A coordinated program control system for a sheet coating production line

Abstract

This invention discloses a collaborative program control system for a sheet coating production line, relating to the field of industrial automation and control technology. It includes a global optimizer connected to the coating control unit, curing scheduling unit, and logistics coordination unit, respectively. This optimizer is used to synchronously acquire wet film state prediction data, heat accumulation load prediction data, and path capacity prediction data at the start of each production cycle. By introducing a global optimizer and its built-in Markov decision process-based dynamic optimization model, this invention innovatively incorporates global states such as the process coupling strength between coating and curing, and the dynamic capacity of the logistics path into an online optimization framework. This model aims to maximize quality consistency and minimize unit energy consumption. It iterates strategies at the start of each production cycle, dynamically generating collaborative adjustment instructions, thereby achieving a fundamental shift from static planning control to dynamic global optimization.

Description

Technical Field

[0001] This invention relates to the field of industrial automation and control technology, and in particular to a collaborative program control system for a sheet metal coating production line. Background Technology

[0002] As modern manufacturing transforms towards high-end and intelligent manufacturing, sheet metal coating, as a key process for improving product appearance, corrosion resistance, and functionality, faces stringent requirements for production lines demanding high-variety, small-batch, high-quality, and low-energy consumption. Traditional coating production lines typically consist of multiple relatively independent sections connected in series, including pretreatment, coating, curing, and material conveying, each equipped with its own independent control system. While this decentralized control model can accomplish basic production functions, the lack of effective information exchange and collaboration mechanisms between subsystems results in the entire production process being essentially in a state of "local optimization, global suboptimal," making it difficult to adapt to dynamic order changes and process fluctuations, thus hindering further improvements in production flexibility, energy efficiency, and product consistency.

[0003] Currently, technological developments in coating production line automation primarily focus on precise control and local optimization of individual process steps. For example, in the coating process, high-precision sensors and closed-loop control algorithms are used to achieve stable control of coating flow and film thickness; in the curing process, advanced temperature control models and multi-temperature zone strategies are employed to ensure curing quality; and in logistics scheduling, programmable logic controllers and simple queuing logic are used to manage the flow of boards. Furthermore, some integrated solutions attempt to connect various sub-units through a supervisory control system (SCADA) to achieve centralized data monitoring and download of static production plans. However, the "coordination" of such systems is largely limited to data collection and command issuance; control decisions between units are still based on their own local information and fixed rules, failing to consider the real-time status of the entire production line for forward-looking dynamic coupling and global resource allocation.

[0004] The core problem facing current technological development lies in the lack of a central decision-making and control architecture capable of truly achieving deep multi-unit collaboration and online dynamic optimization. Specifically, this manifests in several ways: First, key subsystems such as pretreatment, coating, curing, and logistics form "information silos," with their control decisions fragmented and unable to respond in real-time to coupled changes in upstream and downstream process states, such as the dynamic impact of wet film state on curing heat demand and the compression of process windows due to logistics congestion. Second, traditional control strategies or static scheduling rules struggle to cope with potential disturbances during production, such as fluctuations in material properties, temporary equipment failures, and order insertions, leading to delayed adjustments and often sacrificing energy consumption or quality to maintain production continuity. Finally, local optimization objectives often conflict with global optimal objectives. For example, pursuing the highest speed at a single coating station may lead to an imbalance in the curing oven's heat load, while simply accelerating logistics may disrupt the settling time required for coating surface drying.

[0005] Therefore, it is essential to invent a collaborative program control system for sheet coating production lines to solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide a collaborative program control system for a sheet metal coating production line to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a collaborative program control system for a sheet coating production line, comprising a collaborative program control module, a production line execution module, and a data acquisition and communication module;

[0008] The data acquisition and communication module is used to collect real-time operating status data and process parameter data from various physical devices on the production line, and send the operating status data to the logistics coordination unit in the collaborative program control module, and send the process parameter data to the coating control unit and curing scheduling unit in the collaborative program control module respectively.

[0009] The collaborative program control module, as the central coordination unit of the system, includes:

[0010] The preprocessing planning unit is used to generate initial preprocessing instructions based on order information and can update them based on collaborative adjustment instructions;

[0011] The coating control unit is used to generate coating process instructions based on coating quality requirements and received relevant process parameter data, and output coating wet film state prediction data.

[0012] The curing scheduling unit is used to generate curing strategy instructions and output heat accumulation load prediction data based on the characteristics of the coating material, the production cycle and the received relevant process parameter data.

[0013] The logistics coordination unit is used to schedule the transport of the sheet material between the pretreatment station, coating station, curing station and buffer warehouse based on the received operating status data, and output path capacity prediction data.

[0014] A global optimizer is connected to the coating control unit, the curing scheduling unit, and the logistics coordination unit, respectively, and is used to synchronously acquire the wet film state prediction data, the heat accumulation load prediction data, and the path capacity prediction data at the beginning of each production cycle.

[0015] The global optimizer incorporates a dynamic optimization model based on a Markov decision process. This model uses the real-time coupling strength between the coating and curing processes, as well as the dynamic throughput of the logistics path, as state variables. Its action space consists of adjusting pretreatment time, coating parameters, curing strategies, and transport paths. The optimization objectives are to maximize the coating quality consistency index and minimize the production line's unit energy consumption. The global optimizer runs the dynamic optimization model to perform online strategy iteration, generating coordinated adjustment instructions for the pretreatment planning unit, coating control unit, curing scheduling unit, and logistics coordination unit.

[0016] The production line execution module is connected to the pretreatment planning unit, coating control unit, curing scheduling unit, and logistics coordination unit, respectively, and is used to receive and execute the pretreatment instructions, coating process instructions, curing strategy instructions, and logistics scheduling instructions after being corrected by the collaborative adjustment instructions.

[0017] The technical effects and advantages of this invention are as follows:

[0018] 1. This invention constructs an integrated control architecture with a collaborative program control module as the central coordinating unit, deeply integrating pretreatment, coating, curing and logistics scheduling functions. This architecture enables comprehensive centralized factory control of all physical equipment and processes in the board coating production line. The system breaks through the limitations of "information silos" and independent control of each section of the traditional production line, unifying the originally dispersed decision-making power into an intelligent hub with a global perspective, laying a solid system foundation for true full-line collaborative optimization.

[0019] 2. This invention collects real-time data on the operating status and process parameters of the entire production line through a data acquisition and communication module, and establishes a standardized data flow distribution mechanism. This enables multi-unit synchronous perception and status sharing based on real-time data-driven processes. This provides accurate and timely decision input for sub-units such as coating control, curing scheduling, and logistics coordination, ensuring that the control decisions of each local unit can respond to the dynamic changes of the entire production line. This is a prerequisite for achieving collaborative control.

[0020] 3. This invention, by introducing a global optimizer and its built-in dynamic optimization model based on Markov decision process, innovatively incorporates global states such as the coupling strength of coating and curing processes and the dynamic throughput of logistics paths into the online optimization framework. The model takes maximizing quality consistency and minimizing unit energy consumption as a unified goal, performs strategy iteration at the beginning of each production cycle, and dynamically generates collaborative adjustment instructions, thereby realizing a fundamental transformation from static planning control to dynamic global optimization.

[0021] 4. This invention achieves online flexible reconstruction and collaborative linkage of production resources (time, parameters, energy, and path) by accurately decomposing and distributing global optimization instructions to each unit of pretreatment planning, coating control, curing scheduling, and logistics coordination. The system can proactively adjust pretreatment time, coating parameters, curing strategy, and conveying path according to real-time status, effectively cope with disturbances such as order changes, process fluctuations, and equipment abnormalities, and significantly improve the flexibility and robustness of the production system.

[0022] 5. By employing reinforcement learning algorithms such as deep Q-networks, this invention enables the global optimizer to have online learning and adaptive evolution capabilities. The system can accumulate experience during continuous operation and continuously optimize its collaborative decision-making strategy, thereby continuously improving the overall intelligent control level of the production line. It can autonomously find and approach the global optimal or suboptimal operating point under varying operating conditions. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0024] Figure 2 This is a structural diagram of the preprocessing planning unit of the present invention.

[0025] Figure 3 This is a structural diagram of the coating control unit of the present invention.

[0026] Figure 4 This is a structural diagram of the solidified scheduling unit of the present invention.

[0027] Figure 5 This is a structural diagram of the local optimizer of the present invention. Detailed Implementation

[0028] 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.

[0029] This invention provides, for example Figure 1 The diagram shows a collaborative program control system for a sheet coating production line, which includes a collaborative program control module, a production line execution module, and a data acquisition and communication module.

[0030] In one specific embodiment, the hardware foundation of the data acquisition and communication module is a distributed industrial IoT gateway and sensor network, and its software core is a data aggregation and distribution service deployed on a central server. The real-time operational status data acquired by this module comes directly from the controllers and sensors of the physical equipment on the production line, such as the real-time speed and position feedback from the servo drives of the conveyor rollers and elevators, the board arrival signals detected by photoelectric switches, and the fault codes reported by the transfer machine controller. The acquired process parameter data comes from dedicated process monitoring instruments, such as the paint viscosity and flow rate measured by online viscometers and electromagnetic flowmeters installed on the paint circulation pipeline, the temperature distribution measured by thermocouple arrays arranged in each temperature zone of the curing oven, and the wind speed measured by an anemometer installed in the circulation duct.

[0031] All raw signals acquired on-site, including analog and digital signals, are standardized and converted from analog to digital through corresponding input modules. After being uploaded to the central server in real time, this data undergoes a predefined preprocessing process. This process includes using a Kalman filter to remove noise from key process parameters, converting all data to engineering units (e.g., converting analog current values ​​to specific temperatures in degrees Celsius or flow rates in liters per minute), adding a precise timestamp to each data point, and finally encapsulating it into a structured data message in a unified format.

[0032] In terms of communication architecture, field-level devices communicate with IoT gateways using high-speed, real-time industrial fieldbus protocols, such as PROFIBUS-DP or EtherCAT. The central server interacts with the upper-level collaborative program control modules via the industry-standard OPC UA protocol. Data distribution follows a publish-subscribe model: the central server publishes encapsulated operational status data to a logical node, such as an OPC UA node named "Production Line Status," which is received in real-time by the logistics coordination unit as the sole subscriber; simultaneously, it publishes processed process parameter data to another logical node, such as a node named "Process Parameters," for both the coating control unit and the curing scheduling unit to subscribe to. Each control unit can select the specific parameters it needs from the subscribed data stream according to its own requirements.

[0033] To ensure real-time control, the acquisition and update cycles for different data have been carefully configured. For rapidly changing parameters that directly affect control, such as coating head position and paint flow rate, the cycle is set in the hundreds of milliseconds. For slowly changing parameters, such as ambient temperature and humidity, the cycle can be relaxed to several seconds. The module also incorporates a data buffer queue and a reconnection mechanism to ensure that production data is not lost during temporary network interruptions and automatically resumes transmission after communication is restored, guaranteeing the system's robustness in complex industrial environments.

[0034] The collaborative program control module, as the central coordination unit of the system, includes:

[0035] The preprocessing planning unit is used to generate initial preprocessing instructions based on order information and can update them based on collaborative adjustment instructions;

[0036] Furthermore, in the above technical solution, refer to Figure 2 The preprocessing planning unit includes a process knowledge base and a scheduling instruction generation subunit;

[0037] The process knowledge base stores pretreatment process sequences corresponding to different material plates and coating types;

[0038] The scheduling instruction generation subunit is used to match the process sequence in the process knowledge base according to the order information, and combine it with the preprocessing time constraint issued by the global optimizer and adjusted by collaborative decision to generate the final preprocessing instruction.

[0039] In one specific embodiment, the pretreatment planning unit is implemented as follows. The process knowledge base is essentially a relational database table, whose storage structure defines the mapping relationship between "board material," "coating type," and "pretreatment process sequence." A complete process sequence records all the steps and their standard parameters that must be executed sequentially from the board entering the pretreatment station to its departure. For example: step one is "plasma cleaning," with parameters including power, gas flow rate, and processing time; step two is "primer spraying," with parameters including nozzle type, spraying distance, and film thickness requirements; step three is "surface drying," with parameters including settling time and ambient humidity requirements.

[0040] When a new production order arrives at the system, the scheduling instruction generation subunit first parses the order information, extracting the "sheet material code" and "target coating code." Then, using these two codes as a joint query key, the subunit performs a precise match in the process knowledge base to retrieve the corresponding standard preprocessing process sequence. This sequence constitutes the initial preprocessing instruction blueprint.

[0041] The collaborative adjustment instruction issued by the global optimizer is mainly manifested as a "preprocessing time constraint adjustment amount" for the current material to be processed. This adjustment amount is a time offset value with positive and negative values, in seconds. For example, to balance the load of subsequent coating and curing stations, the global optimizer may instruct "shorten the total preprocessing time by 30 seconds". After receiving this adjustment instruction, the scheduling instruction generation subunit will start a built-in process time compression algorithm. This algorithm does not simply shorten the time of each step proportionally, but intelligently adjusts according to predefined step priorities and process flexibility rules. For example, it will prioritize shortening the "surface drying" settling time (because of its greater flexibility), while carefully fine-tuning the time of key steps such as "plasma cleaning" while ensuring the lower limit of quality. Finally, the subunit integrates the adjusted parameters and times of each step to generate a structured final preprocessing instruction that can be directly parsed by the production line execution module. This instruction explicitly includes equipment action commands, process parameter settings, and adjusted timing information, thereby realizing precise and flexible preprocessing planning under the global collaborative optimization framework.

[0042] The coating control unit is used to generate coating process instructions based on coating quality requirements and received relevant process parameter data, and output coating wet film state prediction data.

[0043] Furthermore, in the above technical solution, refer to Figure 3 The coating control unit includes a coating thickness control subunit and a wet film state prediction subunit;

[0044] The coating thickness control subunit integrates a coating calculation model, which is used to calculate and generate coating flow control commands based on the target coating thickness and the received real-time coating viscosity and coating head speed process parameter data.

[0045] The wet film state prediction subunit is connected to the coating thickness control subunit and is used to predict and generate the wet film state prediction data based on the coating flow control command, coating trajectory and coating rheological properties.

[0046] In one specific embodiment, the core function of the coating control unit is jointly performed by a coating thickness control subunit and a wet film state prediction subunit. The coating calculation model integrated within the coating thickness control subunit is a physical model based on fluid dynamics and coating transfer efficiency. The core calculation formula of this model is expressed as: Q=(v*W*H) / η. Where Q represents the coating volume flow rate to be controlled, v is the real-time collected coating head travel speed, W is the coating width preset in the coating process, H is the target wet film thickness required by the order, and η is the transfer efficiency coefficient of the coating under the current operating conditions, obtained through experimental calibration. This subunit receives coating viscosity data from the data acquisition module in real time. When the viscosity fluctuates, the transfer efficiency coefficient η is dynamically corrected using a pre-stored viscosity-efficiency relationship table. The subunit performs calculations several times per second based on the above model, outputting the results as analog signals or set values ​​under a specific communication protocol, which are then sent to the actuator of the coating pump as immediate coating flow control commands.

[0047] The wet film state prediction subunit interacts with the coating thickness control subunit in real time. It receives coating flow control commands Q and coating head speed v from the former, as well as ambient temperature and humidity data from the data acquisition module. This subunit incorporates a differential equation model based on mass and heat transfer to simulate the dynamic changes of the wet film before it enters the curing oven. This model calculates the initial film thickness H using the flow rate Q and speed v. initial Starting with the coating's rheological properties (such as thixotropic index and solvent evaporation rate constant) and coating trajectory information (calculating the time each point on the substrate is exposed to air via the trajectory), the wet film state within a specific future time window (e.g., the next 3 minutes) is numerically solved. The predicted wet film state data output is a structured dataset, containing at least the predicted film thickness, solvent residue, and surface temperature at key points on the substrate in the future time series. This predicted data is provided in real-time to the global optimizer as a key input for evaluating the coupling state between the coating and curing processes.

[0048] The curing scheduling unit is used to generate curing strategy instructions and output heat accumulation load prediction data based on the characteristics of the coating material, the production cycle and the received relevant process parameter data.

[0049] Furthermore, in the above technical solution, refer to Figure 4 The curing scheduling unit includes a curing strategy library, a heat load prediction subunit, and a dynamic correction subunit;

[0050] The curing strategy library stores benchmark curing strategies corresponding to different coating materials;

[0051] The heat load prediction subunit is used to calculate and output the cumulative heat load prediction data based on the received wet film state prediction data and the board queue in the curing oven.

[0052] The dynamic correction subunit is used to dynamically correct the baseline curing strategy retrieved from the curing strategy library based on the thermal cumulative load prediction data and the received process parameter data of curing oven temperature and wind speed, and generate the curing strategy instruction.

[0053] In one specific embodiment, the curing scheduling unit achieves dynamic curing strategy formulation through the coordinated operation of a curing strategy library, a heat load prediction subunit, and a dynamic correction subunit. The curing strategy library is a formula table stored in a database, where each record corresponds to a coating material code and is associated with a baseline curing strategy. This strategy is specifically represented by a temperature-time process curve that varies over time, along with a set of parameters such as the corresponding circulating wind speed setpoints for each temperature zone and the heating power allocation ratio.

[0054] When a batch of boards enters the curing scheduling queue, the heat load prediction subunit begins operation. It receives wet film state prediction data from the coating control unit, which includes the predicted solvent residue, film thickness, and surface temperature of each board in the queue when it enters the curing oven. Simultaneously, the subunit is aware of the current spatial distribution and sequence of board entry within the curing oven. Based on this, the subunit uses a thermodynamic calculation model to predict the total heat load required by the curing oven over a future period. The core of this model is calculating the theoretical total heat required to heat and maintain the wet film of each board from its predicted entry state until its coating is fully cured. This calculation comprehensively considers the latent heat of solvent evaporation, the sensible heat rise of the coating and the board, and the heat loss of the oven itself. The final output of the cumulative heat load prediction data is a time-varying prediction curve, whose value represents the instantaneous heat power required by the curing oven to meet the curing requirements of all boards in the queue at each future time point.

[0055] The dynamic correction subunit is the core of the entire scheduling decision. It first retrieves the corresponding baseline curing strategy (temperature-time curve) from the curing strategy library based on the coating material code of the current production order. Then, it reads two key inputs in real time: the cumulative heat load prediction curve provided by the heat load prediction subunit, and the actual temperature and wind speed of each temperature zone in the curing oven fed back by the data acquisition module. This subunit has a built-in model prediction controller. This controller uses the baseline strategy as the setpoint and the actual oven conditions and predicted heat load as feedforward and feedback to solve the matching problem between "actual heating capacity" and "predicted demand load." For example, when the prediction shows that the heat load will peak in the next 2 minutes, while the current oven temperature is trending downward, the controller will calculate in advance and output adjustment instructions: appropriately increase the set temperature of the relevant temperature zone or increase the circulating wind speed to ensure sufficient heating when the load arrives. These dynamically calculated and adjusted parameters, such as temperature, wind speed, and conveyor belt speed, constitute the final curing strategy instruction adapted to the real-time production status and are sent to the execution mechanism of the curing oven.

[0056] The logistics coordination unit is used to schedule the transport of the sheet material between the pretreatment station, coating station, curing station and buffer warehouse based on the received operating status data, and output path capacity prediction data.

[0057] Furthermore, in the above technical solution, the logistics coordination unit includes a virtual logistics map and a traffic capacity prediction subunit;

[0058] The virtual logistics map is used to model the position and status of the sheet material at each workstation and on the conveying equipment in real time based on the received operating status data of the conveying equipment, elevators, and transfer machines.

[0059] The traffic capacity prediction subunit is connected to the virtual logistics map and is used to predict and output the path traffic capacity prediction data based on the current state of the virtual map and the equipment state.

[0060] In one specific embodiment, the logistics coordination unit, through the cooperation of a virtual logistics map and a capacity prediction subunit, achieves transparent management and proactive scheduling of the logistics system. The virtual logistics map is a topological data model updated in real-time in memory. It abstracts the physical layout of the entire production line into a directed graph structure, where nodes represent the intersections of preprocessing stations, coating stations, curing stations, buffer storage locations, and conveyor paths, and edges represent connecting equipment such as conveyor rollers and elevators. Each node and edge object is associated with a set of dynamic attributes, such as "occupancy status," "currently loaded sheet ID," and "equipment health status."

[0061] The map's real-time update mechanism is event-driven. The logistics coordination unit continuously receives operational status data streams from the data acquisition module, such as the triggering of a photoelectric switch or the start / stop of a roller conveyor motor. Each status data is parsed into an event, driving the status update of the corresponding object in the virtual map. For example, when the photoelectric switch installed at the entrance of the coating station detects the arrival of a sheet material, the node representing that station in the virtual map immediately changes its status from "idle" to "occupied," and its associated sheet material ID is updated to the current sheet material. By integrating the discrete events of all equipment, the system can construct a continuous logistics panorama in the virtual space that is synchronized with the physical world.

[0062] The capacity prediction subunit is deeply integrated with the virtual logistics map. It doesn't just observe the current state, but performs a forward-scrolling discrete event simulation based on the map's current snapshot and logistics rules (such as maximum equipment speed, safety distances, and path interlocking logic). The core of the simulation is a dynamic model based on queuing theory. Starting from the current moment, this model considers the movement of all boards already in transit and planned (e.g., predicting when boards will arrive at a certain branch point based on the upstream process's cycle time), simulating their movement, queuing, and passage through the virtual map network over a future period (e.g., the next 5-10 production cycles). The simulation output, the path capacity prediction data, is quantified as a matrix of the expected travel time or the number of boards that can pass per unit time for the critical path within each future time window. For example, the output data can be expressed as "Path A from the coating outlet to the curing oven inlet, during the next 3rd cycle, is expected to have a travel time of 45 seconds and a capacity of 0.8 boards / minute." This quantified, future-oriented prediction data provides the global optimizer with precise data for assessing logistics bottlenecks and collaboratively adjusting production rhythms.

[0063] A global optimizer is connected to the coating control unit, the curing scheduling unit, and the logistics coordination unit, respectively, and is used to synchronously acquire the wet film state prediction data, the heat accumulation load prediction data, and the path capacity prediction data at the beginning of each production cycle.

[0064] The global optimizer incorporates a dynamic optimization model based on a Markov decision process. This model uses the real-time coupling strength between the coating and curing processes, as well as the dynamic throughput of the logistics path, as state variables. Its action space consists of adjusting pretreatment time, coating parameters, curing strategies, and transport paths. The optimization objectives are to maximize the coating quality consistency index and minimize the production line's unit energy consumption. The global optimizer runs the dynamic optimization model to perform online strategy iteration, generating coordinated adjustment instructions for the pretreatment planning unit, coating control unit, curing scheduling unit, and logistics coordination unit.

[0065] Furthermore, in the above technical solution, refer to Figure 5 The global optimizer includes a state encoder, a policy iteration processor, and an instruction fusion unit;

[0066] The state encoder is used to encode the synchronously acquired wet film state prediction data, heat accumulation load prediction data, and path accessibility prediction data into the state vector of the dynamic optimization model.

[0067] The strategy iteration processor is used to run the dynamic optimization model, perform online strategy evaluation and improvement, and output the optimal action sequence;

[0068] The instruction fusion unit is used to convert the optimal action sequence into a set of collaborative control instructions that can be executed by the preprocessing planning unit, coating control unit, curing scheduling unit, and logistics coordination unit.

[0069] In one specific embodiment, the internal operating logic of the global optimizer is specifically implemented through the collaborative work of the state encoder, the policy iteration processor, and the instruction fusion unit. Its core is to build and run a reinforcement learning decision model that can be updated online.

[0070] The specific implementation of the state encoder is as follows:

[0071] The state encoder receives three prediction data streams from upstream: wet film state prediction data D wet Heat cumulative load forecast data D heat Path capacity prediction data D logistics The encoding process is as follows:

[0072] Data preprocessing: Synchronize and align each data stream and time-slice it to ensure that the data corresponds to the same future time window (e.g., the next 3 production cycles).

[0073] Feature extraction and computation:

[0074] Coupling strength characteristic f couple According to D wet The predicted solvent residue rate r of the board that will enter the curing oven in the kth cycle of the future. solvent (k) and film thickness h(k), and D heat The predicted heat power P that the curing oven can supply at the corresponding time is... supply (k), calculate the instantaneous supply-demand ratio: γ(k) = (α*Σ(r) solvent (k)*h(k))) / P supply (k), where α is the conversion factor for converting film thickness and solvent residue to theoretical heat demand. couple Let N be the mean and variance of the γ(k) sequence over the next N beats, for example, 3.

[0075] Logistics obstruction characteristics f delay From D logistics Extract the predicted travel time t of the critical path (such as the coating-curing path) over the next N cycles. path (k). Calculate its time relative to the standard travel time t. std Relative delay rate: d(k) = (t path (k)-t std ) / t std f delay Take the maximum value and trend (such as the slope of the linear fit) of the d(k) sequence.

[0076] System load characteristics f load : Including the current real-time load rate L of the curing oven oven (between 0 and 1), and the queue length Q of the pretreatment and coating stations. pre Q coat .

[0077] Vectorization and Normalization: In the specific implementation of the state encoder, the construction of the original feature vector Fraw aims to comprehensively quantify the cooperative state of the system. Its seven key parameters are extracted from two dimensions: future prediction and current load, respectively: the mean coupling strength f. couple，mean With variance f couple，var The series statistics, derived from the ratio of the theoretical heat demand of the board material to the predicted heating capacity of the curing oven over several future production cycles, characterize the average heat load pressure and its fluctuation risk faced by the system; the maximum delay rate f in the logistics obstruction characteristics. delay，max With delay trend f delay，trend This stems from the deviation analysis of the future travel time of the critical path from the standard value, used to quantify the most severe expected bottleneck and its dynamics of deterioration or mitigation; simultaneously, the real-time load status of the system is determined by the current load rate L of the curing oven. oven (i.e., the ratio of real-time power to rated power) and the pre-processing station queue length Q pre and coating station queue length Q coat These features are collectively described, directly reflecting the instantaneous operational intensity and material backlog at each core workstation. Then, using pre-stored historical maximum and minimum values ​​for each feature, F... raw Perform min-max normalization to obtain the standardized state vector S. t .

[0078] The specific implementation of the strategy iteration processor:

[0079] The policy iteration processor implements an online optimization model based on a Deep Q-Network (DQN). Its action space, reward function, and training mechanism are designed as follows:

[0080] Discretization of action space A: The system presets M basic adjustment actions, for example: {a1: preprocessing time ± ΔT} pre a2: Coating head speed ±ΔV, a3: Curing reference temperature ±ΔT oven a4: Select path A / B}. The adjustment amount Δ for each action is set to a fixed step size (e.g., ΔT) according to the process allowable range. pre =5 seconds). The global action space is a finite combination of these basic actions. For example, selecting no more than two actions to execute simultaneously constitutes a discrete action set of controllable size.

[0081] The specific calculation of the reward function R: The reward function is calculated after each production cycle and fed back to the policy iteration processor.

[0082] Quality Consistency Index Q Index Based on online testing data of the cured substrate within this cycle (such as film thickness uniformity σ measured by an infrared thickness gauge). thick The gloss difference Δ measured by a gloss meter gloss ) Calculate. Q Index =exp(-(w thick *σ thick ^2+w gloss *Δ gloss ^2)), where w thick w gloss These are the weighting coefficients.

[0083] Unit energy consumption index E Index E is based on the sum of the meter readings of all major energy-consuming equipment (coating pump, curing heater, conveyor motor) within this cycle. total , with the area S of qualified boards produced good The calculation formula is E. Index =E total / S good .

[0084] Production delay penalty P delay : Calculate the actual beat time T of this beat. actual With the theoretical optimal beat T theory The difference.

[0085] Final reward value: R t =β1*Q Index -β2*E Index -β3*max(0,T) actual -T theory β1, β2, and β3 are reward weights that can be adjusted through the human-machine interface. The initial values ​​are set according to production preferences (e.g., β1=100, β2=10, β3=5).

[0086] Online learning mechanism:

[0087] Network structure: Both the main network and the target network of DQN are fully connected neural networks with two hidden layers (e.g., 128 and 64 neurons), and the input is the state vector S. t The output is the Q-value for each possible action a∈A.

[0088] Experience replay: Set a fixed-capacity experience replay buffer (e.g., 10,000 entries). Experience (S) per beat t a t R t S {t+1} The data is stored in the buffer.

[0089] Training Update: At the beginning of each beat, the policy iteration processor first selects and executes action a based on the current main network and the ε-greedy policy (initial ε=0.2, gradually decaying during training). t Then, a small batch (e.g., 64 samples) of experience is randomly sampled from the buffer, and the target Q value y = R + γ*max[Q] is calculated. target (S {t+1} ,a')]. Where, max[Q target (S {t+1} ,a')] indicates using the target network Q target Calculate the next state S {t+1} Find the Q-values ​​for all possible actions a', and take the maximum value, where γ is a discount factor (e.g., 0.9). target This is the output of the target network. The parameters of the main network are updated using backpropagation by minimizing the mean square error between the output Q value and y. The main network parameters are copied to the target network every fixed number of beats (e.g., 100 beats).

[0090] Policy Output: At the beginning of each cycle, the policy iteration processor outputs the current state S. t Input the trained main network and select the action 'a' with the highest Q value. t* This is the optimal action at present.

[0091] The specific implementation of the instruction fusion unit is as follows:

[0092] The instruction fusion unit maintains an "action-instruction mapping table". This table defines each discrete action a. t* How to convert it into a specific set of instructions that can be executed by each downstream unit.

[0093] For example, if a t* For the condition {preprocessing time + 5 seconds, path B selected}, the instruction fusion unit executes:

[0094] Generate an instruction for the preprocessing planning unit: {Instruction type: Adjust time, Target board ID: XXX, Time increment: +5 seconds}.

[0095] Generate an instruction for the logistics coordination unit: {Instruction type: Path redirection, Target board ID: XXX, Target path: Path B}.

[0096] The instruction fusion unit packages these instructions and publishes them to the instruction nodes subscribed to by the corresponding units through the OPC UA service interface of the data acquisition and communication module, thereby completing the issuance of coordinated adjustment instructions.

[0097] Furthermore, in the above technical solution, the calculation method of the real-time coupling strength is as follows: based on the wet film state prediction data and the heat accumulation load prediction data, calculate the weighted sum of the matching error between the theoretical heat input required by each plate in the curing oven and the actual heat supply capacity.

[0098] In one specific embodiment, the calculation of the real-time coupling strength is implemented through a proprietary matching error algorithm. The input to this algorithm is the wet film state prediction data and the thermal cumulative load prediction data. The calculation process is as follows: First, for the i-th board in the current queue within the curing oven, based on the key parameter in its wet film state prediction data—the solvent residue M—is measured. i (Unit: g / m²) and predicted film thickness H i (Unit: μm), calculate the required theoretical heat input E. i，req This calculation is based on a simplified coating curing reaction kinetics and heat transfer model, the core formula of which is as follows:

[0099] E i，req =Q rxn +Q evap +Q sens ,

[0100] Among them, Q rxn The heat required for the curing reaction of the coating resin is expressed by the formula Q. rxn =ρ*H i The formula is estimated as *A*ΔH*(1-exp(-k*t)), where ρ is the coating density, A is the single-sided area of ​​the board, ΔH is the curing reaction heat per unit mass of coating (which can be measured by differential scanning calorimetry), k is the curing reaction rate constant, and t is the planned residence time of the board in the curing oven.

[0101] Q evap The heat required for solvent evaporation is expressed by the formula Q.evap =M i *A*L v Calculate, L v This is the latent heat of vaporization of the solvent.

[0102] Q sens The sensible heat required to raise the temperature of the sheet material and coating from the furnace temperature to the target temperature is calculated using specific heat capacity and temperature rise.

[0103] Secondly, from the heat accumulation load forecast data, obtain the predicted heat supply E that the curing oven system can actually apply to the board during the planned residence time of the board under the current operating strategy. i，sup This value is calculated based on the heat balance model of the curing oven. The inputs to the model are the set temperature, wind speed and oven heat loss coefficient in the curing strategy command, and the output is the integral of the heat flux effectively applied to the board over time and area.

[0104] Subsequently, the individual matching error δ of the board was calculated. i It is defined as the relative error between theoretical demand and actual supply, i.e., δ. i =(E i，req -E i，sup ) / E i，req Finally, the real-time coupling strength I is obtained by weighted summation of all individual errors in the furnace plate, and the calculation formula is: I=Σ(ω i *|δ i |). Where, ω i The weighting coefficient assigned to the i-th plate can be configured based on the plate's priority or the criticality of coating quality. The coupling strength I is a dimensionless scalar indicator; the closer its value is to 0, the better the supply-demand match. A positive value indicates an overall excess of heat demand, while a negative value indicates a surplus of heat supply.

[0105] The production line execution module is connected to the pretreatment planning unit, coating control unit, curing scheduling unit, and logistics coordination unit, respectively, and is used to receive and execute the pretreatment instructions, coating process instructions, curing strategy instructions, and logistics scheduling instructions after being corrected by the collaborative adjustment instructions.

[0106] In one specific embodiment, the production line execution module serves as the physical execution terminal for control commands, and its core is an instruction agent and driver system deployed on an industrial control computer. This system maintains a connection with the upstream coordinating program control module through standard industrial communication protocols (such as OPC UA, Modbus TCP) and subscribes to a dedicated instruction issuing node.

[0107] After the preprocessing planning unit, coating control unit, curing scheduling unit, and logistics coordination unit generate final instructions, these instructions are encapsulated and published to this node. The instruction receiver and parser of the production line execution module monitors this node in real time to obtain structured instruction data packets. Each instruction packet contains an instruction type, a target device identifier, one or more parameter key-value pairs, and a timestamp. After the parser performs format and validity checks on the instructions, it passes them to the instruction dispatch and execution engine.

[0108] The execution engine maintains a device driver library, which configures dedicated drive protocol adapters for each type of physical execution device on the production line (such as pretreatment plasma power supplies, coating head servo axes, paint metering pumps, curing oven temperature controllers, and conveyor frequency converters). The engine invokes the corresponding drive adapter based on the target device identifier in the instruction. The core function of the drive adapter is to convert abstract process parameters into low-level control commands that the device can recognize. For example:

[0109] For a coating process instruction, the parameter "coating flow rate setpoint: 150 ml / min" will be converted by the coating pump drive adapter into a specific analog output value or PROFIBUS control word sent to the pump controller.

[0110] For a "logistics scheduling instruction", the parameter "next path: move to curing oven inlet B line" will be converted into a series of pre-programmed logic control commands by the conveyor line PLC drive adapter, which will trigger the actions of the corresponding roller conveyor motor, elevator and positioning cylinder in sequence.

[0111] The instruction execution engine issues these low-level commands to the target equipment sequentially or in parallel. Simultaneously, the execution status monitor within the module reads feedback signals from the equipment in real time (such as "positioning complete" for the actuator and "opening feedback" for the valve) and compares them with the expected state of the instructions to confirm whether the instructions have been executed correctly. The monitor summarizes important execution results (such as "instruction delivered," "equipment started," "target achieved," or "execution timeout / failure"), generates execution feedback information, and returns it to the relevant units of the collaborative program control module via the communication link. Thus, the production line execution module achieves a reliable, closed-loop transition from collaborative optimization decisions to physical equipment actions, ensuring the accurate implementation of the global optimization strategy on the production line.

[0112] Furthermore, the above technical solution also includes a human-computer interaction and monitoring module, which includes a data visualization engine and an instruction input interface;

[0113] The data visualization engine is used to graphically render system status information;

[0114] The instruction input interface is used to receive verified external instructions and forward them to the collaborative program control module.

[0115] In one specific embodiment, the human-computer interaction and monitoring module serves as a bridge between the operator and the collaborative program control system, and is implemented collaboratively by a data visualization engine and an instruction input interface.

[0116] The data visualization engine is a graphical rendering application developed based on web technology or industrial configuration software. It acquires real-time system status information by subscribing to data nodes published by the global optimizer, various control units, and data acquisition modules within the coordinating program control module. The core function of the engine is to fuse and render this multi-source, heterogeneous real-time data with historical data onto a unified graphical monitoring interface. This interface includes at least the following views:

[0117] Production line panoramic dynamic view: In two-dimensional or pseudo-3D animation, it intuitively displays the real-time location and status of the board material in the pretreatment, coating, curing, buffering and conveying path, and overlays key information of the virtual logistics map.

[0118] In-depth monitoring view of process parameters: Key process parameters such as coating viscosity, coating head speed, temperature of each temperature zone, predicted value of solvent residue rate, and real-time coupling strength are displayed in real time and can be traced back in history in the form of trend curves, digital instruments, color temperature diagrams, etc.

[0119] System Health and Performance View: Displays key performance indicators such as equipment utilization rate, current production cycle time, real-time estimated value of quality consistency index, and cumulative value of unit energy consumption in a Kanban format, as well as a list of equipment alarms and faults.

[0120] The command input interface is a command receiving channel with strict security and logical verification mechanisms. It provides a graphical form, button, or script editor, allowing authorized operators to input external commands, such as "emergency stop at a workstation," "insert a high-priority order," "manually set the batch curing time," or "switch global optimization mode." All input commands must pass through a multi-layered validator before being forwarded.

[0121] Permission verification: Verify the identity and role of the currently logged-in operator to determine whether they have permission to execute this type of command.

[0122] Syntax and range validation: Checks whether the instruction format conforms to the predefined protocol and whether the parameter values ​​are within a reasonable range allowed by equipment safety and process requirements.

[0123] Logical status verification: Based on the real-time status of the current production line, such as whether the equipment is ready or in automatic mode, determine whether the instruction can be safely executed at this moment to avoid conflicts or dangerous operations.

[0124] Only after an instruction passes all checks will the instruction input interface encapsulate it into a standard format message and forward it to the corresponding processing unit in the coordinating program control module via a secure communication link (such as an authorized OPC UA call). For instructions that fail checks, the interface will immediately return a clear error reason to the operator. This module ensures the intuitiveness of personnel monitoring and the security and effectiveness of intervention operations.

[0125] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A collaborative program control system for a sheet metal coating production line, characterized in that, It includes a collaborative program control module, a production line execution module, and a data acquisition and communication module; The data acquisition and communication module is used to collect real-time operating status data and process parameter data from various physical devices on the production line, and send the operating status data to the logistics coordination unit in the collaborative program control module, and send the process parameter data to the coating control unit and curing scheduling unit in the collaborative program control module respectively. The collaborative program control module, as the central coordination unit of the system, includes: The preprocessing planning unit is used to generate initial preprocessing instructions based on order information and can update them based on collaborative adjustment instructions; The coating control unit is used to generate coating process instructions based on coating quality requirements and received relevant process parameter data, and output coating wet film state prediction data. The curing scheduling unit is used to generate curing strategy instructions and output heat accumulation load prediction data based on the characteristics of the coating material, the production cycle and the received relevant process parameter data. The logistics coordination unit is used to schedule the transport of the sheet material between the pretreatment station, coating station, curing station and buffer warehouse based on the received operating status data, and output path capacity prediction data. A global optimizer is connected to the coating control unit, the curing scheduling unit, and the logistics coordination unit, respectively, and is used to synchronously acquire the wet film state prediction data, the heat accumulation load prediction data, and the path capacity prediction data at the beginning of each production cycle. The global optimizer incorporates a dynamic optimization model based on a Markov decision process. This model uses the real-time coupling strength between the coating and curing processes, as well as the dynamic throughput of the logistics path, as state variables. Its action space consists of adjusting pretreatment time, coating parameters, curing strategies, and transport paths. The optimization objectives are to maximize the coating quality consistency index and minimize the production line's unit energy consumption. The global optimizer runs the dynamic optimization model to perform online strategy iteration, generating coordinated adjustment instructions for the pretreatment planning unit, coating control unit, curing scheduling unit, and logistics coordination unit. The production line execution module is connected to the pretreatment planning unit, coating control unit, curing scheduling unit, and logistics coordination unit, respectively, and is used to receive and execute the pretreatment instructions, coating process instructions, curing strategy instructions, and logistics scheduling instructions after being corrected by the collaborative adjustment instructions.

2. The collaborative program control system for a sheet metal coating production line according to claim 1, characterized in that, The preprocessing planning unit includes a process knowledge base and a scheduling instruction generation subunit; The process knowledge base stores pretreatment process sequences corresponding to different material plates and coating types; The scheduling instruction generation subunit is used to match the process sequence in the process knowledge base according to the order information, and combine it with the preprocessing time constraint issued by the global optimizer and adjusted by collaborative decision to generate the final preprocessing instruction.

3. The collaborative program control system for a sheet metal coating production line according to claim 1, characterized in that, The coating control unit includes a coating thickness control subunit and a wet film state prediction subunit; The coating thickness control subunit integrates a coating calculation model, which is used to calculate and generate coating flow control commands based on the target coating thickness and the received real-time coating viscosity and coating head speed process parameter data. The wet film state prediction subunit is connected to the coating thickness control subunit and is used to predict and generate the wet film state prediction data based on the coating flow control command, coating trajectory and coating rheological properties.

4. A collaborative program control system for a sheet metal coating production line according to claim 1, characterized in that, The curing scheduling unit includes a curing strategy library, a heat load prediction subunit, and a dynamic correction subunit; The curing strategy library stores benchmark curing strategies corresponding to different coating materials; The heat load prediction subunit is used to calculate and output the cumulative heat load prediction data based on the received wet film state prediction data and the board queue in the curing oven. The dynamic correction subunit is used to dynamically correct the baseline curing strategy retrieved from the curing strategy library based on the thermal cumulative load prediction data and the received process parameter data of curing oven temperature and wind speed, and generate the curing strategy instruction.

5. A collaborative program control system for a sheet metal coating production line according to claim 1, characterized in that, The logistics coordination unit includes a virtual logistics map and a traffic capacity prediction subunit; The virtual logistics map is used to model the position and status of the sheet material at each workstation and on the conveying equipment in real time based on the received operating status data of the conveying equipment, elevators, and transfer machines. The traffic capacity prediction subunit is connected to the virtual logistics map and is used to predict and output the path traffic capacity prediction data based on the current state of the virtual map and the equipment state.

6. A collaborative program control system for a sheet metal coating production line according to claim 1, characterized in that, The global optimizer includes a state encoder, a policy iteration processor, and an instruction fusion unit; The state encoder is used to encode the synchronously acquired wet film state prediction data, heat accumulation load prediction data, and path capacity prediction data into the state vector of the dynamic optimization model. The strategy iteration processor is used to run the dynamic optimization model, perform online strategy evaluation and improvement, and output the optimal action sequence; The instruction fusion unit is used to convert the optimal action sequence into a set of collaborative control instructions that can be executed by the preprocessing planning unit, coating control unit, curing scheduling unit, and logistics coordination unit.

7. A collaborative program control system for a sheet metal coating production line according to claim 1, characterized in that, The real-time coupling strength is calculated as follows: based on the wet film state prediction data and the heat accumulation load prediction data, the weighted sum of the matching error between the theoretical heat input required by each plate in the curing oven and the actual heat supply capacity is calculated.

8. A collaborative program control system for a sheet metal coating production line according to claim 1, characterized in that, It also includes a human-computer interaction and monitoring module, which includes a data visualization engine and a command input interface; The data visualization engine is used to graphically render system status information; The instruction input interface is used to receive verified external instructions and forward them to the collaborative program control module.

Images