Glass processing production distributed control method and system based on internet of things

Abstract

This invention discloses a distributed control method and system for glass processing production based on the Internet of Things (IoT), belonging to the field of industrial automation technology. The method includes: cross-node correlation and fusion, condition feature extraction, and threshold parsing of structured raw process data; extraction of state parameters and control thresholds to generate multi-source fused control data; using rule-based control methods to perform condition matching, logical judgment, and decision processing on the multi-source fused control data to generate preliminary control commands; after the collaborative control commands are issued, edge intelligence methods are used to coordinate control points to execute control actions, real-time collection of operating status and control response data, and the use of anomaly detection and confidence assessment to filter invalid data and generate a high-quality control response dataset. This invention improves the adaptability of control commands to actual operating conditions, enhances the timeliness and stability of multi-node collaborative control, and ensures the continuous operation of the glass processing production process.

Description

Technical Field

[0001] This invention relates to the field of industrial automation technology, and in particular to a distributed control method and system for glass processing production based on the Internet of Things. Background Technology

[0002] With the continuous development of intelligent manufacturing and industrial automation technologies, glass processing steps such as cutting, edging, cleaning, and tempering are gradually evolving towards networking and collaboration. Existing glass processing production lines typically manage each process node through on-site control equipment, monitoring platforms, and data acquisition units to achieve equipment operation monitoring, process parameter acquisition, and production process control, thereby providing basic support for the automation of glass processing production.

[0003] However, in the glass processing and production process involving multiple process nodes and parallel collaboration of multiple devices, existing control methods mostly rely on preset parameters and local rules for control, lacking a unified fusion, collaborative scheduling, and closed-loop update mechanism for multi-source state data. This makes it difficult for control commands to adapt to changes in operating conditions in a timely manner, thereby affecting the collaborative efficiency between multiple nodes and the continuous and stable operation of the production process. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a distributed control method for glass processing production based on the Internet of Things to solve the problems of slow response speed and low coordination efficiency in the existing distributed control of glass processing production.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, this invention provides a distributed control method for glass processing production based on the Internet of Things (IoT). The method includes: collecting process node data and preprocessing it according to unified identification rules, time synchronization rules, and data cleaning rules to generate structured raw process data; performing cross-node correlation fusion, condition feature extraction, and threshold parsing on the structured raw process data to extract state parameters and control thresholds, generating multi-source fused control data; using rule-based control methods to perform condition matching, logical judgment, and decision processing on the multi-source fused control data to generate preliminary control commands; uploading the preliminary control commands as cloud coordination inputs to a cloud coordinator for unified scheduling and conflict coordination to generate collaborative control commands; after issuing the collaborative control commands, using edge intelligence methods to coordinate control points to execute control actions, collecting real-time operating status and control response data, and using anomaly detection and confidence assessment to filter invalid data and generate a high-quality control response dataset; performing operational deviation and trend analysis on the high-quality control response dataset to generate state diagnostic results, and combining trigger conditions for judgment and optimization to generate updated collaborative control commands.

[0007] As a preferred embodiment of the distributed control method for glass processing production based on the Internet of Things described in this invention, the method involves: collecting the operating status parameters, unique identification information, and control boundary parameters of process nodes to generate raw process data; and using outlier detection and median filtering methods to remove outliers and filter out noise, thereby generating structured raw process data.

[0008] As a preferred embodiment of the distributed control method for glass processing production based on the Internet of Things described in this invention, the specific steps for generating multi-source fusion control data are as follows: The original structured process data is dimensionality reduced, and the principal component vectors of glass processing condition changes are extracted to generate process feature vectors. The process feature vectors are processed for time synchronization and format standardization to generate unified time-series fusion data. Based on unified time-series fusion data, state identification and threshold calculation analysis are performed to extract state parameters and control thresholds, and a set of state and control indicators is generated. By mapping and integrating the set of status and control indicators with the unique identifier information in the structured original process data, multi-source fusion control data is generated.

[0009] As a preferred embodiment of the distributed control method for glass processing production based on the Internet of Things described in this invention, the specific steps for generating preliminary control commands are as follows: Extract control-related fields and unique identification information from multi-source fused control data, and organize them to generate an analysis dataset; Based on a pre-defined rule base, condition matching and logical reasoning are performed on the analysis dataset to generate trigger identifiers; Based on the trigger identifier, a control strategy is selected from the preset rule base, and control parameters are calculated to generate a set of control parameters; The set of control parameters and control thresholds are processed for control decision-making to generate preliminary control commands.

[0010] As a preferred embodiment of the distributed control method for glass processing production based on the Internet of Things described in this invention, the specific steps for generating collaborative control instructions are as follows: Based on the initial control instructions, analyze resource conflicts and coordination needs, and generate global control requirement data. Multi-label classification and clustering algorithms are used to dynamically classify the set of control requests, identify the correlation and conflict of control objectives, and generate global regulation and control demand data. The global control and regulation demand data is uniformly scheduled to obtain the control allocation results; A template mapping and instruction encapsulation method is used to perform structured processing on the control allocation results and generate collaborative control instructions.

[0011] As a preferred embodiment of the distributed control method for glass processing production based on the Internet of Things described in this invention, the specific steps for generating global control demand data are as follows: Extract control target information, control parameter information, execution time information, and priority information from the preliminary control instructions to generate a set of control requests to be coordinated; Based on the time overlap, regulation direction, and resource consumption relationships among the control objectives in the set of control requests to be coordinated, control conflict relationships are identified, and conflict identification results are generated. Based on the conflict identification results, related control requests are prioritized and collaboratively merged to generate a coordinated ranking result. The conflict identification results and coordination and ranking results are integrated to generate global control requirements data.

[0012] As a preferred embodiment of the IoT-based distributed control method for glass processing production described in this invention, the specific steps for generating a high-quality control response dataset are as follows: Edge intelligence methods are used to trigger collaborative control commands and simultaneously collect field responses to generate raw operating status and control response data; Anomaly detection methods are used to identify and remove abnormal data from the original operating status and control response data, generating preliminary valid data. The credibility of the preliminary valid data is assessed, and low-confidence information is filtered out to generate a high-quality control response dataset.

[0013] As a preferred embodiment of the distributed control method for glass processing production based on the Internet of Things described in this invention, the specific steps for generating updated collaborative control instructions are as follows: A multi-scale diagnostic mechanism is adopted to decompose a high-quality control response dataset into multiple scales and extract state change features at different time scales. By utilizing the characteristics of state changes, operational deviation analysis and trend identification are performed on the operating state to generate state diagnosis results; Based on the status diagnosis results and preset trigger conditions, determine whether the status meets the update requirement and output the trigger determination result; When the trigger determination result meets the update requirements, the trigger determination result is combined with the status diagnosis result to adjust and optimize the original operating state and generate updated collaborative control instructions.

[0014] As a preferred embodiment of the distributed control method for glass processing production based on the Internet of Things described in this invention, the specific steps for outputting the trigger determination result are as follows: Extract operational deviation data and trend recognition data from the status diagnosis results to generate diagnostic input data; The diagnostic input data is matched with the deviation threshold condition and trend persistence condition in the preset trigger conditions to generate condition matching results; Based on the condition matching results, determine whether the current running status meets the update requirement and generate a trigger judgment result; When the trigger determination result indicates that the update requirement is met, the trigger determination result is output.

[0015] Secondly, this invention provides a distributed control system for glass processing production based on the Internet of Things (IoT), comprising a process acquisition module, a data fusion module, a rule judgment module, a cloud scheduling module, an edge execution module, and a diagnostic optimization module. The process acquisition module collects process node data and preprocesses it according to unified identification rules, time synchronization rules, and data cleaning rules to generate structured raw process data. The data fusion module performs cross-node correlation fusion, condition feature extraction, and threshold parsing on the structured raw process data to extract state parameters and control thresholds, generating multi-source fused control data. The rule judgment module uses rule-based control methods to perform condition matching on the multi-source fused control data. The system comprises four modules: a configuration module, a logic judgment and decision-making module, and an edge execution module. The edge execution module coordinates control points to perform control actions after the collaborative control commands are issued, collects real-time operating status and control response data, and uses anomaly detection and confidence assessment to filter invalid data and generate a high-quality control response dataset. The diagnostic optimization module performs operational deviation and trend analysis on the high-quality control response dataset, generates status diagnostic results, and combines trigger conditions for judgment and optimization to generate updated collaborative control commands.

[0016] The beneficial effects of this invention are as follows: by preprocessing and fusing process node data, multi-source fused control data containing state parameters and control thresholds is formed. Based on this, combined with rule control, unified cloud scheduling, edge execution feedback, and multi-scale diagnostic updates, a distributed closed-loop control link for the glass processing and production process is constructed. This achieves continuous connection between control decisions, collaborative execution, and state feedback, thereby improving the adaptability of control commands to actual working conditions, enhancing the timeliness and stability of multi-node collaborative control, and ensuring the continuous operation of the glass processing and production process. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a distributed control method for glass processing and production based on the Internet of Things.

[0019] Figure 2 This is a schematic diagram of a distributed control system for glass processing and production based on the Internet of Things.

[0020] Figure 3 A flowchart for generating multi-source fusion control data.

[0021] Figure 4 A flowchart for optimizing the diagnostics of updated control commands. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides a distributed control method for glass processing production based on the Internet of Things, including the following steps: S1. Collect process node data and preprocess it according to unified identification rules, time synchronization rules and data cleaning rules to generate structured raw process data.

[0026] S1.1 Collect the operating status parameters, unique identification information and control boundary parameters of the process nodes, generate raw process data, and use outlier detection method and median filtering method to remove outliers and filter out noise to generate structured raw process data.

[0027] Specifically, operating status parameters are collected from the sensor acquisition interfaces and execution feedback interfaces corresponding to each process node of the glass processing production line. Unique identification information is read from the equipment code, workpiece code, and process code corresponding to the process node. Control boundary parameters are retrieved from the process formula parameter table and equipment control parameter table. During data collection, the data interfaces corresponding to each process node are polled and read. The operating status parameters, unique identification information, and control boundary parameters are recorded according to the same time mark and the same unique identification information to generate raw process data. An outlier detection method is applied to the operating status parameters and control boundary parameters in the raw process data to remove outliers. Median filtering method is applied to the operating status parameters after outlier removal to remove noise. The operating status parameters, unique identification information, and control boundary parameters after outlier removal and noise removal are reintegrated according to the corresponding relationship to generate structured raw process data.

[0028] S2. Perform cross-node correlation fusion, operating condition feature extraction and threshold analysis on the structured raw process data, extract state parameters and control thresholds, and generate multi-source fused control data.

[0029] S2.1. Perform dimensionality reduction processing on the structured original process data, extract the principal component vector of glass processing condition changes, and generate process feature vector.

[0030] Specifically, when performing principal component analysis on the structured raw process data, the operating state parameters and control boundary parameters in the structured raw process data are normalized to eliminate the influence of dimensions. The covariance matrix of the normalized structured raw process data is calculated to obtain the set of covariance values ​​between each process parameter in the structured raw process data. The set of covariance values ​​is used to characterize the degree of linear correlation between each process parameter in the structured raw process data. Eigenvalue decomposition is performed on the covariance matrix of the structured raw process data to extract the eigenvectors and eigenvalues ​​corresponding to the covariance matrix. The eigenvalues ​​of the covariance matrix are sorted in descending order according to their numerical values. The eigenvectors corresponding to the eigenvalues ​​of the top few covariance matrices with a cumulative contribution rate of, for example, 85% or more are selected as the principal components of the structured raw process data. The normalized structured raw process data is projected onto the selected principal component eigenvectors to obtain the principal component vectors of glass processing condition changes, generating the process feature vectors.

[0031] It should also be noted that the specific steps for calculating the covariance matrix of the normalized structured original process data are as follows: After normalizing the structured original process data by mean, let the normalized structured original process data be represented as a matrix. Then the... The and the first The expression for calculating the covariance between process parameters is: ; in, Indicates the first The process parameter and the first Covariance between process parameters Indicates the first A column vector of process parameters, Indicates the first A column vector of process parameters, Represents the total number of samples. Indicates the number of process parameters. This indicates the index number of the current sample. Indicates the first In the nth sample Normalized values ​​of each process parameter Indicates the first The average value of each process parameter, Indicates the first In the nth sample Normalized values ​​of each process parameter Indicates the first The average value of each process parameter, This represents the first element in the structured raw process data. An index of process parameters, This represents the first element in the structured raw process data. An index of process parameters; For all and By combining the results, the covariance matrix can be obtained. ,in, Represents the covariance matrix. Represents the real number field, the first The element is .

[0032] S2.2 Perform time-series synchronization and format standardization processing on the process feature vectors to generate unified time-series fusion data.

[0033] Specifically, timestamp information is extracted from process feature vectors generated from different acquisition sources, and a unified timeline is constructed based on the timestamp information. Interpolation methods are used to interpolate and complete process feature vectors with inconsistent or missing sampling times, so that all process feature vectors have consistent time granularity on the unified timeline. In accordance with the unified data structure format requirements, the field names, data types, and numerical units of the process feature vectors are renamed and units are converted to ensure consistent field dimensions. All process feature vectors after time-series alignment and format standardization are concatenated to generate unified time-series fusion data.

[0034] S2.3. Based on the unified time-series fusion data, perform state identification and threshold calculation analysis, extract state parameters and control thresholds, and generate a set of state and control indicators.

[0035] Specifically, based on the process feature parameters in the unified time-series fusion data, empirical mapping is performed according to the existing glass processing experience parameter range to initially classify and label the operating states corresponding to data points at each time point. For the process feature parameters under each state, a numerical interpolation method is used to select similar process feature parameters within adjacent time periods from the historical process feature parameters in the unified time-series fusion data, construct an interpolation function, and perform numerical calculations on the process feature parameters under boundary conditions to obtain continuous and representative control boundary values. The identified operating state labels are matched with the calculated control boundary values ​​to form operating state parameter and control threshold pairs. Finally, the state parameters and control thresholds corresponding to each time point are recorded uniformly to generate a set of state and control indicators.

[0036] It should also be noted that the specific steps for constructing the interpolation function are as follows: For similar process feature parameters selected from adjacent time periods in the historical process feature parameters of the unified time series fusion data, determine the endpoint values ​​of the interpolation interval according to the time series order; based on the known endpoint values, select either a linear interpolation method or a spline interpolation method from the existing technology as the interpolation function type; based on the selected interpolation method, calculate the interpolation value corresponding to the intermediate time within the interpolation interval, and construct a complete interpolation function to represent the calculation and change process of the process feature parameters at different time points; The control threshold is calculated by using a numerical interpolation method on historical process feature parameters in unified time-series fusion data, combined with the range of glass processing experience parameters. Specifically, it involves selecting similar process feature parameters within adjacent time periods, performing continuous numerical calculations on the process feature parameters under boundary conditions based on the constructed interpolation function, and generating representative control boundary values, which are then used as the control threshold.

[0037] S2.4 Map and integrate the set of status and control indicators with the unique identifier information in the structured original process data to generate multi-source fusion control data.

[0038] Specifically, based on the unique identifier information in the structured raw process data, the corresponding identifier information in the set of state and control indicators is matched. The state parameters and control thresholds are associated one by one with the corresponding unique identifier information in the structured raw process data using a key-value pair method, thus completing the association mapping of multi-source data. The mapped state parameters and control thresholds are then integrated with the operating state parameters and control boundary parameters in the structured raw process data to generate multi-source fusion control data.

[0039] S3. Use rule-based control methods to perform condition matching, logical judgment and decision processing on multi-source fusion control data to generate preliminary control instructions.

[0040] S3.1 Extract control-related fields and unique identification information from multi-source fusion control data, and organize them to generate an analysis dataset.

[0041] Specifically, control-related fields, including state parameters, control thresholds, and operating status parameters, are extracted from multi-source fused control data through field filtering operations. Unique identification information fields are also extracted. These extracted fields are then categorized and sorted according to the unique identification information, and duplicates are eliminated to ensure their uniqueness. Finally, the filtered and organized control-related fields and unique identification information are integrated into a unified format to generate an analysis dataset.

[0042] S3.2 Based on the preset rule base, perform condition matching and logical reasoning on the analysis dataset to generate trigger identifiers.

[0043] Specifically, based on a preset rule base, the system iterates through the state parameter field, control threshold field, and operating state parameter field of each data point in the dataset. Each parameter is compared and judged against the corresponding process characteristic parameter threshold, state parameter threshold, and control threshold in the preset rule base. This includes determining whether the state parameter is greater than, less than, equal to, or between the upper and lower limits of the state parameter threshold; determining whether the control threshold is greater than, less than, or between the upper and lower limits of the control threshold; and determining whether the operating state parameter meets the judgment logic of the operating state parameter threshold. Based on the judgment results, multiple judgment conditions are combined using logical AND, logical OR, and logical NOT operations to determine whether the overall triggering conditions defined in the preset rule base are met. When all judgment conditions meet the requirements in the preset rule base, a trigger identifier is generated.

[0044] The specific steps of the preset rule base should also be explained: The preset rule base collects and organizes historical process data, clarifies control objectives and anomaly judgment criteria, and sets thresholds for process characteristic parameters, state parameters, and control thresholds according to their categories, thus compiling multiple conditional judgment rules. Each judgment rule includes a specific parameter name, judgment type (including greater than, less than, equal to, or falling within a range), corresponding threshold value (i.e., process characteristic parameter threshold, state parameter threshold, or control threshold), and logical operation relationship. All judgment rules are arranged in the preset rule base according to priority or logical judgment order, and the rule base content is periodically supplemented and adjusted based on updated historical process data to ensure the adaptability and completeness of the preset rule base.

[0045] S3.3. Based on the trigger identifier, select a control strategy from the preset rule base, calculate the control parameters, and generate a set of control parameters.

[0046] Specifically, the system searches for the control strategy content corresponding to the trigger identifier field in the preset rule base, locating the control target parameter name, target value range, and calculation method contained in the control strategy. Based on the control target parameter name in the control strategy, the system extracts the corresponding process characteristic parameter field, state parameter field, and operating state parameter field from the analysis dataset, and calculates them according to the parameter estimation method defined in the control strategy, such as weighted average, moving average, or interval fitting, to obtain the estimated value of the control target parameter. Based on the deviation between the estimated value and the target value, the control quantity is adjusted according to the calculation formula or estimation method in the control strategy to calculate the final control parameter. After summarizing all control parameters, a control parameter set is generated.

[0047] S3.4. Perform control decision processing on the set of control parameters and control thresholds to generate preliminary control commands.

[0048] Specifically, for each control parameter name in the control parameter set, the corresponding upper and lower control limits are searched in the control threshold field. Each control parameter value is then checked against its corresponding upper and lower limits. If the value is within the control threshold range, the current control state is maintained. If the value exceeds the upper limit, the control intensity is reduced. If the value is below the lower limit, the control intensity is increased. Based on the results and the control parameter names, a preliminary control instruction is generated, including the control action type, execution direction, and control magnitude information.

[0049] S4. Upload the preliminary control commands as cloud coordination inputs to the cloud coordinator for unified scheduling and conflict coordination, and generate collaborative control commands.

[0050] S4.1 Based on the preliminary control instructions, analyze resource conflicts and coordination needs, and generate global control requirement data.

[0051] S4.1.1 Extract the control target information, control parameter information, execution time information and priority information from the preliminary control instructions to generate a set of control requests to be coordinated.

[0052] Specifically, the preliminary control instructions are parsed according to a preset field order, reading the control target information, control parameter information, execution time information, and priority information from each instruction. The control target information is used as a classification basis to merge the corresponding control parameter information, recording the corresponding execution time information and priority information to form parameter and time correspondences under the same control target. The merged data is sorted according to execution time information and arranged sequentially according to priority information, completing the field organization and ordering of the control requests. The organized control target information, control parameter information, execution time information, and priority information are combined according to a unified record format to generate a set of control requests to be coordinated.

[0053] S4.1.2 Based on the time overlap, regulation direction, and resource usage relationships among the control objectives in the set of control requests to be coordinated, identify control conflict relationships and generate conflict identification results.

[0054] Specifically, the execution time information, control parameter information, and resource occupancy information corresponding to each control objective are extracted from the set of control requests to be coordinated. Each control objective is compared pairwise according to its execution time information to determine if there is an overlap between the execution start and end times, thus establishing a time overlap relationship. Based on the parameter change requirements in the control parameter information, the consistency of the control direction relationship for each control objective is determined, identifying unidirectional and reverse control relationships on the same control object. Combined with resource occupancy information, the corresponding equipment occupancy, workstation occupancy, and processing time slot occupancy for each control objective are checked to determine if multiple control objectives share the same resource within the same time period, thus establishing a resource occupancy relationship. The time overlap relationship, control direction relationship, and resource occupancy relationship are summarized accordingly, and control objectives with overlapping times, conflicting control directions, or overlapping resource occupancy are marked, generating a conflict identification result.

[0055] S4.1.3. Based on the conflict identification results, prioritize and merge related control requests to generate a coordinated ranking result.

[0056] Specifically, related control requests are extracted from the conflict identification results, and the control target information, execution time information, priority information, and conflict type information corresponding to the related control requests are read. Related control requests are sorted according to priority information; if priority information is the same, the order is determined by execution time information to form an initial sorting relationship. Based on conflict type information, related control requests with the same control target information, adjacent execution time information, or related resource usage relationships are collaboratively merged. Related control requests with consistent control directions and continuous execution are merged into a unified coordination processing unit, while related control requests with inconsistent control directions or conflicting execution periods are arranged in a time-sharing manner according to the sorting results. The related control requests after priority sorting and collaborative merging are organized in a unified order to generate a coordinated sorting result.

[0057] S4.1.4 Integrate the conflict identification results and coordination and sorting results to generate global control requirement data.

[0058] Specifically, the conflict identification results and coordination and sorting results are matched item by item according to the control target information and execution time information. The conflict type information, time overlap relationship, control direction relationship and resource occupation relationship in the conflict identification results are associated and matched with the sorting order information, collaborative merging relationship and time-sharing arrangement relationship in the coordination and sorting results. Based on the associated results, the conflict status, coordination order and execution arrangement corresponding to each control target are uniformly organized to form a control request correspondence relationship containing conflict mark, sorting position and coordination arrangement. The unified organized control request correspondence relationship is summarized according to the preset data format to generate global control demand data.

[0059] S4.2. The global control demand data is uniformly scheduled to obtain the control allocation results.

[0060] Specifically, the control objective information, control parameter information, time constraint information, and control conflict type information are extracted from the global control demand data to form the set of input variables required for scheduling optimization. Existing centralized optimization algorithms, such as linear programming, integer programming, or the Lagrange multiplier method, are selected to construct the scheduling optimization objective function. The objective function aims to minimize the number of control conflicts, balance the efficiency of control resource allocation, and satisfy the control objectives. Constraints are set, including the maximum execution time of the control objectives, the allowable range of control parameters, and the minimum time interval requirement between control objectives. The control objective information, control parameter information, time constraint information, and control conflict type information from the global control demand data are used as the set of input variables for scheduling optimization and substituted into the scheduling optimization objective function for calculation. The execution order of the control objectives is solved through iterative calculation or mathematical programming methods. Under the premise of satisfying all constraints, the optimal execution order of the control objectives, the control parameter configuration values ​​corresponding to each control objective, and the corresponding time allocation scheme are obtained. The control allocation result is obtained according to the execution order corresponding to the control objective information, control parameter information, and time constraint information in the global control demand data.

[0061] It should also be explained that the specific steps for constructing the scheduling optimization objective function are as follows: First, determine the optimization direction based on the control objective information in the global control demand data, such as minimizing control conflicts, maximizing resource utilization, or minimizing response time. Then, convert the control parameter information, time constraint information, and control conflict type information into constraints and penalty terms, respectively. Using the priority of each control objective as the weight, construct a weighted sum objective expression containing multiple weighted sub-objectives. Finally, organize the objective expression to form the scheduling optimization objective function for centralized optimization calculation.

[0062] S4.4. Using template mapping and instruction encapsulation methods, the control allocation results are structured to generate collaborative control instructions.

[0063] Specifically, based on predefined template mapping rules, the execution order of control targets, configuration values ​​of control parameters, and time allocation schemes are extracted, and fields are matched and arranged in a structured manner according to the template format. The structured data is then encapsulated into instructions, and the control target identifier, parameter values, timestamps, and priority information are uniformly encoded to generate standardized collaborative control instructions. The encapsulated collaborative control instructions are then format-validated to ensure field integrity and encoding correctness, generating a set of collaborative control instructions.

[0064] It should also be noted that the predefined template mapping rules clarify the correspondence between each field and the corresponding data item in the control allocation result by analyzing the format of historical control instructions and the characteristics of control targets, and formulate matching specifications for field names, data types and encoding methods; the rules include field order, data length, unit conversion and format constraints to ensure the consistency and accuracy of the mapping process; at the same time, different template versions are set according to the control target category and priority, and the template mapping rules are updated and optimized regularly based on actual operation feedback to maintain the timeliness and applicability of the rules.

[0065] S5. After the collaborative control command is issued, the edge intelligence method is used to coordinate the control points to execute the control actions, collect the operating status and control response data in real time, and use anomaly detection and confidence assessment to filter invalid data and generate a high-quality control response dataset.

[0066] S5.1. The edge intelligence method is used to trigger collaborative control commands and simultaneously collect field responses to generate original operating status and control response data.

[0067] Specifically, using edge intelligence, predefined collaborative control commands are invoked based on trigger identifiers to issue control commands to the field control devices. Simultaneously, the field data acquisition devices are activated to collect field operation status data and control response data in real time according to the set sampling frequency and sampling parameters. The collected operation status data and control response data include equipment operating parameters, environmental status parameters, and actuator feedback information. After collection, the original operation status data and control response data are timestamped and uniformly formatted to generate original operation status data and control response data.

[0068] It should also be noted that setting the sampling frequency and sampling parameters involves first determining the specific value of the sampling frequency, for example, 50 samples per second; then determining the type of sampling parameters based on the sampling object, including but not limited to specific parameters such as current, voltage, temperature, and pressure; next, configuring the acquisition device and performing data acquisition operations according to the determined sampling frequency and sampling parameters; finally, verifying the response time of the sampling device and data integrity to ensure the accurate execution of the sampling frequency and sampling parameters and meet the requirements for data real-time performance and integrity.

[0069] S5.2. Using anomaly detection methods, identify and remove abnormal data from the original operating status and control response data to generate preliminary valid data.

[0070] Specifically, using anomaly detection methods, based on pre-set maximum and minimum allowable thresholds for temperature, pressure, and current, each parameter in the original operating status and control response data is scanned one by one to determine whether the parameter value exceeds the corresponding threshold range. For example, the maximum allowable threshold for voltage is 240 volts, and the minimum allowable threshold is 220 volts. Statistical methods are used to identify data outside the mean ± 3 times the standard deviation as anomalies. Combining historical operating status data, a sliding window method is used to detect anomalies where the mutation amplitude exceeds the preset mutation amplitude threshold. Anomalies are marked and removed, and the remaining data are summarized to form preliminary valid data.

[0071] It should also be explained that the specific steps for setting the maximum and minimum allowable thresholds for temperature, pressure, and current parameters are as follows: Based on historical operating data and process specifications, the numerical distribution ranges of temperature, pressure, and current parameters during normal operation are statistically analyzed to determine the maximum and minimum allowable thresholds for temperature, pressure, and current parameters. For example, the maximum and minimum values ​​of temperature parameters can be extracted from historical data and combined with a safety margin to set the maximum and minimum allowable thresholds for temperature. These thresholds are then written into the threshold parameter set as fields and used as the standard for judging whether various parameters exceed the limits during subsequent anomaly identification. The specific steps for determining whether a parameter value exceeds the corresponding threshold range are as follows: For each data point in the original operating status and control response data, read the temperature parameter, pressure parameter, and current parameter one by one. Compare the temperature parameter value with the maximum and minimum allowable threshold values ​​for the temperature parameter to determine if it is greater than or less than the minimum allowable threshold. Then compare the pressure parameter value with the maximum and minimum allowable threshold values ​​for the pressure parameter to determine if it is greater than or less than the minimum allowable threshold. Next, compare the current parameter value with the maximum and minimum allowable threshold values ​​for the current parameter to determine if it is greater than or less than the minimum allowable threshold. If a parameter value exceeds the corresponding threshold range, mark the parameter as abnormal data; otherwise, mark it as normal data. Repeat this process for all parameters and output the abnormal data identification result. The specific steps for setting the threshold for sudden change amplitude are as follows: Based on the variation characteristics of each parameter in the historical operating status data and control response data, calculate the statistical distribution characteristics of the changes in temperature parameter, pressure parameter, and current parameter, and obtain the maximum fluctuation range of each type of parameter under normal operating conditions; on this basis, combined with the control stability requirements, set the threshold for sudden change amplitude of temperature parameter, pressure parameter, and current parameter respectively. For example, the threshold for sudden change amplitude of temperature parameter can be set to 5℃, the threshold for sudden change amplitude of pressure parameter can be set to 0.3MPa, and the threshold for sudden change amplitude of current parameter can be set to 10A; these thresholds will be uniformly recorded as the threshold for sudden change amplitude.

[0072] S5.3. Assess the credibility of the preliminary valid data and filter out low-confidence information to generate a high-quality control response dataset.

[0073] Specifically, each data point in the preliminary valid data is processed individually. Based on the sensor status at the time of data acquisition, the acquisition time interval, the data fluctuation amplitude, and the historical stability score, the confidence level of the temperature parameter is calculated. The expression is as follows: ; in, This indicates the confidence score for the temperature parameter. This indicates the sensor status value corresponding to the temperature parameter acquisition. This indicates the time interval between two consecutive temperature data collections. This indicates the range of change between the current temperature value and the previous temperature value. This represents the stability score of the temperature parameter in historical data. This represents the confidence score function; The confidence level of the pressure parameter is calculated using the following expression: ; in, This indicates the confidence score for the pressure parameter. This indicates the sensor status value corresponding to the pressure parameter acquisition. This indicates the time interval between two consecutive pressure data collections. This indicates the magnitude of change between the current pressure value and the previous pressure value; The confidence level of the current parameter is calculated using the following expression: ; in, This indicates the confidence score for the current parameter. This indicates the sensor status value corresponding to the current parameter acquisition. This indicates the time interval between two consecutive current data acquisitions. This indicates the range of change between the current value and the previous current value; a confidence score function is used to convert each indicator into a confidence value between 0 and 1. For example, when the sampling time interval is stable, the parameter changes smoothly and are consistent with historical values, the confidence score is close to 1; a confidence threshold is set, for example, 0.7, and data below the confidence threshold are judged as low-confidence data; all temperature, pressure, and current parameters with confidence scores below the confidence threshold are removed, and data with confidence scores higher than or equal to the confidence threshold are retained to generate a high-quality control response dataset.

[0074] It should also be noted that the specific steps for setting the confidence threshold are as follows: Select data samples marked as high stability from historical operating status data and control response data, and calculate the statistical distributions of confidence scores for temperature, pressure, and current parameters respectively; then perform percentile analysis on the confidence distribution of each parameter type to obtain the confidence value at the P percentile; for example, select the confidence value at the 70th percentile as the confidence threshold. If the 70th percentile value in the confidence distribution of temperature parameters is 0.72, then the confidence threshold for temperature parameters is set to 0.72; similarly, set the confidence thresholds for pressure and current parameters; finally, form a set of confidence thresholds consisting of the confidence thresholds for temperature, pressure, and current parameters, which will be used for subsequent confidence judgment and low-confidence data removal operations.

[0075] S6. Perform operational deviation and trend analysis on the high-quality control response dataset, generate state diagnosis results, and combine trigger conditions for judgment and optimization to generate updated collaborative control instructions.

[0076] S6.1. A multi-scale diagnostic mechanism is adopted to decompose the high-quality control response dataset into multiple scales and extract state change features at different time scales.

[0077] Specifically, the temperature, pressure, and current parameters in the high-quality control response dataset are extracted as input sequences. Then, wavelet transform is used to decompose each input sequence at multiple scales. For example, with three scale levels, low-frequency and high-frequency coefficients are obtained for scale one, scale two, and scale three, respectively. The high-frequency coefficients extracted from different scale levels are used to characterize rapid change trends, and the low-frequency coefficients are used to characterize slow change trends, recorded sequentially as scale one state change features, scale two state change features, and scale three state change features. Finally, the state change features of temperature, pressure, and current parameters at multiple time scales are extracted, forming state change features at different scales.

[0078] S6.2 Utilize the characteristics of state changes to perform operational deviation analysis and trend identification on the operating state, and generate state diagnosis results.

[0079] Specifically, the state change characteristics of temperature, pressure, and current parameters are extracted from the state change characteristics at scales one, two, and three, respectively. The short-term and long-term mean values ​​of each parameter's state change characteristic are calculated using a moving average method, with window lengths set to 5 and 20, respectively. The deviation between the short-term and long-term mean values ​​is then compared to identify operational deviations in temperature, pressure, and current parameters. Based on time series slope analysis, the growth rate and decline rate of the state change characteristics over a continuous time period are calculated to identify trends in temperature, pressure, and current parameters. Finally, the operational deviation results are combined with the trend identification results to generate a state diagnosis result.

[0080] S6.3. Based on the status diagnosis results and preset trigger conditions, determine whether the status meets the update requirements and output the trigger judgment result.

[0081] S6.3.1 Extract the operational deviation data and trend recognition data from the status diagnosis results to generate diagnostic input data.

[0082] Specifically, the operational deviation data and trend identification data corresponding to each control objective are read item by item from the status diagnosis results. The operational deviation data and trend identification data are matched according to the control objective information and time sequence to form a deviation-trend correspondence for the same control objective in the same analysis period. The operational deviation data and trend identification data after matching are organized according to a unified field order. The deviation value, deviation direction, and deviation position in the operational deviation data are combined and recorded with the trend type, trend change direction, and trend duration in the trend identification data to form a set of diagnostic discrimination fields. The set of diagnostic discrimination fields is summarized according to the data format to generate diagnostic discrimination input data.

[0083] S6.3.2 Match the diagnostic input data with the deviation threshold condition and trend persistence condition in the preset trigger conditions to generate condition matching results.

[0084] Specifically, the system reads operational deviation data and trend identification data from the diagnostic input data according to the control target information and time sequence, and simultaneously reads the deviation threshold conditions and trend persistence conditions from the preset trigger conditions. The deviation values, directions, and locations in the operational deviation data are compared item by item with the corresponding deviation threshold conditions to determine whether the conditions are met. The trend type, direction of trend change, and persistence status in the trend identification data are checked item by item with the corresponding trend persistence conditions to determine whether the conditions are met. The satisfaction of the deviation threshold conditions and the trend persistence conditions are summarized according to the same control target and the same analysis period to form a condition correspondence. The condition correspondence is then organized according to a preset record format to generate condition matching results.

[0085] S6.3.3 Determine whether the current running state meets the update requirement based on the condition matching result, and generate the trigger judgment result.

[0086] Specifically, the system reads the deviation threshold condition satisfaction and trend persistence condition satisfaction status corresponding to each control objective from the condition matching results. It then verifies the correspondence between these conditions according to the control objective information and the analysis period. Based on preset judgment rules, it jointly judges the deviation threshold condition satisfaction and trend persistence condition satisfaction status for the same control objective within the same analysis period. When both conditions are met, the current operating state is determined to meet the update requirement; otherwise, it is determined to not meet the update requirement. Finally, the judgment statuses corresponding to each control objective are organized according to a unified recording format to generate trigger judgment results.

[0087] S6.3.4 When the trigger judgment result indicates that the update requirement is met, output the trigger judgment result.

[0088] Specifically, the system reads the judgment status, analysis period information, and condition fulfillment markers corresponding to each control target in the trigger judgment result. It then checks each judgment status against the control target information and analysis period information, filtering out records that represent the status of meeting update requirements from the trigger judgment result. These records are then organized according to the control target information, analysis period information, and condition fulfillment markers to clarify the scope of control targets and time periods for subsequent update processes. Finally, the filtered and organized records are summarized according to a preset output order, and the trigger judgment result is output.

[0089] S6.4 When the trigger judgment result meets the update requirements, the trigger judgment result is combined with the status diagnosis result to adjust and optimize the original operating status and generate the updated collaborative control command.

[0090] Specifically, when the trigger judgment result meets the update requirements, the operating deviations of temperature, pressure, and current parameters in the trigger judgment result, as well as the trends of temperature, pressure, and current parameters in the state diagnosis result, are extracted as control inputs. These control inputs are then substituted into the objective function of the adaptive optimization algorithm to optimize the initial values ​​of temperature, pressure, and current parameters in the original operating state. The partial derivatives of the objective function with respect to each initial value are calculated using gradient descent or particle swarm optimization, and the target values ​​for temperature, pressure, and current parameters are updated in each iteration until convergence to a local optimum or the termination condition is met. Based on the optimized target values ​​for temperature, pressure, and current parameters, combined with the control target information and control parameter information, updated collaborative control instructions are generated through template mapping rules.

[0091] This embodiment also provides an IoT-based distributed control system for glass processing production, including: a process acquisition module, a data fusion module, a rule judgment module, a cloud scheduling module, an edge execution module, and a diagnostic optimization module; the process acquisition module is used to acquire process node data and preprocess it according to unified identification rules, time synchronization rules, and data cleaning rules to generate structured raw process data; the data fusion module is used to perform cross-node correlation fusion, operating condition feature extraction, and threshold parsing on the structured raw process data, extracting state parameters and control thresholds to generate multi-source fused control data; the rule judgment module is used to perform condition matching on the multi-source fused control data using rule control methods. The system comprises four modules: a logical judgment and decision processing module to generate initial control commands; a cloud scheduling module to upload these initial control commands as input to a cloud coordinator for unified scheduling and conflict coordination, generating collaborative control commands; an edge execution module to coordinate control points to perform control actions using edge intelligence methods after the collaborative control commands are issued, collect real-time operating status and control response data, and filter invalid data using anomaly detection and confidence assessment to generate a high-quality control response dataset; and a diagnostic optimization module to analyze operating deviations and trends in the high-quality control response dataset, generate status diagnostic results, and perform judgment and optimization based on trigger conditions to generate updated collaborative control commands.

[0092] In summary, this invention preprocesses and fuses process node data to form multi-source fused control data containing state parameters and control thresholds. Based on this, it combines rule-based control, unified cloud scheduling, edge execution feedback, and multi-scale diagnostic updates to construct a distributed closed-loop control link for the glass processing and production process. This achieves continuous connection between control decisions, collaborative execution, and state feedback, thereby improving the adaptability of control commands to actual operating conditions, enhancing the timeliness and stability of multi-node collaborative control, and ensuring the continuous operation of the glass processing and production process.

[0093] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A distributed control method for glass processing production based on the Internet of Things, characterized in that: include, Collect process node data and preprocess it according to unified identification rules, time synchronization rules and data cleaning rules to generate structured raw process data; The structured raw process data is cross-node correlation and fusion, operating condition feature extraction and threshold parsing are performed to extract state parameters and control thresholds and generate multi-source fused control data. A rule-based control method is used to perform condition matching, logical judgment, and decision processing on multi-source fused control data to generate preliminary control instructions. The initial control commands are uploaded to the cloud coordinator as cloud coordination input for unified scheduling and conflict coordination, generating collaborative control commands. After the collaborative control command is issued, the edge intelligence method is used to coordinate the control points to execute the control actions, collect the operating status and control response data in real time, and use anomaly detection and confidence assessment to filter invalid data and generate a high-quality control response dataset. Perform operational deviation and trend analysis on high-quality control response datasets, generate state diagnostic results, and combine triggering conditions for judgment and optimization to generate updated collaborative control instructions.

2. The distributed control method for glass processing and production based on the Internet of Things as described in claim 1, characterized in that: The system collects the operating status parameters, unique identifiers, and control boundary parameters of process nodes to generate raw process data. It then uses outlier detection and median filtering methods to remove outliers and filter out noise, generating structured raw process data.

3. The distributed control method for glass processing and production based on the Internet of Things as described in claim 1, characterized in that: The specific steps for generating multi-source fused control data are as follows: The original structured process data is dimensionality reduced, and the principal component vectors of glass processing condition changes are extracted to generate process feature vectors. The process feature vectors are processed for time synchronization and format standardization to generate unified time-series fusion data. Based on unified time-series fusion data, state identification and threshold calculation analysis are performed to extract state parameters and control thresholds, and a set of state and control indicators is generated. By mapping and integrating the set of status and control indicators with the unique identifier information in the structured original process data, multi-source fusion control data is generated.

4. The distributed control method for glass processing and production based on the Internet of Things as described in claim 1, characterized in that: The specific steps for generating the initial control commands are as follows: Extract control-related fields and unique identification information from multi-source fused control data, and organize them to generate an analysis dataset; Based on a pre-defined rule base, condition matching and logical reasoning are performed on the analysis dataset to generate trigger identifiers; Based on the trigger identifier, a control strategy is selected from the preset rule base, and control parameters are calculated to generate a set of control parameters; The set of control parameters and control thresholds are processed for control decision-making to generate preliminary control commands.

5. The distributed control method for glass processing and production based on the Internet of Things as described in claim 1, characterized in that: The specific steps for generating the coordinated control instructions are as follows: Based on the initial control instructions, analyze resource conflicts and coordination needs, and generate global control requirements data; The global control and regulation demand data is uniformly scheduled to obtain the control allocation results; A template mapping and instruction encapsulation method is used to perform structured processing on the control allocation results and generate collaborative control instructions.

6. The distributed control method for glass processing and production based on the Internet of Things as described in claim 5, characterized in that: The specific steps for generating global control demand data are as follows: Extract control target information, control parameter information, execution time information, and priority information from the preliminary control instructions to generate a set of control requests to be coordinated; Based on the time overlap, regulation direction, and resource consumption relationships among the control objectives in the set of control requests to be coordinated, control conflict relationships are identified, and conflict identification results are generated. Based on the conflict identification results, related control requests are prioritized and collaboratively merged to generate a coordinated ranking result. The conflict identification results and coordination and ranking results are integrated to generate global control requirements data.

7. The distributed control method for glass processing production based on the Internet of Things as described in claim 1, characterized in that: The specific steps for generating a high-quality control response dataset are as follows. Edge intelligence methods are used to trigger collaborative control commands and simultaneously collect field responses to generate raw operating status and control response data; Anomaly detection methods are used to identify and remove abnormal data from the original operating status and control response data, generating preliminary valid data. The credibility of the preliminary valid data is assessed, and low-confidence information is filtered out to generate a high-quality control response dataset.

8. The distributed control method for glass processing and production based on the Internet of Things as described in claim 1, characterized in that: The specific steps for generating the updated cooperative control instructions are as follows: A multi-scale diagnostic mechanism is adopted to decompose a high-quality control response dataset into multiple scales and extract state change features at different time scales. By utilizing the characteristics of state changes, operational deviation analysis and trend identification are performed on the operating state to generate state diagnosis results; Based on the status diagnosis results and preset trigger conditions, determine whether the status meets the update requirement and output the trigger determination result; When the trigger determination result meets the update requirements, the trigger determination result is combined with the status diagnosis result to adjust and optimize the original operating state and generate updated collaborative control instructions.

9. The distributed control method for glass processing production based on the Internet of Things as described in claim 8, characterized in that: The specific steps for determining the output trigger result are as follows. Extract operational deviation data and trend recognition data from the status diagnosis results to generate diagnostic input data; The diagnostic input data is matched with the deviation threshold condition and trend persistence condition in the preset trigger conditions to generate condition matching results; Based on the condition matching results, determine whether the current running status meets the update requirement and generate a trigger judgment result; When the trigger determination result indicates that the update requirement is met, the trigger determination result is output.

10. A distributed control system for glass processing production based on the Internet of Things (IoT), based on the distributed control method for glass processing production based on the IoT as described in any one of claims 1 to 7, characterized in that: It includes a process acquisition module, a data fusion module, a rule judgment module, a cloud scheduling module, an edge execution module, and a diagnostic optimization module; The process acquisition module is used to collect process node data and preprocess it according to unified identification rules, time synchronization rules and data cleaning rules to generate structured raw process data. The data fusion module is used to perform cross-node correlation fusion, operating condition feature extraction, and threshold parsing on structured raw process data, extract state parameters and control thresholds, and generate multi-source fused control data. The rule judgment module is used to perform condition matching, logical judgment and decision processing on multi-source fused control data using rule control methods, and generate preliminary control instructions. The cloud scheduling module is used to upload preliminary control commands as cloud coordination inputs to the cloud coordinator for unified scheduling and conflict coordination, and generate collaborative control commands. The edge execution module is used to coordinate the control points to perform control actions after the collaborative control commands are issued, and to collect the operating status and control response data in real time. It also uses anomaly detection and confidence assessment to filter invalid data and generate a high-quality control response dataset. The diagnostic optimization module is used to perform operational deviation and trend analysis on high-quality control response datasets, generate state diagnostic results, and combine trigger conditions for judgment and optimization to generate updated collaborative control instructions.

Images