Intelligent manufacturing production line collaborative control system based on artificial intelligence

Abstract

This invention relates to the field of intelligent manufacturing technology and discloses an artificial intelligence-based intelligent manufacturing production line collaborative control system, including a multi-source data sensing module, a digital twin virtual-real mapping module, a multi-agent collaborative decision-making module, and a human-machine interaction module. The multi-source data sensing module collects and preprocesses multi-source heterogeneous data across the entire production line, constructing a real-time database. The digital twin virtual-real mapping module builds a 1:1 high-precision digital twin model of the production line based on real-time data, enabling bidirectional real-time data interaction between the virtual and real worlds. It also determines the disturbance level of the production line by calculating multi-dimensional disturbance coefficients and introducing a time decay function to accumulate the effective disturbance influence coefficient. The multi-agent collaborative decision-making module matches and executes hierarchical collaborative control strategies according to the disturbance level, generating a comprehensive collaborative control scheme. This achieves accurate quantification and hierarchical collaborative response to production disturbances, significantly improving the adaptive capability and operational stability of the intelligent manufacturing production line.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology, and more specifically to a collaborative control system for intelligent manufacturing production lines based on artificial intelligence. Background Technology

[0002] With the deepening of intelligent manufacturing, modern production lines are developing towards high automation, flexibility, and intelligence. However, in actual production processes, production lines often face numerous technical challenges, such as the difficulty in integrating multi-source heterogeneous data, lagging disturbance state identification, and the disconnect between control decisions and production status. Traditional manufacturing execution systems mainly rely on preset fixed rules and human experience for production management, making it difficult to cope with complex and ever-changing disturbance scenarios.

[0003] Currently, some intelligent manufacturing control systems are disclosed in existing technologies, such as those that collect production data by deploying sensors, construct virtual factories using digital twin technology, or employ multi-agent task scheduling. However, these technical solutions have the following shortcomings: First, existing control strategies are mostly based on single threshold triggering mechanisms, failing to comprehensively consider the continuous impact of historical disturbances, and are prone to misjudgment or omission due to instantaneous fluctuations; Second, there is a lack of clear causal mapping between the decision-making module and the state perception module, and the selection of control strategies is disconnected from the actual disturbance level of the production line, resulting in system response lag and low collaborative efficiency.

[0004] Therefore, this invention proposes a collaborative control system for intelligent manufacturing production lines based on artificial intelligence. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a collaborative control system for intelligent manufacturing production lines based on artificial intelligence.

[0006] To achieve the above-mentioned objectives, the present invention provides the following technical solution: The AI-based intelligent manufacturing production line collaborative control system includes a multi-source data sensing module, a digital twin virtual-real mapping module, a multi-agent collaborative decision-making module, and a human-computer interaction module. The multi-source data sensing module is used to collect multi-source heterogeneous data across the entire production line, and to perform outlier detection, missing value filling and normalization processing on the data to build a real-time database across the entire region. The digital twin virtual-real mapping module is connected to the multi-source data sensing module. It is used to construct a 1:1 high-precision digital twin model of the production line based on the full-domain real-time database, realize two-way real-time data interaction between the actual production line and the virtual model, calculate the comprehensive disturbance impact coefficient at the current moment based on the full-domain multi-source heterogeneous data, introduce a time decay function to accumulate the historical disturbance impact coefficient, obtain the effective disturbance impact coefficient, and determine the disturbance level of the current production line based on the effective disturbance impact coefficient. The multi-agent collaborative decision-making module is connected to the digital twin virtual-real mapping module and is used to match and execute the corresponding collaborative control strategy according to the disturbance level to generate a production line-wide collaborative control scheme. The human-computer interaction module is connected to the multi-agent collaborative decision-making module and is used to display the production line operation status and control scheme in real time, and supports remote human intervention and parameter setting.

[0007] As a further description of the technical solution of the present invention, the sensing devices deployed in the multi-source data sensing module include industrial sensors, PLC controllers, machine vision modules, RFID logistics identification units, AGV positioning modules, energy consumption monitoring instruments, and environmental monitoring equipment; the collected multi-source heterogeneous data includes equipment operation data, production process data, logistics flow data, quality inspection data, order and scheduling data, energy consumption and cost data, and environmental data.

[0008] As a further description of the technical solution of the present invention, the outlier detection adopts an outlier detection algorithm based on the 3σ criterion to identify and remove outliers from the collected continuous data. The missing value imputation uses an LSTM neural network-based missing value imputation algorithm, which utilizes the correlation of time series data to accurately imput missing values ​​and avoid data deviation caused by missing values. The data normalization uses the min-max normalization algorithm to map all preprocessed data to the [0,1] interval, thereby achieving data standardization.

[0009] As a further description of the technical solution of the present invention, the digital twin model adopts a layered modeling approach, which is divided into four levels: geometric model, physical model, behavioral model, and rule model; the geometric accuracy of the geometric model is consistent with that of the actual production line, with an error of ≤0.1mm; the physical model is constructed based on multibody dynamics, thermodynamics, and fluid mechanics theory, using the finite element analysis method; the replication delay of the virtual-real two-way data interaction is ≤200ms.

[0010] As a further description of the technical solution of the present invention, the disturbance judgment process includes: Step S1: Collect multi-dimensional operating data of the production line, including equipment operating data, production process data, logistics flow data, quality inspection data, order and scheduling data, energy consumption and cost data, and environmental data; Step S2: Calculate the disturbance sub-coefficients for each dimension, including equipment operation disturbance sub-coefficient, process disturbance sub-coefficient, logistics disturbance sub-coefficient, quality disturbance sub-coefficient, order disturbance sub-coefficient, energy consumption cost disturbance sub-coefficient, and environmental disturbance sub-coefficient. Step S3: Sum the perturbation coefficients of each dimension according to the preset dimension weights to obtain the comprehensive perturbation influence coefficient at the current moment; Step S4: Introduce a time decay function to accumulate the disturbance influence coefficients at historical moments to obtain the effective disturbance influence coefficient; Step S5: Compare the effective disturbance impact coefficient with the preset early warning threshold and emergency threshold to determine the disturbance level of the current production line.

[0011] As a further description of the technical solution of the present invention, the calculation formula for the effective disturbance influence coefficient in step S4 is as follows: In the formula, The length of the historical time window. For time offset index, The attenuation coefficient set for the system. The comprehensive disturbance impact coefficient at the current moment. This is the effective disturbance influence coefficient.

[0012] As a further description of the technical solution of the present invention, the specific working process of step S5 includes: when If the disturbance level of the current production line is determined to be Level 1, the production line is operating normally and no intervention is required; when At this time, the disturbance level of the current production line is determined to be level two, indicating that there is a slight abnormality in the production line. It is recommended to pay attention and record it. when When the disturbance level of the current production line is determined to be level three, the production line has obvious abnormalities, triggering an early warning and recommending manual intervention; when When the disturbance level of the current production line is determined to be level four, the production line has a serious abnormality, an alarm is triggered and immediate action is required; when When the disturbance level of the current production line is determined to be Level 5, the production line faces the risk of shutdown, triggering the emergency response mechanism.

[0013] As a further description of the technical solution of the present invention, the disturbance judgment process also includes trend analysis and preventive control: Calculate the rate of change of the perturbation coefficients in each dimension. ,in, Let be the coefficient of the i-th perturbation type; When the rate of change of the perturbation coefficient in any dimension When the value is greater than A, it is determined that the disturbance in this dimension is showing a worsening trend, and preventive control measures are implemented in advance. When the rate of change of the perturbation coefficient in any dimension When the value is less than B, it is determined that the disturbance in this dimension is showing an improving trend, and control resources are gradually released. Where A>0, B<0, and A and B are preset thresholds.

[0014] As a further description of the technical solution of the present invention, the working process of the multi-agent collaborative decision-making module includes: When the disturbance level is Level 1, execute the normal monitoring strategy, maintain the existing control strategy, and record the operating data. When the disturbance level is level two, a local adjustment strategy is implemented to adjust the parameters of the abnormal local components and strengthen monitoring. When the disturbance level is level three, a dynamic rescheduling strategy is executed to reallocate production tasks and activate standby resources. When the disturbance level is level four, an emergency response strategy is implemented, which involves local line shutdown in the fault area and initiation of emergency dispatch. When the disturbance level is level 5, the disaster recovery strategy is implemented, including a plant-wide emergency shutdown and the activation of the disaster recovery plan.

[0015] The beneficial effects of this invention are: 1. This invention embeds a disturbance quantification algorithm into a digital twin model. By calculating disturbance sub-coefficients in seven dimensions and introducing a time decay function to accumulate historical disturbances, an effective disturbance influence coefficient is obtained, enabling dynamic perception and structured output of complex disturbance states on the production line. This mechanism considers both the current instantaneous disturbance and the persistent impact of historical disturbances, avoiding misjudgments caused by instantaneous fluctuations and significantly improving the accuracy and robustness of disturbance assessment.

[0016] 2. This invention significantly shortens the time delay from disturbance occurrence to control response through an integrated architecture of perception, mapping, decision-making and execution. It realizes the collaborative optimization of multiple elements such as equipment, logistics, quality and orders, effectively reduces the risk of production interruption caused by disturbances, improves the overall equipment efficiency of the production line, product quality stability and order delivery capability, and provides reliable technical support for intelligent manufacturing. Attached Figure Description

[0017] The present invention will now be further described with reference to the accompanying drawings.

[0018] Figure 1 This is a partial structural schematic diagram of the intelligent manufacturing production line collaborative control system based on artificial intelligence provided by the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 As shown, the AI-based intelligent manufacturing production line collaborative control system includes a multi-source data sensing module, a digital twin virtual-real mapping module, a multi-agent collaborative decision-making module, and a human-computer interaction module. The multi-source data sensing module is used to collect multi-source heterogeneous data across the entire production line, and to perform outlier detection, missing value filling and normalization processing on the data to build a real-time database across the entire region. The sensing devices deployed in the multi-source data sensing module include industrial sensors, PLC controllers, machine vision modules, RFID logistics identification units, AGV positioning modules, energy consumption monitoring instruments, and environmental monitoring equipment; the collected multi-source heterogeneous data includes equipment operation data, production process data, logistics flow data, quality inspection data, order and scheduling data, energy consumption and cost data, and environmental data.

[0021] The outlier detection employs an outlier detection algorithm based on the 3σ criterion to identify and remove outliers from the collected continuous data. The missing value imputation uses an LSTM neural network-based missing value imputation algorithm, which utilizes the correlation of time series data to accurately imput missing values ​​and avoid data deviation caused by missing values. The data normalization uses the min-max normalization algorithm to map all preprocessed data to the [0,1] interval, thereby achieving data standardization.

[0022] After the above preprocessing, the module builds a real-time database covering the entire production line, providing a unified, reliable, and real-time data foundation for the subsequent construction of digital twin models, disturbance quantification analysis, and collaborative decision-making. It serves as the data entry point for the entire system to achieve precise perception and closed-loop control.

[0023] The digital twin virtual-real mapping module is connected to the multi-source data sensing module. It is used to construct a 1:1 high-precision digital twin model of the production line based on the full-domain real-time database, realize two-way real-time data interaction between the actual production line and the virtual model, calculate the comprehensive disturbance impact coefficient at the current moment based on the full-domain multi-source heterogeneous data, introduce a time decay function to accumulate the historical disturbance impact coefficient, obtain the effective disturbance impact coefficient, and determine the disturbance level of the current production line based on the effective disturbance impact coefficient. The digital twin model adopts a hierarchical modeling approach, divided into four levels: geometric model, physical model, behavioral model, and rule model. The geometric accuracy of the geometric model is consistent with that of the actual production line, with an error of ≤0.1mm. The physical model is constructed using the finite element analysis method based on multibody dynamics, thermodynamics, and fluid mechanics theories. The replication latency of the virtual-real two-way data interaction is ≤200ms.

[0024] The disturbance determination process includes: Step S1: Collect multi-dimensional operating data of the production line, including equipment operating data, production process data, logistics flow data, quality inspection data, order and scheduling data, energy consumption and cost data, and environmental data; Step S2: Calculate the disturbance sub-coefficients for each dimension, including equipment operation disturbance sub-coefficient, process disturbance sub-coefficient, logistics disturbance sub-coefficient, quality disturbance sub-coefficient, order disturbance sub-coefficient, energy consumption cost disturbance sub-coefficient, and environmental disturbance sub-coefficient. Equipment operation disturbance coefficient The calculation formula is: ; In the formula, The total number of devices. For device index, j belongs to , , and For device dimension sub-weights, Let j be the standard mean time between failures (MTBF). Let j be the actual mean time between failures (MTBF) of device j. Let j be the standard comprehensive equipment efficiency. Let j be the actual overall equipment efficiency. These are the actual key operating parameters of device j. These are the standard key operating parameters for device j. Here are the maximum permissible critical operating parameters for device j. Let be the minimum allowable critical operating parameters for equipment j, where the overall equipment efficiency is the ratio of the actual output speed to the designed output speed of the equipment.

[0025] Process disturbance coefficient The calculation formula is: ; In the formula, This represents the total number of process parameters. For process parameter index, belong , , and For the process dimension sub-weight, This is the actual temperature value. To set the temperature value, The temperature tolerance set for the system. This is the actual pressure value. To set the pressure value, The pressure tolerance set for the system. The alarm indication coefficient is the factor used when the process... Alarm triggered =1, when the process No alarm occurred. =0.

[0026] Logistics disturbance coefficient The calculation formula is: ; In the formula, This represents the total number of logistics links. For the index of logistics links, belong , This is the actual delay time. For standard beat time, This is the current queue length. The maximum allowed queue length, This refers to the actual logistics distance. To plan logistics distance, , and For the logistics dimension, sub-weights.

[0027] Mass perturbation coefficient The calculation formula is: ; In the formula, The number of defective products. This represents the total number of tests. The number of products requiring rework. Where m is the number of quality characteristic parameters, and m is the index of the quality characteristic. For the m-th quality characteristic, For the m-th quality characteristic specification value, For the m-th quality characteristic, the allowable deviation tolerance is... , and For quality dimension sub-weights.

[0028] Order Disturbance Coefficient The calculation formula is: ; In the formula, To increase the quantity of urgent orders, Total number of orders For changes in order quantity, To plan order volume, Due to order delays, For order delivery deadline, In order to deliver the order number on time, , , and Sub-weights for the order dimension.

[0029] Energy consumption cost disturbance coefficient The calculation formula is: ; In the formula, This represents the actual energy consumption per unit of product. Energy consumption per standard unit of product. This represents the actual unit product cost. Cost per unit of product Additional penalty costs, which are extra expenses exceeding normal production costs. , and These are the sub-weights of the energy consumption cost dimension.

[0030] Environmental disturbance coefficient The calculation formula is: ; In the formula, For ambient temperature, For ambient humidity, and For the best working environment, and For the maximum permissible deviation, Dust concentration, For noise level, and This is the maximum allowed value. , , and These are the sub-weights of the environment dimension.

[0031] Step S3: Sum the perturbation coefficients of each dimension according to the preset dimension weights to obtain the comprehensive perturbation influence coefficient at the current moment; Step S4: Introduce a time decay function to accumulate the disturbance influence coefficients at historical moments to obtain the effective disturbance influence coefficient; Step S5: Compare the effective disturbance impact coefficient with the preset early warning threshold and emergency threshold to determine the disturbance level of the current production line.

[0032] The formula for calculating the effective disturbance influence coefficient in step S4 is as follows: In the formula, The length of the historical time window. For time offset index, The attenuation coefficient set for the system. The comprehensive disturbance impact coefficient at the current moment. This is the effective disturbance influence coefficient.

[0033] The specific working process of step S5 includes: when If the disturbance level of the current production line is determined to be Level 1, the production line is operating normally and no intervention is required; when At this time, the disturbance level of the current production line is determined to be level two, indicating that there is a slight abnormality in the production line. It is recommended to pay attention and record it. when When the disturbance level of the current production line is determined to be level three, the production line has obvious abnormalities, triggering an early warning and recommending manual intervention; when When the disturbance level of the current production line is determined to be level four, the production line has a serious abnormality, an alarm is triggered and immediate action is required; when When the disturbance level of the current production line is determined to be Level 5, the production line faces the risk of shutdown, triggering the emergency response mechanism.

[0034] The disturbance assessment process also includes trend analysis and preventive control: Calculate the rate of change of the perturbation coefficients in each dimension. ,in, Let be the coefficient of the i-th perturbation type; When the rate of change of the perturbation coefficient in any dimension When the value is greater than A, it is determined that the disturbance in this dimension is showing a worsening trend, and preventive control measures are implemented in advance. When the rate of change of the perturbation coefficient in any dimension When the value is less than B, it is determined that the disturbance in this dimension is showing an improving trend, and control resources are gradually released. Where A>0, B<0, and A and B are preset thresholds.

[0035] The multi-agent collaborative decision-making module is connected to the digital twin virtual-real mapping module and is used to match and execute the corresponding collaborative control strategy according to the disturbance level to generate a production line-wide collaborative control scheme. The aforementioned technical solution aims to transform the complex state of the production line into quantifiable disturbance levels, providing precise input for subsequent collaborative decision-making. This process first collects multi-dimensional operational data across seven categories: equipment operation, production process, logistics flow, quality inspection, order scheduling, energy consumption costs, and environment. Then, it calculates disturbance sub-coefficients for each dimension: the equipment operation disturbance sub-coefficient is evaluated based on the deviation from mean time between failures (MTBF) and overall equipment efficiency; the process disturbance sub-coefficient is quantified by the degree of deviation from tolerances for key parameters such as temperature and pressure, and alarm status; the logistics disturbance sub-coefficient considers deviations in delay time, queue length, and logistics distance; the quality disturbance sub-coefficient considers the overall defect rate, quality characteristic deviations, and rework rate; the order disturbance sub-coefficient assesses the increase in urgent orders, changes in order volume, and delivery delays; the energy consumption cost disturbance sub-coefficient measures unit product energy consumption, cost overruns, and additional penalty costs; and the environmental disturbance sub-coefficient considers environmental factors such as temperature, humidity, dust concentration, and noise levels. Subsequently, the disturbance sub-coefficients are weighted and summed according to preset dimensional weights to obtain the overall disturbance impact coefficient at the current moment. To further improve the accuracy of the assessment, a time decay function is introduced into the process. The disturbance impact coefficients at each moment within the historical time window are accumulated according to an exponential decay law to form the effective disturbance impact coefficient. The calculation formula is as follows: This mechanism enables the system to reflect both the current instantaneous disturbance state and the persistent impact of historical disturbances, avoiding misjudgments caused by instantaneous fluctuations. Finally, the effective disturbance impact coefficient is compared with preset multi-level thresholds to classify five disturbance levels: Level 1 is normal operation requiring no intervention; Level 2 is minor anomaly requiring attention and recording; Level 3 is significant anomaly triggering an early warning and recommending manual intervention; Level 4 is severe anomaly triggering an alarm and requiring immediate handling; and Level 5 is a risk of line stoppage triggering an emergency response mechanism. Furthermore, the process includes trend analysis functionality. By calculating the rate of change of disturbance sub-coefficients in each dimension, preventative control measures are implemented in advance when a disturbance in a certain dimension shows a worsening trend, and control resources are gradually released when it shows an improving trend, thereby achieving dynamic perception, precise quantification, and proactive early warning of the production line status.

[0036] The working process of the multi-agent collaborative decision-making module includes: When the disturbance level is Level 1, execute the normal monitoring strategy, maintain the existing control strategy, and record the operating data. When the disturbance level is level two, a local adjustment strategy is implemented to adjust the parameters of the abnormal local components and strengthen monitoring. When the disturbance level is level three, a dynamic rescheduling strategy is executed to reallocate production tasks and activate standby resources. When the disturbance level is level four, an emergency response strategy is implemented, which involves local line shutdown in the fault area and initiation of emergency dispatch. When the disturbance level is level 5, the disaster recovery strategy is implemented, including a plant-wide emergency shutdown and the activation of the disaster recovery plan.

[0037] The human-computer interaction module is connected to the multi-agent collaborative decision-making module and is used to display the production line operation status and control scheme in real time, and supports remote human intervention and parameter setting.

[0038] In summary, the working principle of this invention is as follows: A multi-source data sensing module constructs a sensing network covering all elements of the production line, collecting multi-dimensional heterogeneous data such as equipment operation, production process, logistics flow, quality inspection, order scheduling, energy consumption costs, and environmental data. The data is preprocessed using outlier detection based on the 3σ criterion, missing value imputation via LSTM neural network, and min-max normalization algorithms to construct a full-domain real-time database. Based on this, a digital twin virtual-real mapping module utilizes the preprocessed data to build a 1:1 high-precision production line with a geometric accuracy error not exceeding 0.1mm, layered and encompassing geometric models, physical models, behavioral models, and rule models. The digital twin model achieves synchronous mapping between the physical production line and the virtual model through two-way real-time data interaction, with replication latency controlled within 200ms. It also performs quantitative analysis of disturbances in the production line's operating status. Specifically, this module calculates disturbance sub-coefficients for seven dimensions: equipment operation, process, logistics, quality, orders, energy consumption costs, and environment. Each sub-coefficient measures disturbances from the perspectives of equipment mean time between failures (MTBF) and overall equipment efficiency deviation, process parameter deviation tolerance, logistics delays and queue backlogs, product defect rate and quality characteristic deviations, order changes and delivery delays, unit product energy consumption and cost exceeding limits, and environmental temperature, humidity, dust, and noise exceeding standards. The system calculates the degree of anomaly in each dimension, then weights and sums the results to obtain the comprehensive disturbance impact coefficient at the current moment. A time decay function is then introduced to accumulate the disturbance coefficients within the historical window, forming an effective disturbance impact coefficient that considers the continued impact of historical disturbances. This coefficient is compared with preset multi-level thresholds to determine the disturbance level of the production line (Level 1: Normal; Level 2: Minor Anomaly; Level 3: Significant Anomaly; Level 4: Severe Anomaly; Level 5: Production Line Stoppage Risk). Simultaneously, the system calculates the rate of change of each dimension's disturbance sub-coefficient to predict the disturbance development trend. When a disturbance in a certain dimension shows a worsening trend, preventative control measures are implemented in advance. The multi-agent collaborative decision-making module... The digital twin module determines the disturbance level and generates a global collaborative control scheme using a hierarchical collaborative control strategy. When the disturbance level is Level 1, a normal monitoring strategy is executed, maintaining the existing control strategy and recording operational data. When the disturbance level is Level 2, a local adjustment strategy is executed, adjusting parameters in abnormal local components and increasing the data acquisition frequency. When the disturbance level is Level 3, a dynamic rescheduling strategy is executed, reallocating production tasks and activating backup resources. When the disturbance level is Level 4, an emergency response strategy is executed, performing a partial shutdown of the faulty area and initiating emergency dispatch. When the disturbance level is Level 5, a disaster recovery strategy is executed, performing a plant-wide emergency shutdown and activating the disaster recovery plan. The entire control process is displayed in real-time through a human-machine interface module, showing the production line operating status, disturbance level determination results, collaborative control scheme, and optimization data. It supports remote manual intervention and parameter setting, thus forming a complete closed loop from data acquisition, twin modeling, disturbance quantification and level determination, to hierarchical collaborative decision-making, and finally returning to the physical production line for execution control.The core innovation of this system lies in the deep integration of digital twin technology with multi-agent collaborative decision-making. The digital twin model not only performs the static function of mapping between the virtual and real worlds, but also embeds a disturbance quantification algorithm based on time decay accumulation. This enables dynamic perception and structured output of complex disturbance states on the production line. The multi-agent collaborative decision-making module then establishes a hierarchical mapping mechanism between disturbance levels and collaborative strategies based on this structured output, achieving end-to-end automated closed-loop control from state identification to control execution. This integrated "perception-mapping-decision-execution" mechanism effectively solves the technical problems of lagging state perception, coarse disturbance judgment, and disconnect between decision-making and state in traditional production line control systems, significantly improving the adaptability, robustness, and collaborative efficiency of intelligent manufacturing production lines under complex operating conditions.

[0039] For ease of understanding, the following are specific examples of the present invention: A new energy vehicle component manufacturer has deployed an intelligent production line for motor housings. This line includes key units such as a CNC machining center, cleaning machine, airtightness testing equipment, automated assembly line, AGV logistics system, and automated warehouse, undertaking the tasks of precision machining, cleaning, testing, and assembly of motor housings. Due to the fast production cycle, high quality requirements, and large order fluctuations, traditional control systems struggle to cope with complex disturbances such as equipment failures, process fluctuations, and material delays. Therefore, the artificial intelligence-based intelligent manufacturing production line collaborative control system described in this invention is adopted.

[0040] The system deploys sensing devices throughout the entire production line: Industrial sensors: Vibration sensors and temperature sensors are installed on the spindle of a CNC machining center to monitor the spindle status in real time; PLC controller: Collects the operating status, cycle time, and alarm codes of each device; Machine vision module: Deploy a high-precision camera at the airtightness inspection station to automatically identify defects on the shell surface and QR code information; RFID logistics identification unit: installed on pallets and AGVs to track the location and flow status of materials in real time; AGV positioning module: Employs laser SLAM navigation to report location and task execution status in real time; Energy consumption monitoring instruments: monitor the instantaneous power and cumulative energy consumption of each device; Environmental monitoring equipment: collects data such as workshop temperature, humidity, and dust concentration.

[0041] During the period from 8:00 to 10:00 on a certain weekday, the system collected the following key data: The spindle temperature of CNC machining center No. 3 rose from the normal value of 45℃ to 68℃, exceeding the warning threshold. Three products at the airtightness testing station were found to have excessive leakage, causing the defect rate to rise from 0.5% to 4%. Due to congestion in the aisle, the delivery time of a certain batch of materials was delayed by 85 seconds in the AGV logistics process, exceeding the standard cycle time by 60 seconds. The order system received an urgent order at 9:30, adding 50 urgent orders, reducing the original order delivery cycle by 20%.

[0042] The multi-source data sensing module collects these data in real time, removes occasional noise from the sensors by 3σ outlier detection, fills in the temporary data loss caused by communication interruption by using LSTM neural network, and finally unifies all data to the [0,1] interval by min-max normalization to build a real-time database for the entire domain.

[0043] Based on the aforementioned real-time data, the digital twin module constructed a 1:1 high-precision digital twin model of the motor housing production line: Geometric model: The 3D model has an accuracy of 0.08mm, which is completely consistent with the actual production line layout; Physical Model: A CNC spindle thermal deformation model is established based on multibody dynamics, and a pressure decay model for airtightness testing is established based on fluid mechanics; Behavioral model: Simulates AGV path planning, equipment start-up and shutdown cycle time, and material flow logic; Rule model: Built-in process specifications, quality judgment standards, and safety interlock rules.

[0044] The system enables bidirectional data interaction between the virtual and physical worlds, with synchronization latency controlled within 150ms. Between 8:00 and 10:00, the digital twin module performs disturbance detection. Calculate the perturbation coefficients for each dimension: Equipment operation disturbance factor: The calculated value is 0.42 due to the temperature deviation of the No. 3 CNC spindle and the overall equipment efficiency decreasing from 85% to 72%. Process disturbance factor: The pressure curve of the airtightness test deviates from the set value and alarms multiple times, and the calculated value is 0.38; Logistics disturbance coefficient: AGV delays cause queue backlog, calculated to be 0.35; Quality disturbance factor: When the defect rate rises to 4% and the number of reworks increases, the calculated value is 0.52; Order disturbance factor: calculated as 0.45 due to urgent order insertions and delivery pressure; Energy consumption cost disturbance coefficient: 0.22 is calculated when the energy consumption per unit product increases by 8%; Environmental disturbance factor: With a workshop temperature of 28℃ (optimal temperature of 24℃) and dust concentration close to the upper limit, the calculated value is 0.28.

[0045] The weighted summation yields the comprehensive disturbance impact coefficient: Calculate based on preset dimension weights (equipment 0.25, process 0.20, logistics 0.15, quality 0.20, orders 0.10, energy consumption 0.05, environment 0.05). =0.415.

[0046] Considering historical disturbances over the past hour ( The values ​​are 0.32, 0.35, 0.38, and 0.40 respectively. Taking the attenuation coefficient k=0.5, the effective disturbance influence coefficient is calculated: =0.918.

[0047] System preset thresholds: 1 = 0.3, 2 = 0.5 3 = 0.8 4 = 1.2. =0.918 is between 0.8 and 1.2, and the disturbance level is judged to be level four, which is a serious anomaly.

[0048] The multi-agent collaborative decision-making module receives the disturbance level (level four) and trend analysis results output by the digital twin module, and then matches the control strategy according to the level: For Level 4 severe anomalies: Implement an emergency response strategy. The module generates instructions to partially halt the production line in the faulty area (CNC 3 and the airtightness testing station), while simultaneously initiating emergency scheduling—allocating the remaining processing tasks to CNCs 1 and 2, adjusting AGV paths to bypass the congested area, and issuing material preparation instructions to the automated warehouse to ensure the continued operation of other parts of the production line.

[0049] Calculate the rate of change of mass perturbation coefficients =0.0047 / min > preset threshold A = 0.003 / min, indicating a worsening trend in quality disturbance, triggering a preventative control warning. In response to the worsening quality disturbance trend: preventative control measures take effect. The system automatically retrieves historical pressure curve data from the airtightness testing equipment, identifies seal wear characteristics, and triggers a maintenance work order in advance, scheduling an operator to replace the testing fixture at 10:30.

[0050] The multi-agent collaborative decision-making module adopts a multi-agent architecture based on reinforcement learning, in which the equipment agent, logistics agent, quality agent, and order agent complete the above scheduling decisions through collaborative negotiation, and the whole process is completed within 200ms.

[0051] It should be noted that the formulas in this application are all dimensionless and numerical calculations. The formulas are obtained by software simulation based on a large amount of data and are the closest to the real situation. The thresholds, coefficients, standard values ​​and allowable values ​​involved in this application are all empirical values ​​and are selected by those skilled in the art according to the actual situation.

[0052] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A collaborative control system for intelligent manufacturing production lines based on artificial intelligence, characterized in that, It includes a multi-source data perception module, a digital twin virtual-real mapping module, a multi-agent collaborative decision-making module, and a human-computer interaction module; The multi-source data sensing module is used to collect multi-source heterogeneous data across the entire production line, and to perform outlier detection, missing value filling and normalization processing on the data to build a real-time database across the entire region. The digital twin virtual-real mapping module is connected to the multi-source data sensing module. It is used to construct a 1:1 high-precision digital twin model of the production line based on the full-domain real-time database, realize two-way real-time data interaction between the actual production line and the virtual model, calculate the comprehensive disturbance impact coefficient at the current moment based on the full-domain multi-source heterogeneous data, introduce a time decay function to accumulate the historical disturbance impact coefficient, obtain the effective disturbance impact coefficient, and determine the disturbance level of the current production line based on the effective disturbance impact coefficient. The multi-agent collaborative decision-making module is connected to the digital twin virtual-real mapping module and is used to match and execute the corresponding collaborative control strategy according to the disturbance level to generate a production line-wide collaborative control scheme. The human-computer interaction module is connected to the multi-agent collaborative decision-making module and is used to display the production line operation status and control scheme in real time, and supports remote human intervention and parameter setting.

2. The collaborative control system for intelligent manufacturing production lines based on artificial intelligence according to claim 1, characterized in that, The sensing devices deployed in the multi-source data sensing module include industrial sensors, PLC controllers, machine vision modules, RFID logistics identification units, AGV positioning modules, energy consumption monitoring instruments, and environmental monitoring equipment; the collected multi-source heterogeneous data includes equipment operation data, production process data, logistics flow data, quality inspection data, order and scheduling data, energy consumption and cost data, and environmental data.

3. The collaborative control system for intelligent manufacturing production lines based on artificial intelligence according to claim 2, characterized in that, The outlier detection employs an outlier detection algorithm based on the 3σ criterion to identify and remove outliers from the collected continuous data. The missing value imputation uses an LSTM neural network-based missing value imputation algorithm, which utilizes the correlation of time series data to accurately imput missing values ​​and avoid data deviation caused by missing values. The data normalization uses the min-max normalization algorithm to map all preprocessed data to the [0,1] interval, thereby achieving data standardization.

4. The collaborative control system for intelligent manufacturing production lines based on artificial intelligence according to claim 1, characterized in that, The digital twin model adopts a hierarchical modeling approach, consisting of four layers: geometric model, physical model, behavioral model, and rule model. The geometric accuracy of the geometric model is consistent with that of the actual production line, with an error of ≤0.1mm. The physical model is constructed based on multibody dynamics, thermodynamics, and fluid mechanics theories. The synchronization delay of the virtual-real two-way data interaction is ≤200ms.

5. The collaborative control system for intelligent manufacturing production lines based on artificial intelligence according to claim 4, characterized in that, The disturbance determination process includes: Step S1: Collect multi-dimensional operating data of the production line, including equipment operating data, production process data, logistics flow data, quality inspection data, order and scheduling data, energy consumption and cost data, and environmental data; Step S2: Calculate the disturbance sub-coefficients for each dimension, including equipment operation disturbance sub-coefficient, process disturbance sub-coefficient, logistics disturbance sub-coefficient, quality disturbance sub-coefficient, order disturbance sub-coefficient, energy consumption cost disturbance sub-coefficient, and environmental disturbance sub-coefficient. Step S3: Sum the perturbation coefficients of each dimension according to the preset dimension weights to obtain the comprehensive perturbation influence coefficient at the current moment; Step S4: Introduce a time decay function to accumulate the disturbance influence coefficients at historical moments to obtain the effective disturbance influence coefficient; Step S5: Compare the effective disturbance impact coefficient with the preset early warning threshold and emergency threshold to determine the disturbance level of the current production line.

6. The collaborative control system for intelligent manufacturing production lines based on artificial intelligence according to claim 5, characterized in that, The formula for calculating the effective disturbance influence coefficient in step S4 is as follows: In the formula, The length of the historical time window. For time offset index, The attenuation coefficient set for the system. The comprehensive disturbance impact coefficient at the current moment. This is the effective disturbance influence coefficient.

7. The collaborative control system for intelligent manufacturing production lines based on artificial intelligence according to claim 5, characterized in that, The specific working process of step S5 includes: when If the disturbance level of the current production line is determined to be Level 1, the production line is operating normally and no intervention is required; when At this time, the disturbance level of the current production line is determined to be level two, indicating that there is a slight abnormality in the production line. It is recommended to pay attention and record it. when When the disturbance level of the current production line is determined to be level three, the production line has obvious abnormalities, triggering an early warning and recommending manual intervention; when When the disturbance level of the current production line is determined to be level four, the production line has a serious abnormality, an alarm is triggered and immediate action is required; when When the disturbance level of the current production line is determined to be Level 5, the production line faces the risk of shutdown, triggering the emergency response mechanism.

8. The collaborative control system for intelligent manufacturing production lines based on artificial intelligence according to claim 4, characterized in that, The disturbance assessment process also includes trend analysis and preventive control: Calculate the rate of change of the perturbation coefficients in each dimension. ,in, Let be the coefficient of the i-th perturbation type; When the rate of change of the perturbation coefficient in any dimension When the value is greater than A, it is determined that the disturbance in this dimension is showing a worsening trend, and preventive control measures are implemented in advance. When the rate of change of the perturbation coefficient in any dimension When the value is less than B, it is determined that the disturbance in this dimension is showing an improving trend, and control resources are gradually released. Where A>0, B<0, and A and B are preset thresholds.

9. The collaborative control system for intelligent manufacturing production lines based on artificial intelligence according to claim 7, characterized in that, The working process of the multi-agent collaborative decision-making module includes: When the disturbance level is Level 1, execute the normal monitoring strategy, maintain the existing control strategy, and record the operating data. When the disturbance level is level two, a local adjustment strategy is implemented to adjust the parameters of the abnormal local components and strengthen monitoring. When the disturbance level is level three, a dynamic rescheduling strategy is executed to reallocate production tasks and activate standby resources. When the disturbance level is level four, an emergency response strategy is implemented, which involves local line shutdown in the fault area and initiation of emergency dispatch. When the disturbance level is level 5, the disaster recovery strategy is implemented, including a plant-wide emergency shutdown and the activation of the disaster recovery plan.

Images