Intelligent production control system and method for flexible disassembly of decommissioned products

Abstract

This invention discloses an intelligent production control system and method for the flexible dismantling of decommissioned products. The system includes a dismantling execution control unit, a logistics control unit, a warehousing control unit, and a rolling time-domain coordination optimizer. This invention collects data on the dismantling station status, logistics transportation status, warehousing status, and working condition tags obtained through visual recognition. It uses a time prediction model to predict the remaining completion time and uncertainty variance of the dismantling task. Based on the prediction results, it coordinates the control parameters of the dismantling execution mechanism, logistics transportation tasks, and warehousing inbound and outbound tasks, achieving dynamic coordinated control between dismantling, logistics, and warehousing. Simultaneously, by constructing a rolling optimization model that considers time uncertainty, it achieves real-time optimization and closed-loop control of task scheduling instructions, solving the problems of large differences in working conditions, strong fluctuations in operation time, and difficulty in coordinating the rhythms of dismantling, logistics, and warehousing during the dismantling of decommissioned products.

Description

Technical Field

[0001] This invention relates to the field of solid waste resource recycling technology, and in particular to an intelligent production control system and method for the flexible dismantling of decommissioned products. Background Technology

[0002] In the process of recycling retired products, especially for the dismantling of complex products such as electronics, automobiles, and batteries, intelligentization and flexibility have become key directions for improving recycling efficiency and scale. Existing systems typically use automated storage and retrieval systems (AS / RS), RGVs, AGVs, and other logistics equipment, combined with robotic arms and other execution units to build dismantling production lines, achieving flexible scheduling and automated processing to a certain extent.

[0003] However, the practical application of such systems still faces significant constraints. Currently, the inconsistent states of retired products (different models, retired statuses) and varying operator skill levels lead to a high degree of uncertainty in dismantling operation time. The time required by each operator to handle products in different states fluctuates significantly. This dynamic and unpredictable fluctuation in working hours far exceeds the adaptability of existing dismantling scheduling control systems. Existing systems' scheduling methods are largely based on ideal assumptions such as fixed working hours and predetermined work sequences, lacking the ability to perceive and respond to real-time changes in working hours. This makes it impossible to dynamically adapt to the uncertainty of dismantling operation time, leading to a series of problems: delayed system scheduling response, inability to adjust task allocation and logistics rhythm according to real-time working hours; mismatch between equipment and personnel configuration and actual operational needs, resulting in low equipment utilization efficiency, idle personnel, and uneven workloads; ultimately, insufficient on-time order completion rates, severely restricting the large-scale and efficient operation of retired product dismantling production lines.

[0004] Among the existing published patents, CN202311592266 discloses a disassembly sequence planning method and system for products with unpredictable states. Its core lies in dynamically solving the disassembly sequence through tracking optimization and time-domain robust optimization algorithms, adjusting the disassembly scheme by updating state information while disassembling. However, its optimization objective is limited to adjusting the disassembly order, without involving online perception and quantitative prediction of real-time disassembly time, nor does it use time fluctuation information to drive the linkage adjustment of logistics conveying speed and underlying control parameters, making it difficult to solve the production line-level efficiency bottleneck problem caused by the high uncertainty of time. CN116109299B discloses a dynamic disassembly line balancing method based on support vector regression and Gaussian inverse model. It predicts environmental target values ​​through support vector regression and uses Gaussian inverse model to generate an initial population for evolutionary calculation to improve optimization efficiency. However, its prediction results only affect the initial solution space of the offline disassembly line balancing scheme, without using the disassembly time prediction information for logistics or warehousing scheduling decisions, nor establishing a real-time feedback link between prediction uncertainty and logistics conveying speed or control parameters. Summary of the Invention

[0005] The present invention aims to at least partially solve one of the technical problems in the related art.

[0006] The main objective of this invention is to provide an intelligent production control system for the flexible dismantling of retired products.

[0007] Another objective of this invention is to propose an intelligent production control method for the flexible dismantling of retired products.

[0008] To achieve the above objectives, a first aspect of the present invention provides an intelligent production control system for the flexible dismantling of decommissioned products, comprising:

[0009] The on-site distributed control station includes a dismantling execution control unit, a logistics control unit, and a warehousing control unit. The dismantling execution control unit controls the motion parameters of the dismantling station's execution mechanism and adaptively adjusts these parameters based on working condition tags obtained through visual recognition. The dismantling execution control unit includes a time prediction submodule, used to predict the remaining completion time of the dismantling task and the corresponding uncertainty variance based on current working condition information. The logistics control unit acquires the task execution progress of the target dismantling station and the remaining completion time output by the time prediction submodule, and schedules logistics equipment to perform inter-station transport based on the acquired results. The warehousing control unit schedules warehousing equipment to perform inbound and outbound operations of the dismantled products based on the estimated completion time of the dismantling task output by the time prediction submodule and the logistics transport status. The operator workstation is used to monitor the status of the dismantling station, the logistics transportation status, the time prediction results and the scheduling execution, and to receive manual intervention instructions. The rolling time-domain coordination optimizer is used to periodically collect status data of each dismantling station, logistics and warehousing, build a rolling optimization model that takes into account the uncertainty of time based on the time prediction results, and output task allocation instructions and execution mechanism settings. The field distributed control station, operator workstation, and rolling time-domain coordinator are connected via the OPC UA industrial communication network.

[0010] Optionally, the control loop of the disassembly execution control unit adopts an improved RBF neural network dynamic adaptive PID controller; The neural network identifier is used to obtain the input-output relationship of the disassembly actuator and output sensitivity information. ; PID controller weights , , Corresponding to proportionality coefficients Integral coefficient Differential coefficients Its update rules are as follows:

[0011] in, The deviation between the set value and the actual output. , , , This is the dynamic learning rate.

[0012] Optionally, the dynamic learning rate The adjustment is made dynamically based on the change range of the working condition labels obtained by visual recognition and the uncertainty variance output by the working time prediction submodule. Specifically, when the working condition label changes across levels or the uncertainty variance exceeds a preset threshold, the dynamic learning rate is increased. This improves the controller's ability to adapt to changes in the status of retired products and fluctuations in operating time; the increase in the dynamic learning rate is positively correlated with the magnitude of changes in operating condition levels.

[0013] Optionally, the working condition label includes product type, disassembly requirements, corrosion level, degree of obstruction of connectors, and degree of structural damage; The working condition label is synchronously input to the working time prediction submodule to construct a dismantling operation time prediction model that is strongly correlated with the status of retired products, providing a basis for adaptive adjustment of control loop parameters and advance scheduling of logistics.

[0014] Optionally, the time prediction submodule predicts the percentage of task completion for the current dismantling task based on the current visual recognition work condition label and the operator's proficiency using a Gaussian process regression model. Remaining working hours and uncertainty variance .

[0015] Optionally, the logistics control unit acquires the target dismantling station task execution progress data from the time prediction submodule in real time, and displays the task completion percentage. ≥90% and predicted remaining working hours If the time is ≤1 minute, the logistics equipment should be scheduled to be ready at the workstation in advance.

[0016] Optionally, the position control loop of the warehouse control unit adopts a fuzzy PID control algorithm to measure the position deviation. and the rate of change of deviation As input, the PID parameters are adjusted online; The warehouse control unit schedules warehouse equipment to perform inbound and outbound operations in advance based on the expected completion time of the dismantling task and the arrival time of the logistics transportation.

[0017] Optionally, the rolling time-domain coordinated optimizer, within each sampling period, calculates the current task progress of each dismantling station and the remaining completion time output by the time prediction submodule. With variance The task scheduling scheme aims to minimize the total completion time by rolling the solution within the finite prediction time domain, and only issues instructions for the current moment. When the uncertainty variance corresponding to a certain disassembly station When the preset threshold is exceeded, the rolling time-domain coordination optimizer reserves buffer time in the subsequent task scheduling of the dismantling station.

[0018] To achieve the above objectives, a second aspect of the present invention proposes an intelligent production control method for the flexible dismantling of retired products, comprising the following steps: S1. Collect the status of the dismantling station, logistics and transportation, storage, and working condition labels obtained by visual recognition. S2. Based on the working condition labels collected in step S1, predict the remaining completion time and corresponding uncertainty variance of the current dismantling task, and generate a working time prediction result. S3. Based on the working condition labels collected in step S1 and the working time prediction results generated in step S2, adaptively adjust the control parameters of the disassembly actuator. S4. Based on the task execution progress of the target dismantling station and the remaining completion time predicted in step S2, schedule logistics equipment to perform the transportation task. S5. Determine the estimated completion time of the dismantling task based on the remaining completion time predicted in step S2, and schedule the warehousing equipment to perform inbound and outbound operations in conjunction with the logistics transportation status collected in step S1. S6. Based on the dismantling station status, logistics transportation status and warehousing status collected in step S1, and the time prediction results generated in step S2, construct a rolling optimization model that considers the uncertainty of time, and generate the task scheduling instructions and actuator settings for the current control cycle. S7. Send the task scheduling instructions and execution mechanism settings generated in step S6 to the corresponding dismantling execution mechanism, logistics equipment and warehousing equipment, and return to step S1 to enter the next control cycle.

[0019] To achieve the above objectives, a third aspect of this application provides an electronic device, including a processor and a memory; wherein the processor runs a program corresponding to the executable program code stored in the memory to implement the method described in the first aspect.

[0020] To achieve the above objectives, a fourth aspect of this application provides a non-transitory computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect.

[0021] The embodiments of this invention have the following beneficial effects: By integrating the field control layer, monitoring and operation layer, and optimization scheduling layer into a multi-level closed-loop control structure through the OPC UA communication network, the information silos of dismantling, logistics, and warehousing in traditional systems are broken down. This achieves full-process cycle coordination and seamless connection of dismantling, logistics, and warehousing, eliminating production line downtime and effectively improving overall throughput efficiency. This invention employs a rolling time-domain coordinated optimizer to periodically collect real-time progress and predicted remaining working hours at each workstation, and to continuously solve for the optimal task allocation scheme with the goal of minimizing the total order completion time. Simultaneously, the dismantling unit uses an improved RBF neural network dynamic adaptive PID controller, which dynamically adjusts control parameters based on changes in working condition labels and uncertainties in working hour prediction. This enhances the adaptability and robustness of the dismantling process to differences in product status, significantly improving equipment utilization and dismantling efficiency. Attached Figure Description

[0022] The above-described and additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, in which: Figure 1 A schematic diagram of an intelligent production control system for the flexible dismantling of decommissioned products provided in an embodiment of the present invention; Figure 2 A flowchart of an intelligent production control method for the flexible dismantling of retired products provided in an embodiment of the present invention; Figure 3 This is a plan view of the intelligent production line for flexible dismantling of scrapped electric vehicles provided in Example 3; Figure 4 This is a Gantt chart of the traditional dismantling production control scheme in Example 4; Figure 5 Gantt chart of the intelligent production control scheme for flexible disassembly in Example 4; Figure 6 This is a schematic diagram comparing the workstation utilization rates of the present invention and the traditional solution in Example 4. Detailed Implementation

[0023] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0025] This invention aims to provide an intelligent production control system and method for the flexible dismantling of retired products, addressing the problems of low dismantling efficiency, equipment idleness, and delayed scheduling response caused by the high uncertainty of the retired product's condition in existing technologies. By constructing a multi-level closed-loop control architecture, and driving dynamic coordination of logistics transportation and warehousing scheduling based on real-time work hour prediction results, adaptive control and rolling optimization scheduling of the dismantling process are achieved, thereby minimizing the total order completion time and improving production line flexibility and efficiency. The intelligent production control system and method for the flexible dismantling of retired products according to embodiments of the present invention are described below with reference to the accompanying drawings.

[0026] Example 1 This embodiment provides an intelligent production control system for the flexible dismantling of retired products. For example... Figure 1 As shown, the system includes a field distributed control station, an operator workstation, a rolling time-domain coordinating optimizer, and an engineer station. The field distributed control station, operator workstation, rolling time-domain coordinating optimizer, and engineer station communicate with each other through the OPC UA industrial communication network to form a multi-level closed-loop control architecture for the flexible dismantling process of retired products.

[0027] (1) Field distributed control station.

[0028] In this embodiment of the invention, a field distributed control station is deployed at the bottom control layer of the flexible dismantling production line for decommissioned products, and is used for distributed control of the real-time execution process at the dismantling site. This field distributed control station is communicatively connected to the dismantling workstations, logistics conveying nodes, and warehousing equipment, and is divided into dismantling execution control units, logistics control units, and warehousing control units according to the functional attributes of different controlled objects. Through the coordinated operation of these units, the field distributed control station can achieve synchronous control of dismantling operations, material conveying, and warehousing inbound and outbound processes, thereby ensuring the continuity of processes and the coordination of cycle time during the flexible dismantling of decommissioned products.

[0029] ① The disassembly execution control unit is installed at each disassembly station to control the action parameters of the disassembly execution mechanism. The disassembly execution mechanism includes, but is not limited to, industrial robotic arms, electric wrenches, cutting mechanisms, clamping and positioning mechanisms, pressing or pulling mechanisms, etc. Based on the process requirements of the current disassembly task, the disassembly execution control unit controls the disassembly execution mechanism to complete actions such as positioning, clamping, loosening, separation, moving, and detection, and collects feedback information in real time, including displacement, speed, torque, current, vibration, and action completion status of the execution mechanism.

[0030] In this embodiment, the control loop of the disassembly execution control unit employs an improved RBF neural network dynamic adaptive PID controller. Specifically, this controller includes a neural network identifier and a PID controller. The neural network identifier is used to dynamically identify the input-output relationship of the disassembly execution mechanism under different retired product states, different disassembly resistances, and different operating conditions based on the real-time input-output data of the disassembly execution mechanism, and outputs sensitivity information reflecting the degree of influence of disassembly resistance on the output of the execution mechanism. This sensitivity information is used to characterize the direction and magnitude of the impact of changes in control input on actual output changes, thereby enabling the PID controller to adaptively adjust parameters based on resistance changes and state disturbances during the disassembly process.

[0031] PID controller weights , , Corresponding to the proportionality coefficients proportionality coefficients Integral coefficient Differential coefficients The proportional coefficient is used to enhance the system's response to current deviations, the integral coefficient is used to eliminate steady-state errors, and the derivative coefficient is used to suppress oscillations caused by rapid deviation changes. To enable the controller to adapt to the characteristics of dismantling retired products, such as large structural differences, inconsistent corrosion levels, uncertain connector conditions, and random variations in dismantling resistance, this embodiment uses online parameter tuning rules based on RBF neural network identification results to update the PID parameters.

[0032] Specifically, the update rule for the PID controller weights is as follows:

[0033] in, , For the first The time setting value, For the first The actual output at any given moment. The deviation between the set value and the actual output; This is used to reflect the current control deviation; This is used to reflect the cumulative amount of deviation; This is used to reflect the rate of change of deviation; For the first The weight at the th th The dynamic learning rate at any given time. Through the above parameter update method, the controller can use the current deviation, cumulative deviation, deviation change rate, and sensitivity information identified by the neural network to make online corrections to the proportional, integral, and derivative parameters, so that the disassembly actuator can maintain good tracking accuracy and control stability when facing different disassembly resistances and changes in operating conditions.

[0034] Furthermore, dynamic learning rate This is a time-varying parameter; its value is not fixed but is adjusted by the combined effect of the change in the working condition label obtained from visual recognition and the variance of the uncertainty in working time prediction. Specifically, when visual recognition results indicate that the working condition label of the currently retired product has changed across levels compared to the previous moment or the previous task, or when the variance of the uncertainty output by the working time prediction submodule exceeds a preset threshold, the dynamic learning rate... The learning rate is temporarily increased from the base learning rate to improve the update speed of PID parameters, enabling the controller to adapt more quickly to differences in the state of retired products, sudden changes in disassembly resistance, and fluctuations in operating time. The increase in the learning rate is positively correlated with the magnitude of the change in operating condition level; that is, the greater the change in operating condition level, the greater the increase in the learning rate. When the change in operating condition level is small, the learning rate is only slightly adjusted or the base learning rate is maintained to avoid frequent and large changes in control parameters that could cause system oscillations.

[0035] The operational condition labels obtained through visual recognition include product type, dismantling requirements, corrosion level, degree of obstruction by connectors, and degree of structural damage. Product type indicates the category or model of the product to be dismantled and decommissioned; dismantling requirements indicate whether the current task is complete dismantling, selective dismantling, priority dismantling of hazardous components, or priority dismantling of recyclable components; corrosion level indicates the degree of corrosion on the surfaces of connectors, shells, or structural components; degree of obstruction by connectors indicates the extent to which bolts, clips, welds, and other connecting parts are obstructed by oil, deformed structures, or other components; degree of structural damage indicates whether the decommissioned product has shell damage, missing components, deformation, breakage, or other defects. These operational condition labels reflect the differences between the actual condition of the decommissioned product and the standard product condition, providing a basis for parameter control and operation time prediction for the dismantling actuator.

[0036] The work condition label is synchronously input to the work time prediction submodule. This work time prediction submodule is integrated into the dismantling execution control unit and is used to predict the completion progress and remaining operation time of the current dismantling task. In this embodiment, the work time prediction submodule predicts the current task completion percentage based on the current visually recognized work condition label and the operator's proficiency using a Gaussian process regression model. Remaining working hours and uncertainty variance Operator proficiency is used to characterize the experience level and work efficiency of operators in manual disassembly or manual verification processes. It can be determined based on historical work data, job level, or average work time recorded by the system.

[0037] Gaussian process regression models can probabilistically predict dismantling operation time under conditions of limited sample size and significant differences in the state of decommissioned products, and output the uncertainty variance of the prediction results. The current task completion percentage *p* represents the proportion of the dismantling task already completed, the predicted remaining time *μ* represents the estimated time the current task will continue, and the uncertainty variance *σ²* represents the reliability of the prediction results. A large uncertainty variance indicates that the current state of decommissioned products is complex or that historical samples are insufficient, requiring the system to improve the adaptive capability of controller parameters and reserve greater adjustment space for logistics and warehousing scheduling. Therefore, the time prediction submodule not only provides a basis for the dynamic adaptive PID parameter adjustment of the dismantling execution control unit but also provides real-time prediction information for subsequent advance scheduling of logistics equipment, pre-action of warehousing equipment, and rolling time-domain coordination optimization.

[0038] ② The logistics control unit is deployed at AGVs, RGVs, or fixed material transport nodes to control the transport of products to be dismantled, dismantled components, recyclable products, and materials to be stored between dismantling stations. The logistics control unit can control the operating route, conveying speed, start / stop status, obstacle avoidance actions, arrival waiting, and material handover actions of the logistics equipment. The logistics control unit obtains the task execution progress and current task completion percentage of the target dismantling station in real time through the OPC UA communication network. And predicting remaining working hours The system will dynamically adjust the operating status of logistics equipment based on the above information.

[0039] Specifically, when the disassembly execution control unit outputs the current task completion percentage. And predict the remaining working hours When the time is ≤1 minute, the logistics control unit determines that the target dismantling station is about to complete its current dismantling task and pre-schedules the corresponding logistics equipment to that station for standby. Once the dismantling task is completed, the logistics equipment can immediately perform material pick-up, transfer, or delivery to the next station, thereby reducing station waiting time and empty-running time, achieving seamless transitions between stations. When the predicted remaining time is long or the uncertainty variance is large, the logistics control unit can lower the priority of the logistics equipment or schedule it to other more pressing stations to avoid prolonged waiting times for logistics resources before the station completes its task.

[0040] In actual operation, the logistics control unit can dynamically adjust the conveying speed and start / stop status based on production line congestion, remaining power of logistics equipment, path occupancy, material priority, and target workstation buffer capacity. For example, when the target workstation is not yet ready to receive materials, the logistics control unit controls the AGV or RGV to wait in the buffer zone; when the target workstation is about to complete dismantling and downstream warehousing equipment has completed pre-scheduling, the logistics control unit increases the conveying response priority to ensure timely transfer of dismantled products. In this way, the logistics control unit can transform the real-time progress prediction results of the dismantling workstation into a pre-scheduling strategy for logistics equipment, thereby improving the overall flow efficiency of the dismantling production line.

[0041] ③ The warehouse control unit is deployed in automated warehouses and flat warehouses to control stacker cranes, shuttle cars, conveyor lines, lifting mechanisms, and inbound / outbound identification devices to complete the inbound, outbound, transfer, and buffering operations of disassembled products. The warehouse control unit receives product information and completion status uploaded by the disassembly system, as well as delivery arrival time uploaded by the logistics system, and pre-schedules stacker cranes and shuttle cars to perform corresponding warehousing actions based on the type of disassembled product, storage location, inventory status, inbound priority, and outbound task requirements.

[0042] In this embodiment, the position control loop of the warehouse control unit adopts a fuzzy PID control algorithm. This algorithm takes the position deviation ep(k) and the deviation change rate Δep(k) as inputs, and adjusts the PID control parameters online through fuzzy inference rules, enabling stable and accurate positioning control of warehouse equipment such as stacker cranes and shuttles under different loads, operating speeds, and positioning distances. The position deviation ep(k) characterizes the difference between the current position and the target position of the warehouse equipment, and the deviation change rate Δep(k) characterizes the trend of position deviation changing over time. When the position deviation is large, the fuzzy PID control algorithm increases the proportional action to accelerate the equipment's approach to the target position; when the equipment approaches the target position and the deviation changes rapidly, the derivative action is enhanced to suppress overshoot; when the equipment has a continuous small deviation, the integral action is appropriately enhanced to improve positioning accuracy.

[0043] The warehouse control unit can pre-position stacker cranes, shuttle cars, or target storage locations to prepare based on the completion status of dismantled products and the estimated arrival time of logistics equipment. For example, when the logistics control unit reports that a dismantled product will arrive at the warehouse entrance in a short time, the warehouse control unit can pre-allocate target storage locations and control the stacker crane to move to the vicinity of the corresponding aisle, and control the shuttle car to move to the predetermined pick-up position, thereby shortening the waiting time after the materials arrive. For tooling, containers, or parts awaiting repair that require priority outbound processing, the warehouse control unit can also pre-execute outbound preparations based on the task plan issued by the rolling time-domain coordination optimizer to ensure rhythm coordination between warehousing, dismantling, and logistics.

[0044] Meanwhile, the warehouse control unit provides real-time feedback on inventory status to the rolling time-domain coordination optimizer. This inventory status includes inventory quantity, storage location occupancy, product category, inbound time, outbound tasks, buffer capacity, warehouse equipment operating status, and abnormal alarm information. Based on the inventory status feedback from the warehouse control unit, the rolling time-domain coordination optimizer can coordinate and optimize subsequent dismantling task scheduling, logistics route planning, and warehouse resource allocation. When inventory approaches its capacity limit, the system can reduce the production cycle time of relevant dismantled products or prioritize outbound shipments; when the inventory of a certain type of recyclable product is insufficient or downstream processing tasks are urgent, the system can increase the priority of the corresponding dismantling task.

[0045] Through the above setup, the on-site distributed control station can integrate dismantling execution, logistics transportation, and warehousing operations into a unified real-time control system. At the dismantling execution level, the improved RBF neural network dynamic adaptive PID controller can adjust control parameters online based on differences in the state of decommissioned products and changes in dismantling resistance, improving the control accuracy and adaptability of the dismantling execution mechanism. At the logistics transportation level, the logistics control unit can pre-schedule AGVs, RGVs, or fixed conveyor equipment using time prediction results, achieving continuous connection between workstations. At the warehousing operation level, the warehousing control unit can pre-schedule warehousing equipment based on product information and logistics arrival times, and improve the positioning accuracy of warehousing equipment through fuzzy PID control. Therefore, this embodiment achieves real-time perception, dynamic control, and cycle time coordination at the flexible dismantling site layer of decommissioned products, providing a stable and reliable underlying control foundation for upper-level rolling time-domain coordinated optimization.

[0046] (2) Operator workstation.

[0047] In this embodiment of the invention, the operator workstation is located in the monitoring and human-machine interface layer of the flexible dismantling production line for decommissioned products. It is used for centralized display, status monitoring, and operational interaction of various real-time operating data uploaded by the on-site distributed control station. The operator workstation communicates with the on-site distributed control station, the rolling time-domain coordinating optimizer, and the engineering workstation via the OPC UA industrial communication network to achieve bidirectional transmission of on-site status information, scheduling execution information, abnormal alarm information, and manual intervention information.

[0048] Specifically, the operator workstation can monitor the operational status of each disassembly station in real time, including the currently executing disassembly task, task start time, task completion percentage, actuator operating status, fixture positioning status, disassembly action completion status, equipment alarm status, and manual assistance operation status. For stations using robotic arms, electric wrenches, cutting mechanisms, or other disassembly actuators, the operator workstation can also display the actuator's action mode, current position, operating speed, torque feedback, control deviation, and control loop status, enabling operators to intuitively grasp the disassembly operation process.

[0049] The operator workstation is also used to monitor the location of logistics transportation and the execution of logistics scheduling. It can display the current location, target workstation, running path, conveying speed, start / stop status, task queue, waiting status, and abnormal lane occupancy information of AGVs, RGVs, or fixed material transport nodes. When the logistics control unit pre-schedules logistics equipment to the workstation based on time prediction results, the operator workstation can simultaneously display the pre-scheduling instruction and its execution status, enabling operators to determine whether an effective connection has been achieved between the logistics system and the dismantling workstation.

[0050] Meanwhile, the operator workstation displays the time prediction results. These results include the current task completion percentage, predicted remaining time, and uncertainty variance, output by the time prediction submodule in the disassembly execution control unit. Operators can view the predicted remaining work time and prediction reliability for different workstations through the operator workstation. When the uncertainty variance for a particular workstation is large, the operator workstation can provide an alert through color, pop-up windows, sound, or text prompts, indicating that the current workstation may have complex product conditions, abnormal disassembly resistance, unstable visual recognition results, or significant fluctuations in manual assistance time.

[0051] In terms of scheduling execution monitoring, the operator workstation can receive task scheduling plans issued by the rolling time-domain coordinator and display the workstation tasks, logistics tasks, and warehousing tasks that have been issued at the current moment. For subsequent tasks that are still within the prediction time domain but have not yet been actually issued, the operator workstation can display them as preview information so that operators can understand the subsequent operation trend of the production line. Since the rolling time-domain coordinator only issues instructions for the current moment, the actual execution instructions displayed by the operator workstation can be distinguished from the prediction scheduling plan, avoiding operators from mistakenly treating predicted tasks that have not yet taken effect as executed tasks.

[0052] Furthermore, the operator workstation integrates a human-machine interface, supporting manual intervention by the operator in abnormal situations. These abnormal situations include, but are not limited to, alarms from disassembly actuators, visual recognition failures, severe corrosion of connectors, obstruction of components preventing automatic disassembly, blocked logistics equipment paths, abnormal positioning of storage equipment, insufficient inventory capacity, and the inability of the rolling optimization scheduling scheme to meet actual on-site constraints. When these abnormalities occur, the operator can use the operator workstation to pause the current task, adjust task priorities, reallocate disassembly workstations, designate logistics equipment, confirm the results of manual assisted disassembly, release or lock workstation resources, and send a rescheduling request to the rolling time-domain coordinating optimizer.

[0053] Upon receiving a manual intervention command, the operator workstation transmits the command to the corresponding field distributed control station or rolling time-domain coordinator via the OPC UA industrial communication network. For commands that directly affect equipment safety or field actions, such as pausing equipment, emergency stop reset, workstation release, and material flow cessation, the operator workstation sends the command to the corresponding field control unit for execution. For commands that affect global task scheduling, such as task priority adjustment, workstation reallocation, abnormal workstation masking, or task reassignment, the operator workstation sends the command to the rolling time-domain coordinator, which then regenerates the scheduling plan based on the field status in the next optimization cycle. Thus, the operator workstation not only achieves visualized monitoring of the production site but also enables the synergy between human experience and automated optimization scheduling under complex and abnormal conditions.

[0054] (3) Rolling time-domain coordinating optimizer.

[0055] The rolling time-domain coordination optimizer, as the core device of the optimization scheduling layer, is used to globally coordinate and optimize the dismantling, logistics, and warehousing tasks during the flexible dismantling process of retired products. This rolling time-domain coordination optimizer can be deployed on a standalone server or integrated into an operator workstation. When deployed on a standalone server, it provides higher computing power and better system scalability; when integrated into an operator workstation, it reduces the complexity of system hardware configuration and is suitable for smaller-scale or computationally less demanding flexible dismantling production lines.

[0056] The rolling time-domain coordinated optimizer periodically collects global status data through the OPC UA communication network. This global status data includes the current task type, task execution progress, task completion percentage, and predicted remaining time for each dismantling station. Uncertainty and variance The system collects data on equipment operating status, workstation occupancy status, logistics equipment location, logistics equipment task queue, warehouse inventory status, storage location occupancy, inbound and outbound status, buffer capacity, and abnormal alarm information. Through periodic collection of this data, the rolling time-domain coordination optimizer can monitor the real-time operating status of workstations, logistics, and warehousing resources within the decomposition system, providing a data foundation for subsequent optimization and scheduling.

[0057] In this embodiment, the rolling time-domain coordination optimizer aims to minimize the total order completion time and constructs a rolling optimization model that considers the uncertainty of work time. The total order completion time refers to the total time required to complete the dismantling, logistics transfer, and necessary warehousing tasks of all retired products in the current order set. Due to the significant differences in the state of retired products, different products may have significant differences in corrosion level, degree of connector obstruction, degree of structural damage, and dismantling requirements. Therefore, the actual working time of each workstation is uncertain. The rolling time-domain coordination optimizer incorporates the predicted remaining work time and uncertainty variance output by the work time prediction submodule into the optimization model, so that the scheduling scheme considers not only the average task completion time but also the reliability of the work time prediction results.

[0058] Specifically, the rolling optimization model schedules and solves for tasks within a finite prediction time domain. This finite prediction time domain can be pre-defined based on production line cycle time, order size, number of devices, and computing resources, for example, set to a future number of minutes or task cycles. Within each optimization cycle, the rolling time-domain coordinator updates the optimization model based on the latest collected global state data and solves for the task scheduling scheme within the prediction time domain. This task scheduling scheme includes the workstation allocation results for each subdivided task, the task start order, the logistics equipment connection plan, the warehousing or outbound plan, and alternative arrangements for abnormal workstations.

[0059] Unlike one-time static scheduling, the rolling time-domain coordination optimizer in this embodiment does not issue all tasks for the entire prediction time domain at once after each solution, but only issues the instructions that need to be executed at the current moment. As the production process continues, the system re-collects the field status in the next optimization cycle and solves the scheduling scheme again based on the new operating condition information and the new prediction results. Through this rolling solution and rolling execution method, the system can absorb field disturbance information in a timely manner and dynamically correct subsequent task arrangements, avoiding the failure of the original scheduling scheme due to uncertainties in the status of retired products, fluctuations in dismantling time, or abnormalities in logistics and warehousing.

[0060] When the time forecasting uncertainty variance for a certain workstation When a preset threshold is exceeded, the rolling time-domain coordination optimizer determines that the current or subsequent tasks at that workstation have a high risk of time fluctuation. In this case, the optimizer reserves additional buffer time for subsequent tasks at that workstation in the scheduling scheme. This additional buffer time can be positively correlated with the magnitude of the uncertainty variance; that is, the larger the uncertainty variance, the longer the reserved buffer time; the smaller the uncertainty variance, the shorter the reserved buffer time. By setting a buffer time, the impact of high-uncertainty workstations on downstream logistics connections, warehousing, and the connection of tasks at other workstations can be reduced, improving the robustness of the overall scheduling scheme.

[0061] Furthermore, the rolling time-domain coordination optimizer can also collaboratively modify the scheduling plan based on logistics location and warehouse inventory status. For example, when a logistics device is close to a workstation where a task is about to be completed, the optimizer can prioritize scheduling that device for a connecting task to reduce empty travel distance and waiting time. When warehouse space corresponding to a certain type of dismantling product is scarce, the optimizer can lower the priority of dismantling tasks that produce that type of product, or prioritize scheduling warehouse equipment to complete outbound and transfer operations. When a dismantling workstation is temporarily disabled by an operator due to an anomaly, the optimizer can reassign tasks originally planned for that workstation to other workstations with the same capabilities. Thus, the rolling time-domain coordination optimizer can achieve global coordination among dismantling, logistics, and warehousing resources.

[0062] (4) Engineer station.

[0063] The engineering workstation is used to complete the engineering configuration, parameter maintenance, model training, and system deployment of the intelligent production control system. Typically used by authorized engineering technicians, the engineering workstation communicates with field distributed control stations, operator workstations, and the rolling time-domain coordinating optimizer via the OPC UA industrial communication network. It is used to configure and manage the system before commissioning, during operation, and during maintenance.

[0064] Specifically, the engineering workstation is used to complete the system hardware configuration. Engineering technicians can use the engineering workstation to configure hardware objects such as disassembly stations, robotic arms, electric wrenches, AGVs, RGVs, fixed conveyors, stacker cranes, shuttles, vision recognition devices, sensors, and safety protection devices, and establish the correspondence between each hardware object and the control unit. Through hardware configuration, the system can clearly define the workstation, communication address, control permissions, operating mode, and interlocking relationships of each device, providing a basic configuration for subsequent real-time control and optimized scheduling.

[0065] The engineering workstation is also used to tune control loop parameters. For the improved RBF neural network dynamic adaptive PID controller in the disassembled execution control unit, the engineering workstation can set the basic PID parameters, initial weights, basic learning rate, learning rate increase limit, operating condition level change threshold, time prediction uncertainty variance threshold, and control output limiting parameters. For the fuzzy PID control algorithm in the warehouse control unit, the engineering workstation can set the fuzzy universe of discourse, membership function, fuzzy rule table, PID parameter adjustment range, and allowable positioning error range for position deviation and deviation change rate. Through the above parameter tuning, different equipment can achieve stable and reliable control effects under different operating conditions.

[0066] The engineering workstation is also used for communication network configuration. Engineers can configure OPC UA server and client nodes, data variable names, sampling periods, data access permissions, communication security policies, abnormal reconnection mechanisms, and alarm data mapping relationships through the engineering workstation. Through unified communication network configuration, standardized data interaction can be achieved between the field distributed control station, operator workstation, rolling time-domain coordinator, and engineering workstation, avoiding data silos caused by inconsistent communication protocols between different devices or control levels.

[0067] In addition, the engineering station is used for offline training and initial deployment of the time prediction model. The time prediction model employs a Gaussian process regression model. The engineering station can extract training samples from historical dismantling operation data. These training samples include data such as product type, dismantling requirements, corrosion level, degree of obstruction by connectors, degree of structural damage, operator proficiency, actual operation time, task completion progress, and anomaly handling records. After cleaning, labeling, and normalizing these samples, the engineering station trains the time prediction model using the Gaussian process regression method and deploys the trained model parameters to the time prediction submodules in each dismantling execution control unit.

[0068] After the initial deployment of the model, the engineering workstation can periodically update the time prediction model based on new data accumulated during system operation. For new models of decommissioned products, new dismantling processes, or new operator work data, the engineering workstation can supplement them into the training sample library and retrain or incrementally update the Gaussian process regression model to improve the model's adaptability to complex decommissioned product conditions. After completing the model update, the engineering workstation will distribute the new model parameters to the corresponding dismantling execution control unit, so that the on-site time prediction results can continuously approach the actual operation conditions.

[0069] Example 2 This invention also provides an intelligent production control method for the flexible dismantling of retired products, which is implemented through the intelligent production control system for the flexible dismantling of retired products shown in Embodiment 1.

[0070] like Figure 2 As shown, the method includes: S1. Collect the status of the dismantling station, logistics transportation, warehousing, and working condition labels obtained through visual recognition.

[0071] In this embodiment of the invention, real-time status data of the field distributed control station is collected through the OPC UA industrial communication network. The real-time status data includes the action parameters of the actuators of each dismantling station, the working condition tags obtained by visual recognition, the task execution progress, the logistics transportation location, and the warehouse inventory status.

[0072] S2. Based on the working condition labels collected in step S1, predict the remaining completion time and corresponding uncertainty variance of the current dismantling task, and generate the working time prediction result.

[0073] In this embodiment of the invention, based on the working condition label and operator proficiency, a Gaussian process regression model is used to predict the remaining completion time of the current dismantling task. and uncertainty variance Based on this, the predicted working hours for each workstation are generated.

[0074] S3. Based on the working condition labels collected in step S1 and the working time prediction results generated in step S2, adaptively adjust the control parameters of the disassembly actuator.

[0075] In this embodiment of the invention, the control parameters of the adaptive PID controller in the disassembly execution control unit are dynamically adjusted based on the predicted working hours and the change range of the visually recognized working condition label. The adjustment employs an improved RBF neural network dynamic adaptive PID control rule, wherein the PID weights are based on the deviation. and sensitivity information Iterative updates are performed, and when the operating condition label changes across levels or uncertainty variance occurs... When the threshold is exceeded, the dynamic learning rate Temporarily increased.

[0076] S4. Based on the task execution progress of the target dismantling station and the remaining completion time predicted in step S2, schedule logistics equipment to perform the transportation task.

[0077] In this embodiment of the invention, the logistics control unit obtains the task execution progress and predicted remaining working hours of the target dismantling station in real time through the OPC UA communication network. When the task completion percentage... ≥90%, predicted remaining working hours When the time is ≤1 minute, AGVs, RGVs or material transfer nodes are pre-scheduled to the workstation side to stand by, so as to achieve seamless connection between workstations.

[0078] S5. Determine the estimated completion time of the dismantling task based on the remaining completion time predicted in step S2, and schedule the warehousing equipment to perform inbound and outbound operations in conjunction with the logistics transportation status collected in step S1.

[0079] In this embodiment of the invention, the warehouse control unit receives the estimated off-line time of the dismantled products and the delivery arrival time of the logistics system, and uses a fuzzy PID position control algorithm to schedule the stacker crane and shuttle car in advance to complete the inbound and outbound operations, thereby realizing the rhythm coordination of warehousing, dismantling and logistics.

[0080] S6. Based on the dismantling station status, logistics transportation status, and warehousing status collected in step S1, and the time prediction results generated in step S2, construct a rolling optimization model that considers the uncertainty of time, and generate the task scheduling instructions and actuator settings for the current control cycle.

[0081] In this embodiment of the invention, a rolling time-domain coordination optimizer periodically collects global state data, constructs a rolling optimization model considering the variance of time uncertainty with the objective of minimizing the total order completion time, and solves for task allocation instructions and actuator settings within a finite prediction time domain; when the variance of a certain workstation... When the threshold is exceeded, the optimizer reserves extra buffer time in the subsequent task scheduling of that workstation.

[0082] S7. Send the task scheduling instructions and execution mechanism settings generated in step S6 to the corresponding dismantling execution mechanism, logistics equipment and warehousing equipment, and return to step S1 to enter the next control cycle.

[0083] In this embodiment of the invention, the current time instruction output by the optimizer is sent to the field distributed control station through the OPC UA communication network, and the process returns to step S1 to perform closed-loop control for the next cycle.

[0084] It should be noted that the specific methods of performing each step in the above embodiments have been described in detail in the embodiments of the system, and will not be elaborated here.

[0085] Example 3 To illustrate in detail the specific implementation process of the intelligent production control system and method for flexible dismantling of retired products proposed in Examples 1 and 2, this example uses scrapped electric vehicles as a typical retired product. The floor plan of the intelligent production line for flexible dismantling of scrapped electric vehicles on which this example is based is attached. Figure 3 As shown, the system adopts a flexible, flow-line layout, which can support continuous batch dismantling operations of retired vehicles of multiple models and in multiple states.

[0086] Reference Figure 3The on-site distributed control station includes a dismantling execution control unit, a logistics control unit, and a warehousing control unit. Among them: Disassembly execution control units: These are deployed at six independent control stations: body disassembly stations A1 and A2, chassis disassembly stations B1 and B2, and demolition stations C1 and C2. Each control station uses a Siemens S7-1500 series PLC, directly linked to the disassembly tools at each station to complete closed-loop control of the underlying actuators. The disassembly tools equipped at each station are as follows: Body disassembly workstations A1 / A2 are equipped with pneumatic wrenches (for bolt removal), hydraulic tools (for separating body panels), a plasma cutter (for cutting non-removable connectors), and a lifting disassembly platform (for adjusting vehicle height and operating angle). The PLC controls the torque and speed of the pneumatic wrenches, the pressure and stroke of the hydraulic tools, the current and cutting speed of the plasma cutter, and the lifting height and tilt angle of the lifting disassembly platform.

[0087] Chassis disassembly stations B1 / B2: Equipped with pneumatic wrenches, hydraulic tools, plasma cutters, and posture adjustment robotic arms (used to flip or tilt chassis components for easy disassembly of battery packs, subframes, etc.). PLC controls the joint motion trajectory, posture maintenance, and coordinated movements of the robotic arms.

[0088] Demolition stations C1 / C2: Equipped with an automotive dismantling machine (hydraulic shears, gripping arm) and a compactor (used to compact the shredded vehicle body scrap into blocks). The PLC controls the shearing force and opening distance of the dismantling machine, and the pressure and holding time of the compactor.

[0089] Logistics Control Unit: This unit is deployed as an onboard controller within the AGV (Automated Guided Vehicle), as well as at RGV (Railway Shuttle Vehicle) and roller conveyor nodes. The AGV is responsible for transporting vehicles to be dismantled or dismantled materials between workstations. The Logistics Control Unit uses OPC UA to obtain real-time task execution progress and predicted remaining working hours for each target workstation, dynamically adjusting the AGV's speed, path, and start / stop status to achieve on-demand material supply.

[0090] Warehouse Control Unit: Deployed in the intelligent automated warehouse, it connects to stacker cranes and shuttles. The warehouse control unit controls the stacker cranes to move up and down and horizontally along the aisles to complete the storage and retrieval of goods, and controls the shuttles to transfer material boxes between racks. Its control loop schedules the stacker cranes to the warehouse entrance in advance based on the expected off-line time of the dismantled products, and feeds back the real-time inventory status to the rolling time-domain coordinator.

[0091] Operator workstation: Used for full-process status monitoring, data visualization, and manual intervention. It allows real-time viewing of the operating parameters of dismantling tools at each workstation, AGV position and speed, stacker crane status, predicted working hours, and scheduling instructions, and supports operators in manually adjusting task allocation.

[0092] Rolling Time-Domain Coordination Optimizer: As the core of system scheduling, it is deployed on an independent server or operator workstation. It periodically collects global status data (including the busy / idle status of each workstation, tool load, AGV position, stacker crane status, time prediction, etc.) through OPC UA. With the goal of minimizing the total completion time, it constructs and solves a rolling optimization model that considers the uncertainty of time, and outputs task allocation instructions and actuator settings to each field control unit.

[0093] Engineer Station: Used for system hardware configuration, control loop parameter tuning, communication network configuration, and offline training and initial deployment of time prediction models, providing engineering support for stable system operation.

[0094] In this embodiment, the time prediction submodule for each dismantling station uses a Gaussian Process Regression (GPR) model to predict the remaining completion time of the current dismantling task online and quantify its uncertainty. The model input feature vector includes: real-time work condition labels (vehicle model, dismantling requirement level, corrosion level, degree of connector occlusion, and degree of structural damage) obtained by the visual recognition system, and the current operator's proficiency code (updated online through historical work data). The output is the percentage of the current task completed. Remaining working hours and its variance .

[0095] Taking vehicle body dismantling station A1 as an example, when a scrapped electric vehicle (model A-class car of a certain brand) enters the station, the vision system acquires images and outputs condition labels: rust level 3 (moderate to severe, bolts and body panels show obvious signs of rusting), connector obstruction level "high" (multiple wiring harnesses and pipes cover key bolts), and structural damage level "minor" (partial damage to the front bumper). The operator proficiency code is 0.85 (range 0-1, higher values ​​represent higher proficiency). These input features are then fed into a pre-trained GPR model to obtain the current task completion percentage. (New task just started), predict remaining working hours Uncertainty variance The prediction results are synchronously uploaded to the rolling time-domain coordinated optimizer via the OPCUA network and sent to the logistics control unit and the adaptive PID controller at this workstation.

[0096] As the dismantling operation proceeds, the prediction model is recalculated every 10 seconds and updated in real time. , and For example, after the pneumatic wrench has removed six bolts, the vision system reconfirms the number of remaining bolts and their corrosion status. At this point, the model outputs... , , If at a certain moment Exceeding a preset threshold (in this embodiment, it is set to...) The system determines that the current task has a high degree of uncertainty and triggers the buffer time reservation mechanism in subsequent scheduling.

[0097] In this embodiment, the control loop in the disassembly execution control unit adopts an improved RBF neural network dynamic adaptive PID controller. Taking the pneumatic wrench torque control loop of the vehicle body disassembly station A1 as an example, its control objective is to adjust the output torque in real time according to the current bolt specifications and corrosion status, so that the actual torque tracks the set value.

[0098] The RBF neural network identifier acquires the input (control voltage) of the pneumatic wrench in real time. ) and output (actual torque) Output sensitivity information PID controller weights Corresponding to proportionality coefficients Integral coefficient Differential coefficients The weight update rules are as follows:

[0099] Among them, deviation ( To set the torque, (actual torque) , , Dynamic learning rate The initial value is set to 0.01.

[0100] When the visual recognition of the working condition label changes across levels (for example, the current bolt corrosion level suddenly changes from level 2 in the previous task to level 3), the system automatically calculates the magnitude of the working condition level change. And according to preset rules Temporarily increased to Meanwhile, if the variance of the current task's time prediction... If the variance exceeds the threshold of 0.5, the learning rate is further increased to 0.03. This mechanism allows the PID controller to quickly adapt to drastic differences in the condition of retired products, avoiding torque overshoot or response delay caused by parameter mismatch. Once the operating conditions stabilize and the variance falls back below the threshold, the learning rate gradually returns to its baseline value.

[0101] In this embodiment, the logistics control unit subscribes to the task progress data of each dismantling station in real time through the OPCUA network. Taking the material flow between the body dismantling station A1 and the chassis dismantling station B1 as an example: after the body panels and interior of a scrapped car are dismantled at station A1, the remaining body-chassis assembly needs to be transported to station B1 via AGV for chassis component dismantling.

[0102] The logistics control unit continuously monitors the task completion percentage of workstation A1. and forecast remaining working hours .when and Upon arrival, the system immediately issues a dispatch command to the nearest idle AGV to workstation A1. The AGV departs from the standby area and travels at a speed of 0.5 m / s to the designated pick-up point at workstation A1, taking approximately 45 seconds. At this point, workstation A1 has just completed its remaining disassembly work, and the AGV can load the assembly without waiting, then proceeds to workstation B1 via an optimized route. During the AGV's journey, the logistics control unit simultaneously monitors the task progress at workstation B1: if workstation B1 still has tasks in progress, the AGV temporarily stores in the upstream buffer zone of workstation B1; if workstation B1 predicts remaining time... Then the AGV will drive directly into the unloading point, realizing "no waiting connection" for material transportation between workstations.

[0103] Actual testing showed that this mechanism reduced the average waiting time between workstations from 12 seconds for call response under the traditional scheduling mode to 0 seconds of no idle waiting time, significantly improving the production line cycle efficiency.

[0104] In this embodiment, the storage control unit is deployed in an intelligent automated warehouse, responsible for the storage of dismantled products and the supply of products to downstream processes. Taking the shredded vehicle body waste produced at dismantling station C1 as an example: after station C1 completes the crushing and compaction of a vehicle body, it produces a 0.8 cubic meter briquette, which needs to be transported by shuttle car to the designated storage location in the automated warehouse.

[0105] The warehouse control unit receives the estimated time off from the rolling time-domain coordinating optimizer via OPC UA (e.g., the remaining time predicted for the current task at workstation C1). The warehouse control unit receives the arrival time of the AGV from the logistics control unit (the AGV's transport time from workstation C1 to the inlet is approximately 1.5 minutes). Based on this, the warehouse control unit schedules the stacker crane to move from its current position to the inlet to stand by 2 minutes in advance, and schedules the shuttle to prepare to receive the material boxes.

[0106] The stacker crane's position control loop uses a fuzzy PID control algorithm, with the input being the position deviation. and the rate of change of deviation The output is the frequency command of the inverter. The fuzzy rule base adjusts the PID parameters online based on the deviation and the rate of change of the deviation, enabling the stacker crane to respond quickly when moving over a wide range and decelerate smoothly when approaching the target position, with a positioning accuracy of ±2mm. When the AGV arrives at the warehouse entrance, the stacker crane is already in standby mode, immediately completing the retrieval and storage of goods in the designated location. The entire inbound and outbound connection time does not exceed 10 seconds, avoiding queuing and waiting in the warehousing process.

[0107] In this embodiment, the rolling temporal coordinated optimizer performs a global optimization every 30 seconds. Let the current time be... The order involves the dismantling of 12 end-of-life electric vehicles, distributed across six workstations: A1, A2, B1, B2, C1, and C2. The optimizer collects the real-time status of each workstation. Workstation A1: Remaining work hours for the current task ,variance Workstation utilization rate is 80%; Workstation A2: Available. ; Workstation B1: Remaining work hours for the current task ,variance (Exceeding the threshold by 0.5); Workstation B2: Remaining work hours for the current task ,variance ; Stations C1 and C2: 5.0 min and 3.5 min remaining respectively.

[0108] The optimizer aims to minimize the total order completion time in the prediction time domain. A mixed-integer programming model is established internally. Due to the variance of workstation B1... The optimizer allocates an additional 0.5 minutes of buffer time for subsequent tasks at this workstation in the scheduling scheme. After solving, the optimizer outputs the current-time instruction: assign the next vehicle to be dismantled to the idle A2 workstation, and send the set torque, pressure, and other parameters of the A2 workstation to its dismantling execution control unit; at the same time, it recommends that the AGV prioritize transporting the materials completed at the B1 workstation to the B2 workstation (to avoid downstream congestion caused by fluctuations in the B1 work time). The optimizer only issues the instruction for the current moment, and the next cycle will re-roll and optimize, forming a closed loop.

[0109] Finally, the current-time command output by the optimizer is sent to the field distributed control station through the OPC UA communication network, and the system returns to the initial steps to perform closed-loop control for the next cycle.

[0110] Example 4 To quantitatively verify the effectiveness of the flexible dismantling intelligent production control system proposed in this invention, this embodiment uses 20 scrapped electric vehicles as simulation objects and conducts a comparative experiment under the same production line configuration.

[0111] The production line comprises three core disassembly stations: body disassembly station (A), chassis disassembly station (B), and demolition station (C). Each station has two parallel operation units: A1 / A2, B1 / B2, and C1 / C2. Vehicles must flow sequentially through stations A → B → C; skipping stations or operating in reverse order is prohibited. Based on visual recognition results, the 20 vehicles are categorized by disassembly difficulty: 8 vehicles of light difficulty (L), 8 vehicles of medium difficulty (M), and 4 vehicles of heavy difficulty (H). The disassembly time for each station is randomly generated according to the distribution shown in Table 1. Different time ranges are set to simulate the differences in disassembly time caused by vehicles of different difficulty (output by visual recognition conditions) and the randomness of operators.

[0112] Table 1. Range of dismantling time for each workstation based on visual recognition and operator randomness.

[0113] The comparison scheme is set as follows: Current traditional solutions (worker call-based): such as Figure 4 As shown, vehicles enter the production line in the order of V01→V20; the two parallel units of each workstation are allocated by polling, without flexible scheduling or precision scheduling; the logistics scheduling system adopts a call response mode (after the workstation is completed, a call is issued and the AGV departs from the standby area).

[0114] The present invention solution (flexible disassembly intelligent control): as follows Figure 5 As shown, by integrating the field control layer, monitoring and operation layer and optimization scheduling layer through the OPC UA industrial communication network, a closed-loop linkage control structure of dismantling-logistics-transportation is constructed to realize the rhythm coordination and waiting-free connection of the entire process of dismantling, logistics and warehousing. At the same time, the rolling time domain coordination optimizer dynamically decides the vehicle entry order and workstation allocation with the goal of minimizing the total order completion time.

[0115] Based on the simulation verification results shown in Tables 2 and 3, and Figure 6 The diagram showing the comparison of workstation utilization rates between the present invention and traditional solutions illustrates that the present invention has the following advantages over current traditional solutions: Table 2 Comparison of Total Completion Time and Equipment Utilization Rate

[0116] Table 3. Operation time and equipment utilization rate of each workstation

[0117] From Table 2 and Figure 6As can be seen, the present invention reduces the total completion time of 20 scrapped electric vehicles from 101.17 minutes to 94.46 minutes, a reduction of approximately 6.6%. Simultaneously, the overall equipment utilization rate increases from 71.0% to 75.0%, an increase of 4 percentage points, indicating a significant enhancement in overall capacity utilization efficiency. Table 3 shows that the equipment utilization rate of all workstations under the present invention is no lower than that of the traditional solution. Specifically, the increases for A1, A2, B1, and C1 are 6.23, 5.09, 8.26, and 6.62 percentage points, respectively. The standard deviation of the utilization rate for each workstation decreases from 15.2% in the traditional solution to 7.8%, indicating a significant improvement in load balance.

[0118] In summary, without increasing hardware resources, this invention effectively alleviates efficiency fluctuations caused by differences in vehicle difficulty and operational randomness through flexible scheduling and closed-loop collaborative control, significantly improving the overall dismantling efficiency and equipment utilization balance, and verifying the superiority and engineering practical value of the proposed system and method in complex dismantling scenarios.

[0119] Example 5 To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0120] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0121] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0122] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. An intelligent production control system for the flexible dismantling of retired products, characterized in that, include: The field distributed control station includes a disassembly execution control unit, a logistics control unit, and a warehousing control unit; The disassembly execution control unit is used to control the action parameters of the disassembly station's execution mechanism and adaptively adjust the control parameters based on the working condition tags obtained through visual recognition. The disassembly execution control unit includes a time prediction submodule, used to predict the remaining completion time of the disassembly task and the corresponding uncertainty variance based on the current working condition information. The logistics control unit is used to acquire the task execution progress of the target disassembly station and the remaining completion time output by the time prediction submodule, and to schedule logistics equipment to perform inter-station transport based on the acquired results. The warehousing control unit is used to schedule warehousing equipment to perform inbound and outbound operations of the disassembled products based on the estimated completion time of the disassembly task output by the time prediction submodule and the logistics transport status. The operator workstation is used to monitor the status of the dismantling station, the logistics transportation status, the time prediction results and the scheduling execution, and to receive manual intervention instructions. The rolling time-domain coordination optimizer is used to periodically collect status data of each dismantling station, logistics and warehousing, build a rolling optimization model that takes into account the uncertainty of time based on the time prediction results, and output task allocation instructions and execution mechanism settings. The field distributed control station, operator workstation, and rolling time-domain coordinator are connected via the OPC UA industrial communication network.

2. The intelligent production control system according to claim 1, characterized in that, The control loop of the disassembly execution control unit adopts an improved RBF neural network dynamic adaptive PID controller; The neural network identifier is used to obtain the input-output relationship of the disassembly actuator and output sensitivity information. ; PID controller weights , , Corresponding to proportionality coefficients Integral coefficient Differential coefficients Its update rules are as follows: in, The deviation between the set value and the actual output. , , , This is the dynamic learning rate.

3. The intelligent production control system according to claim 2, characterized in that, The dynamic learning rate The adjustment is made dynamically based on the change range of the working condition labels obtained by visual recognition and the uncertainty variance output by the working time prediction submodule. Specifically, when the working condition label changes across levels or the uncertainty variance exceeds a preset threshold, the dynamic learning rate is increased. This improves the controller's ability to adapt to changes in the status of retired products and fluctuations in operating time; the increase in the dynamic learning rate is positively correlated with the magnitude of changes in operating condition levels.

4. The intelligent production control system according to claim 1, characterized in that, The operating condition label includes product type, disassembly requirements, corrosion level, degree of obstruction of connectors, and degree of structural damage; The working condition label is synchronously input to the working time prediction submodule to construct a dismantling operation time prediction model that is strongly correlated with the status of retired products, providing a basis for adaptive adjustment of control loop parameters and advance scheduling of logistics.

5. The intelligent production control system according to claim 1, characterized in that, The time prediction submodule uses a Gaussian process regression model to predict the percentage of task completion for the current dismantling task based on the current visual recognition work condition labels and operator proficiency. Remaining working hours and uncertainty variance .

6. The intelligent production control system according to claim 1, characterized in that, The logistics control unit acquires real-time task execution progress data of the target dismantling workstation from the time prediction submodule and displays the task completion percentage. ≥90% and predicted remaining working hours If the time is ≤1 minute, the logistics equipment should be scheduled to be ready at the workstation in advance.

7. The intelligent production control system according to claim 1, characterized in that, The position control loop of the warehouse control unit adopts a fuzzy PID control algorithm to measure position deviation. and the rate of change of deviation As input, the PID parameters are adjusted online; The warehouse control unit schedules warehouse equipment to perform inbound and outbound operations in advance based on the expected completion time of the dismantling task and the arrival time of the logistics transportation.

8. The intelligent production control system according to claim 1, characterized in that, The rolling time-domain coordination optimizer, within each sampling period, calculates the current task progress of each dismantling station and the remaining completion time output by the time prediction submodule. With variance The task scheduling scheme aims to minimize the total completion time by rolling the solution within the finite prediction time domain, and only issues instructions for the current moment. When the uncertainty variance corresponding to a certain disassembly station When the preset threshold is exceeded, the rolling time-domain coordination optimizer reserves buffer time in the subsequent task scheduling of the dismantling station.

9. A smart production control method for the flexible dismantling of retired products, characterized in that, Includes the following steps: S1. Collect the status of the dismantling station, logistics and transportation, storage, and working condition labels obtained by visual recognition. S2. Based on the working condition labels collected in step S1, predict the remaining completion time and corresponding uncertainty variance of the current dismantling task, and generate a working time prediction result. S3. Based on the working condition labels collected in step S1 and the working time prediction results generated in step S2, adaptively adjust the control parameters of the disassembly actuator. S4. Based on the task execution progress of the target dismantling station and the remaining completion time predicted in step S2, schedule logistics equipment to perform the transportation task. S5. Determine the estimated completion time of the dismantling task based on the remaining completion time predicted in step S2, and schedule the warehousing equipment to perform inbound and outbound operations in conjunction with the logistics transportation status collected in step S1. S6. Based on the dismantling station status, logistics transportation status and warehousing status collected in step S1, and the time prediction results generated in step S2, construct a rolling optimization model that considers the uncertainty of time, and generate the task scheduling instructions and actuator settings for the current control cycle. S7. Send the task scheduling instructions and execution mechanism settings generated in step S6 to the corresponding dismantling execution mechanism, logistics equipment and warehousing equipment, and return to step S1 to enter the next control cycle.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in claim 9.

Images