AI-driven industrial automatic production line intelligent scheduling method and system

Abstract

The invention discloses an AI-driven industrial automatic production line intelligent scheduling method and system, and relates to the technical field of production line intelligent scheduling, and the method comprises the steps: carrying out the multi-source real-time collection based on an industrial automatic production line, obtaining a multi-source heterogeneous data set, carrying out the deep analysis and prediction, and recognizing a production dynamic prediction data set; positioning and identifying production equipment, determining production bottleneck position information, and drawing a demand change trend chart; carrying out production operation analysis, generating production bottleneck early warning information, carrying out control scheduling, and formulating a production scheduling strategy; and scheduling monitoring is carried out on the industrial automatic production line, continuous optimization response is carried out, and a real-time production scheduling optimization strategy is obtained. The technical problems of low scheduling efficiency and insufficient adaptability caused by lack of dynamic prediction capability of the industrial automatic production line in the prior art are solved, and the technical effects of realizing dynamic prediction of the industrial automatic production line and improving the scheduling efficiency and adaptability of the production line are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent scheduling of production lines, and in particular to an AI-driven intelligent scheduling method and system for industrial automation production lines. Background Art

[0002] In the field of industrial automation, efficient production line scheduling is crucial for improving productivity and reducing costs. Traditional scheduling methods, which rely heavily on pre-set rules or manual experience, struggle to cope with complex and volatile production environments, such as sudden equipment failures and fluctuating production demand. This often leads to difficulties in predicting production bottlenecks and delayed scheduling strategy adjustments, impacting the overall operational efficiency and adaptability of the production line.

[0003] The existing technology has the technical problem that industrial automation production lines lack dynamic prediction capabilities, resulting in low scheduling efficiency and insufficient adaptability. Summary of the Invention

[0004] This application provides an AI-driven intelligent scheduling method and system for industrial automation production lines, which is used to solve the technical problem in the existing technology that industrial automation production lines lack dynamic prediction capabilities, resulting in low scheduling efficiency and insufficient adaptability.

[0005] In view of the above problems, this application provides an AI-driven intelligent scheduling method and system for industrial automation production lines.

[0006] In a first aspect, the present application provides an AI-driven intelligent scheduling method for an industrial automation production line, the method comprising: Based on the industrial automation production line, multi-source real-time data collection is carried out to obtain a multi-source heterogeneous data set, and an AI algorithm is integrated to perform in-depth analysis and prediction on the multi-source heterogeneous data set to identify a production dynamic prediction data set, which includes production bottleneck prediction parameters and production demand prediction parameters; the production equipment is located and identified according to the production bottleneck prediction parameters, and the production bottleneck location information is determined. The demand change analysis is performed according to the production demand prediction parameters, and a demand change trend chart is drawn; the production operation of the production equipment is analyzed according to the demand change trend chart and the production bottleneck location information, and production bottleneck early warning information is generated for control scheduling and production scheduling strategies are formulated; the production scheduling strategies are executed to schedule and monitor the industrial automation production line, and monitoring feedback data is generated to continuously optimize the production scheduling strategies and obtain real-time production scheduling optimization strategies.

[0007] A second aspect of the present application provides an AI-driven intelligent scheduling system for industrial automation production lines, the system comprising: An analysis and prediction module is used to perform multi-source real-time acquisition based on an industrial automation production line to obtain a multi-source heterogeneous data set, integrate an AI algorithm to perform in-depth analysis and prediction on the multi-source heterogeneous data set, and identify a production dynamic prediction data set, wherein the production dynamic prediction data set includes production bottleneck prediction parameters and production demand prediction parameters; a demand change analysis module is used to locate and identify production equipment according to the production bottleneck prediction parameters, determine the production bottleneck location information, perform demand change analysis according to the production demand prediction parameters, and draw a demand change trend chart; a control and scheduling module is used to perform production operation analysis on production equipment according to the demand change trend chart in combination with the production bottleneck location information, generate production bottleneck warning information for control scheduling, and formulate a production scheduling strategy; an optimization response module is used to execute the production scheduling strategy to perform scheduling monitoring on the industrial automation production line, generate monitoring feedback data to continuously optimize the production scheduling strategy, and obtain a real-time production scheduling optimization strategy.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: Based on the industrial automation production line, multi-source real-time data collection is performed to obtain multi-source heterogeneous data sets. AI algorithms are integrated to conduct in-depth analysis and prediction of these multi-source heterogeneous data sets, identifying production dynamic prediction data sets. Production equipment is located and identified, production bottleneck location information is determined, and demand change trend charts are drawn. The production operation of production equipment is analyzed based on the production bottleneck location information, and production bottleneck warning information is generated for control and scheduling, and a production scheduling strategy is formulated. The production scheduling strategy is executed to schedule and monitor the industrial automation production line, and monitoring feedback data is generated to continuously optimize the production scheduling strategy and obtain a real-time production scheduling optimization strategy. The technical effect of achieving dynamic prediction of industrial automation production lines and improving the efficiency and adaptability of production line scheduling is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A flowchart of an AI-driven intelligent scheduling method for industrial automation production lines provided in an embodiment of the present application; Figure 2 Schematic diagram of the structure of the AI-driven intelligent scheduling system for industrial automation production lines provided in an embodiment of the present application.

[0011] Description of the reference numerals: analysis and prediction module 10 , demand change analysis module 20 , control and scheduling module 30 , optimization and response module 40 . DETAILED DESCRIPTION

[0012] This application provides an AI-driven intelligent scheduling method and system for industrial automation production lines, which is used to solve the technical problem in the existing technology that industrial automation production lines lack dynamic prediction capabilities, resulting in low scheduling efficiency and insufficient adaptability.

[0013] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative work are within the scope of protection of this application.

[0014] Example 1, as Figure 1 As shown, the present application provides an AI-driven intelligent scheduling method for industrial automation production lines, the method comprising: Step S100: Perform multi-source real-time acquisition based on the industrial automation production line to obtain a multi-source heterogeneous data set, integrate an AI algorithm to perform in-depth analysis and prediction on the multi-source heterogeneous data set, and identify a production dynamic prediction data set, which includes production bottleneck prediction parameters and production demand prediction parameters.

[0015] Specifically, the production and operation status of the industrial automation production line is collected in parallel in real time from multiple sources to obtain a multi-source heterogeneous data set covering equipment operation parameters, material flow data, environmental information, etc.; the data set is spatiotemporally aligned to generate a standardized time series data stream, and then the AI ​​algorithm is integrated to build an AI prediction model group (including an AI production capacity prediction model using a spatiotemporal convolutional network and an AI demand prediction model using a multi-head self-attention mechanism), and the standardized time series data stream is synchronized to the model group for production operation prediction to generate an initial prediction data set; at the same time, a digital twin verification layer is built based on the multi-source heterogeneous data set, and the initial prediction data set is verified. Then, the AI ​​prediction model group is incrementally learned according to the verification results, and the initial prediction data set is prioritized and rearranged according to the learning results, and finally a production dynamic prediction data set containing production bottleneck prediction parameters and production demand prediction parameters is identified.

[0016] Step S200: positioning and identifying the production equipment according to the production bottleneck prediction parameters, determining the production bottleneck location information, performing demand change analysis according to the production demand prediction parameters, and drawing a demand change trend graph.

[0017] Specifically, for the production bottleneck prediction parameters, the position analysis of the robotic arm as the production equipment is carried out to determine the bottleneck station identifier, and the risk analysis is carried out in combination with the operation data of the robotic arm to obtain the risk level value. The equipment topology map is constructed through three-dimensional topology analysis and the identifier is synchronized to locate the target equipment coordinate set of the robotic arm; based on the coordinate set, the production load analysis of the robotic arm is carried out to construct a production load time series matrix, and after extracting N cycle load peaks, the random forest algorithm is used to obtain the fault signal-operation time series correlation coefficient, which is mapped to a two-dimensional grid coordinate system for identification to determine the production bottleneck position information of the robotic arm; for the production demand prediction parameters, the product demand sequence is analyzed and determined and grouped by category to obtain multiple demand sequence groups, and demand confidence analysis is carried out to generate multiple demand confidence levels. The confidence interval boundaries are set to filter out sequence groups that exceed the boundaries to construct the product dimension, and the time dimension is constructed in combination with the production line operation process data. A dual-axis line chart is constructed and the demand change trend chart is drawn through demand interaction response.

[0018] Step S300: performing production operation analysis on production equipment according to the demand change trend diagram in combination with the production bottleneck location information, generating production bottleneck warning information for control scheduling, and formulating a production scheduling strategy.

[0019] Specifically, the demand change trend graph is traversed to extract the set of demand mutation time points, and the demand increase detection is performed on the time point set to obtain the demand increment data; at the same time, the production bottleneck location information is traversed to perform bottleneck risk analysis, determine the high-risk location coordinate set, and map the demand mutation time points to the high-risk location coordinate set for time-space coupling to calculate the bottleneck impact factor; based on the demand increment data, an impact critical analysis is performed and an impact risk threshold is set. When the bottleneck impact factor exceeds the threshold, the impact range information is delineated according to the bottleneck impact factor and the warning level is set to generate production bottleneck warning information including the warning level and impact range; then, the warning information is combined with the production line process flow to conduct a comprehensive analysis of the production operation status of the production equipment, and a production scheduling strategy is formulated including equipment load adjustment, process priority sorting, material supply linkage, etc., to achieve control and scheduling of the production process.

[0020] Step S400: executing the production scheduling strategy to perform scheduling monitoring on the industrial automation production line, generating monitoring feedback data to continuously optimize the production scheduling strategy in response, and obtaining a real-time production scheduling optimization strategy.

[0021] Specifically, a production scheduling strategy is executed to control the operation of the industrial automation production line, and the production equipment and production line are monitored and collected simultaneously to obtain equipment response data and production line output data; based on the production scheduling strategy, an instruction parameter vector is defined, which is correlated and analyzed with the equipment response data to generate an instruction-response deviation component, and the key deviation factor is analyzed and identified; reinforcement learning is performed on the key deviation factor and the production line output data to obtain data reinforcement learning results, and based on this, conflict detection is performed on the production scheduling strategy to obtain production scheduling conflict parameters; the production scheduling conflict parameters are injected into the digital twin sandbox to continuously verify the effectiveness of the strategy. If the verification fails, conflict replacement scheduling is performed according to the conflict parameters, and finally a real-time production scheduling optimization strategy is constructed.

[0022] In one possible implementation, step S100 further includes: Step S110: Parallel collection of production operations of an industrial automation production line is performed to obtain a multi-source heterogeneous data set.

[0023] Step S120: performing spatiotemporal alignment processing based on the multi-source heterogeneous data sets to generate a standardized time series data stream.

[0024] Step S130: Integrate the AI ​​algorithm to build an AI prediction model group, synchronize the standardized time series data stream to the AI ​​prediction model group to perform production operation prediction, and generate an initial prediction data set.

[0025] Step S140: construct a digital twin verification layer based on the multi-source heterogeneous data set, perform data verification on the initial prediction data set through the digital twin verification layer, and generate a data verification result.

[0026] Step S150: performing incremental learning on the AI ​​prediction model group according to the data verification results, performing priority screening and rearrangement on the initial prediction data set according to the learning results, and identifying the production dynamic prediction data set.

[0027] Specifically, by deploying sensors, intelligent monitoring equipment, and data acquisition terminals at key nodes of industrial automation production lines, real-time, parallel data collection is performed on the production process. The collected information includes the operating parameters of production equipment, the flow status of materials, product quality inspection data, and workshop environmental data. This data, encompassing both structured and unstructured data, together constitutes a multi-source, heterogeneous dataset. The operating parameters of production equipment reflect the dynamic performance of the equipment during the production process, the flow status of materials reflects the delivery of materials through each link, product quality inspection data is used to assess whether production results meet standards, and workshop environmental data relates to the external conditions of production. This combination of structured and unstructured data, in various forms and dimensions, provides comprehensive raw information for subsequent analysis.

[0028] We perform spatiotemporal alignment on the acquired multi-source heterogeneous datasets. This aligns the time records of different acquisition devices and nodes, ensuring consistency across all data. Spatially, we correlate and match each type of data with the corresponding equipment locations and workstation information, ensuring that the data accurately corresponds to the spatial layout of the production line. This process transforms the multi-source heterogeneous datasets into standardized time-series data streams, providing a standardized and unified data foundation for subsequent AI algorithm analysis and prediction.

[0029] Integrate AI algorithms to build an AI prediction model group. First, divide the standardized time series data stream into capacity time series data stream and demand time series data stream according to the production line operation process data. Among them, use the spatiotemporal convolutional network as the algorithm architecture to pre-train the production equipment for capacity and build an AI capacity prediction model. Use the multi-head self-attention mechanism as the algorithm architecture to pre-train the production line for demand and build an AI demand prediction model. Then synchronize the capacity time series data stream to the AI ​​capacity prediction model for loading training to obtain the initial parameters for production bottleneck prediction. Synchronize the demand time series data stream to the AI ​​demand prediction model for supply training to obtain the initial parameters for production demand prediction. Finally, integrate the initial parameters for production bottleneck prediction with the initial parameters for production demand prediction to generate an initial prediction data set.

[0030] A digital twin verification layer is constructed based on multi-source heterogeneous data sets. This verification layer uses digital twin technology to build a virtual model that accurately corresponds to the physical entity of the industrial automation production line. The initial prediction data set is input into the virtual model to simulate the operating status of the production line under the working conditions corresponding to the prediction data. The simulation operation results are then compared and analyzed with the actual production data in the multi-source heterogeneous data sets to determine the degree of deviation between the initial prediction data and the actual production situation, and then generate data verification results to evaluate the accuracy and reliability of the initial prediction data set.

[0031] According to the data verification results, the prediction deviation data determined in the verification process is input into the AI ​​prediction model group as a new training sample, and incremental learning is performed by adjusting the network weights and parameters of the model to optimize the prediction performance of the model; according to the prediction results output by the model after incremental learning, combined with the size of the prediction error and the degree of impact on production scheduling, the initial prediction data set is prioritized and screened and rearranged, retaining data with small errors and high impact, and reducing the priority of data with large errors and low impact, and finally identifying the production dynamic prediction data set containing production bottleneck prediction parameters and production demand prediction parameters from the screened and rearranged data set.

[0032] In one possible implementation, step S130 further includes: Step S131: Segment the standardized time series data flow according to the production line operation process data to obtain the production capacity time series data flow and the demand time series data flow.

[0033] Step S132: Use the spatiotemporal convolutional network as the algorithm architecture to pre-train the production capacity of production equipment and build an AI production capacity prediction model.

[0034] Step S133: Use the multi-head self-attention mechanism as the algorithm architecture to pre-train the production line demand and build an AI demand forecasting model.

[0035] Step S134: Synchronize the production capacity time series data stream to the AI ​​production capacity prediction model for loading training to obtain initial parameters for production bottleneck prediction.

[0036] Step S135: Synchronize the demand time series data stream to the AI ​​demand forecasting model for supply training to obtain initial parameters for production demand forecasting.

[0037] Step S136: Integrate the initial parameters for production bottleneck prediction and the initial parameters for production demand prediction to construct an initial prediction data set.

[0038] Specifically, the standardized time-series data stream is structured and segmented based on production line operational process data. Following the complete production line process, from raw material input and processing to finished product output, time-series data related to production equipment operating efficiency, processing capacity, and load status are extracted and integrated to form a capacity time-series data stream. Simultaneously, time-series data related to product order volume, material requirements, and market demand fluctuations are extracted and integrated to form a demand time-series data stream. This targeted segmentation based on process nodes allows for precise separation of the two types of data, providing targeted data input for subsequent capacity and demand forecasting.

[0039] A spatiotemporal convolutional network is adopted as the algorithm architecture, and the historical capacity data of production equipment is used as training samples. The capacity characteristics of different equipment in different time periods are extracted through the convolutional layer in the network. The spatiotemporal convolution operation is used to capture the changing trend of equipment capacity in the time dimension and the correlation between the spatial dimension, and the capacity of production equipment is pre-trained. During the training process, the network parameters are continuously adjusted to minimize the error between the predicted value and the actual capacity data, and finally an AI capacity prediction model is constructed that can accurately predict changes in production equipment capacity and identify potential capacity bottlenecks.

[0040] A multi-head self-attention mechanism is adopted as the algorithm architecture, and the historical demand data of the production line is used as the training sample. Through multiple parallel self-attention heads, the correlation features of the demand data at different time scales are captured respectively. The features output by each attention head are spliced ​​and fused to extract the dependency and change pattern of the production line demand in the time dimension. In the pre-training process, the deviation between the predicted demand and the actual demand is used as the loss function, and the model parameters are optimized through back-propagation iteration. The attention weight is continuously adjusted to improve the prediction accuracy. Finally, an AI demand forecasting model is constructed that can effectively capture the demand fluctuation pattern and accurately predict the changes in production line demand.

[0041] The segmented capacity time series data stream is divided according to the preset time window to form training samples and verification samples, which are simultaneously input into the constructed AI capacity prediction model; the capacity features in the samples are deeply extracted through the model's spatiotemporal convolution layer to capture the capacity fluctuation patterns and spatial correlations of the equipment in different time periods. During the loading and training process, the error between the predicted capacity and the actual capacity is used as the optimization goal, and the convolution kernel parameters and network weights of the model are continuously adjusted; when the model prediction accuracy reaches the preset threshold, the initial production bottleneck prediction parameters that can reflect the future capacity bottleneck situation of the production equipment are output. The parameters contain information such as the equipment identification, time node and bottleneck degree where the bottleneck may occur.

[0042] The demand time series data stream is sorted in chronological order and synchronously input into the AI ​​demand forecasting model. The model extracts the correlation features of demand data in different time periods through a multi-head self-attention mechanism, focusing on capturing the changing patterns and potential patterns of product demand. During the supply training process, the deviation between predicted demand and actual demand is used as the optimization goal, and the model parameters are continuously adjusted to improve the forecast accuracy. The final output includes the initial parameters for production demand forecasting, including future product demand scale, demand change trends, and key time nodes.

[0043] The initial parameters for production bottleneck prediction and production demand prediction are integrated through a data fusion algorithm. The data formats of the two types of parameters are first standardized, and the timestamps and parameter units are unified. Then, an association mapping is established based on the time dimension, so that the time nodes when production bottlenecks occur correspond to the demand forecast data of the corresponding time period. At the same time, the bottleneck equipment information is matched with the product demand information involved in the equipment through spatial dimension association, and finally integrated into an initial prediction data set containing multi-dimensional associated data such as time, equipment, bottleneck degree, and demand scale.

[0044] In one possible implementation, step S200 further includes: Step S210: performing position analysis on production equipment based on the production bottleneck prediction parameters to determine a bottleneck workstation identifier.

[0045] Step S220: performing risk analysis on the industrial automation production line according to the bottleneck workstation identifier in combination with production equipment to determine a risk level value.

[0046] Step S230: performing a three-dimensional topological analysis based on the production equipment, constructing an equipment topology map, synchronizing the bottleneck station identifier to the equipment topology map for positioning, and determining a target equipment coordinate set.

[0047] Step S240: constructing a two-dimensional grid coordinate system of the production line, synchronizing the target device coordinate set to the two-dimensional grid coordinate system for identification, and determining the production bottleneck location information.

[0048] Specifically, based on the production bottleneck prediction parameters, the position analysis of the robotic arms in the production equipment is carried out. By extracting information reflecting the bottleneck characteristics from the robotic arm operation data, such as decreased motion accuracy and insufficient load capacity, combined with its specific installation position and work station in the production line, a unique identifier is assigned to the work station where each robotic arm with a bottleneck problem is located, thereby determining the bottleneck work station identifier.

[0049] Based on the identified bottleneck workstation identifier, the corresponding robotic arm is precisely located. Detailed operational data for the robotic arm over the past period is then collected, including real-time operational errors of each joint, gripping accuracy values ​​accurate to the millimeter level, and motor load curves during different operating periods. Historical fault records are also compiled, covering the specific time of occurrence, fault type (such as joint jamming, positioning deviation), and frequency. Considering the role of the robotic arm in the production line, its importance score increases if it participates in demanding processes such as precision assembly. Next, a Failure Mode and Effects Analysis (FMEA) approach is used to quantify the risks of the industrial automation production line. The potential production interruption duration caused by a robotic arm failure is estimated in minutes; the product rework rate caused by the failure is calculated to the nearest percentage; and the scope of the cascading impact on downstream equipment is assessed, including the number and specific equipment affected. These indicators are assigned different weights: the production interruption duration is weighted 0.4, the rework rate is weighted 0.3, and the scope of the cascading impact is weighted 0.3. A weighted calculation is used to determine the risk assessment value. Finally, the system compares the pre-defined grading criteria. For example, levels 1-3 are low risk, indicating that the robot arm failure has a minimal impact on the production line and can be resolved within the regular maintenance cycle; levels 4-6 are medium risk, meaning that dedicated personnel must be assigned to closely monitor the situation and a maintenance plan must be developed in the near future; and levels 7-10 are high risk, indicating that the robot arm failure severely impacts production and requires immediate shutdown for repairs. By comparing the risk assessment values, the risk level of the bottleneck workstation where the robot arm is located is determined.

[0050] For the robotic arms in production equipment, a 3D laser scanner collects their spatial position data within the production workshop. This includes information such as the coordinates of each robotic arm's base, its range of motion, and its relative distance to surrounding equipment. This data is then used to construct a 3D equipment topology map that shows the spatial distribution and connectivity of the robotic arms. Bottleneck station identifiers are then associated with the corresponding robotic arms. Coordinate matching technology is used to locate the 3D spatial coordinates of the robotic arms within the topology map, ultimately integrating them into a target equipment coordinate set.

[0051] A two-dimensional grid coordinate system is constructed based on the physical layout of the production line. The origin is set at the production line starting point, and the horizontal and vertical axes are divided into grid units of fixed length. The three-dimensional coordinates in the target device coordinate set are mapped to this two-dimensional grid coordinate system using a projection transformation algorithm to obtain the corresponding two-dimensional grid coordinates. These coordinates are then highlighted in the coordinate system and the corresponding bottleneck station identifier and risk level are annotated. This visually displays the specific location distribution of production bottlenecks on the production line and determines the location information of production bottlenecks.

[0052] In one possible implementation, step S240 further includes: Step S241: performing production load analysis on the production equipment according to the target equipment coordinate set, and constructing a production load time series matrix.

[0053] Step S242: Perform cycle prediction on the production equipment based on the production load time series matrix, and extract N cycle load peaks, where N is an integer greater than 1.

[0054] Step S243: performing correlation analysis on the production equipment using a random forest algorithm according to the N periodic load peaks to obtain a fault signal-operation sequence correlation coefficient.

[0055] Step S244: mapping the fault signal-operation timing correlation coefficient to the two-dimensional grid coordinate system for identification, and determining the production bottleneck location information.

[0056] Specifically, the corresponding robotic arm is locked to the target device coordinate set. Sensors installed at each joint of the robotic arm collect real-time operational data, including the number of joint movements, load weight, motor output power, and pause duration every 10 minutes, to conduct production load analysis. The collected data is organized chronologically, with the robotic arm number as the row and the time node as the column, to construct a two-dimensional matrix. Each element in the matrix represents the ratio of the actual load to the maximum load of the robotic arm at the corresponding time point as a percentage, thus forming a production load time series matrix.

[0057] Based on the production load time series matrix, a time series cycle analysis algorithm is used to extract periodic features from the load data of production equipment. The duration of the production cycle is determined by identifying recurring patterns in load variations. The load data in the time series matrix is ​​then segmented according to the cycle duration. Within each cycle segment, an extreme value detection algorithm is used to locate the maximum load value, which is used as the load peak value for that cycle. Load peak values ​​corresponding to N cycles (N is an integer greater than 1) are continuously extracted to form a set of N cycle load peak values.

[0058] The extracted N-cycle load peaks are used as input feature variables for the random forest algorithm. Fault signal data from production equipment within the corresponding cycles (such as fault occurrence time, fault type code, and fault duration) are also collected as label variables. Correlation analysis is performed by constructing a random forest model consisting of multiple decision trees. During model training, each decision tree is independently trained based on randomly sampled sample data and feature subsets. The outputs of multiple trees are integrated through a voting mechanism to quantify the correlation between the temporal changes in the N-cycle load peaks and the fault signals. The final output is the fault signal-operation sequence correlation coefficient, which reflects the correlation between the two.

[0059] The obtained fault signal-operation sequence correlation coefficient is used as the mapping basis. It is mapped to the established production line's two-dimensional grid coordinate system according to pre-set mapping rules (such as the correspondence between the correlation coefficient value range and the color / size of the coordinate system identifier). By correlating the target device coordinates, the two-dimensional grid coordinates of the production equipment are concentrated. These equipment locations are visually identified in the coordinate system (e.g., high correlation coefficients are highlighted in red, medium correlation coefficients are highlighted in yellow). The density and intensity of the identifiers are then combined to determine the specific location of production bottlenecks.

[0060] In one possible implementation, step S200 further includes: Step S250: performing product demand analysis based on the production demand forecast parameters to determine a product demand sequence.

[0061] Step S260: grouping the demand categories according to the product demand sequence to obtain a plurality of demand sequence groups, performing demand confidence analysis based on the plurality of demand sequence groups to generate a plurality of demand confidence levels.

[0062] Step S270: setting a confidence interval boundary to determine the multiple demand confidences, extracting the demand confidences that do not exceed the confidence interval boundary to screen out the multiple demand sequence groups, and constructing a product dimension based on the screening results.

[0063] Step S280: constructing a time dimension based on the production line operation process data, and constructing a dual-axis line chart of the production line according to the time dimension and the product dimension.

[0064] Step S290: Perform interactive demand response according to the dual-axis line chart and draw the demand change trend chart.

[0065] Specifically, based on the production demand forecast parameters, various types of product demand information (such as product model, production quantity, delivery cycle, process standards, etc.) are disassembled and sorted, and the core demand items related to production scheduling are extracted. These demand items are arranged in order according to the sequence of the production process or the time nodes of the demand, forming a product demand sequence with time correlation and production logic, so as to clearly present the demand sequence and specific requirements of different products in the production process.

[0066] According to the attributes of the demand items in the product demand sequence (such as the product category, the process steps required for production, the delivery time window, etc.), the K-means clustering algorithm is used to group the demand categories, and the demand items with similar attributes are classified into the same group to obtain multiple demand sequence groups; then, for each demand sequence group, the deviation rate between its historical forecast data and actual production data, the completeness of data collection, the goodness of fit of the forecast model and other indicators are extracted, and the credibility of each group of demand forecast results is calculated through the Bayesian probability model to generate multiple corresponding demand confidence levels, each of which is represented by a value between 0 and 1 (the closer the value is to 1, the higher the confidence level).

[0067] According to the accuracy requirements of production demand forecasts and the fluctuation range of historical data, reasonable confidence interval boundaries are set (such as an 80%~95% confidence interval), and the demand confidence levels corresponding to multiple demand sequence groups are compared with the boundaries. Sequence groups whose demand confidence levels do not exceed the confidence interval boundaries are screened out and eliminated, and valid demand sequence groups whose confidence levels meet the requirements are retained. Then, based on the product types, specifications, models, custom attributes and other information involved in these valid sequence groups, an integrated product dimension covering all valid demands is constructed.

[0068] Production progress time nodes (such as hours, shifts, dates, etc.) are extracted from the production line operation process data, and the time dimension is constructed through timestamp alignment and serialization processing. It is used as the horizontal axis of the line chart and the specific time scale is marked. At the same time, the demand quantitative indicators corresponding to each product in the product dimension (such as planned production quantity and actual completion quantity) are used as the vertical axis data. A dual-axis design is adopted to correspond to the numerical range of different product categories. Through the data visualization tool Matplotlib, the associated data of the time dimension and the product dimension are converted into a line chart, where different colored lines correspond to different products. In this way, a dual-axis line chart of the production line is constructed that can intuitively reflect the changes in demand for each product over time.

[0069] Relying on a dual-axis line chart, the visualization tool ECharts is called through a data interaction interface to achieve interactive demand response. Users can drag and drop the time axis to zoom in and out to view data for a specific time period, and click on the product line to pop up the demand details at the corresponding time point, such as the demand quantity and process requirements for a specific product model. It also automatically identifies mutation points in the line chart where demand fluctuations exceed the preset threshold. For example, if the demand for a certain product increases by more than 30% within 24 hours, it will be marked with a special mark (such as a red triangle symbol) in the chart and associated with the trigger reason label (such as order addition, process adjustment). Based on these interactive analysis results, the demand lines of different products are integrated according to the time dimension, and data noise is eliminated through a smoothing algorithm. The trend line equation and fit parameters are supplemented to finally generate a demand change trend chart that includes trend direction arrows, key mutation point annotations, and product demand comparison legends.

[0070] In one possible implementation, step S300 further includes: Step S310: traverse the demand change trend graph to perform demand mutation analysis and extract a set of demand mutation time points.

[0071] Step S320: performing demand increase detection according to the demand mutation time point set to obtain demand increment data.

[0072] Step S330: traverse the production bottleneck location information to perform bottleneck risk analysis and determine a high-risk location coordinate set.

[0073] Step S340: Mapping the demand mutation time point set to the high-risk location coordinate set for spatiotemporal coupling, and calculating the bottleneck impact factor.

[0074] Step S350: Perform impact criticality analysis based on the incremental demand data and set an impact risk threshold. When the bottleneck impact factor exceeds the impact risk threshold, define the impact range information based on the bottleneck impact factor, set an early warning level according to the impact range information, and add the early warning level to the production bottleneck early warning information.

[0075] Specifically, we traverse the drawn demand change trend chart, analyze the broken lines of the time-varying demand of different products in the chart point by point, and by setting a reasonable mutation judgment threshold, such as when the demand difference between adjacent time nodes exceeds 2 times the historical average fluctuation value, identify the time points when abnormal demand fluctuations (including sudden increases or decreases), summarize and organize these time points that meet the mutation characteristics to form a demand mutation time point set, so as to accurately locate the specific moment when the demand changes significantly.

[0076] According to the set of demand mutation time points, the trend direction of the product demand data corresponding to each time point is judged, and the mutation time points where demand shows an upward trend are screened out. The difference in demand values ​​before and after the mutation is calculated for these time points, and combined with the demand growth rate per unit time, the demand increment data that can reflect the increase in demand and the growth rate are integrated to form.

[0077] Traverse the production bottleneck location information, extract the equipment operation data corresponding to each bottleneck location (such as the number of historical failures, current load rate, maintenance cycle, etc.), build a bottleneck risk assessment model through the hierarchical analysis method, quantify each data into risk indicators and assign weights, and calculate the comprehensive risk value of each bottleneck location; set a risk threshold (such as a comprehensive risk value greater than 0.7), filter out locations where the comprehensive risk value exceeds the threshold, and summarize the coordinates of these locations in the two-dimensional grid coordinate system to form a high-risk location coordinate set.

[0078] Each time point in the demand mutation time point set is converted into a timestamp format and mapped to the two-dimensional coordinates of each location in the high-risk location coordinate set. Using a spatiotemporal coupling algorithm, a time decay factor and spatial distance weight are introduced to calculate the strength of the correlation between the two. In the temporal dimension, a weight is assigned based on the time difference between the demand mutation and the current moment (the closer the time, the higher the weight). In the spatial dimension, an impact coefficient is set based on the distance between the high-risk location and the key production node (the closer the distance, the larger the coefficient). The time weight is multiplied by the spatial coefficient and then normalized to obtain a bottleneck impact factor that comprehensively reflects the degree of spatiotemporal correlation.

[0079] Statistical analysis methods are used to conduct an in-depth analysis of incremental demand data. Based on historical incremental demand data, the probability distribution is calculated and maximum likelihood estimation is employed to accurately determine the critical threshold at which incremental demand impacts the production line. This threshold is then used to establish an impact risk threshold. The calculated bottleneck impact factor is then carefully compared with this threshold. Once the bottleneck impact factor exceeds the threshold, a spatial analysis-based action process is initiated. The global Moran index is used to determine the approximate extent of the affected area. Kriging, a spatial interpolation algorithm, is then used to refine the impact area based on the bottleneck impact factor value, precisely defining the specific scope and area of ​​affected production equipment. This provides comprehensive and accurate information on the impact area. Next, the Analytic Hierarchy Process (AHP) is used to construct a judgment matrix encompassing multiple dimensions, including the size of the impact area, the number of key equipment involved, and the importance of production links. After a consistency check, the weights of each dimension are calculated, resulting in a scientific classification of warning levels: mild, moderate, and severe. Finally, the determined warning levels are embedded in specific fields within the production bottleneck warning information, enabling rapid and accurate assessment of the risk level of production bottlenecks. In one possible implementation, step S400 further includes: Step S410: executing the production scheduling strategy to control the operation of the industrial automation production line, monitoring and synchronously collecting data on the production equipment, and obtaining equipment response data.

[0080] Step S420: Execute the production scheduling strategy to control the operation of the industrial automation production line, monitor the production line and synchronously collect data to obtain production line output data.

[0081] Step S430: defining an instruction parameter vector based on the production scheduling strategy, performing correlation analysis on the instruction parameter vector and the equipment response data, and generating an instruction-response deviation component.

[0082] Step S440: performing analytical identification based on the command-response deviation component to determine a key deviation factor.

[0083] Step S450: Continuously optimizing the production scheduling strategy based on the key deviation factor and the production line output data to obtain a real-time production scheduling optimization strategy.

[0084] Specifically, production scheduling strategies are implemented to control the operation of industrial automation production lines. For the robotic arm, a production equipment, real-time parameters during its operation are synchronously collected through monitoring devices such as torque sensors installed at the joints of the robotic arm, force sensors of the end effector, and position encoders. These monitoring devices, such as the rotation angle of each joint, movement speed, output torque, gripping force of the end effector, and spatial position coordinates, are used to aggregate and process these data reflecting the execution of scheduling instructions by the robotic arm to obtain equipment response data.

[0085] In the process of implementing production scheduling strategies to control the operation of industrial automation production lines, by deploying visual sensors, counting devices and quality inspection equipment at key nodes of the production line (such as the end of the conveyor belt, finished product inspection area, etc.), the overall operation results of the production line are monitored and collected synchronously in real time, including data such as product output per unit time, the pass rate of each batch of products, the length of the production cycle and the material consumption rate. After summarizing and standardizing these data reflecting the output of the production line, the production line output data is obtained.

[0086] Based on the production scheduling strategy, key command parameters for equipment operation (such as the target motion angle, speed, and operation duration of the robot arm) are extracted, and these parameters are combined according to preset dimensions to construct a command parameter vector. The Pearson correlation coefficient algorithm is used to calculate the correlation between this vector and the equipment response data vector (such as the vector composed of the actual motion angle, speed, and operation duration of the robot arm). At the same time, the mean square error algorithm is used to calculate the numerical difference between the two in each dimension. The correlation result and the mean square error result are weightedly fused to generate a command-response deviation component that can quantify the degree of deviation between the command and the response.

[0087] The generated command-response deviation components are feature extracted, and the principal component analysis algorithm is used for dimensionality reduction to screen out the main components that contribute most to the deviation. The decision tree algorithm is then used to classify and identify these main components to determine which deviation components have the most significant impact on the production scheduling effect. These deviation components with significant impact are marked as key deviation factors, such as robot arm motion trajectory deviation, speed response delay, etc.

[0088] According to the key deviation factors combined with the production line output data, reinforcement learning is performed based on the key deviation factors and the production line output data to obtain data reinforcement learning results. According to the results, conflict detection is performed on the production scheduling strategy to obtain the production scheduling conflict parameters. The production scheduling conflict parameters are then injected into the digital twin sandbox to continuously verify the effectiveness of the production scheduling strategy. If the verification fails, the production scheduling strategy is conflict-replaced according to the production scheduling conflict parameters, thereby constructing a real-time production scheduling optimization strategy.

[0089] In one possible implementation, step S450 further includes: Step S451: Reinforcement learning is performed based on the key deviation factor and the production line output data to obtain a data reinforcement learning result.

[0090] Step S452: Conflict detection is performed on the production scheduling strategy according to the data reinforcement learning result to obtain production scheduling conflict parameters.

[0091] Step S453: Based on the production scheduling conflict parameters injected into the digital twin sandbox, the effectiveness of the production scheduling strategy is continuously verified. If the verification fails, the production scheduling strategy is conflict-replaced and scheduled according to the production scheduling conflict parameters to construct the real-time production scheduling optimization strategy.

[0092] Specifically, the deep reinforcement learning algorithm (DQN) is employed. Key deviation factors (such as device response delays and parameter execution errors) and production line output data (such as output per unit time and product qualification rate) are used as input features in the state space to construct a reinforcement learning model consisting of states, actions, and rewards. A reward function is designed with the goal of minimizing the key deviation factors and maximizing production line output data. A neural network is used to assess the value of states. The agent continuously updates its action strategy as it interacts with the production environment. After multiple rounds of iterative training, a data-driven reinforcement learning result is output, containing the optimal scheduling parameter adjustment plan.

[0093] Based on the results of data reinforcement learning, optimization suggestions for production scheduling are extracted, such as equipment operating parameter adjustment values, production process change plans, etc.; these suggestions are compared one by one with the various instructions in the current production scheduling strategy, including equipment start and stop times, robot arm operation paths, material distribution sequences, etc.; by setting conflict judgment criteria, such as parameter difference thresholds, timing overlap thresholds, etc., the inconsistencies between the two in key parameters and execution logic are identified, and these conflict points are presented in a quantitative form, such as parameter deviation amount, timing conflict duration, etc., to finally form production scheduling conflict parameters.

[0094] The production scheduling conflict parameters are input into the digital twin sandbox, which constructs a virtual environment consistent with the actual production scenario based on the physical model and real-time data of the production line, simulates the operation process of the production scheduling strategy under the influence of conflict parameters, and continuously verifies the effectiveness of the strategy by comparing the output data of the production line, equipment operating status, etc. in the virtual environment with the expected goals. When the verification results show that the strategy cannot meet the production requirements, the conflicting parts of the scheduling strategy are replaced and adjusted according to the type and degree of conflict indicated by the conflict parameters, such as reallocating equipment operation tasks, adjusting the production process sequence, etc., and finally constructing a real-time production scheduling optimization strategy that adapts to the current production status.

[0095] Example 2, based on the same inventive concept as the AI-driven industrial automation production line intelligent scheduling method in the above embodiment, Figure 2 As shown, this application provides an AI-driven intelligent scheduling system for industrial automation production lines. The system and method embodiments in the embodiments of this application are based on the same inventive concept. The system includes: The analysis and prediction module 10 is used to perform multi-source real-time acquisition based on the industrial automation production line, obtain a multi-source heterogeneous data set, integrate an AI algorithm to perform in-depth analysis and prediction on the multi-source heterogeneous data set, and identify a production dynamic prediction data set, which includes production bottleneck prediction parameters and production demand prediction parameters.

[0096] The demand change analysis module 20 is used to locate and identify production equipment according to the production bottleneck prediction parameters, determine the production bottleneck location information, perform demand change analysis according to the production demand prediction parameters, and draw a demand change trend chart.

[0097] The control and scheduling module 30 is used to analyze the production operation of the production equipment according to the demand change trend diagram in combination with the production bottleneck location information, generate production bottleneck warning information for control and scheduling, and formulate a production scheduling strategy.

[0098] The optimization response module 40 is used to execute the production scheduling strategy to perform scheduling monitoring on the industrial automation production line, generate monitoring feedback data to continuously optimize the production scheduling strategy, and obtain a real-time production scheduling optimization strategy.

[0099] Furthermore, the system is also used to implement the following functions: Perform parallel collection of production operations on industrial automation production lines to obtain multi-source heterogeneous data sets; perform spatiotemporal alignment processing based on the multi-source heterogeneous data sets to generate a standardized time series data stream; integrate AI algorithms to build an AI prediction model group, synchronize the standardized time series data stream to the AI ​​prediction model group to perform production operation prediction, and generate an initial prediction data set; build a digital twin verification layer based on the multi-source heterogeneous data set, perform data verification on the initial prediction data set through the digital twin verification layer, and generate data verification results; perform incremental learning on the AI ​​prediction model group according to the data verification results, perform priority screening and rearrangement on the initial prediction data set according to the learning results, and identify the production dynamic prediction data set.

[0100] Furthermore, the system is also used to implement the following functions: The standardized time series data stream is divided according to the production line operation process data to obtain the capacity time series data stream and the demand time series data stream; the spatiotemporal convolutional network is used as the algorithm architecture to pre-train the capacity of the production equipment and build an AI capacity prediction model; the multi-head self-attention mechanism is used as the algorithm architecture to pre-train the demand of the production line and build an AI demand prediction model; the capacity time series data stream is synchronized to the AI ​​capacity prediction model for loading training to obtain the initial parameters for production bottleneck prediction; the demand time series data stream is synchronized to the AI ​​demand prediction model for supply training to obtain the initial parameters for production demand prediction; the initial parameters for production bottleneck prediction are integrated with the initial parameters for production demand prediction to build an initial prediction data set.

[0101] Furthermore, the system is also used to implement the following functions: Based on the production bottleneck prediction parameters, the production equipment is positionally analyzed to determine the bottleneck station identifier; according to the bottleneck station identifier combined with the production equipment, the industrial automation production line is risk analyzed to determine the risk level value; based on the production equipment, a three-dimensional topological analysis is performed to construct an equipment topology map, the bottleneck station identifier is synchronized to the equipment topology map for positioning, and the target equipment coordinate set is determined; a two-dimensional grid coordinate system of the production line is constructed, the target equipment coordinate set is synchronized to the two-dimensional grid coordinate system for identification, and the production bottleneck location information is determined.

[0102] Furthermore, the system is also used to implement the following functions: Perform production load analysis on the production equipment based on the target equipment coordinate set and construct a production load timing matrix; perform cycle prediction on the production equipment based on the production load timing matrix and extract N cycle load peaks, where N is an integer greater than 1; perform correlation analysis on the production equipment using a random forest algorithm according to the N cycle load peaks to obtain a fault signal-operation timing correlation coefficient; map the fault signal-operation timing correlation coefficient to the two-dimensional grid coordinate system for identification, and determine the production bottleneck location information.

[0103] Furthermore, the system is also used to implement the following functions: Perform product demand analysis based on the production demand forecast parameters to determine a product demand sequence; group demand categories according to the product demand sequence to obtain multiple demand sequence groups; perform demand confidence analysis based on the multiple demand sequence groups to generate multiple demand confidence levels; set confidence interval boundaries to judge the multiple demand confidence levels, extract demand confidence levels that do not exceed the confidence interval boundaries to screen out the multiple demand sequence groups, and construct a product dimension based on the screening results; construct a time dimension based on the production line operation process data, and construct a dual-axis line chart of the production line according to the time dimension and the product dimension; perform demand interaction response based on the dual-axis line chart, and draw the demand change trend chart.

[0104] Furthermore, the system is also used to implement the following functions: Traverse the demand change trend graph to perform demand mutation analysis and extract the demand mutation time point set; perform demand increase detection according to the demand mutation time point set to obtain demand increment data; traverse the production bottleneck location information to perform bottleneck risk analysis and determine the high-risk location coordinate set; map the demand mutation time point set to the high-risk location coordinate set for spatiotemporal coupling and calculate the bottleneck impact factor; perform impact criticality analysis based on the demand increment data and set an impact risk threshold; when the bottleneck impact factor exceeds the impact risk threshold, define the impact range information based on the bottleneck impact factor, set a warning level according to the impact range information, and add the warning level to the production bottleneck warning information.

[0105] Furthermore, the system is also used to implement the following functions: Execute the production scheduling strategy to control the operation of the industrial automation production line, monitor the production equipment and synchronously collect data to obtain equipment response data; execute the production scheduling strategy to control the operation of the industrial automation production line, monitor the production line and synchronously collect data to obtain production line output data; define an instruction parameter vector based on the production scheduling strategy, correlate and analyze the instruction parameter vector with the equipment response data to generate an instruction-response deviation component; analyze and identify the instruction-response deviation component to determine the key deviation factor; continuously optimize the production scheduling strategy in response to the key deviation factor in combination with the production line output data to obtain a real-time production scheduling optimization strategy.

[0106] Furthermore, the system is also used to implement the following functions: Reinforcement learning is performed based on the key deviation factors in combination with the production line output data to obtain data reinforcement learning results; conflict detection is performed on the production scheduling strategy according to the data reinforcement learning results to obtain production scheduling conflict parameters; the effectiveness of the production scheduling strategy is continuously verified by injecting the production scheduling conflict parameters into the digital twin sandbox; when the verification fails, conflict replacement scheduling is performed on the production scheduling strategy according to the production scheduling conflict parameters to construct the real-time production scheduling optimization strategy.

[0107] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0108] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

[0109] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. AI-driven intelligent scheduling method for industrial automation production lines, characterized by: The method comprises: Based on the industrial automation production line, multi-source real-time acquisition is performed to obtain a multi-source heterogeneous data set, and the AI ​​algorithm is integrated to perform in-depth analysis and prediction on the multi-source heterogeneous data set to identify the production dynamic prediction data set, which includes production bottleneck prediction parameters and production demand prediction parameters; Positioning and marking the production equipment according to the production bottleneck prediction parameters, determining the production bottleneck location information, performing demand change analysis according to the production demand prediction parameters, and drawing a demand change trend chart; Analyze the production operation of production equipment based on the demand change trend diagram and the production bottleneck location information, generate production bottleneck warning information for control scheduling, and formulate production scheduling strategies; The production scheduling strategy is executed to perform scheduling monitoring on the industrial automation production line, and monitoring feedback data is generated to continuously optimize the production scheduling strategy to obtain a real-time production scheduling optimization strategy.

2. The AI-driven intelligent scheduling method for industrial automation production lines according to claim 1, characterized in that: Based on the industrial automation production line, multi-source real-time acquisition is performed to obtain a multi-source heterogeneous data set, and an AI algorithm is integrated to perform in-depth analysis and prediction on the multi-source heterogeneous data set to identify the production dynamic prediction data set. The method includes: Parallel collection of production operations on industrial automation production lines to obtain multi-source heterogeneous data sets; Performing spatiotemporal alignment processing based on the multi-source heterogeneous data sets to generate a standardized time series data stream; Integrate AI algorithms to build an AI prediction model group, synchronize the standardized time series data stream to the AI ​​prediction model group to perform production operation prediction, and generate an initial prediction data set; Building a digital twin verification layer based on the multi-source heterogeneous data set, performing data verification on the initial prediction data set through the digital twin verification layer, and generating a data verification result; Incremental learning is performed on the AI ​​prediction model group according to the data verification results, and priority screening and rearrangement of the initial prediction data set are performed based on the learning results to identify the production dynamic prediction data set.

3. The AI-driven intelligent scheduling method for industrial automation production lines according to claim 2, characterized in that: Integrating AI algorithms to build an AI prediction model group, synchronizing the standardized time series data stream to the AI ​​prediction model group for production operation prediction, and generating an initial prediction data set, the method includes: According to the standardized time series data flow, the production line operation process data is divided to obtain the production capacity time series data flow and the demand time series data flow; Use spatiotemporal convolutional networks as the algorithm architecture to pre-train production equipment capacity and build an AI capacity prediction model; Use the multi-head self-attention mechanism as the algorithm architecture to pre-train production line demand and build an AI demand forecasting model; Synchronize the production capacity time series data stream to the AI ​​production capacity prediction model for loading training to obtain initial parameters for production bottleneck prediction; Synchronize the demand time series data stream to the AI ​​demand forecasting model for supply training to obtain initial parameters for production demand forecasting; The initial parameters for production bottleneck prediction are integrated with the initial parameters for production demand prediction to construct an initial prediction data set.

4. The AI-driven intelligent scheduling method for industrial automation production lines according to claim 1, characterized in that: Positioning and identifying production equipment according to the production bottleneck prediction parameters to determine production bottleneck location information, the method includes: Performing position analysis on production equipment based on the production bottleneck prediction parameters to determine a bottleneck workstation identifier; Perform risk analysis on the industrial automation production line according to the bottleneck workstation identifier and production equipment to determine a risk level value; Performing a three-dimensional topological analysis based on the production equipment to construct an equipment topology map, synchronizing the bottleneck station identifier to the equipment topology map for positioning, and determining a target equipment coordinate set; A two-dimensional grid coordinate system of the production line is constructed, the target device coordinate set is synchronized to the two-dimensional grid coordinate system for identification, and the production bottleneck location information is determined.

5. The AI-driven intelligent scheduling method for industrial automation production lines according to claim 4, characterized in that: Synchronizing the target device coordinate set to the two-dimensional grid coordinate system for identification, and determining the production bottleneck location information, the method includes: Performing production load analysis on the production equipment according to the target equipment coordinate set, and constructing a production load time series matrix; Performing cycle prediction on production equipment based on the production load time series matrix, extracting N cycle load peaks, where N is an integer greater than 1; Performing correlation analysis on the production equipment using a random forest algorithm according to the N cycle load peaks to obtain a fault signal-operation timing correlation coefficient; The fault signal-operation timing correlation coefficient is mapped to the two-dimensional grid coordinate system for identification, and the production bottleneck location information is determined.

6. The AI-driven intelligent scheduling method for industrial automation production lines according to claim 3, characterized in that: Performing demand change analysis according to the production demand forecast parameters and drawing a demand change trend chart, the method includes: Perform product demand analysis based on the production demand forecast parameters to determine the product demand sequence; Grouping the demand categories according to the product demand sequence to obtain multiple demand sequence groups, performing demand confidence analysis based on the multiple demand sequence groups to generate multiple demand confidence levels; Setting a confidence interval boundary to determine the multiple demand confidences, extracting demand confidences that do not exceed the confidence interval boundary to screen out the multiple demand sequence groups, and constructing a product dimension based on the screening results; Constructing a time dimension based on the production line operation process data, and constructing a dual-axis line chart of the production line according to the time dimension and the product dimension; Demand interaction response is performed according to the dual-axis line graph, and the demand change trend graph is drawn.

7. The AI-driven intelligent scheduling method for industrial automation production lines according to claim 1, characterized in that: The method includes: performing a production operation analysis on the production equipment according to the demand change trend graph and the production bottleneck location information to generate production bottleneck warning information; Traversing the demand change trend graph to perform demand mutation analysis and extract a set of demand mutation time points; Perform demand increase detection according to the demand mutation time point set to obtain demand increment data; Traversing the production bottleneck location information to perform bottleneck risk analysis and determine a high-risk location coordinate set; Mapping the demand mutation time point set to the high-risk location coordinate set for spatiotemporal coupling, and calculating the bottleneck impact factor; An impact criticality analysis is performed based on the incremental demand data, and an impact risk threshold is set. When the bottleneck impact factor exceeds the impact risk threshold, the impact range information is defined based on the bottleneck impact factor, and an early warning level is set according to the impact range information. The early warning level is added to the production bottleneck early warning information.

8. The AI-driven intelligent scheduling method for industrial automation production lines according to claim 1, characterized in that: Executing the production scheduling strategy to monitor the industrial automation production line, generating monitoring feedback data to continuously optimize the production scheduling strategy, and obtaining a real-time production scheduling optimization strategy, the method includes: Execute the production scheduling strategy to control the operation of the industrial automation production line, monitor and collect synchronous data on the production equipment, and obtain equipment response data; Execute the production scheduling strategy to control the operation of the industrial automation production line, monitor the production line and collect data synchronously to obtain the production line output data; defining an instruction parameter vector based on the production scheduling strategy, performing correlation analysis on the instruction parameter vector and the device response data to generate an instruction-response deviation component; Analyze and identify the command-response deviation component to determine a key deviation factor; The production scheduling strategy is continuously optimized and responded to according to the key deviation factors in combination with the production line output data to obtain a real-time production scheduling optimization strategy.

9. The AI-driven intelligent scheduling method for industrial automation production lines according to claim 8, characterized in that: The production scheduling strategy is continuously optimized and responded to based on the key deviation factor and the production line output data to obtain a real-time production scheduling optimization strategy, the method comprising: Performing reinforcement learning based on the key deviation factor and the production line output data to obtain a data reinforcement learning result; Performing conflict detection on the production scheduling strategy according to the data reinforcement learning result to obtain production scheduling conflict parameters; The effectiveness of the production scheduling strategy is continuously verified based on the injection of the production scheduling conflict parameters into the digital twin sandbox. If the verification fails, the production scheduling strategy is conflict-replaced and scheduled according to the production scheduling conflict parameters to construct the real-time production scheduling optimization strategy.

10. AI-driven intelligent scheduling system for industrial automation production lines, characterized by: The system is used to implement the AI-driven intelligent scheduling method for industrial automation production lines according to any one of claims 1 to 9, and the system includes: An analysis and prediction module is used to perform multi-source real-time acquisition based on industrial automated production lines to obtain multi-source heterogeneous data sets, integrate AI algorithms to perform in-depth analysis and prediction on the multi-source heterogeneous data sets, and identify production dynamic prediction data sets, which include production bottleneck prediction parameters and production demand prediction parameters; a demand change analysis module, configured to locate and identify production equipment according to the production bottleneck prediction parameters, determine production bottleneck location information, perform demand change analysis according to the production demand prediction parameters, and draw a demand change trend graph; A control and scheduling module is used to analyze the production operation of production equipment based on the demand change trend diagram and the production bottleneck location information, generate production bottleneck warning information for control and scheduling, and formulate production scheduling strategies; The optimization response module is used to execute the production scheduling strategy to schedule and monitor the industrial automation production line, generate monitoring feedback data to continuously optimize the production scheduling strategy, and obtain a real-time production scheduling optimization strategy.

Images