Distributed control logic dynamic deployment method for multi-variety mixed-line production

Abstract

The application provides a kind of distributed control logic dynamic deployment method for multi-variety mixed line production, belongs to mixed line production deployment technical field.Method includes: collecting production conditions, distributed node resources, process constraint data, real-time acquisition mixed line timing operation data, analysis obtains production efficiency index;Combined with node resources and process constraints, initial deployment parameters are formulated, and real-time deployment scheme is generated;Judge whether the efficiency meets the standard, if not, optimize dynamic parameters;Obtain correlation data by condition-performance correlation analysis, extract abnormal features to correct dynamic parameters, and generate deployment correction parameters.The application realizes multi-dimensional data synchronous acquisition and hierarchical analysis, automatically adjusts control configuration, eliminates manual adjustment lag and experience dependence;Through composite condition recognition core disturbance, parameter adjustment is ordered according to priority, process coordination is stabilized, operation and maintenance cost is reduced, and flexible production line adaptive operation capability is improved.

Description

Technical Field

[0001] This invention belongs to the field of mixed-line production deployment technology, specifically a method for dynamic deployment of distributed control logic for multi-product mixed-line production. Background Technology

[0002] With the popularization of flexible manufacturing, production lines can process multiple products of different specifications at the same time. Various control units on the production line are distributed and rely on a distributed architecture to achieve full-line collaborative management.

[0003] Existing production line control configurations are mostly fixed in the early stages of equipment commissioning, only adaptable to a single stable production condition. Once the on-site production status changes, the entire set of control parameters cannot be adjusted autonomously and synchronously. During on-site production, various external interferences continuously occur, including process changes due to frequent product model switching, load fluctuations caused by long-term equipment operation, and continuous fluctuations in workshop temperature, humidity, and electromagnetic environment. The superposition of multiple interferences can change the line transmission status and hardware operating capabilities. When complex interferences occur on-site, existing adjustment methods can only correct one type of control parameter at a time, failing to take into account multi-dimensional constraints and making it difficult to meet the stable, efficient, and manual intervention-free control requirements of multi-product simultaneous processing scenarios. Summary of the Invention

[0004] This invention provides a method for dynamically deploying distributed control logic for multi-product mixed-line production, in order to overcome the deficiencies in the existing technology.

[0005] This invention provides a method for dynamically deploying distributed control logic for multi-product mixed-line production, including:

[0006] Production condition data, distributed node resource data, and process constraint data are collected, and real-time time-series operation data of mixed-line production is acquired. A multi-dimensional time-series efficiency extraction method is used to analyze the time-series operation data to obtain production efficiency indicators.

[0007] Initial deployment parameters are formulated based on distributed node resource data and process constraint data combined with mixed-line production task requirements. Dynamic deployment parameters are calculated based on production disturbance type and production line efficiency depth index. The initial deployment parameters are then adjusted based on the dynamic deployment parameters to obtain a real-time operation deployment plan.

[0008] Determine whether the current production efficiency indicators have reached the preset efficiency threshold. If so, maintain the real-time operation deployment plan; otherwise, optimize the dynamic deployment parameters.

[0009] The working condition-efficiency correlation analysis method is used to analyze production working condition data, distributed node resource data and production efficiency indicators to obtain correlation impact data. The current production efficiency indicators are feature extracted to obtain abnormal operation characteristics. Combined with the correlation impact data, the dynamic deployment parameters are accurately corrected to obtain the deployment correction parameters.

[0010] According to the distributed control logic dynamic deployment method for multi-product mixed-line production provided by this invention, the production condition data includes order switching data, equipment load data, and environmental disturbance data. Distributed node resource data includes computing power margin data, communication bandwidth data, storage capacity data, and interface compatibility data. Process constraint data includes process path dependency constraints, workstation cycle time matching constraints, quality control threshold constraints, and safety interlocking rule constraints.

[0011] The process of analyzing and obtaining production efficiency indicators according to the distributed control logic dynamic deployment method for multi-product mixed-line production provided by the present invention includes: Collect basic operating signals, production line efficiency related data, and deployment control parameters, and preprocess them as time-series operating data.

[0012] Production cycle time, equipment utilization rate, and product qualification rate are extracted from basic operation signals as basic production performance indicators. Workstation-level cycle time matching degree, work-in-process flow delay, and process synchronization stability are extracted from production line efficiency correlation data as production line efficiency depth indicators.

[0013] Logical redeployment time, resource scheduling response, and node adaptability metrics are extracted from deployment control parameters as deployment adaptability metrics.

[0014] The trend characteristics and disturbance correlation indicators of basic production performance indicators, production line efficiency depth indicators and deployment adaptation indicators are extracted as production efficiency indicators.

[0015] The method for dynamically deploying distributed control logic for multi-product mixed-line production provided by the present invention includes the following process for determining initial deployment parameters: The task requirements of different mixed-line production scenarios are transformed into technical indicators. Performance constraint boundaries are formulated based on distributed node resource data, and process constraint data is combined to determine process dependencies, cycle time matching requirements and safety interlock rules.

[0016] The optimal set of deployment nodes is selected by combining the characteristics of node computing power distribution and the granularity of control logic slices, and the weight of logical slice allocation is determined based on the upper limit of computing power of a single node.

[0017] The baseline deployment granularity is determined based on cross-node transmission latency and node computing power range, and the control logic splitting method is selected according to technical indicators and real-time requirements.

[0018] The scheduling strategy is configured based on the node's parallel processing capability and the timing matching requirements. The deployment switching trigger condition is set based on the node communication latency threshold, and the initial sensitivity coefficient of each deployment parameter is determined.

[0019] The method for dynamically deploying distributed control logic for multi-product mixed-line production provided by the present invention includes the following process for calculating dynamic deployment parameters: Production condition data is categorized into disturbance types based on order switching disturbances, equipment load disturbances, and environmental disturbances, and then quantified into dimensionless disturbance level values ​​according to the degree of impact to obtain disturbance data.

[0020] Determine the preset threshold for the production line efficiency depth index, calculate the index deviation value, classify the production line efficiency level according to the degree of deviation, and perform time synchronization and correlation mapping with the disturbance data.

[0021] Establish a disturbance type-deployment parameter mapping table to identify the dominant type in composite disturbances and determine the strategy strength based on the disturbance level.

[0022] The priority of parameter adjustment is determined based on the strength of the strategy, the adjustment range is determined based on the production line efficiency deviation, and the single-step adjustment amount is limited. Specific deployment parameters are added for the types of production disturbances to obtain dynamic deployment parameters.

[0023] The distributed control logic dynamic deployment method for multi-product mixed-line production provided by the present invention includes the following process for adjusting and obtaining a real-time operational deployment scheme: Based on distributed node resource data and historical deployment records, the impact of each deployment parameter on production efficiency is updated to form an iterative sensitivity coefficient.

[0024] Prioritize adjustments based on the directness of their impact on production efficiency and the speed of response.

[0025] A smooth transition mechanism is established for setting transition times and effective conditions for continuous deployment parameters.

[0026] The current dynamic deployment parameters are superimposed with the initial deployment parameters according to the adjustment priority and the current sensitivity coefficient. The rate of change of parameters before and after superposition and the reasons for the change are recorded to obtain the real-time deployment plan.

[0027] The distributed control logic dynamic deployment method for multi-product mixed-line production provided by the present invention includes the following process for obtaining correlated impact data: Multi-mode disturbance features are extracted from production condition data, node status features are extracted from distributed node resource data, and efficiency change features are extracted from production efficiency indicators.

[0028] Correlation analysis was performed on the characteristics of multi-mode disturbances, node state characteristics, and efficiency changes to identify lagging production disturbances.

[0029] Based on the combination of disturbance types, scenarios are divided. In different scenarios, regression models of node state characteristics, efficiency change characteristics and multi-mode disturbance characteristics are constructed, the goodness of fit is calculated, and the core disturbance factors are screened.

[0030] Analyze the correlation between order switching and logical redeployment, the correlation between load fluctuation and computing power allocation, and the correlation between environmental disturbances and node stability to generate statistical analysis reports.

[0031] The statistical analysis report is transformed into execution rules. The confidence and support of all execution rules are calculated, and reliable execution rules that meet the preset thresholds are retained to generate a rule base for correlation effects.

[0032] The distributed control logic dynamic deployment method for multi-product mixed-line production provided by the present invention includes the following process for generating a statistical analysis report: Draw a scatter plot of order switching complexity and control logic deployment latency, identify switching nodes and deployment conflict areas, and calculate the mapping relationship between switching complexity and cycle time loss to obtain cycle time correlation.

[0033] The relationship between equipment load rate and node computing power utilization is analyzed to generate a fitted relationship curve. Resource correlation is obtained by combining the correlation coefficient between load mutation rate and task scheduling delay.

[0034] The frequency of node failures under a preset temperature threshold is statistically analyzed, and the communication packet loss rate under a preset humidity threshold is combined to obtain the working condition correlation.

[0035] The statistical analysis report is obtained by standardizing the expression of cycle time correlation, resource correlation, and working condition correlation in the format of influencing factors-affected parameters-quantified relationship-scenario adaptation suggestions.

[0036] The distributed control logic dynamic deployment method for multi-product mixed-line production provided by the present invention includes the following process for obtaining operational anomaly characteristics: For the preset production efficiency threshold, determine the abnormal threshold range of each efficiency indicator, and for time-series fluctuations, set dynamic thresholds to adapt to the time-varying characteristics of the production line.

[0037] Iterate through all time windows, and if any performance indicator exceeds the abnormal threshold range or dynamic threshold, it is determined to be an abnormal segment.

[0038] Extract corresponding abnormal segments from the time-series abnormality dimension, resource abnormality dimension, and deployment adaptation abnormality dimension respectively.

[0039] By combining associated production status data and distributed node resource data, the anomaly type of abnormal segments is marked.

[0040] Calculate the Pearson correlation coefficient between the performance change index and the corresponding disturbance factor within each abnormal segment, retain the abnormal segments that reach the preset coefficient threshold, and combine them with the corresponding abnormal type as operational abnormality characteristics.

[0041] The distributed control logic dynamic deployment method for multi-product mixed-line production provided by the present invention includes the following process for adjusting and obtaining deployment correction parameters: Based on the associated impact data, the abnormal causes of abnormal features are located through feature matching, and an adjustment mapping table is established by combining the priority adjustment parameters and strategy direction.

[0042] Identify parameter adjustment conflicts, coordinate them through preset constraint priorities, and plan the parameter adjustment sequence from fastest to slowest response speed.

[0043] By combining the correlation impact data to calculate the power allocation correction range, the logical slice node allocation adjustment amount, and the cross-node synchronization interface parameter correction value as the adjustment parameter correction range, the dynamic deployment parameters are adjusted to obtain the deployment correction parameters.

[0044] This invention provides a method for dynamically deploying distributed control logic for multi-product mixed-line production. Through synchronous collection and hierarchical analysis of multi-dimensional operational data, it can capture various disturbances, distributed hardware resources and production operation status in real time. It can autonomously complete the dynamic adjustment of control configuration without continuous manual intervention, eliminating the lag and experience dependence caused by manual adjustment.

[0045] By combining hardware limits and rigid constraints of production processes to generate basic configurations, and by generating adaptive adjustment quantities layer by layer based on real-time disturbances on site, a parameter smooth transition mechanism is set up to avoid production line operation oscillations caused by sudden configuration changes.

[0046] By exploring the inherent correlation between various disturbances and equipment operating status, and constructing stable correlation judgment rules, when abnormalities such as production line imbalance or resource overload occur, abnormal features can be extracted autonomously and the root cause can be located, eliminating the need for manual trial and error to troubleshoot.

[0047] Faced with complex operating conditions with multiple interferences, it can automatically identify the core disturbance factors affecting production, and adjust various control configurations in an orderly manner according to response speed and constraint priority. This ensures stable process coordination during multi-variety mixed production, reduces on-site debugging and daily operation and maintenance investment, and improves the long-term stability and autonomous adaptation capability of flexible production lines. Attached Figure Description

[0048] The invention will now be further described with reference to the accompanying drawings.

[0049] Figure 1 This is a flowchart illustrating the dynamic deployment method of distributed control logic for multi-product mixed-line production in this invention. Figure 2 This is a flowchart illustrating the process of setting initial deployment parameters in this invention; Figure 3 This is the method used in this invention to calculate dynamic deployment parameters. Detailed Implementation

[0050] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0051] like Figures 1 to 3 As shown in the figure, the distributed control logic dynamic deployment method for multi-product mixed-line production provided by the embodiments of the present invention includes: Collect production condition data, distributed node resource data, and process constraint data. Production condition data includes order changeover data, equipment load data, and environmental disturbance data.

[0052] Distributed node resource data includes computing power margin data, communication bandwidth data, storage capacity data, and interface compatibility data. Process constraint data includes process path dependency constraints, workstation cycle time matching constraints, quality control threshold constraints, and safety interlocking rule constraints.

[0053] Production status data is collected uniformly through the production line-level industrial IoT platform. Order switching data comes from the manufacturing execution system, including order number, product model, process route, start and end time of changeover, number of processes involved in changeover, sampling trigger time is the moment the order switching instruction is issued, and material code and tooling change information are recorded.

[0054] Equipment load data is read in real time through the programmable logic controller registers of each workstation. The collected parameters include spindle load rate, motor operating current, equipment operating status, and fault alarm signals.

[0055] Environmental disturbance data is collected by temperature and humidity sensors and dust sensors deployed at key nodes of the production line, recording the values ​​of ambient temperature, relative humidity, and dust concentration, and correlated with the operating status of the workshop's air conditioning and ventilation systems.

[0056] The distributed nodes specifically include a workstation-level embedded programmable logic controller, a production line-level edge computing gateway, a servo drive control unit, and an industrial input / output acquisition module. All nodes are clock-synchronized through an industrial Ethernet bus. Computing power margin data is read through the node's built-in performance monitoring interface to record the real-time communication rate and bandwidth utilization between nodes.

[0057] Process constraint data is parsed and extracted from product process documents. Process path dependency constraints clearly define the execution order and preconditions of each process, prohibiting skipping steps. Station cycle time matching constraints stipulate that the cycle time deviation between upstream and downstream stations shall not exceed 10% of the baseline cycle time to avoid work-in-process accumulation or material shortage. Quality control threshold constraints set upper and lower limits for key quality parameters of each process. When the threshold is exceeded, quality traceability and shutdown logic are triggered. Safety interlock rule constraints define the trigger response logic of equipment safety doors, emergency stop buttons, and light curtain protection devices, which have higher priority than all production control logic.

[0058] Real-time acquisition of time-series operational data from mixed-line production, followed by analysis of the data using a multi-dimensional time-series efficiency extraction method to obtain production efficiency indicators. The process includes: Collect basic operating signals, production line efficiency related data, and deployment control parameters, and preprocess them as time-series operating data.

[0059] Basic operating signals are collected by photoelectric sensors, barcode scanning equipment, and programmable logic controllers at the end of the production line, including product output time, equipment operating status, and quality inspection results.

[0060] Production line efficiency data is collected through RFID readers and work-in-process inspection devices at each workstation, including work-in-process location, process completion time, workstation waiting time, and recording the operating status of the transmission mechanism between processes.

[0061] Deployment control parameters are read through the configuration interface of the deployment scheduling unit, including the current logical slice allocation scheme, scheduling cycle, node running parameters, and synchronization interface configuration.

[0062] The data preprocessing steps include outlier removal, time series alignment, and sliding smoothing. Outlier removal uses the 3σ criterion, identifying collected values ​​that exceed three times the standard deviation as outliers and replacing them with linear interpolation of adjacent data. Time series alignment involves uniformly interpolating and resampling data from different sampling frequencies to a 1Hz time series. Sliding smoothing uses a sliding window of length 5 to take the mean value, suppressing the interference of acquisition noise on subsequent analysis.

[0063] Production cycle time, equipment utilization rate, and product qualification rate are extracted from basic operation signals as basic production performance indicators. Workstation-level cycle time matching degree, work-in-process flow delay, and process synchronization stability are extracted from production line efficiency correlation data as production line efficiency depth indicators.

[0064] The production cycle time is the arithmetic mean of the time intervals between the production of ten consecutive qualified products. The smaller the value, the higher the production efficiency.

[0065] Equipment utilization rate is the ratio of actual equipment operating time to planned production time within the statistical window, expressed as a percentage, excluding planned downtime and equipment failure downtime.

[0066] The product qualification rate is the ratio of the number of qualified products to the total number of products produced within the statistical window, expressed as a percentage. Unqualified products include reworked products and scrapped products.

[0067] The workstation-level cycle time matching degree is calculated by the ratio of the actual cycle time difference between upstream and downstream adjacent workstations to the benchmark cycle time. The higher the matching degree, the better the capacity balance between workstations. The value range is from 0 to 1.

[0068] Work-in-process turnover delay is the difference between the total time a single product takes from entering the first process to leaving the last process and the actual processing time of each process, reflecting the waiting and transmission time between processes.

[0069] The process synchronization stability is measured by the standard deviation of the completion time of each process within the same production batch. The smaller the standard deviation, the better the synchronization of multiple processes and the more stable the production rhythm.

[0070] Logical redeployment time, resource scheduling response, and node adaptability metrics are extracted from deployment control parameters as deployment adaptability metrics.

[0071] The time taken for logical redeployment is the total time it takes for the deployment command to be issued to all nodes to complete the logical update and enter a stable running state.

[0072] Resource scheduling response metrics are the time interval between the generation of scheduling instructions and the start of node execution of scheduling actions, reflecting the response speed.

[0073] Node adaptability is the degree of matching between node computing power, bandwidth, and storage resources under the current deployment scheme. The value ranges from 0 to 1, and the higher the value, the more balanced the resource utilization.

[0074] The trend characteristics and disturbance correlation indicators of basic production performance indicators, production line efficiency depth indicators and deployment adaptation indicators are extracted as production efficiency indicators.

[0075] Trend feature extraction uses a sliding window with a length of 300 seconds and a step size of 60 seconds. The mean, variance and rate of change of each indicator within each window are calculated. The rate of change is the ratio of the difference between the values ​​at the beginning and end of the window to the window duration, which reflects the trend of the indicator.

[0076] The perturbation correlation index calculates the Pearson correlation coefficient between each performance index and various perturbation data. The closer the absolute value of the correlation coefficient is to 1, the more significant the impact of the perturbation on performance.

[0077] Among them, the comprehensive production efficiency score The calculation formula is:

[0078] In the formula, , , The weight coefficients are dimensionless and satisfy the following conditions: ; This refers to the actual production cycle time; Based on the production cycle time; This represents the total number of distributed nodes. For the first Resource utilization of distributed nodes; For the first The rated computing power of each distributed node; Deployment latency for end-to-end control logic; This represents the maximum allowable deployment latency; the weighting coefficient under standard operating conditions has a value of [value missing]. , , It can be adjusted according to the needs of different production scenarios.

[0079] Initial deployment parameters are determined based on distributed node resource data, process constraint data, and mixed-line production task requirements. The process includes: The task requirements of different mixed-line production scenarios are transformed into technical indicators. Performance constraint boundaries are formulated based on distributed node resource data, and process constraint data is combined to determine process dependencies, cycle time matching requirements and safety interlock rules.

[0080] Mixed-line production scenarios are mainly divided into three categories: multi-variety small-batch scenarios, large-batch mixed-flow scenarios, and customized flexible production scenarios. In multi-variety small-batch scenarios, the output of a single batch is less than 50 pieces, the changeover frequency is high, and the core requirements are rapid changeover and deployment response. In large-batch mixed-flow scenarios, the output of a single batch is greater than 500 pieces, the changeover frequency is low, and the core requirements are high production efficiency and stable operation. In customized flexible production scenarios, the product processes vary greatly, and the core requirements are flexible logical adaptation and strong robustness.

[0081] The technical indicators are broken down into three dimensions: reliability, real-time performance, and resource utilization. The reliability indicator requires a communication packet loss rate of less than 0.1% and no control logic conflicts. The real-time performance indicator requires an end-to-end control response latency of less than 10 milliseconds. The resource utilization indicator requires the average computing power utilization of the node to be in the optimal range of 60% to 80%.

[0082] Performance constraints are determined based on node hardware parameters. The upper limit of computing power per node is set according to different node types: 500 MIPS for embedded programmable logic controllers, 8 TOPS for edge computing gateways, and 200 MIPS for servo drive control units. The upper limit of communication bandwidth is 100 Mbps, and the upper limit of storage capacity is 8 GB. All deployment schemes must not exceed the hardware constraints.

[0083] Process dependencies are represented by a directed acyclic graph, where each node represents a process and directed edges represent the order of processes. During deployment, it is essential to ensure that the control logic of the preceding process is deployed on the upstream node first, and the logic of the subsequent process is deployed on the downstream node to ensure the shortest signal transmission path.

[0084] Cycle time matching requires that the control logic operation cycle of each workstation must match the production cycle of that workstation, and the operation cycle must not be less than one-tenth of the cycle time to avoid control lag. Safety interlock rules are the highest priority constraints.

[0085] The optimal set of deployment nodes is selected by combining the characteristics of node computing power distribution and the granularity of control logic slices, and the weight of logical slice allocation is determined based on the upper limit of computing power of a single node.

[0086] The control logic is divided into five categories according to functional granularity: data acquisition slices, logic operation slices, motion control slices, safety interlock slices, and data upload slices. Each type of slice has a pre-defined baseline computing power requirement: 20 MIPS for data acquisition slices, 100 MIPS for logic operation slices, 150 MIPS for motion control slices, 80 MIPS for safety interlock slices, and 30 MIPS for data upload slices.

[0087] The control logic is divided based on the principle of functional modularization. Each slice is an independent executable logic unit with standardized input and output interfaces. The interface data includes three categories: process variables, control instructions, and status flags. Slices interact with each other through shared memory or bus messages.

[0088] The optimal deployment node set is selected using a greedy algorithm. Control logic slices are sorted from highest to lowest real-time priority: safety interlocking slices, motion control slices, logic operation slices, data acquisition slices, and data upload slices. During allocation, all available nodes are traversed sequentially according to priority. The total transmission latency after deploying the current slice to a candidate node is calculated. The total transmission latency is the sum of the cross-node communication latency between this slice and its upstream and downstream deployed slices. The cross-node communication latency is the sum of the historical average communication latency between nodes and the current network load correction value. The network load correction value is the product of the current bandwidth utilization rate and the baseline bandwidth latency coefficient. If the remaining computing power and bandwidth of the candidate node meet the slice requirements, the slice is allocated to the node with the minimum total transmission latency. If all nodes do not meet the constraints, the computing power quota of non-real-time slices is reduced. If the requirements are still not met, a backup node is activated to participate in the deployment.

[0089] The allocation weight is directly proportional to the node's computing power margin; the larger the computing power margin, the more slices are allocated to the node.

[0090] The baseline deployment granularity is determined based on cross-node transmission latency and node computing power range, and the control logic splitting method is selected according to technical indicators and real-time requirements.

[0091] The baseline deployment granularity is determined based on the cross-node transmission latency. When the average transmission latency between adjacent nodes is less than 5 milliseconds, a fine-grained splitting method is adopted, which splits the control logic into independent slices according to functional modules and flexibly deploys them on different nodes. When the transmission latency is greater than 10 milliseconds, a coarse-grained splitting method is adopted, which deploys all control logic of a single process as a whole on the same node to reduce cross-node communication overhead.

[0092] The scheduling strategy is configured based on the node's parallel processing capability and the timing matching requirements. The deployment switching trigger condition is set based on the node communication latency threshold, and the initial sensitivity coefficient of each deployment parameter is determined.

[0093] The scheduling strategy configuration adopts a fixed-priority preemptive scheduling, with the safety interlock logic having the highest priority, followed by the motion control logic, and the data acquisition and uploading logic having the lowest priority. The scheduling cycle is set according to the beat matching requirements, with a scheduling cycle of 1 millisecond for high-speed workstations, 5 milliseconds for medium-speed workstations, and 10 milliseconds for low-speed workstations.

[0094] The deployment switchover trigger conditions are set to two threshold levels. When the node's computing power utilization exceeds 90% for 5 consecutive seconds or the communication latency exceeds 15 milliseconds for 20 consecutive milliseconds, a level 1 warning is triggered and the pre-deployment process is started. When the computing power utilization exceeds 95% for 10 consecutive seconds or the communication latency exceeds 20 milliseconds for 50 consecutive milliseconds, a level 2 threshold is triggered and the formal deployment switchover is executed.

[0095] The initial sensitivity coefficient was obtained through offline calibration. Under standard test conditions, each deployment parameter was adjusted step by step, and the change in production efficiency after each adjustment was recorded. The change in efficiency corresponding to a unit parameter change was used as the initial sensitivity coefficient.

[0096] The dynamic deployment parameters are calculated based on the type of production disturbance and the depth of production line efficiency indicators. The process includes: Production condition data is categorized into disturbance types based on order switching disturbances, equipment load disturbances, and environmental disturbances, and then quantified into dimensionless disturbance level values ​​according to the degree of impact to obtain disturbance data.

[0097] Order changeover disturbances are categorized into five levels based on the number of processes involved and the degree of technological difference: Level 1 involves single-process tooling change with no change in process parameters; Level 2 involves multi-process tooling change with minor adjustments to process parameters; Level 3 involves full-process change with partial changes in process parameters; Level 4 involves product model change with partial changes in process route; and Level 5 involves the introduction of a completely new product with a completely different process route. Equipment load disturbances are categorized into five levels based on load rate and fluctuation amplitude: Level 1 involves a load rate below 60% and a fluctuation amplitude of less than 10%; Level 2 involves a load rate of 60% to 80% and a fluctuation amplitude of 10% to 20%; Level 3 involves a load rate of 80% to 90% and a fluctuation amplitude of 20% to 30%; Level 4 involves a load rate of 90% to 95% and a fluctuation amplitude of 30% to 40%; and Level 5 involves a load rate exceeding 95% and a fluctuation amplitude greater than 40%.

[0098] Environmental disturbances are classified into five levels based on the degree of deviation of temperature and humidity from standard operating conditions: Level 1 is 20 to 25 degrees Celsius and 40% to 60% RH; Level 2 is 15 to 30 degrees Celsius and 30% to 70% RH; Level 3 is 10 to 35 degrees Celsius and 20% to 80% RH; Level 4 is 5 to 40 degrees Celsius and 10% to 85% RH; and Level 5 is below 5 degrees Celsius or above 40 degrees Celsius and below 10% RH or above 85% RH. The disturbance level values ​​are dimensionless integers ranging from 1 to 5, with larger values ​​indicating higher disturbance intensity.

[0099] Determine the preset threshold for the production line efficiency depth index, calculate the index deviation value, classify the production line efficiency level according to the degree of deviation, and perform time synchronization and correlation mapping with the disturbance data.

[0100] The preset thresholds for production line efficiency depth indicators are based on design indicators. The threshold for workstation-level cycle time matching is 0.9, the threshold for work-in-process flow delay is 30 seconds, and the threshold for process synchronization stability is 0.8 seconds.

[0101] The indicator deviation value is the relative deviation between the actual indicator value and the threshold. The calculation formula is that the deviation value is equal to the difference between the actual value and the threshold value, divided by the threshold value. Positive indicators take negative values ​​to represent deviation, and negative indicators take positive values ​​to represent deviation.

[0102] The production line efficiency level is divided into four levels: excellent, good, qualified, and unqualified. Excellent corresponds to a deviation of less than -10%, good corresponds to a deviation between -10% and 0, qualified corresponds to a deviation between 0 and 10%, and unqualified corresponds to a deviation greater than 10%.

[0103] Time synchronization adopts a unified precision time protocol clock synchronization mechanism for the production line, which aligns the timestamps of disturbance data and performance data to the millisecond level. The time window matching method is used to achieve correlation mapping, and the disturbance data in each 5-second window corresponds one-to-one with the performance data in the same window.

[0104] Establish a disturbance type-deployment parameter mapping table to identify the dominant type in composite disturbances and determine the strategy strength based on the disturbance level.

[0105] In the disturbance type-deployment parameter mapping table, the core impact indicators of order switching disturbances are deployment latency and cycle time matching. The priority adjustment parameters are logical slice preloading and scheduling priority. The strategy is to preload the target process logic in advance and increase the scheduling priority of changeover-related tasks. The core impact indicators of equipment load disturbances are computing power utilization and response latency. The priority adjustment parameters are computing power allocation and task offloading. The strategy is to dynamically adjust node computing power quotas and offload non-real-time tasks to the edge gateway. The core impact indicators of environmental disturbances are communication reliability and node stability. The priority adjustment parameters are synchronization cycle and redundant switching. The strategy is to adjust interface synchronization parameters and enable redundant backup nodes.

[0106] The identification of the dominant type of composite disturbance employs the entropy weight method to calculate the contribution weight of each disturbance factor to the performance deviation. First, a multi-factor evaluation matrix is ​​constructed, with rows corresponding to disturbance types and columns corresponding to evaluation indicators. These indicators include three categories: performance deviation magnitude, disturbance duration, and indicator correlation coefficient. The evaluation matrix is ​​then normalized and normalized using the following normalization formula:

[0107] In the formula, For the first The first type of perturbation Evaluation index values, The total number of disturbance types; calculate the number of disturbance types. The entropy value of the item indicator, the formula for calculating the entropy value is:

[0108] Calculate the first Coefficient of difference of the items Finally, the number was obtained. The overall weight of the disturbance type is calculated using the following formula:

[0109] In the formula The total number of evaluation indicators; the disturbance type with the highest weight is determined as the dominant disturbance, and the rest are secondary disturbances.

[0110] The strategy strength corresponds one-to-one with the disturbance level. Level 1 disturbance corresponds to a strategy strength of 0.5, Level 2 to 0.8, Level 3 to 1.0, Level 4 to 1.3, and Level 5 to 1.5. The strategy strength is a dimensionless coefficient used to adjust the magnitude of parameter correction.

[0111] The priority of parameter adjustment is determined based on the strength of the strategy, the adjustment range is determined based on the production line efficiency deviation, and the single-step adjustment amount is limited. Specific deployment parameters are added for the types of production disturbances to obtain dynamic deployment parameters.

[0112] Parameter adjustment priorities are sorted from high to low based on response speed and direct impact, in the following order: computing power allocation adjustment, scheduling cycle adjustment, logical slice migration, interface parameter adjustment, and node redundancy switching. Parameters with higher priority are adjusted first.

[0113] The single-step adjustment amount is subject to upper limit constraints. The single-step adjustment amount of computing power allocation shall not exceed 20% of the node's rated computing power, the single-step adjustment amount of the scheduling cycle shall not exceed 50% of the current cycle, and the number of single-step migrations of logical slices shall not exceed 2 functional slices to prevent excessive parameter adjustment from causing system oscillation.

[0114] The specific deployment parameters are set according to the type of disturbance. Under order switching disturbances, supplementary process logic preload parameters are added to deploy the target process logic to the backup node in advance before the changeover instruction is issued, and the operation can be switched directly during the changeover. Under equipment load disturbances, supplementary task unloading parameters are added to unload non-real-time data analysis and quality statistics tasks from the programmable logic controller node to the edge computing gateway, freeing up the controller's computing power. Under environmental disturbances, supplementary redundant backup parameters are added to automatically activate the backup node into hot standby state when environmental conditions approach the critical threshold, and seamlessly switch over when the primary node fails.

[0115] Among them, the adjustment range of continuous deployment parameters The calculation formula is:

[0116] In the formula, This represents the initial sensitivity coefficient for the corresponding deployment parameters; This is the quantized value of the current disturbance intensity; This represents the deviation between the baseline value and the current value of production efficiency. The production efficiency benchmark score is taken as the stable operating efficiency value under standard working conditions and is set to 0.9.

[0117] The real-time deployment plan is obtained by adjusting the initial deployment parameters based on the dynamic deployment parameters. The process includes: Based on distributed node resource data and historical deployment records, the impact of each deployment parameter on production efficiency is updated to form an iterative sensitivity coefficient.

[0118] The sensitivity coefficient is updated every hour, using a recursive least squares method with a forgetting factor for iterative updates. The update formula is as follows:

[0119]

[0120]

[0121] In the formula, For the first The sensitivity coefficient after the latest update. This is the sensitivity coefficient from the previous test. For Kalman gain, This refers to the change in performance corresponding to this parameter adjustment. This is the amount of parameter adjustment for this test. Let covariance matrix be the variance matrix. This is the forgetting factor, with a value ranging from 0.9 to 0.99, and 0.95 under standard operating conditions.

[0122] Adjustments are prioritized based on their direct impact on production efficiency and response speed. Adjustments to computing power allocation directly change the available computing power of nodes, with a response time of less than 10 milliseconds, having the most direct impact on efficiency and thus the highest priority. Adjustments to scheduling cycles change the task execution frequency, with a response time of less than 50 milliseconds, and are the next highest priority. Logical slice migration requires data synchronization and state transitions, with a response time of less than 100 milliseconds, and is the third highest priority. Adjustments to interface parameters require renegotiation of communication parameters, with a response time of less than 200 milliseconds, and are the fourth highest priority. Node redundancy switching requires node initialization and business takeover, with a response time of less than 500 milliseconds, and is the lowest priority.

[0123] A smooth transition mechanism is established for setting transition times and effective conditions for continuous deployment parameters.

[0124] The smooth transition mechanism sets a linear transition time for continuously changing parameters such as computing power allocation and scheduling cycle. The transition time for computing power adjustment is 10 scheduling cycles, and the transition time for scheduling cycle adjustment is 5 scheduling cycles. The parameter values ​​gradually change to the target values ​​linearly within the transition time, avoiding control logic oscillations and production cycle fluctuations caused by sudden parameter changes.

[0125] For discrete adjustment actions such as logical slice migration and node switching, a step-by-step transition mechanism is adopted. Only one functional slice is migrated at a time, and the next slice migration is only executed after the previous slice migration is completed and has been running stably for at least ten scheduling cycles. Node switching uses a dual-node parallel operation transition method. The primary and backup nodes synchronously receive input data, the primary node outputs control commands, and the backup node performs synchronous calculations. Output permissions are switched only after ten consecutive cycles of consistent calculation results, avoiding system oscillations caused by excessively large single adjustment. A disturbance duration threshold is set for parameter activation conditions. Parameter adjustment is only formally triggered when the disturbance duration exceeds 500 milliseconds and the performance deviation exceeds 10%, avoiding erroneous adjustments caused by instantaneous disturbances.

[0126] The current dynamic deployment parameters are superimposed with the initial deployment parameters according to the adjustment priority and the current sensitivity coefficient. The rate of change and the reasons for the change are recorded before and after the superposition to obtain the real-time deployment plan. The parameter superposition adopts an incremental adjustment method, adding the calculated adjustment amount to the current operating parameters. Multiplicative adjustment is used for proportional parameters, and additive adjustment is used for absolute value parameters.

[0127] After the deployment plan is generated, security rules are verified. Plans that fail the verification are discarded and rolled back to the previous stable running plan.

[0128] The deployment plan records include all parameter values ​​before and after the adjustment, parameter change rate, type and level of disturbance that triggered the adjustment, performance deviation value, adjustment strategy basis, parameter effective time and rollback conditions, and also records the configuration address and encoding value of the corresponding hardware register to ensure that deployment instructions can be directly issued and executed.

[0129] Determine whether the current production efficiency indicators have reached the preset efficiency threshold. If so, maintain the real-time operation deployment plan; otherwise, optimize the dynamic deployment parameters.

[0130] The preset performance threshold is set at a comprehensive production performance score of 0.8. When the performance scores of three consecutive detection windows are all higher than the threshold, it is judged as meeting the standard and the current deployment plan remains unchanged; when the performance score of any window is lower than the threshold, it is judged as not meeting the standard and the dynamic deployment parameter feedback optimization process is initiated.

[0131] Feedback optimization is applied to the calculation benchmark of dynamically deployed parameters. When the target is not met for the first time, the original disturbance-parameter mapping relationship is maintained, the strategy strength coefficient is increased by 30%, and the single-step adjustment limit is reduced by 20%. The response speed is improved through small-step high-frequency adjustments. When the target is not met for two consecutive times, the collaborative adjustment of secondary adjustment parameters is enabled. On the basis of the primary adjustment parameter, the correction amount of the secondary parameter is superimposed, and the correction amount is 50% of the adjustment amount of the primary parameter. When the target is not met for three consecutive times, the mode is switched to fuzzy adaptive adjustment mode. This mode uses performance deviation and deviation change rate as input variables and parameter adjustment magnitude as output variables. The input variables are divided into five fuzzy subsets: negative large, negative small, zero, positive small, and positive large. The output variables are divided into five adjustment levels. The triangular membership function is used for fuzzification. The pre-set fuzzy rules are used for inference. Finally, the centroid method is used to defuzzify and obtain the parameter adjustment amount, realizing adaptive adjustment in nonlinear scenarios.

[0132] After the parameter adjustment takes effect, three consecutive verification windows are set, each lasting 10 seconds. At the end of each window, the production efficiency score is verified. If the efficiency score in any window is lower than 95% of the level before the adjustment, the adjustment is deemed invalid, and a rollback process is immediately triggered. The rollback process restores the parameters step by step in the reverse order of the adjustment, with an interval of at least five scheduling cycles between each step. The safety constraints are verified synchronously during the rollback process. After the rollback is completed, the reason and scenario for the invalid adjustment are recorded, and the sensitivity coefficient and rule base confidence are updated.

[0133] The working condition-efficiency correlation analysis method is used to analyze production working condition data, distributed node resource data, and production efficiency indicators to obtain correlation impact data. The process includes: Multi-mode disturbance features are extracted from production condition data, node status features are extracted from distributed node resource data, and efficiency change features are extracted from production efficiency indicators.

[0134] Multimodal disturbance characteristics include changeover time, process change percentage, and number of material change types in the order change dimension; average load rate, peak load, load fluctuation variance, and load change rate in the equipment load dimension; and mean temperature, mean humidity, temperature change rate, and humidity change rate in the environmental disturbance dimension.

[0135] Node status characteristics include node computing power utilization, memory usage, communication bandwidth utilization, interface bit error rate, message transmission latency, and task queue length. Performance change characteristics include performance degradation rate, performance fluctuation amplitude, duration of continuous performance degradation, and performance recovery time.

[0136] Correlation analysis was performed on the characteristics of multi-mode disturbances, node state characteristics, and efficiency changes to identify lagging production disturbances.

[0137] Linear correlation features were calculated using the Pearson correlation coefficient, while nonlinear correlation features were calculated using the Spearman rank correlation coefficient. For each set of disturbance and performance features, correlation coefficients were calculated for lags of 0, 5, 10, 20, and 30 seconds. The lag time with the largest absolute value of the correlation coefficient was taken as the lag duration of the disturbance. When the absolute value of the correlation coefficient was greater than 0.6, it was marked as a strongly correlated disturbance. Disturbances with a lag duration greater than 5 seconds were marked as lagging production disturbances.

[0138] Based on the combination of disturbance types, scenarios are divided. In different scenarios, regression models of node state characteristics, efficiency change characteristics and multi-mode disturbance characteristics are constructed, the goodness of fit is calculated, and the core disturbance factors are screened.

[0139] The disturbance scenarios are divided into six categories: single order switching scenario, single device load disturbance scenario, single environmental disturbance scenario, order switching plus load disturbance scenario, load disturbance plus environmental disturbance scenario, and composite disturbance scenario. For each scenario, a multiple linear regression model is constructed with performance change characteristics as the dependent variable and multi-mode disturbance characteristics and node state characteristics as independent variables. For scenarios with significant nonlinear relationships, a random forest regression model is used.

[0140] Calculate the coefficient of determination of the model As a goodness-of-fit index Models with an importance greater than 0.6 are considered effective models. The independent variables in the effective models are ranked by importance, and the top three features are identified as core perturbation factors, while the rest are considered secondary perturbation factors.

[0141] The process involves analyzing the correlation between order switching and logical redeployment cycles, the correlation between load fluctuations and resource allocation for computing power, and the correlation between environmental disturbances and node stability to generate statistical analysis reports. The process includes: Draw a scatter plot of order switching complexity and control logic deployment latency, identify switching nodes and deployment conflict areas, and calculate the mapping relationship between switching complexity and cycle time loss to obtain cycle time correlation.

[0142] The order switching complexity uses a continuous quantification value from 0 to 5, which is calculated by weighting the proportion of changeover processes and the degree of process difference. The horizontal axis of the scatter plot is the order switching complexity, and the vertical axis is the control logic deployment delay. Each sample point corresponds to one actual order switching process.

[0143] The deployment conflict area is a non-linear region where deployment latency increases sharply after the switching complexity exceeds level 3. In this region, the multi-process logic involved in the transformation has resource competition and dependency conflicts, resulting in a significant increase in deployment latency.

[0144] The mapping relationship is obtained through linear fitting. When the complexity is below level 3, the deployment latency increases with each level increase in complexity; when the complexity is above level 3, the latency increase is significantly amplified.

[0145] The relationship between equipment load rate and node computing power utilization is analyzed to generate a fitted relationship curve. Resource correlation is obtained by combining the correlation coefficient between load mutation rate and task scheduling delay.

[0146] Plotting device load rate on the horizontal axis and node computing power utilization on the vertical axis, steady-state data under different loads were collected, and a quadratic polynomial was used to fit the relationship curves. When the device load rate is below 60%, the computing power utilization rate increases slowly and linearly with the load rate; when the load rate is between 60% and 90%, the computing power utilization rate increases rapidly; after the load rate exceeds 90%, the computing power utilization rate approaches saturation, and the upward trend slows down. The load mutation rate is the change in device load rate per unit time, expressed as a percentage per second. Task scheduling latency is the time difference between the issuance of scheduling instructions and task execution. The two are positively correlated; the higher the load mutation rate, the greater the task scheduling latency. When the mutation rate exceeds a threshold, scheduling queue congestion is likely to occur.

[0147] The frequency of node failures under a preset temperature threshold is statistically analyzed, and the communication packet loss rate under a preset humidity threshold is combined to obtain the working condition correlation.

[0148] The preset temperature threshold is 40 degrees Celsius. The average failure frequency of nodes when the ambient temperature is higher than 40 degrees Celsius is statistically analyzed. Compared with normal temperature conditions, the increase in temperature will lead to an increase in the frequency of node hardware failures and a decrease in operational stability. The preset humidity threshold is 80%RH. When the ambient humidity is higher than 80%RH, the average packet loss rate of the communication interface increases and the attenuation of high-frequency band signals increases.

[0149] The statistical analysis report is obtained by standardizing the expression of cycle time correlation, resource correlation, and working condition correlation in the format of influencing factors-affected parameters-quantified relationship-scenario adaptation suggestions.

[0150] Each correlation in the statistical analysis report uses a unified format. For example, the cycle time correlation is described as follows: the influencing factor is order switching complexity greater than level 3, the affected parameters are control logic deployment latency and production cycle time, the quantitative relationship is a significant increase in deployment latency and a rise in cycle time loss rate, and the scenario adaptation suggestion is to preload the target process logic to the backup node before the model change and complete resource allocation in advance. The resource correlation is described as follows: the influencing factor is equipment load rate greater than 90% and mutation rate greater than 20% per second, the affected parameters are node computing power utilization and scheduling latency, the quantitative relationship is that computing power utilization is close to saturation and scheduling latency increases, and the scenario adaptation suggestion is to start the computing power offloading mechanism and migrate non-real-time tasks to the edge gateway. The operating condition correlation is described as follows: the influencing factor is ambient temperature greater than 40 degrees Celsius or humidity greater than 80%RH, the affected parameters are node failure rate and communication packet loss rate, the quantitative relationship is that failure frequency increases and packet loss rate increases, and the scenario adaptation suggestion is to enable redundant hot standby nodes, reduce interface transmission rate and improve anti-interference capability.

[0151] The statistical analysis report is converted into execution rules. The confidence and support of all execution rules are calculated, and reliable execution rules meeting preset thresholds are retained to generate a rule base for related impacts. Execution rules adopt a condition-conclusion production rule format. The condition part is a combination threshold of perturbation features, and the conclusion part is the change in affected parameters and adjustment suggestions. The preset retention threshold is a confidence level greater than or equal to 80% and a support level greater than or equal to 10%. Rules meeting these thresholds are considered reliable rules and included in the rule base for related impacts. The rule base is divided into four subcategories based on perturbation type: order switching rules, load fluctuation rules, environmental perturbation rules, and composite perturbation rules. Each rule stores the rule number, triggering condition, impact conclusion, confidence level, support level, applicable scenario, and suggested adjustment strategy.

[0152] Among them, the confidence level of association rules With support The calculation formulas are as follows:

[0153]

[0154] In the formula, For production disturbance events, This refers to an abnormal event affecting production efficiency. The support level represents the simultaneous occurrence of both disturbance and performance anomaly events, and is a dimensionless ratio. For disturbance events The number of samples that occurred. This represents the total number of samples.

[0155] The process of extracting features from current production efficiency indicators to obtain operational anomaly features includes: For the preset production efficiency threshold, determine the abnormal threshold range of each efficiency indicator, and for time-series fluctuations, set dynamic thresholds to adapt to the time-varying characteristics of the production line.

[0156] The static anomaly threshold range is set based on the lower limit of the design indicators. An overall efficiency score below 0.7, a production cycle time exceeding 120% of the baseline cycle time, equipment utilization rate below 80%, and product qualification rate below 95% are all considered abnormal. The dynamic threshold uses a sliding window adaptive calculation, taking the average of the corresponding indicators over the past hour minus three times the standard deviation as the dynamic anomaly threshold. When operating conditions change, such as production line shift changes or order switching, the baseline average and standard deviation are automatically updated to adapt to the efficiency fluctuation characteristics of different production stages.

[0157] Iterate through all time windows, and if any performance indicator exceeds the abnormal threshold range or dynamic threshold, it is determined to be an abnormal segment.

[0158] The time window length is set to 60 seconds, the sliding step is 10 seconds, and all continuous time windows are traversed. If any performance indicator in the window exceeds the static threshold or dynamic threshold, the window is marked as an abnormal segment. Adjacent abnormal segments are merged to form continuous abnormal time periods.

[0159] Extract corresponding abnormal segments from the time-series abnormality dimension, resource abnormality dimension, and deployment adaptation abnormality dimension respectively.

[0160] Three types of anomalies are extracted from the timing anomaly dimension: trend anomalies are characterized by a continuous unidirectional decline in performance indicators exceeding a threshold rate; volatility anomalies are characterized by performance indicator variance exceeding the normal range; and sudden anomalies are characterized by a rapid drop in performance indicators exceeding a threshold rate within a short period. Two types of anomalies are extracted from the resource anomaly dimension: computing power overload anomalies are characterized by node computing power utilization consistently exceeding 95%; and bandwidth insufficiency anomalies are characterized by node communication bandwidth occupancy consistently exceeding 90%. Three types of anomalies are extracted from the deployment adaptation anomaly dimension: ineffective adjustment anomalies are characterized by no significant performance improvement after parameter adjustment; parameter oscillation anomalies are characterized by repeated switching of parameters between two values; and logical conflict anomalies are characterized by conflicting execution sequences of different control logics.

[0161] By combining associated production status data and distributed node resource data, the anomaly type of abnormal segments is marked.

[0162] Anomalies are categorized into four main types: Disturbance-related anomalies, caused by sudden external disturbances, characterized by a sharp drop in performance that is highly synchronized with the disturbance event; Resource bottleneck anomalies, caused by insufficient node resources, characterized by a slow decline in performance accompanied by resource utilization saturation; Deployment adaptation anomalies, caused by parameter mismatch after process switching, characterized by persistently low performance after the change without significant disturbance; and Hardware failure anomalies, caused by node hardware failures, characterized by single-node data interruption accompanied by alarm signals.

[0163] Anomaly type labeling uses an association matching method, matching the operating condition data and node data within anomaly segments with anomaly patterns. The type with the highest matching degree is the type of the anomaly segment. If the correlation coefficients of multiple disturbance factors all reach the threshold, they are sorted from high to low weight, with the highest weight being the primary cause and the rest being secondary causes. When adjusting parameters, the adjustment amount corresponding to the primary cause is executed at 100%, and the adjustment amounts corresponding to the secondary causes are calculated proportionally and then added together. The total adjustment amount must not exceed the upper limit of the single-step adjustment amount. If it exceeds this limit, the adjustment amount of each cause is reduced proportionally.

[0164] Calculate the Pearson correlation coefficient between the performance change index and the corresponding disturbance factor within each abnormal segment, retain the abnormal segments that reach the preset coefficient threshold, and combine them with the corresponding abnormal type as operational abnormality characteristics.

[0165] For each anomalous segment, the Pearson correlation coefficients between the efficiency change sequence and three types of factors—order switching, equipment load, and environmental disturbances—are calculated. When the absolute value of the correlation coefficient is greater than or equal to 0.7, the disturbance factor is determined to be the primary cause of the anomaly. Anomalous segments with clearly defined primary causes are retained, while random fluctuation anomaly segments without significant correlation are discarded. The final operational anomaly characteristics include five items: anomaly type, anomaly duration, anomaly severity, primary cause, and correlation coefficient.

[0166] The deployment correction parameters are obtained by accurately adjusting the dynamic deployment parameters by combining the correlation impact data. The process includes: Based on the associated impact data, the abnormal causes of abnormal features are located through feature matching, and an adjustment mapping table is established by combining the priority adjustment parameters and strategy direction.

[0167] Feature matching employs a fuzzy proximity algorithm, calculating the membership degree for each indicator of the abnormal feature and the triggering condition of each rule in the rule base. The membership degree is calculated using a triangular membership function, with values ​​ranging from 0 to 1. Overall similarity is calculated using a weighted average proximity degree, with the following formula:

[0168] In the formula, For the first Membership degree of a feature The weights are assigned to the corresponding features; when the similarity is greater than or equal to 0.7, the match is considered successful, and the rule with the highest similarity corresponds to the core abnormality cause.

[0169] The mapping table is adjusted and constructed according to the anomaly type. For disturbance and impact anomalies, the priority adjustment parameters are computing power allocation and scheduling priority, with the strategy being to quickly increase the computing power of critical nodes and increase the priority of control tasks. For resource bottleneck anomalies, the priority adjustment parameters are task offloading and slice migration, with the strategy being to offload non-real-time tasks and migrate some logic to idle nodes. For deployment adaptation anomalies, the priority adjustment parameters are logical slice reallocation and interface parameter correction, with the strategy being to rematch process logic with node resources and optimize cross-node synchronization parameters. For hardware failure anomalies, the priority adjustment parameter is redundant node switching, with the strategy being to seamlessly switch to hot standby nodes and isolate faulty nodes.

[0170] Identify parameter adjustment conflicts, coordinate them through preset constraint priorities, and plan the parameter adjustment sequence from fastest to slowest response speed.

[0171] Parameter adjustment conflicts refer to contradictions arising when parameters with different adjustment directions are executed simultaneously, such as conflicts between increasing computing power and reducing power consumption, or shortening scheduling cycles and reducing bandwidth usage. Constraint priorities, from highest to lowest, are: security constraints, process constraints, performance constraints, and power consumption constraints. When conflicts occur, higher-priority constraints are prioritized, sacrificing lower-priority objectives. Parameter adjustments are strictly executed according to response speed, from fastest to slowest: first, computing power allocation adjustments are performed; second, scheduling cycle adjustments are performed; third, logical slice migrations are performed; and finally, interface parameter corrections and node switching are performed. This ensures that fast-adjusting parameters take effect first, quickly suppressing performance degradation, while slower-adjusting parameters take effect later, achieving deep optimization.

[0172] By combining the correlation impact data to calculate the power allocation correction range, the logical slice node allocation adjustment amount, and the cross-node synchronization interface parameter correction value as the adjustment parameter correction range, the dynamic deployment parameters are adjusted to obtain the deployment correction parameters.

[0173] The logical slice node allocation adjustment amount is calculated based on the computing power gap of abnormal nodes. The computing power gap is the difference between the current computing power load and the optimal computing power load, and the unit is consistent with the computing power unit. The rated computing power of each non-real-time slice is a fixed value. The number of slices to be migrated is the computing power gap divided by the rated computing power of a single slice and rounded up. The migration priority is from low to high as follows: data upload slices, data acquisition slices, and logical operation slices. Slices to be migrated are selected in order of priority from low to high until the cumulative migration computing power meets the gap requirement.

[0174] Slice migration adopts a load-then-switch execution flow. First, the program image and current running status data of the target slice are loaded on the target node, and it enters a hot standby state. Simultaneously, the interface compatibility and resource reserves of the target node are verified. After successful verification, at the end of the current scheduling cycle, the input and output permissions are switched from the source node to the target node, while the image on the source node is kept running for at least two scheduling cycles. If the data verification is error-free for three consecutive scheduling cycles after the switch, the corresponding slice on the source node is officially unloaded, and the migration is completed. The pre-loading process is the same as the migration loading process, except that the permission switch is not performed. The pre-loaded slice is stored in the local spare memory area of ​​the node and does not occupy the running computing power. It is only activated when the switch is triggered.

[0175] The cross-node synchronization interface parameters adopt a tiered adjustment method, with four tiers for the synchronization period: 1 millisecond, 2 milliseconds, 5 milliseconds, and 10 milliseconds, and three tiers for the number of retransmissions: 1, 3, and 5. When the packet loss rate is between 0.1% and 0.5%, the synchronization period is increased by one tier; when the packet loss rate is between 0.5% and 1%, the synchronization period is increased by two tiers and the number of retransmissions is increased by one tier; when the packet loss rate is higher than 1%, the synchronization period and the number of retransmissions are increased to the highest tier; when the packet loss rate is below 0.05% for 30 seconds, the parameters are gradually reduced back to the baseline tier.

[0176] Among them, the adjustment range of node computing power allocation The calculation formula is:

[0177] In the formula, The computing power adaptation factor is set to 100 MIPS for embedded programmable logic controller nodes and 2 TOPS for edge computing gateway nodes. This is the quantized value of the current disturbance intensity. For the first The current resource utilization rate of each node. The optimal utilization threshold for nodes is uniformly set to 0.75.

[0178] In summary, this embodiment provides a method for dynamically deploying distributed control logic for multi-product mixed-line production. Through synchronous collection and hierarchical analysis of multi-dimensional operational data, it can capture various disturbances, distributed hardware resources, and production operation status in real time. It can autonomously complete the dynamic adjustment of control configuration without continuous manual intervention, eliminating the lag and experience dependence caused by manual adjustment.

[0179] By combining hardware limits and rigid constraints of production processes to generate basic configurations, and by generating adaptive adjustment quantities layer by layer based on real-time disturbances on site, a parameter smooth transition mechanism is set up to avoid production line operation oscillations caused by sudden configuration changes.

[0180] By exploring the inherent correlation between various disturbances and equipment operating status, and constructing stable correlation judgment rules, when abnormalities such as production line imbalance or resource overload occur, abnormal features can be extracted autonomously and the root cause can be located, eliminating the need for manual trial and error to troubleshoot.

[0181] Faced with complex operating conditions with multiple interferences, it can automatically identify the core disturbance factors affecting production, and adjust various control configurations in an orderly manner according to response speed and constraint priority. This ensures stable process coordination during multi-variety mixed production, reduces on-site debugging and daily operation and maintenance investment, and improves the long-term stability and autonomous adaptation capability of flexible production lines.

[0182] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0183] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for dynamically deploying distributed control logic for multi-product mixed-line production, characterized in that, include: Production condition data, distributed node resource data, and process constraint data are collected, and real-time time-series operation data of mixed-line production is collected. The time-series operation data is analyzed using a multi-dimensional time-series efficiency extraction method to obtain production efficiency indicators. Initial deployment parameters are formulated based on the distributed node resource data and the process constraint data combined with the requirements of mixed-line production tasks. Dynamic deployment parameters are calculated based on the production disturbance type and production line efficiency depth index. The initial deployment parameters are adjusted based on the dynamic deployment parameters to obtain a real-time operation deployment plan. Determine whether the current production efficiency indicators have reached the preset efficiency threshold. If so, maintain the real-time operation deployment plan; otherwise, optimize the dynamic deployment parameters. The working condition-efficiency correlation analysis method is used to analyze the production working condition data, distributed node resource data and production efficiency indicators to obtain the correlation impact data. The current production efficiency indicators are feature extracted to obtain the abnormal operation characteristics. Combined with the correlation impact data, the dynamic deployment parameters are accurately corrected to obtain the deployment correction parameters.

2. The method for dynamically deploying distributed control logic for multi-product mixed-line production according to claim 1, characterized in that, The production status data includes order switching data, equipment load data, and environmental disturbance data; the distributed node resource data includes computing power margin data, communication bandwidth data, storage capacity data, and interface compatibility data; the process constraint data includes process path dependency constraints, workstation cycle time matching constraints, quality control threshold constraints, and safety interlocking rule constraints.

3. The method for dynamically deploying distributed control logic for multi-product mixed-line production according to claim 1, characterized in that, The process of analyzing and obtaining the aforementioned production efficiency indicators includes: Collect basic operating signals, production line efficiency-related data, and deployment control parameters, and preprocess them to obtain the time-series operating data; Production cycle time, equipment utilization rate and product qualification rate are extracted from the basic operation signals as basic production performance indicators, and workstation-level cycle time matching degree, work-in-process flow delay and process synchronization stability indicators are extracted from the production line efficiency correlation data as production line efficiency depth indicators. Logical redeployment time, resource scheduling response metrics, and node adaptability metrics are extracted from the deployment control parameters as deployment adaptation metrics. The trend characteristics and disturbance correlation indicators of the basic production performance indicators, the production line efficiency depth indicators, and the deployment adaptation indicators are extracted as the production efficiency indicators.

4. The method for dynamically deploying distributed control logic for multi-product mixed-line production according to claim 1, characterized in that, The process of determining the initial deployment parameters includes: The task requirements of different mixed-line production scenarios are transformed into technical indicators. Performance constraint boundaries are formulated based on the distributed node resource data. Combined with the process constraint data, process dependencies, cycle time matching requirements and safety interlocking rules are determined. The optimal set of deployment nodes is selected by combining the characteristics of node computing power distribution and the granularity of control logic slices, and the weight of logical slice allocation is determined based on the upper limit of computing power of a single node. The baseline deployment granularity is determined based on cross-node transmission latency and node computing power range, and the control logic splitting method is selected according to the aforementioned technical indicators and real-time requirements. The scheduling strategy is configured based on the node's parallel processing capability and the timing matching requirements. The deployment switching trigger condition is set based on the node communication latency threshold, and the initial sensitivity coefficient of each deployment parameter is determined.

5. The method for dynamically deploying distributed control logic for multi-product mixed-line production according to claim 1, characterized in that, The process of calculating the dynamic deployment parameters includes: The production condition data is classified into disturbance types according to order switching disturbance, equipment load disturbance and environmental disturbance, and the disturbance data is obtained by quantifying the degree of impact into dimensionless disturbance level values. Determine the preset threshold of the production line efficiency depth index, calculate the index deviation value, classify the production line efficiency level according to the degree of deviation, and perform time synchronization and correlation mapping processing with the disturbance data; Establish a disturbance type-deployment parameter mapping table to identify the dominant type in composite disturbances and determine the strategy strength based on the disturbance level; The parameter adjustment priority is determined based on the strength of the strategy, the adjustment range is determined based on the production line efficiency deviation, and the single-step adjustment amount is limited. The dynamic deployment parameters are obtained by supplementing specific deployment parameters for the production disturbance type.

6. The method for dynamically deploying distributed control logic for multi-product mixed-line production according to claim 1, characterized in that, The process of adjusting to obtain the real-time deployment scheme includes: Based on the distributed node resource data and historical deployment records, the impact of each deployment parameter on production efficiency is updated to form an iterative sensitivity coefficient. Prioritize adjustments based on the directness of their impact on production efficiency and the speed of response. A smooth transition mechanism is established for setting transition times and effective conditions for continuous deployment parameters; The current dynamic deployment parameters are superimposed with the initial deployment parameters according to the adjustment priority and the current sensitivity coefficient. The parameter change rate and the reason for the change are recorded before and after superposition to obtain the real-time operation deployment scheme.

7. The method for dynamically deploying distributed control logic for multi-product mixed-line production according to claim 1, characterized in that, The process of obtaining the aforementioned correlation impact data includes: Multi-mode disturbance features are extracted from the production condition data, node status features are extracted from the distributed node resource data, and efficiency change features are extracted from the production efficiency indicators. Correlation analysis was performed on the multi-mode disturbance characteristics, node state characteristics, and efficiency change characteristics to identify lagging production disturbances; Based on the combination of disturbance types, scenarios are divided. In different scenarios, regression models of node state characteristics, efficiency change characteristics and multi-mode disturbance characteristics are constructed, the goodness of fit is calculated, and core disturbance factors are screened. Analyze the correlation between order switching and logical redeployment, the correlation between load fluctuation and computing power allocation, and the correlation between environmental disturbances and node stability to generate statistical analysis reports; The statistical analysis report is converted into execution rules, the confidence and support of all execution rules are calculated, and reliable execution rules that meet the preset threshold are retained to generate a rule base for correlation effects.

8. The method for dynamically deploying distributed control logic for multi-product mixed-line production according to claim 7, characterized in that, The process of generating the statistical analysis report includes: Draw a scatter plot of order switching complexity and control logic deployment latency, identify switching nodes and deployment conflict areas, and calculate the mapping relationship between switching complexity and cycle time loss to obtain the cycle time association; The relationship between equipment load rate and node computing power utilization is analyzed to generate a fitting relationship curve. The resource association is obtained by combining the correlation coefficient between load mutation rate and task scheduling delay. The operating condition correlation is obtained by statistically analyzing the node failure frequency under a preset temperature threshold and combining it with the communication packet loss rate under a preset humidity threshold. The statistical analysis report is obtained by standardizing the expression of the beat association, the resource association, and the working condition association in the format of influencing factors-affected parameters-quantified relationship-scenario adaptation suggestions.

9. The method for dynamically deploying distributed control logic for multi-product mixed-line production according to claim 1, characterized in that, The process of obtaining the aforementioned operational anomaly characteristics includes: For the preset production efficiency threshold, determine the abnormal threshold range of each efficiency indicator, and for time-series fluctuations, set dynamic thresholds to adapt to the time-varying characteristics of the production line. Traverse all time windows, and when any performance indicator exceeds the abnormal threshold range or the dynamic threshold, it is determined to be an abnormal segment; Extract the corresponding abnormal segments from the dimensions of time-series anomalies, resource anomalies, and deployment adaptation anomalies, respectively. By combining the associated production condition data and the distributed node resource data, the anomaly type of the abnormal segment is marked; Calculate the Pearson correlation coefficient between the performance change index and the corresponding disturbance factor within each abnormal segment, retain the abnormal segments that reach the preset coefficient threshold, and combine them with the corresponding abnormal type as the operational abnormality feature.

10. The method for dynamically deploying distributed control logic for multi-product mixed-line production according to claim 1, characterized in that, The process of adjusting the deployment correction parameters includes: Based on the associated impact data, the abnormal causes of the abnormal operation characteristics are located through feature matching, and an adjustment mapping table is established by combining the priority adjustment parameters and strategy direction; Identify parameter adjustment conflicts, coordinate them through preset constraint priorities, and plan the parameter adjustment sequence from fastest to slowest response speed; The deployment correction parameters are obtained by combining the computational power allocation correction magnitude, logical slice node allocation adjustment amount, and cross-node synchronization interface parameter correction value with the aforementioned correlation impact data to calculate the adjustment parameter correction magnitude and adjusting the dynamic deployment parameters.

Images