collaboration layer, and an equipment execution layer. The global collaboration layer receives production order information and all equipment operating condition data collected by the data acquisition module, and is responsible for the decomposition and global scheduling of the overall production task. The subgroup collaboration layer receives subtask instructions issued by the global collaboration layer and the collaboration weight matrix of the corresponding collaborative control subgroup generated by the subgroup partitioning module, and is responsible for the collaborative optimization calculation of process parameters of each device within the subgroup. The equipment execution layer receives process parameter adjustment instructions issued by the subgroup collaboration layer, is responsible for the real-time adjustment of process parameters of specific devices, and feeds back the adjusted equipment operating condition data to the data acquisition module in real time.

9. The intelligent control system for collaborative management of multiple textile production equipment according to claim 8, characterized in that: The specific calculation of the collaborative optimization of process parameters within the subgroup is as follows: Combining the collaborative weight matrix corresponding to the collaborative control subgroup and the equipment independent operating condition component matrix of all equipment within the subgroup generated by the topology reconstruction module, a sum-of-squares form objective function for process parameter optimization is constructed using a multi-objective weighted summation method; using the adjustable process parameters of each equipment within the subgroup as decision variables, the rated value range and unit adjustable step size of each decision variable are pre-set; a set of constraints including equipment safe operation boundary constraints, process parameter coupling constraints, and production task capacity constraints are established; the three optimization objectives of maximizing the overall production efficiency of the subgroup, minimizing product quality deviation, and minimizing unit product energy consumption are transformed into deviation terms relative to the baseline operating condition; using the corresponding elements of the collaborative weight matrix as weight coefficients, the three deviation terms of each equipment are weighted and a sum-of-squares form overall objective function is constructed; a sequential quadratic programming algorithm is used to iteratively solve the constrained overall objective function to obtain the optimal process parameter adjustment values ​​for each equipment within the subgroup.

10. A collaborative intelligent control system for multiple textile production equipment according to claim 9, characterized in that: The specific steps of performing global parameter optimization are as follows: perform global parameter optimization once every 24 hours; extract product quality inspection results and equipment operating condition data from the past 24 hours, and calculate product quality deviation rate, equipment failure rate and production efficiency indicators. The product quality deviation rate threshold, equipment failure rate threshold, and production efficiency benchmark threshold are preset as judgment benchmarks. When the real-time calculated index exceeds the corresponding judgment benchmark, the first spectral entropy threshold, the second spectral entropy threshold, the neighborhood radius of the DBSCAN algorithm, and the minimum number of samples of the DBSCAN algorithm are finely adjusted step by step according to a fixed fine-tuning step size until the real-time calculated index falls back to the corresponding judgment benchmark range. The fixed parameters after fine-tuning are updated in real time.

Images