Intelligent scheduling and coordination system for glass bottle deep processing flexible production line

Abstract

This application relates to the field of industrial automation and production scheduling technology, and discloses an intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles. This system acquires the initial physicochemical state variables and work order characteristic parameters of the production line, and predicts the concentration of the solution during processing. When the predicted concentration of the solution is lower than the minimum allowable concentration of the main production work order, a secondary priority work order is selected as the filling batch. The system calculates the temperature adjustment response time based on the difference in reference baking temperature between the main production work order and the filling batch, and generates a virtual physical gap without workpiece occupation between batches by combining the linear speed of the conveyor belt. The system calculates the thermal radiation absorption conversion coefficient based on the predicted solution concentration and material characteristic parameters, and determines the temperature feedforward bias. The system controls the conveyor belt to operate according to this virtual physical gap, and the temperature feedforward bias is superimposed on the equipment temperature control. This invention reduces unplanned downtime of continuous production lines and lowers the product scrap rate during temperature zone switching.

Description

Technical Field

[0001] This invention relates to the field of industrial automation and production scheduling technology, specifically to an intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles. Background Technology

[0002] Continuous production lines for deep processing of glass bottles typically include preceding chemical treatment processes and subsequent heat treatment processes, such as frosting and baking. In actual production, as processing batches progress, the solution concentration at the chemical treatment station is continuously consumed and decreases, while different batches of products also have different requirements for the processing temperature at the heat treatment station.

[0003] Existing production scheduling systems mostly rely on preset static production plans and control each processing station as an independent unit. When the solution concentration decreases below the minimum physicochemical standard of the current production order, the conventional approach is to directly interrupt the feeding and stop the machine, forcing physical replenishment or liquid replacement. This rigid execution logic between processes leads to frequent start-ups and shutdowns of continuous production lines, reducing the continuous operating rate of the equipment. During the switching between batches of products with different temperature requirements, the furnace temperature rise and fall of the heat treatment equipment exhibits an inherent thermodynamic response lag. Because the existing transmission line control system fails to achieve physical decoupling from the equipment's temperature adjustment time, the preceding and following batches of products are closely connected during the temperature zone switching phase, easily resulting in uneven heating of workpieces in the transition zone and increasing the product scrap rate. In addition, changes in the concentration of the preceding solution directly alter the physical properties of the glass bottle surface, such as roughness, thereby affecting the infrared thermal radiation absorption rate of the product in the subsequent heat treatment station. The existing control logic lacks a calculation and compensation mechanism for such cross-process physicochemical parameter coupling changes, failing to provide effective temperature feedforward intervention in the early stages of processing, further affecting the consistency of product quality under multi-batch flexible production conditions. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent scheduling and collaborative system for a flexible production line in the deep processing of glass bottles. This system solves the problems in existing continuous production lines for deep processing of glass bottles, where the solution concentration decay in the preceding chemical treatment process and the furnace temperature switching in the subsequent heat treatment process are independent. When the solution concentration falls below the production standard, forced shutdown and replenishment are required, and there is a thermodynamic lag in furnace temperature adjustment when switching between different batches of products. This independent control between processes leads to frequent production line downtime, increased scrap rates, and reduced overall equipment utilization.

[0005] To address the above problems, the present invention provides the following technical solution: The first aspect of this invention provides an intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles, comprising at least one processor and a memory, wherein the memory stores a computer program, and the computer program, when executed by the processor, performs the following steps: Obtain the initial physicochemical state variables of the flexible production line and the characteristic parameters of the production orders to be scheduled; Based on the initial concentration of the frosting solution in the initial physicochemical state variables and the preset concentration decay model, the predicted concentration of the solution during the processing is predicted. When the predicted concentration of the solution is lower than the minimum allowable concentration corresponding to the main production order, a secondary priority work order with a minimum allowable concentration not higher than the current predicted concentration of the solution is selected from the work orders to be scheduled and used as the filling batch. Based on the difference in baseline baking temperature between the main production order and the filling batch, the temperature adjustment response time is calculated, and combined with the linear speed of the conveyor belt, a virtual physical gap without workpiece occupation is generated between the main production order and the filling batch. Based on the predicted concentration of the solution and the material characteristic parameters of the filling batch, the surface thermal radiation absorption conversion factor of the batch of products is calculated, and the temperature feedforward bias is determined accordingly. The flexible production line's mesh belt conveyor operates according to a virtual physical gap, and before the filling batch enters the baking station, the temperature feedforward bias is superimposed on the reference baking temperature of the filling batch to perform equipment temperature control.

[0006] Furthermore, the initial physicochemical state variables also include: the linear velocity of the conveyor belt and the normal physical gap time between consecutive batches; the characteristic parameters of the work orders to be scheduled include: the surface area of ​​a single product, the number of work orders, the minimum allowable concentration, the reference baking temperature, and the material characteristic parameters including the empirical coefficient of consumption per unit area and the material coupling attenuation factor.

[0007] Furthermore, based on the difference in baseline baking temperature between the main production order and the seam filling batch, the temperature adjustment response time is calculated. Combined with the linear speed of the conveyor belt, a virtual physical gap without workpiece occupation is generated between the main production order and the seam filling batch. Specifically, this includes: The temperature adjustment response time is determined by dividing the absolute value of the difference between the reference baking temperature between the main production order and the filling batch by the preset temperature adjustment rate of the baking station; the length of the virtual physical gap is the product of the linear speed of the mesh belt transmission line and the effective response time, where the effective response time is the time difference after deducting the normal physical gap time from the temperature adjustment response time, and when the time difference is less than zero, the effective response time is zero.

[0008] This step calculates the time required for heating and cooling and converts it into the distance on the transmission line, creating a physical interval between batches with different temperature requirements, thus reducing uneven heating caused by equipment temperature regulation lag.

[0009] Furthermore, based on the predicted concentration of the solution and the material characteristic parameters of the grouting batch, the surface thermal radiation absorption reduction coefficient of the batch of products is calculated, and the temperature feedforward bias is determined accordingly, specifically including: The surface thermal radiation absorption conversion factor is obtained by subtracting the attenuation product from the preset value of 1. The attenuation product is the product of the material coupling attenuation factor of the filling batch and the difference between the preset calibration concentration and the predicted concentration of the solution. The temperature feedforward bias is the product of the thermodynamic proportional compensation coefficient of the baking station, the conversion deviation rate, and the reference baking temperature of the filling batch. The conversion deviation rate is obtained by subtracting the surface thermal radiation absorption conversion factor from the preset value of 1.

[0010] This step establishes the calculated relationship between solution concentration and the thermal radiation absorption coefficient of the material surface, and generates a temperature offset before processing to compensate for the change in product thermal absorption rate caused by the decrease in concentration.

[0011] Furthermore, when a computer program is executed by a processor, the following closed-loop feedback steps are also implemented: Obtain the actual surface temperature of the product after the grouting batch leaves the baking station; based on the actual surface temperature of the product, the temperature feedforward bias, and the material characteristic parameters of the grouting batch, back-calculate the observed value of the material coupling attenuation factor under the processing state of the corresponding batch; using the Kalman filter algorithm, take the observed value of the material coupling attenuation factor as the observation input, and perform a posteriori iterative update on the original material coupling attenuation factor in the memory.

[0012] Furthermore, the observed values ​​of the material coupling attenuation factor under the corresponding batch processing conditions are calculated by reverse engineering, specifically including: The observed value of the material coupling attenuation factor is the quotient of the residual temperature deviation and the theoretical maximum compensation. The residual temperature deviation is the sum of the temperature feedforward bias and the reference baking temperature of the filling batch, minus the actual surface temperature of the product. The theoretical maximum compensation is the product of the thermodynamic proportional compensation coefficient of the baking station, the difference between the preset calibration concentration and the predicted concentration of the solution, and the reference baking temperature of the filling batch.

[0013] Furthermore, the Kalman filter algorithm is used to perform a posteriori iterative updates on the original material coupling attenuation factor in the memory, specifically including: Calculate the Kalman gain, which is the ratio of the error covariance of the previous batch to the sum of the error covariance of the previous batch and the preset measurement noise variance. Obtain the compensation adjustment term by multiplying the difference between the observed material coupling attenuation factor and the prior material coupling attenuation factor by the Kalman gain. Add the compensation adjustment term to the prior material coupling attenuation factor to obtain the updated a posteriori material coupling attenuation factor, and overwrite it in memory. Subtract the Kalman gain by a preset value of 1 to obtain the gain margin. Multiply the gain margin by the error covariance of the previous batch to obtain the synchronously reduced updated error covariance.

[0014] This step uses the actual temperature deviation to iteratively update the attenuation factor, and reduces measurement noise interference through the covariance reduction mechanism, thereby improving the accuracy of material characteristic parameter calculation.

[0015] Furthermore, based on the initial concentration of the frosting solution in the initial physicochemical state variables and the preset concentration decay model, the predicted concentration of the solution during the processing is predicted, specifically including: For each batch of work orders, the theoretical solution consumption of the corresponding batch of work orders is obtained by multiplying the empirical coefficient of unit area consumption, the surface area of ​​the single product, and the number of work orders by the physical volume of the frosting tank. The predicted solution concentration at any production node is obtained by successively subtracting the cumulative value of the theoretical solution consumption of the scheduled work orders from the initial concentration of the frosting solution.

[0016] Furthermore, when a computer program is executed by a processor, the following exception handling steps are also implemented: When the predicted concentration of the solution is lower than the minimum allowable concentration corresponding to the main production work order, and no candidate work order with a minimum allowable concentration not higher than the current predicted concentration of the solution is found when traversing the work orders to be scheduled, a physical replenishment or liquid replacement shutdown command is issued to the field control equipment.

[0017] Furthermore, the control of the flexible production line's mesh belt transmission line to operate according to virtual physical idle time includes: After the last workpiece in the main production schedule enters the conveyor belt, the linear speed of the conveyor belt is kept constant, and the entry of subsequent workpieces is suspended. After a preset delay time, the suspension is released, and the first workpiece of the filling batch is scheduled to enter the conveyor belt. The preset delay time is the quotient obtained by dividing the length of the virtual physical gap by the linear speed of the conveyor belt.

[0018] This step, while maintaining a constant speed of the conveyor belt, creates intervals in the workpiece sequence for temperature regulation transition by controlling the delay time of the production line entry.

[0019] A second aspect of this invention provides an intelligent scheduling and coordination method for a flexible production line for deep processing of glass bottles, comprising the following steps: The process involves: acquiring the initial physicochemical state variables of the flexible production line and the characteristic parameters of the production orders to be scheduled; predicting the solution concentration during processing based on the initial concentration of the frosting solution in the initial physicochemical state variables and the preset concentration decay model; when the predicted solution concentration is lower than the minimum allowable concentration corresponding to the main production order, selecting a secondary priority production order from the production orders to be scheduled whose minimum allowable concentration is not higher than the current predicted solution concentration, as the filling batch; calculating the temperature adjustment response time based on the difference between the reference baking temperature of the main production order and the filling batch, and generating a virtual physical gap without workpiece occupation between the main production order and the filling batch by combining the linear speed of the conveyor belt; calculating the surface thermal radiation absorption conversion coefficient of the batch of products based on the predicted solution concentration and the material characteristic parameters of the filling batch, and determining the temperature feedforward bias accordingly; controlling the conveyor belt of the flexible production line to run according to the virtual physical gap, and superimposing the temperature feedforward bias onto the reference baking temperature of the filling batch before the filling batch enters the baking station to perform equipment temperature control.

[0020] This invention provides an intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles. It offers the following advantages: 1. This invention establishes a matching mechanism between the predicted solution concentration and work orders. When the predicted solution concentration is lower than the requirements of the main production work order, a secondary priority work order with a minimum allowable concentration not higher than the predicted concentration is scheduled as a filler batch. This technical feature utilizes the solution decay period, which would normally require downtime for replenishment or replacement, to process products with lower physicochemical requirements. This avoids frequent start-ups and shutdowns of continuous production lines, reduces unplanned downtime, and improves the overall continuous operation rate and equipment utilization of the production line.

[0021] 2. This invention calculates the difference in baseline baking temperature between consecutive production batches and the corresponding temperature adjustment response time, and combines this with the linear speed of the conveyor belt to generate a virtual physical gap without workpiece occupation between the main production order and the filling batch. This technical feature, while maintaining a constant speed of the main conveyor belt mechanism, reserves a physical transition space for the heating or cooling process of the baking equipment, preventing uneven heating of workpieces in the transition zone between batches due to equipment temperature adjustment lag, and reducing the scrap rate during temperature zone switching.

[0022] 3. This invention calculates the impact of predicted solution concentration changes on the thermal radiation absorption reduction coefficient of the product surface, generates a temperature feedforward bias in advance, and adds it to the equipment control. Furthermore, it uses a Kalman filter algorithm combined with the actual surface temperature after processing to retrospectively update the material coupling attenuation factor. This technical feature establishes a parameter compensation correlation between the state of the preceding chemical solution and the temperature of the subsequent heat treatment, correcting the error caused by the attenuation of the product surface heat absorption rate due to concentration decrease. Simultaneously, it filters temperature measurement noise in industrial settings through a covariance reduction mechanism, ensuring the accuracy of parameter self-calibration during long-term production. Attached Figure Description

[0023] Figure 1 This is an architecture diagram of the adaptive scheduling and collaborative control system for a flexible production line for deep processing of glass bottles, according to one embodiment of the present invention. Figure 2 This is a flowchart of an adaptive scheduling and collaborative control method for a flexible production line for deep processing of glass bottles, as described in one embodiment of the present invention. Figure 3 This is a flowchart of data perception and feature parsing processing in one embodiment of the present invention; Figure 4 This is a flowchart illustrating the state prediction and time-domain modeling process in one embodiment of the present invention. Figure 5 This is a processing logic diagram of physical and chemical coupling scheduling and spatial decoupling in one embodiment of the present invention; Figure 6 This is a processing logic diagram for cross-domain conversion and temperature feedforward bias generation in one embodiment of the present invention; Figure 7 This is a timing diagram of the execution of dynamic coordinate tracking and temperature feedforward control in one embodiment of the present invention; Figure 8 This is a processing logic diagram of closed-loop feedback and model parameter update in one embodiment of the present invention; Figure 9 This is a comparison chart of solution concentration decay and adaptive scheduling time series in an embodiment of the present invention; Figure 10 This is a graph showing the virtual physical idle time and feedforward temperature control response curves of the baking oven in this embodiment of the invention. Figure 11 This is a scatter plot of the convergence process of the local dimension-reduced Kalman filter parameters under multiple materials in an embodiment of the present invention. Figure 12 This is a bar chart comparing the overall yield and defect distribution of the traditional method and the system of the present invention in the embodiments of the present invention.

[0024] Among them, 10 is the physical execution layer; 20 is the device IoT layer; 30 is the edge control layer; 40 is the planning and scheduling layer; 100 is the data perception module; 200 is the status prediction module; 300 is the collaborative scheduling module; 400 is the instruction generation module; 500 is the device execution module; and 600 is the closed-loop correction module. Detailed Implementation

[0025] 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.

[0026] See attached document Figure 1 This invention relates to an adaptive scheduling and collaborative control system for a flexible production line for deep processing of glass bottles. This system can be deployed on the basis of existing hardware of deep processing production lines for glass bottles without changing the basic process layout of the production line.

[0027] In this embodiment, the underlying hardware environment of the system includes a physical execution layer 10, a device IoT layer 20, an edge control layer 30, and a planning and scheduling layer 40.

[0028] The physical execution layer 10 includes a frosting station, a screen printing station, and a baking station connected sequentially by a mesh belt conveyor. The frosting station is equipped with a frosting acid etching tank and a liquid replenishment mechanism. Under normal production conditions, the mesh belt conveyor runs at a preset constant linear speed; when the deviation between the actual temperature at the baking station and the target set value exceeds the process tolerance, the system enters an abnormal compensation mode and temporarily reduces the speed of the mesh belt conveyor within a range not lower than the minimum linear speed threshold.

[0029] A liquid flow meter is installed on the raw solution replenishment pipeline of the frosted sand tank. The liquid flow meter is used to collect the actual replenishment volume or cumulative replenishment flow corresponding to the replenishment action. When no replenishment action occurs, the system does not use the flow meter data to infer the solution consumption, but instead uses manual titration, online concentration detection, or skips the current cycle consumption coefficient update.

[0030] The baking station employs a multi-temperature zone continuous mesh belt kiln structure, with each zone equipped with a thermocouple and heating actuator. Thermocouples are used to collect the actual furnace temperature of the corresponding zone and participate in temperature control feedback. When the system performs online updates of the material coupling attenuation factor, it also acquires product surface temperature detection data located at the baking station outlet, observation window, or furnace exit detection position. Even without a product surface temperature detection device, it can still read quality inspection feedback data such as adhesion and curing status after exiting the furnace.

[0031] When using adhesion or curing status detection feedback data, the closed-loop correction module 600 converts the adhesion value, curing grade, or quality judgment result into an equivalent product thermal state value through a preset quality feedback-equivalent thermal state mapping table. and with Replacement product surface temperature detection value The calculation is as follows. The quality feedback-equivalent thermal state mapping table can be pre-calibrated using curing tests and adhesion test results under different oven temperatures and material samples. If no product surface temperature detection device is configured on-site and no quality feedback test data is provided, the system will only use thermocouple data to correct the oven temperature control residual and will skip the online iterative update of the material coupling attenuation factor.

[0032] The device IoT layer 20 transmits the cumulative flow data of the liquid flow meter and the measured furnace temperature data of the thermocouple to the upper layer through the industrial communication protocol. When the corresponding detection device is configured on site, the device IoT layer 20 also transmits the product surface temperature detection data to the upper layer, or transmits the quality detection feedback data such as adhesion and curing status after the product is taken out of the furnace.

[0033] The edge control layer 30 is equipped with a programmable logic controller (PLC). This PLC has built-in tooling position tracking logic, which is used to map discrete physical production batches to spatial coordinates on the conveyor belt based on the belt's operating status.

[0034] The planning and scheduling layer 40 is deployed on an industrial server and is used to perform work order data parsing, physicochemical state prediction, scheduling calculation, and control command generation. The planning and scheduling layer 40 establishes a bidirectional communication connection with the edge control layer 30. Specifically, the planning and scheduling layer 40 sends scheduling and control data such as work order sequence, virtual physical gaps, temperature feedforward bias, and safety time windows to the edge control layer 30; the edge control layer 30 sends back batch position, actual linear velocity, actual temperature of the temperature zone, abnormal deceleration status, and liquid replenishment execution status to the planning and scheduling layer 40 for subsequent scheduling correction and model parameter updates.

[0035] Based on the aforementioned hardware environment, the adaptive scheduling and collaborative control system for the flexible production line of deep processing of glass bottles includes a data sensing module 100, a status prediction module 200, a collaborative scheduling module 300, an instruction generation module 400, an equipment execution module 500, and a closed-loop correction module 600.

[0036] See attached document Figure 2 The present invention also provides an adaptive scheduling and collaborative control method for a flexible production line for deep processing of glass bottles. This method can be executed by the aforementioned system and includes the following steps.

[0037] S100: Obtain the production order to be scheduled and its initial physical and chemical state.

[0038] The system receives a set of discrete work orders to be scheduled from the enterprise resource planning system, parses the 3D geometric model file, bill of materials file and process route file corresponding to each work order, and extracts the frosting surface area, target processing depth threshold, reference baking temperature and related process constraint data of a single glass bottle; constructs the material feature vector corresponding to each work order based on the process constraint data, and obtains the initial solution concentration of the frosting tank and the initial temperature status of the baking station through the equipment IoT layer.

[0039] S200 is used for concentration decay prediction and thermal response time modeling.

[0040] Based on the frosting surface area, batch quantity, and corresponding empirical coefficient of unit area consumption for each work order, the predicted changes in frosting solution concentration under different production sequences are calculated. At the same time, based on the difference in baseline baking temperature between adjacent work orders, the asymmetric thermodynamic response time cost of the baking station during the heating and cooling processes is calculated respectively.

[0041] S300 generates physical and chemical coupling constraints and determines the caulking batch.

[0042] A physicochemical state coupling constraint matrix is ​​constructed using the predicted solution concentration and the cost of asymmetric thermodynamic response time. When the predicted solution concentration is lower than the preset target concentration and no physical liquid replenishment action is currently performed, a secondary priority work order that can adapt to the current solution concentration and meets the thermal response time margin requirement after insertion is selected from the remaining discrete work orders to be scheduled as a gap-filling batch; and virtual physical gaps are calculated and injected into the scheduling sequence according to the temperature switching requirements between adjacent batches.

[0043] S400 generates cross-domain thermal compensation instructions.

[0044] Based on the material feature vector of the grouting batch and the attenuation deviation of the predicted solution concentration relative to the preset target concentration, the surface radiation absorptivity conversion factor is calculated; then, based on this surface radiation absorptivity conversion factor, a temperature feedforward bias command is generated to compensate for the thermal radiation absorption loss caused by changes in the micro-roughness of the glass surface.

[0045] The S500 performs space tracking and feedforward temperature control.

[0046] The production sequence, including the grouting batch and the virtual physical gap, is sent to the edge control layer. The edge control layer tracks the position of the grouting batch based on the spatial coordinates on the conveyor belt. When the grouting batch reaches the entrance of the baking station and the baking station is within the safe time window corresponding to the virtual physical gap, the edge control layer superimposes the temperature feedforward bias command onto the original target temperature of the corresponding temperature zone, so that the baking station completes temperature compensation before the grouting batch enters.

[0047] S600 performs feedback corrections and updates model parameters.

[0048] The system collects valid measured data from liquid flow meters, concentration detection data, thermocouple furnace temperature data, product surface temperature data, or quality feedback data, and calculates the system prediction residual. Based on the material feature vector, the valid observation residuals are mapped to the corresponding material's local parameter space, and the empirical coefficient of consumption per unit area and the material coupling attenuation factor are iteratively updated using a Kalman filter. When some observation data is missing or invalid, updates are only performed on parameters with valid observation data.

[0049] See attached document Figure 3 In step S100, the data sensing module 100 is used to complete the parsing of basic work order data, extraction of process parameters, and acquisition of initial on-site conditions, and to provide input data for subsequent physicochemical evolution prediction. Its processing includes the following sub-steps.

[0050] S101 receives and parses the set of discrete production orders to be scheduled.

[0051] The data sensing module 100 receives a set of discrete work orders to be scheduled from the enterprise resource planning system. Let the set of discrete work orders to be scheduled be: ; in This represents the total number of work orders awaiting production. The value of is a positive integer. For any work order in the set... The data sensing module 100 reads the 3D computer-aided design model file and bill of materials file associated with the work order.

[0052] The data perception module 100 parses the geometric and topological data of the 3D model file and extracts the surface area of ​​the frosted glass bottle that needs acid etching treatment. The data sensing module 100 parses the bill of materials and process route files to obtain the batch quantity corresponding to the work order. Target processing depth threshold and the benchmark baking temperature .

[0053] Furthermore, the data sensing module 100 reads the delivery date, customer level, manually set priority or urgency level identifier of the work order from the enterprise resource planning system or production management system, and generates work order priority parameters. Work order priority parameters This is used by the subsequent collaborative scheduling module to determine the primary production order, the highest priority work order, and the candidate order for the secondary priority seam filling batches.

[0054] The data sensing module 100 also uses the target processing depth threshold. Based on the glass substrate type, frosting contact time, and the concentration-processing depth mapping relationship in the process route document, determine the minimum allowable frosting concentration for this work order. ;in Used to determine whether the current, decayed solution concentration still meets the minimum processing quality requirements of the work order.

[0055] It should be noted that the surface area of ​​the frosted glass bottle is... With batch quantity This directly determines the total consumption of frosting solution during the processing of this batch, while the target processing depth threshold... This reflects the minimum process tolerance limit for the solution concentration of this batch of products. Typically, this is the target processing depth threshold. The specific numerical range can be preset in the process route file according to the haze or roughness level required for the product's appearance.

[0056] For parsing and extracting the geometric surface area of ​​3D computer-aided design model files, those skilled in the art can use existing general surface area integration algorithms in computer graphics. Model boundary identification and mesh area calculation are well-known technologies in this field and will not be elaborated upon here. Furthermore, to ensure the completeness of data parsing, the data sensing module 100 performs routine numerical verification after parameter extraction. If any parameters are missing or abnormally exceed the standard process settings, the system will generate a prompt message and wait for manual confirmation or re-issuance of data to avoid errors in subsequent calculations.

[0057] S102, construct the material feature vector corresponding to the work order.

[0058] In the deep processing of glass bottles, different glass substrates consume frosting solution at different rates, and different adhering consumables also exhibit varying changes in their absorption of thermal radiation after frosting treatment. To enable these material properties to be incorporated into subsequent model calculations, the data sensing module 100 extracts the properties of the glass substrate and the adhering consumables, combining them to construct a two-dimensional material feature vector. .

[0059] The material feature vector The specific mathematical expression is: ; In the formula, The parameters characterizing the glass substrate material correspond to the preset codes associated with the glass material type in the database. For example, a specific range of integer numbers can be used to identify soda-lime glass or borosilicate glass respectively. Characterizes the attribute parameters of the attached consumables, corresponding to the preset codes associated with the pigments or gold pastes used in the processing technology.

[0060] Material feature vector These serve as index labels for subsequent calculations, used to match specific material coupling attenuation factors within the system's pre-set material coefficient matrix. Each set of material feature vectors within the material coefficient matrix... The system should store at least two local model parameters, including the empirical coefficient of energy consumption per unit area. and material coupling attenuation factor ;in Used for predicting the concentration decay of frosted sand solutions. Used for cross-domain conversion from chemical decay state to thermal radiation absorption conversion state.

[0061] S103, read the initial physical and chemical state variables of the site.

[0062] The data sensing module 100 uses the industrial Ethernet communication protocol to call the underlying data interface and collect the current initial solution concentration scalar value of the frosted sand tank. And the current actual temperature of each temperature zone in the baking flower station. .

[0063] Among them, the initial solution concentration scalar The concentration can be manually measured and input into the system by on-site operators after solution preparation, or it can be directly fed back by concentration detection instruments.

[0064] Considering the physical disturbances present in the industrial environment, the data sensing module 100 typically performs a moving average filtering process after receiving the initial collected data to smooth out and filter out high-frequency noise points caused by liquid surface fluctuations or thermocouple electromagnetic interference. The data sensing module 100 then determines the filtered values ​​as the initial physicochemical state variables of the system and writes them into the memory database of the industrial server as input boundary conditions for the subsequent evolution prediction model.

[0065] See attached document Figure 4 In step S200, the state prediction module 200 discretizes the changes in frosting solution concentration and the temperature switching process of the baking station during continuous production based on the work order characteristics, material parameters, and initial physicochemical state obtained in step S100, to form state prediction data that can be called upon for subsequent scheduling calculations. Its processing includes the following sub-steps.

[0066] S201, calculate the predicted concentration of the frosting solution after the work order is scheduled.

[0067] During the glass bottle frosting and acid etching process, the concentration of the frosting solution gradually decreases as the reaction area on the glass surface increases. To predict the solution state under different work order sequences during the scheduling stage, the state prediction module 200 uses the frosting surface area of ​​the work order, the batch quantity, and the empirical coefficient of unit area consumption of the corresponding material as inputs to establish a forward iterative prediction model.

[0068] For any work order in the work order set The predicted concentration of the solution after processing The calculation formula is: ; In the formula, For the preceding work order The predicted concentration of the solution after processing. This is used when calculating the first work order in the scheduling sequence. The initial solution concentration scalar obtained in step S100 is used. . For the current work order Batch quantity, is the unfolded surface area of ​​a single glass bottle covered with frosting; both values ​​are positive integers or positive real numbers.

[0069] For the current work order Corresponding material feature vector The empirical coefficient of consumption per unit area, obtained by matching in the material coefficient matrix, takes the value of a positive real number and its physical meaning is the equivalent concentration of active solution consumed per unit area of ​​glass to achieve the target acid etching effect. To establish an initial benchmark that conforms to actual working conditions, during the system initialization phase... The average value can be preset based on historical production data of work orders of similar materials, and then dynamically updated by the closed-loop correction module 600 in subsequent production.

[0070] This refers to the theoretical increase in concentration resulting from the physical replenishment action. During the scheduling simulation phase, if the system determines that a shutdown and chemical dosing action will not be performed before the current work order is processed, then... The value is zero; if it is determined that a drug dosing action will be performed, then The concentration increment parameter is calculated based on the preset dosing volume and the original solution concentration. Specifically, those skilled in the art can calculate the concentration increment parameter after mixing by using the total existing solution volume in the frosted tank, the current predicted concentration value, and the added original solution concentration and dosing volume, based on the law of conservation of solute in the mixed solution. This will not be elaborated further here.

[0071] Through the aforementioned forward iterative calculation, the system transforms the continuous physical and chemical evolution states of the equipment into discrete state parameters that can be called by subsequent scheduling algorithms.

[0072] S202, calculate the time cost of the asymmetric thermodynamic response when switching between adjacent work orders.

[0073] The baking station is typically a continuous mesh belt kiln with significant thermal inertia. During temperature switching, heating mainly relies on active heating elements, while cooling depends more on furnace heat dissipation and heat exchange. Therefore, the time required for heating and cooling processes is not symmetrical. To improve the consistency between scheduling results and actual equipment response capabilities, the status prediction module 200 models the temperature switching time between adjacent work orders.

[0074] For any adjacent work order sequence, the status prediction module 200 sets the baseline baking temperature of the preceding work order to the current temperature. Set the baseline baking temperature of the pending production orders as the target temperature. And establish asymmetric response time cost Piecewise function model: ; In the formula, For the forced heating rate of the flower baking station, This represents the natural heat dissipation cooling rate of the equipment; both values ​​are positive real numbers. In actual physical equipment, this typically satisfies... Conditions. Heating rate. With cooling rate All of these can be obtained through the equipment's factory calibration parameters, or by fitting the historical temperature control curves under no-load conditions collected by the equipment's IoT layer.

[0075] when equal At that time, response time cost The value of is zero. Furthermore, considering that the baking station is typically divided into multiple independently controlled temperature zones, to avoid timing interference dead zones when the algorithm handles synchronous temperature switching across multiple zones, the state prediction module 200 calculates the response time cost of each temperature zone according to the aforementioned piecewise function, and selects the maximum value among the response time costs of each temperature zone as the global asymmetric thermodynamic response time cost for the current work order switch. The state prediction module 200 calculates the theoretical waiting time required for any two temperature switches in the work order sequence using this piecewise function, and uses it as the basic input parameter for subsequent scheduling optimization and spatial decoupling calculations.

[0076] Compared to conventional scheduling methods that ignore temperature variations, this calculation model better reflects the actual temperature characteristics of the physical equipment on site, which helps improve the feasibility of subsequent scheduling results.

[0077] See attached document Figure 5 In step S300, the collaborative scheduling module 300 evaluates the constraints of the production order to be scheduled based on the predicted solution concentration and asymmetric thermodynamic response time cost obtained in step S200, and reserves idle space in the scheduling sequence to meet the requirements of thermal switching. The process includes the following sub-steps.

[0078] S301, construct the physical and chemical state coupling constraint matrix.

[0079] In the continuous glass bottle deep processing, the concentration of frosting acid reflects the processability of the chemical station, while the temperature of the baking oven reflects the response of the thermal station. These two stations are located at different points and each exhibits time characteristics of concentration decay and temperature lag, respectively. The collaborative scheduling module 300 incorporates these two types of state parameters into the same scheduling evaluation system to characterize the comprehensive switching cost generated when different work orders are arranged adjacently.

[0080] The collaborative scheduling module 300 uses the predicted solution concentration values ​​corresponding to each work order. As the first dimension of the constraint coordinates, the asymmetric thermodynamic response time between adjacent work order sequences is used as the cost. As the second dimension of the constraint coordinates, a two-dimensional physical and chemical state coupling constraint matrix is ​​constructed. Specifically, assuming the set of production orders to be scheduled is m in size, the collaborative scheduling module 300 constructs a constraint matrix with dimension m×m. .

[0081] Due to the predicted value of solution concentration As the order of scheduled work orders changes cumulatively, the collaborative scheduling module 300 constructs the constraint matrix using a rolling update method; that is, after each new scheduling node is determined, the predicted value of the solution concentration under the current sequence is recalculated, and the concentration matching cost corresponding to the candidate work orders is dynamically updated. The thermal response time cost between adjacent work orders can be pre-calculated or calculated in real time during the order insertion evaluation.

[0082] Elements in the matrix If the work order is represented Arranged in work order The subsequent comprehensive state cost. This cost element includes the work order. For the current predicted concentration The matching degree coefficient and the thermal response time cost between the two This matrix quantifies the switching process between discrete work orders into a data structure that includes the costs of chemical concentration deviation and thermodynamic time waiting, providing boundary constraints and computational basis for the system to search for the globally optimal processing sequence.

[0083] S302, Select the caulking batch that is suitable for the current physical and chemical state.

[0084] In regular production, when the concentration of the frosting solution drops to an insufficient level to meet the processing requirements of the current high-priority work orders, it is usually necessary to replenish the solution or stop the machine to change the solution. To reduce waiting time, the collaborative scheduling module 300 prioritizes finding work orders from the remaining work orders that can continue processing using the current reduced concentration, provided that the minimum processing quality of the product is met.

[0085] In conventional production, when the frosting solution concentration decreases to a level that does not meet the current production process requirements, it is usually necessary to shut down the plant to change the solution or replenish the chemicals. To reduce downtime and improve equipment utilization, the collaborative scheduling module 300 monitors the predicted solution concentration in real time. When determining the predicted solution concentration for a specific production node... Less than the system's preset target concentration standard At this time, the collaborative scheduling module 300 pauses triggering the underlying physical fluid replenishment command. Target concentration standard. The process requirements are pre-set by the system based on the current highest priority work order. The collaborative scheduling module 300 traverses and searches the remaining unscheduled subset of the work orders to be scheduled, looking for a second-priority work order that can adapt to the current attenuation concentration.

[0086] The specific screening criteria include two aspects: First, the target processing depth threshold for candidate work orders. Able to match the current low predicted solution concentration. This means that the product corresponding to this work order has relatively low requirements for surface roughness or etching depth; and the current predicted solution concentration value No less than the candidate work order The corresponding minimum allowable sand concentration That is, satisfying: ; when If the current concentration of the frosting solution is insufficient to meet the minimum acid etching quality requirements of the candidate work order, the candidate work order shall not be used as a filler batch. Secondly, after a candidate work order is inserted into the current scheduling sequence, its baseline baking temperature and the asymmetric thermodynamic response time cost generated by the preceding and following batches... The minimum condition is met. The subsequent temperature feedforward bias is only used to compensate for the thermal radiation absorption loss caused by changes in the roughness of the frosting, and not to compensate for situations where the frosting processing depth does not meet the minimum allowable requirement. The specific determination method for this minimum condition is as follows: the collaborative scheduling module 300 calculates the time cost between adjacent batches before the candidate work order is inserted, and the sum of the time costs of two adjacent batches in the new sequence formed after the insertion.

[0087] When the sum of costs after order insertion is within the system's allowed time margin, the condition is deemed met to avoid drastic temperature fluctuations at the baking station caused by the insertion. The collaborative scheduling module 300 sets work orders that simultaneously meet both of the above conditions as filling batches and inserts them into the current production sequence. To ensure the completeness of the algorithm, if no work order meeting the above conditions is found after traversing the remaining unscheduled subset, the collaborative scheduling module 300 issues a physical replenishment command according to conventional logic, resets the solution concentration state, and continues scheduling according to the original planned priority order. For the specific search process of set traversal and multi-objective optimization, those skilled in the art can use existing heuristic algorithms.

[0088] As one possible implementation, the collaborative scheduling module 300 pairs each candidate work order. The concentration matching cost, the increase in thermal response time, and the priority delay cost were calculated separately, and a comprehensive evaluation value was constructed. .

[0089] The concentration matching cost is used to characterize the current solution concentration prediction value. Minimum allowable sand concentration for candidate work orders The degree of matching between them, the increase in hot response time is used to characterize the increase in the hot switching time cost of adjacent batches after the insertion of an order, and the priority delay cost is used to characterize the degree of deviation of the candidate work order relative to the original production priority. The collaborative scheduling module 300 satisfies: ; Among candidate work orders where the increase in thermal response time does not exceed the preset time margin, the comprehensive evaluation value is selected. The smallest work order is considered the caulking batch. Its iterative solution mechanism is a well-known technique in this field and will not be elaborated upon here.

[0090] S303, calculate and inject virtual physical gaps.

[0091] Because the conveyor belt runs continuously at a constant linear speed during production, if different production batches with set temperature differences are directly connected in the scheduling sequence, the baking station cannot complete the heating or cooling within the limited physical transmission time when performing temperature switching, easily causing overheating or underheating of products in successive batches. To decouple the thermal conditioning time from the transmission line space, the collaborative scheduling module 300 calculates the required virtual physical gap length for adjacent batches with temperature differences in the scheduling sequence. Virtual physical gap length The specific calculation formula is as follows: ; In the formula, The fixed operating linear speed of the mesh belt transmission line; This is the time cost of the asymmetric thermodynamic response between the current batch and the previous batch; This is the normal standard physical gap time parameter when adjacent tooling plates or products enter the baking station. This parameter is usually determined in advance by the equipment based on the mechanical dimensions of the tooling fixtures and the transmission line speed.

[0092] when When this time is reached, it indicates that the normal physical interval has been sufficient for the baking station to complete the set temperature switch. The value is zero; when At that time, the collaborative scheduling module 300 converts the time difference between the two into the unused space length on the mesh transmission line. All three parameters mentioned above are positive real numbers. When the calculated... When the time is less than or equal to zero, it indicates that the normal physical gap time is sufficient for the baking station to complete the switching of the set temperature. At this time, the virtual physical gap length is determined. The value is zero. When the calculated value is... When the value is greater than zero, the collaborative scheduling module 300, when generating the final scheduling sequence, forcibly inserts a length of [value missing] between the previous batch and the current gap-filling batch. The empty space instruction.

[0093] The idle space command is generated by controlling the feeding mechanism to pause product delivery, deliver an idle tooling plate, or create a product-free occupying section in the production scheduling data. The conveyor belt itself continues to run continuously at the preset linear speed under normal production conditions. The edge control layer 30 writes this idle section as a virtual batch into the position tracking logic and records its head coordinates, tail coordinates, and corresponding safety time window. The safety time window is determined based on the time when the head of the virtual batch arrives at the entrance of the baking station, the time when the tail of the virtual batch leaves the exit of the baking station, and the response time requirements of each temperature zone of the baking station.

[0094] As one possible implementation method, the safe time window can be represented as: ; In the formula, This indicates the moment when the first end of the virtual batch arrives at the entrance of the baking station. This indicates the moment when the virtual batch leaves the exit of the baking station. Within this safe time window, the baking station is allowed to perform temperature switching, temperature feedforward bias overlay, or temperature feedforward bias cancellation actions.

[0095] This mechanism reserves a portion of the space on the conveyor belt, thereby securing an independent and safe temperature-changing time window for the baking station and cutting off thermodynamic interference between work orders with different process parameters.

[0096] See attached document Figure 6In step S400, the instruction generation module 400, for the caulking batch determined by the collaborative scheduling module 300, converts the surface state change caused by the decrease in frosting liquid concentration into the heat compensation amount for the baking station, and generates a temperature feedforward bias instruction for the underlying temperature control loop to call. The processing includes the following sub-steps.

[0097] S401, calculate the surface radiation absorptivity conversion factor.

[0098] During the deep processing of glass bottles, the attenuation of the frosting solution concentration affects the acid etching depth of the glass surface, thereby altering the surface micro-roughness. When the surface roughness decreases, the absorption efficiency of infrared thermal radiation by the glass bottle in the subsequent baking process may decrease. To quantify this effect, the instruction generation module 400 converts the chemical state deviation of the current caulking batch into a thermal absorption state deviation.

[0099] The instruction generation module 400 extracts the material feature vector of the current grouting batch. (It includes the material parameters of the glass substrate) With attachment consumable property parameters The instruction generation module 400 uses the glass substrate material parameters as an addressing index for matching and retrieval within a pre-set material coefficient matrix within the system. Specifically, the instruction generation module 400 uses the glass substrate material parameters as an addressing index for matching and retrieval. The encoding is mapped to the row index of the material coefficient matrix, which represents the attached consumable property parameters. The encoding is mapped to column indices, thereby extracting the material coupling attenuation factor at the intersection of rows and columns. .

[0100] To avoid abnormal program termination due to addressing failure, if the material feature vector Since no matching term was found in the material coefficient matrix, instruction generation module 400 assigns a material coupling attenuation factor. A default security value of zero means that cross-domain associations are not allowed.

[0101] exist When = 0, the surface radiation absorptivity conversion factor Maintain the standard baseline state, that is: ; At this time, the instruction generation module 400 does not generate additional temperature feedforward bias caused by material coupling attenuation, but only performs control according to the original benchmark baking temperature of the work order, thereby avoiding overcompensation under unknown material conditions.

[0102] Instruction generation module 400 calculates the predicted value of the current solution concentration. Compared with the preset target concentration standard The attenuation deviation between them is determined by the material coupling attenuation factor. Converting chemical attenuation deviation into a surface radiation absorptivity conversion factor in the thermodynamic domain Surface radiation absorptivity conversion factor The calculation formula is ; In the formula, and The values ​​of are all positive real numbers, and in combination with the aforementioned logic, they satisfy . ; The constant 1 in the formula represents the ideal thermal radiation absorption reference state under the standard process concentration. It is a positive real number that was fitted and calibrated by engineers in the early stage through the spectral absorption rate test of glass blocks at different concentrations and then fixed in the system database.

[0103] Furthermore, to ensure the rationality of the physical meaning and avoid the negative value dead zone of the conversion factor, the system's underlying layer presets a minimum absorption rate cutoff value. The cutoff value is typically set between 0.3 and 0.5. When the calculated... Less than At that time, instruction generation module 400 enforces the following command: ; When the calculation yields When the value is greater than 1, instruction generation module 400 forces an instruction: ; This is to characterize the ideal absorption reference state where the current surface thermal radiation absorption state of the product does not exceed the standard process concentration.

[0104] Furthermore, the surface radiation absorptivity conversion factor It can be uniformly represented as: ; This step converts the state deviation of the chemical workstation into physical quantitative parameters for the subsequent thermal workstation.

[0105] S402 generates a temperature feedforward bias command.

[0106] Obtaining the surface radiation absorptivity conversion factor Subsequently, due to the reduced actual absorption rate of heat radiation on the glass surface, if the curing station continues to control the curing temperature according to the original set benchmark temperature, the processing consumables adhering to the glass surface are prone to poor adhesion due to insufficient curing heat, resulting in quality defects. The instruction generation module 400 calculates the required additional temperature offset real number based on proportional conversion logic. Temperature feedforward offset instruction. The specific conversion formula is as follows: ; In the formula, The preset baseline baking temperature for this caulking batch is a positive real number; The thermodynamic proportional compensation coefficient is a positive real number. This coefficient depends on the heat exchange transfer efficiency constant of a specific temperature zone in the baking oven. It can be calibrated by measuring the actual temperature rise of the product surface caused by a unit temperature increase in the baking oven.

[0107] To prevent the theoretically calculated bias value from being too large and exceeding the actual heating limit of the physical device, or causing the glass to crack due to excessive temperature gradient, the instruction generation module 400 introduces a maximum bias threshold at the algorithm level. As a safety truncation constraint. When the calculated Greater than At that time, the instruction generation module 400 will The value is forcibly set to For the maximum bias threshold The method for determining the thermal stress limit parameter of the glass substrate used and the rated power of the heating element of the baking oven can be calculated by those skilled in the art. The mechanical and thermal calculation process is a well-known technology in the field and will not be described in detail here.

[0108] The instruction generation module 400 will determine the final temperature feedforward bias instruction. The data is stored in system memory, awaiting distribution to the underlying control hardware for subsequent precise overlay operations. When the baking station contains multiple independent temperature control zones, the instruction generation module 400 calculates the zone-level temperature feedforward bias according to the reference set temperature, heat exchange efficiency constant, and maximum bias threshold of each zone. And form a temperature zone bias vector; the device execution module 500 will assign the corresponding temperature zone to the bias vector during execution. These values ​​are respectively added to the target setpoints of the proportional-integral-derivative controllers in each temperature zone.

[0109] See attached document Figure 7 In step S500, the equipment execution module 500 receives the production sequence, virtual physical gap, and temperature feedforward bias instruction issued by the planning and scheduling layer, and converts the above discrete production scheduling information into spatial tracking and temperature control actions that can be executed by the underlying programmable logic controller. The processing includes the following sub-steps.

[0110] S501, Establish dynamic coordinates for caulking batches.

[0111] The device execution module 500 receives a scheduling sequence containing instructions for filling batches and virtual physical gaps, and sends the scheduling sequence to the underlying programmable logic controller.

[0112] Under continuous operation of the conveyor belt, discrete production batches need to establish a correspondence with their actual spatial positions on the conveyor belt. When the first product of the filling batch triggers the reference position sensor at the material end of the conveyor belt, the equipment execution module 500 uses this trigger moment as the starting point for position tracking of the batch and obtains the real-time running line speed based on the pulse data fed back by the incremental encoder at the end of the conveyor belt drive motor.

[0113] When the first product in the caulking batch triggers the reference position sensor at the material end of the conveyor line, the equipment execution module 500 uses this trigger moment as the time zero point and extracts the real-time running linear velocity using the pulse data fed back by the incremental encoder at the end of the mesh belt drive motor, thereby calculating the one-dimensional dynamic physical coordinates of the first end of the caulking batch on the conveyor line. The formula for calculating the physical coordinates of the head end is: ; In the formula, The time point at which the product triggers the reference position sensor; This is the current system real-time clock time; This refers to the real-time operating line speed of the mesh belt transmission line.

[0114] The equipment execution module 500 calculates the total physical length of the caulking batch based on the production quantity and product layout spacing of the batch. Alternatively, the device execution module 500 can also read and update the position coordinates at key physical nodes along the line by deploying RFID nodes on the vehicle tooling plate.

[0115] The device execution module 500 will display the one-dimensional dynamic head physical coordinates. The data is written into the internal shift register of the programmable logic controller to achieve dynamic anchoring of the product in physical space. The specific implementation methods for encoder pulse acquisition or RFID node communication can be accomplished by those skilled in the art using existing industrial automation standard hardware architectures. Data acquisition and address mapping are well-known technologies in this field and will not be elaborated upon here.

[0116] S502 performs temperature feedforward control based on dynamic coordinates.

[0117] The heating and cooling of the caulking station requires a certain physical time period. Significantly adjusting the temperature while the product is within the designated temperature zone can easily lead to uneven heating within the same batch. To utilize the virtual physical gap injected in step S300 for safe temperature adjustments, the equipment execution module 500 continuously updates the one-dimensional dynamic physical coordinates of the first end of the caulking batch. Fixed physical coordinates corresponding to the entrance of the baking flower station Compare them.

[0118] The device execution module 500 tracks the dynamic tail physical coordinates of the preceding batch. And compare it with the fixed physical coordinates corresponding to the current baking station exit. Perform a comparison. When a determination is made... This means that the tail end of the preceding batch of products has left the corresponding temperature control zone and meets the position constraint difference. At that time, the device execution module 500 determines that the current device has entered the physical isolation interval corresponding to the virtual time window.

[0119] After confirming entry into the isolation zone, the device execution module 500 will execute the temperature feedforward bias command generated in step S400. This value is superimposed on the target setpoint of the proportional-integral-derivative (PID) controller corresponding to the current temperature zone. The PID controller receives the new target setpoint and, based on the actual feedback temperature, outputs an adjustment signal to drive the heating element to perform its action. When the coordinates of the end of the caulking batch... Beyond the exit coordinates of the baking flower station Or, when the temperature switching safety window for the next batch begins, the equipment execution module 500 cancels the temperature feedforward bias command corresponding to the current caulking batch. The target setpoint of the proportional-integral-derivative controller is then restored to the reference baking temperature for the next batch, or to the feedforward correction target temperature for the next batch. This execution mechanism utilizes the physical gap in the conveyor belt between the preceding batch and the filling batch as a buffer zone.

[0120] Before the grouting batch actually enters the baking oven, an offset command is issued in advance, so that the heating system has relatively sufficient response time to overcome thermal inertia and complete temperature adjustment, thereby reducing the thermodynamic interference caused by batch switching in terms of timing.

[0121] Through the above control method, the temperature feedforward bias is superimposed before the filling batch enters the baking station and is promptly removed after the batch leaves, thereby using the virtual physical gap to reserve response time for furnace temperature adjustment and reducing thermal interference caused by temperature switching between adjacent batches.

[0122] Furthermore, considering that the heating element may age and cause the temperature change rate to be lower than the theoretical value in the physical environment, the equipment execution module 500 reaches the first end of the caulking batch. Temperature tolerance checks are performed continuously. If the deviation between the current actual temperature and the target set value exceeds the allowable tolerance range of the process, the equipment execution module 500 judgment system enters the abnormal compensation mode.

[0123] In abnormal compensation mode, the equipment execution module 500 temporarily reduces the operating speed of the conveyor belt within a preset speed adjustment range to extend the heat residence time of the product in the baking station. This temporary speed reduction is only a compensation measure for insufficient temperature response and does not change the control principle of the conveyor belt operating at a preset constant linear speed under normal production conditions.

[0124] Since abnormal speed reduction will change the actual arrival time of the batch on the transmission line, the edge control layer 30 no longer calculates the batch position based on a preset constant linear velocity, but rather based on the real-time linear velocity fed back by the encoder. The coordinates of the first and last ends of the caulking batch, as well as the safety time window corresponding to the virtual physical gap, are dynamically updated using integration.

[0125] The coordinates of the first end of the grouting batch can be updated according to the following formula: ; In the formula, The time when the reference position sensor is triggered for the caulking batch. For the current system time, For mesh belt transmission lines at time The real-time operating linear velocity. After the current abnormal batch ends, the planning and scheduling layer 40 adjusts the estimated arrival time and temperature switching window of subsequent batches based on the actual linear velocity record.

[0126] To prevent a logical dead zone where excessive temperature deviation causes the system to reduce the linear speed to an extremely low level, thus halting the entire production line, the equipment execution module 500 incorporates a minimum permissible linear speed threshold into its speed reduction logic. As a safety constraint. When the required compensation speed is lower than... At that time, the system maintains Run and trigger an anomaly alarm. This mechanism helps reduce the risk of batch quality defects under abnormal operating conditions.

[0127] See attached document Figure 8 In step S600, the closed-loop correction module 600 iteratively corrects the local parameters in the state prediction model based on the actual operational feedback from the underlying equipment, in order to reduce the impact of sensor noise, equipment aging, and field environmental drift on the scheduling prediction results. Its processing includes the following sub-steps.

[0128] S601, collect valid feedback data and calculate observation residuals.

[0129] By comparing the physical deviations between actual output and theoretical predictions, the system can inversely perceive objective changes in environmental variables. After the actual processing of the caulking batch is completed, the underlying sensors generate measured data reflecting the real physical process. The closed-loop correction module 600 synchronously collects the actual cumulative flow of the liquid flow meter in the frosting liquid replenishment system. And the actual temperature feedback value of the thermocouple in the corresponding temperature zone of the baking station. .

[0130] Among them, the actual temperature feedback value of the thermocouple Used for furnace temperature control residual correction; when the system needs to implement material coupling attenuation factor. During online updates, the closed-loop correction module 600 also reads the product surface temperature detection values ​​set at the baking station exit, observation window, or oven exit detection position. Alternatively, the equivalent product thermal state value can be obtained by reading the quality test results such as adhesion and curing status after the product exits the oven. To map heterogeneous physical measurements to model internal parameters, the closed-loop correction module 600 calculates the single observation value of the empirical coefficient consumed per unit area. Single observation of the coupling attenuation factor with material .

[0131] Empirical coefficient for consumption per unit area When a physical replenishment action corresponding to the current batch occurs within the current batch processing cycle, and the replenishment action is marked by the system as a restorative replenishment action for the concentration decay of the current batch, and non-productive flow such as cleaning, drainage, and manual maintenance has been excluded, and the liquid flow meter records the effective actual cumulative replenishment flow rate, At that time, the closed-loop correction module 600 according to Calculate the concentration consumption corresponding to this batch and calculate the single observation value. : ; In the formula, The flow-to-concentration conversion factor solidified within the system is a positive real number, which can be obtained through prior titration experiments with different volumes of stock solution. This represents the total number of products included in the current batch. This refers to the surface area parameter of a single product in contact with acid.

[0132] When no physical replenishment occurs during the current batch processing cycle, the closed-loop correction module 600 does not use flow meter data from the replenishment line to infer solution consumption. Instead, when manual titration data or online concentration detection data is available, the closed-loop correction module 600 calculates the solution consumption based on the concentration difference before and after the current batch processing. : ; In the formula, This is the concentration value of the frosting solution measured before processing the current batch. This refers to the concentration detection value of the frosting solution after processing in the current batch. When concentration detection data is missing or the detection result exceeds the preset confidence range, the closed-loop correction module 600 will adjust the concentration of the current batch... Mark as invalid observations and skip this batch. Update.

[0133] When no liquid replenishment action is performed and manual titration or online concentration detection is used, the closed-loop correction module 600 calculates the concentration detection value before processing, the concentration detection value after processing, and the total processing area of ​​the current batch. When concentration detection data is missing or the detection results exceed the preset confidence range, the closed-loop correction module 600 will adjust the current batch's... Mark as invalid observations and skip this batch. Update.

[0134] For the thermal domain, the closed-loop correction module 600 extracts the product at the set reference baking temperature. Product surface temperature readings after reaching thermal equilibrium Alternatively, extract the equivalent product thermal state parameters obtained from the quality inspection results. Calculate the actual temperature deviation. Combined with the feedforward bias logic in step S400, the closed-loop correction module 600 back-calculates the single observation value. If no product surface temperature detection device is installed on-site, and no quality feedback detection data such as adhesion and curing status is provided, the closed-loop correction module 600 will not be based on the thermocouple furnace temperature. Reverse push Instead, it only performs furnace temperature control residual correction and skips the material coupling attenuation factor for the current batch. Online iterative updates.

[0135] When using product surface temperature detection data, it should be considered that a temperature feedforward bias has already been superimposed on the current batch at the baking station. The actual total thermodynamic loss consists of the compensated amount and the residual deviation. The calculation formula is: ; When using quality inspection feedback data The calculation formula is: ; To avoid the division-by-zero anomaly that could cause a computational dead zone in the algorithm due to the denominator approaching zero, the closed-loop correction module 600 performs a boundary validity check before executing the formula.

[0136] When the concentration difference Smaller than the preset small positive real number When, that is, the following condition is met: ; The system determines that the current chemical decay has a negligible impact on thermophysical properties, and forcibly skips the current batch's impact on these properties. The calculation, and Mark as invalid observations or keep as historical forecasts from the previous period, retaining only those related to... Effective calculation. When the product surface temperature is detected... or equivalent product thermal state quantity When there are missing, abnormal, or values ​​exceeding the preset confidence range, the closed-loop correction module 600 also skips the current batch. The calculation.

[0137] when and When all observations are valid, the closed-loop correction module 600 will calculate the... and Constitutes the current iteration cycle The following is the actual two-dimensional observation vector: ; The difference between this vector and the system's original prediction parameters constitutes the system prediction residual, which includes multi-source data. or When any observation in the Kalman filter is invalid, the closed-loop correction module 600 constructs a dimension-reduced observation vector using only the valid observations, and updates only the valid parameters in the subsequent Kalman filter process.

[0138] S602, the closed-loop correction module 600 performs Kalman filtering iterations based on local dimensionality reduction of material features to update model parameters.

[0139] In continuous production systems involving work orders with various material properties, using a globally uniform parameter update logic can easily lead to unexpected shifts in global parameters due to measurement deviations in a single specific batch. The closed-loop correction module 600 utilizes the material feature vector of the current batch. As a condition for dimensionality reduction mapping, the system's global parameter space is projected onto a two-dimensional local subspace that is only relevant to the material. Local dimensionality reduction projection refers to the closed-loop correction module 600 applying the material feature vectors of the current batch... Addressing and selecting the material corresponding to the given material in the material coefficient matrix. and For each parameter unit, filtering and updates are performed only on that unit; the parameter units corresponding to other material feature vectors remain unchanged. This process does not alter the physical dimension of the sensor's measured data, but rather limits the scope of the model parameter updates to prevent measurement biases in a single material batch from affecting the parameters of other materials.

[0140] That is, the closed-loop correction module 600 only extracts the previous period's state vector corresponding to the current material from the material coefficient matrix for updating. This state vector can be represented as: ; In the formula, Represents the current material feature vector The corresponding empirical coefficient for consumption per unit area was estimated in the previous period. Represents the current material feature vector The estimated value of the corresponding material coupling attenuation factor in the previous period.

[0141] If the material feature vector For the new material encountered during the system's first run, the closed-loop correction module 600 initializes the state vector. The empirical mean is used, and an initial error covariance matrix is ​​assigned. Larger diagonal elements characterize extremely low initial trust levels.

[0142] The core role of Kalman filtering here is to utilize the statistical properties of measurement data to find a dynamic balance between the noisy underlying measured feedback and the prior predictions based on historical data. Within the local subspace, the closed-loop correction module 600 constructs a discrete Kalman filter, which converts the single observation vector... Introducing the state update equations. The specific mathematical equations for the Kalman filter iteration are as follows: State-one-step prediction equation: ; Error covariance prediction equation: ; Kalman gain calculation equation: ; State update equation: ; Error covariance update equation: ; Among them, and When all observations are valid, the observation matrix can be a second-order identity matrix, making the observation vector... With state vector Corresponding to the physical dimension; and to the state vector Corresponding to the physical dimension; when only When effective, the closed-loop correction module 600 constructs a one-dimensional observation vector: ; and only for the corresponding state vector Perform the update; when only When effective, the closed-loop correction module 600 constructs a one-dimensional observation vector: ; and only for the corresponding state vector Perform the update.

[0143] In the formula, For the current period The prior state estimation vector; This is the error covariance matrix of the previous period; Pk|k-1 is the system process noise covariance matrix, which characterizes the inherent uncertainty of the model; Pk|k-1 is the prior error covariance matrix for the current period. The noise covariance matrix is ​​used to characterize the measurement error of liquid flow meters, concentration detection devices, product surface temperature detection devices, or quality inspection feedback conversion results. For thermocouple furnace temperature data used only for furnace temperature control residual correction, its error parameter can be used alone for furnace temperature control loop correction, and not as a material coupling attenuation factor. Directly observed sources; This is the calculated Kalman gain matrix.

[0144] The posterior state vector is the smoothly updated version, containing the current material feature vector. The corresponding updated parameters can be represented as: ; In the formula, This is the empirical coefficient for consumption per unit area after the current cycle update. This is the material coupling attenuation factor updated in the current cycle.

[0145] This is the updated posterior error covariance matrix; It is a second-order identity matrix. In terms of data structure, , , , All are 2×1 dimensional column vectors, with a covariance matrix. , , , , and the second-order identity matrix , All are 2×2 square matrices. and The diagonal elements are preset as positive real constants based on the factory accuracy calibration value of the underlying sensor and the historical statistical variance.

[0146] Using the Kalman filter equation described above, the closed-loop correction module 600 adjusts the observed vector... The mechanism seeks the optimal estimate under the condition of minimum mean square error between the estimate and the prior prediction. This mechanism utilizes Kalman gain. By dynamically adjusting the proportion of the residual from a single measurement to correct the model parameters, an anti-interference and smoothing iteration mechanism is formed to suppress sporadic sensor noise and severe parameter oscillations caused by individual abnormal measurements.

[0147] The closed-loop correction module 600 will solve for the posterior state vector. The elements in the matrix are rewritten back into the material vector corresponding to the material coefficient matrix. The storage address is [address]. For the basic mathematical operations involved in Kalman filtering, such as matrix inversion and multiplication, those skilled in the art can implement them using standard numerical algebra libraries. The underlying computer instruction execution logic is well-known in this field and will not be elaborated upon here.

[0148] To facilitate understanding, we will take a real production scenario of a daily-use glass products factory as an example to demonstrate the core numerical calculation logic of the system in steps S100 to S600.

[0149] 1. Initial state and work order parameter extraction (corresponding to S100): Initial state: Current initial solution concentration scalar value in the frosted sand tank. =15.0%, the preset target concentration standard =15.0%. Constant linear velocity of mesh belt transmission line. =0.5 m / min. Normal physical interval time =2 min.

[0150] Production schedule pending: Work Order A (High Priority Master Scheduler): Quantity =1000, surface area of ​​a single bottle =0.02 m 2 Minimum allowable concentration =14.0%; Baseline baking temperature =580℃.

[0151] Work Order B (Second Priority Candidate): Quantity =500, surface area of ​​a single bottle =0.015 m 2 Minimum allowable concentration =12.0% (for applications with lower surface roughness requirements); reference baking temperature =550℃.

[0152] Material characteristic parameters: Local parameters extracted from work order B in the system database: empirical coefficient of consumption per unit area. =0.04% / m 2 Material coupling attenuation factor =0.04. Thermodynamic proportional compensation coefficient of the baking oven. =0.5Kth=0.5. Natural cooling rate =2℃ / min Concentration decay prediction and crack filling batch triggering (corresponding to S200~S300): After processing multiple batches of work orders A, the system calculates using a prediction model (S201) that the predicted solution concentration will decay to a certain level when a certain production scheduling node is reached. =13.5%.

[0153] Because 13.5% < (14.0%), the originally scheduled high-priority work order A could not be processed.

[0154] The system iterates through the caulking batch: work order B. =12.0%≤13.5%, which meets the concentration requirements, so set it as the caulking batch and insert it into the current schedule.

[0155] Spatial decoupling calculation: The temperature of the preceding batch (A) is 580°C, and the target temperature of work order B is 550°C.

[0156] Thermodynamic response time cost: =(580-550) / 2=15 min.

[0157] Virtual physical gap length: =0.5·(15-2)=6.5 m.

[0158] Action to be performed: The conveyor belt will not slow down, but a 6.5m unloaded distance will be reserved between work orders A and B.

[0159] Cross-domain conversion and feedforward compensation (corresponding to S400): Because filler batch B was processed at an attenuation concentration (13.5%), the surface roughness decreased, leading to a reduction in thermal radiation absorption rate. The system calculates the conversion factor: =1-0.04·(15.0-13.5)=1-0.06=0.94.

[0160] Calculate the temperature feedforward bias: =0.5·(1-0.94)·550=16.5℃.

[0161] Execution Action (S500): When the edge control layer detects that work order B is about to enter the baking oven and is within the safety window, the target temperature of the baking oven is temporarily raised from the baseline of 550°C to 550 + 16.5 = 566.5°C.

[0162] Closed-loop feedback and parameter update (for S600): After work order B is completed, the actual surface temperature of the product is measured using an infrared thermometer. =548℃ (still 2℃ lower than the ideal benchmark of 550℃, indicating that the current model's compensation is insufficient and the actual material degradation has a greater impact).

[0163] Calculate the true observations (Substitute into the previously revised formula): ; =(16.5+550-548) / [0.5·(15.0-13.5)·550]=18.5 / 412.5≈0.0448.

[0164] Execution action: The system will include The observation vector with a value of 0.0448 is input into the Kalman filter to filter the original data in the database. (0.04) Upward correction iteration (e.g., updating to 0.042) will make the feedforward compensation amount more accurate when the next batch encounters the same working conditions.

[0165] To verify the effectiveness of the adaptive scheduling and collaborative control system of this invention, a 30-day production verification was conducted on a flexible production line that includes a frosted acid etching station and a continuous multi-temperature zone baking station. During the test, a total of 450 discrete work orders were processed, involving 8 different combinations of glass substrates and attachment consumables.

[0166] Experimental group setup: Control group (traditional control strategy): Static sequential scheduling is adopted. When the concentration of frosting solution decreases to below the current work order threshold, the machine is directly stopped to wait for replenishment or replacement of solution; when the baking oven switches the set temperature, manual intervention is used to reduce the speed or reserve a large physical gap; no cross-domain correlation of physicochemical properties is introduced, and temperature feedforward compensation is not implemented.

[0167] Experimental group (adaptive strategy of this invention): Deploy the system of this invention, fully enabling virtual physical gap injection, automatic scheduling of filling batches, temperature feedforward bias compensation based on surface radiation absorptivity, and online correction of model parameters based on local dimension-reduced Kalman filtering.

[0168] (1) Adaptive physical and chemical scheduling and improvement of overall equipment efficiency (OEE): See attached document Figure 9 The horizontal axis of the graph represents the production scheduling time series, and the vertical axis represents the concentration of the frosted sand tank solution. For example... Figure 9 As shown by the red line, in the control group, when the concentration decay reached the limit threshold of the main production order (e.g., 14.0%), a vertical upward jump occurred, indicating that the system had to be shut down for physical replenishment, resulting in a shutdown waiting time; while in the case of... Figure 9As shown by the green broken line, the experimental group did not trigger a shutdown at the same concentration node. The system automatically searched for and scheduled the "filling batch (marked by the shaded area)" with lower roughness requirements to be inserted into production, so that the concentration of the drug solution smoothly decreased to 12.0%, which significantly extended the life cycle of the single drug solution.

[0169] See attached document Figure 10 , combined Figure 9 The scheduling of grouting batches Figure 10 The lower secondary coordinate system (product placement area) visually illustrates the "virtual physical gap" without product placement injected by the system between the preceding batch and the caulking batch. Due to the existence of this gap, the temperature control curve (solid line) of the upper primary coordinate system can perform the heating action in advance using this spatiotemporal safety window before the caulking batch actually enters the baking oven, eliminating the thermal interference between the preceding and following batches caused by thermodynamic lag.

[0170] Data Results: In the control group, the cumulative ineffective waiting time due to waiting for solution concentration to recover, downtime for replenishment, and interference from temperature changes in the baking oven reached as high as 58 hours within 30 days. Through the aforementioned collaborative scheduling, the experimental group significantly reduced the ineffective waiting time to 14 hours, and the overall equipment efficiency (OEE) jumped from 79.2% in the traditional scheme to 93.5%.

[0171] (2) Comparison of cross-domain thermal compensation and product yield (solving curing defects): When using a lower concentration (e.g., 13.5%) of the chemical solution after decay to process the grouting batch, the reduced surface roughness of the glass leads to a decrease in its absorption rate of infrared thermal radiation.

[0172] See attached document Figure 12 ,like Figure 12 As shown, the control group, lacking cross-domain thermal compensation, suffered from insufficient absorption of radiant heat on the product surface, resulting in a very high percentage (6.8%) of "ink removal / poor adhesion" defects in the final product. The experimental group, however, incorporated the temperature feedforward bias calculated in step S400, which... Figure 10 The response curves shown precisely raised the target set temperature, compensating for the thermal loss caused by changes in the surface physical state. The proportion of qualified products in the experimental group increased significantly, and the rate of poor curing defects was reduced sharply to less than 0.4%, essentially eliminating thermal defects caused by chemical state shifts.

[0173] (3) Convergence verification of closed-loop parameter self-learning: When introducing a new material work order, the system assigns an initial error to the material coupling attenuation factor.

[0174] See attached document Figure 11 In the figure, the horizontal axis represents the iteration period (batch k), and the vertical axis represents the value of the material coupling attenuation factor β within the system. For example... Figure 11As shown, due to sensor noise and measurement bias in the industrial environment, the gray single actual observation points exhibit a discrete distribution with drastic fluctuations in real time. However, through the dynamic smoothing adjustment of the Kalman gain in step S600, the posterior model parameters represented by the blue solid line converge quickly and stably to the true physical characteristic values ​​(such as the benchmark value shown by the dashed line in the figure, around 0.045) after the first 5-8 batches of iterations.

[0175] Data Results: Figure 11 The verification results show that the filtering correction mechanism proposed in this invention has excellent noise resistance and self-learning convergence speed, ensuring that the system can adaptively overcome equipment aging and environmental drift during long-term continuous production, and maintain high-precision scheduling and temperature control prediction capabilities.

[0176] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles, characterized in that, It includes at least one processor and a memory, wherein the memory stores a computer program that, when executed by the processor, performs the following steps: Obtain the initial physicochemical state variables of the flexible production line and the characteristic parameters of the production orders to be scheduled; Based on the initial concentration of the frosting solution in the initial physicochemical state variables and the preset concentration decay model, the predicted concentration of the solution during the processing is predicted. When the predicted concentration of the solution is lower than the minimum allowable concentration corresponding to the main production order, a secondary priority work order with a minimum allowable concentration not higher than the current predicted concentration of the solution is selected from the work orders to be scheduled and used as the filling batch. Based on the difference in baseline baking temperature between the main production order and the filling batch, the temperature adjustment response time is calculated, and combined with the linear speed of the conveyor belt, a virtual physical gap without workpiece occupation is generated between the main production order and the filling batch. Based on the predicted concentration of the solution and the material characteristic parameters of the filling batch, the surface thermal radiation absorption conversion factor of the batch of products is calculated, and the temperature feedforward bias is determined accordingly. The flexible production line is controlled to operate according to the virtual physical gap, and before the filling batch enters the baking station, the temperature feedforward bias is superimposed on the reference baking temperature of the filling batch to perform equipment temperature control.

2. The intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles according to claim 1, characterized in that, The initial physicochemical state variables also include: the linear velocity of the mesh belt transmission line and the normal physical interval time between consecutive batches; The characteristic parameters of the production order to be scheduled include: the surface area of ​​the single product, the number of orders, the minimum allowable concentration, the reference baking temperature, and the material characteristic parameters including the empirical coefficient of consumption per unit area and the material coupling attenuation factor.

3. The intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles according to claim 2, characterized in that, The method involves calculating the temperature adjustment response time based on the difference in baseline baking temperature between the main production order and the seam filling batch, and combining this with the linear speed of the conveyor belt to generate a virtual physical gap without workpiece occupation between the main production order and the seam filling batch. Specifically, this includes: The temperature adjustment response time is determined by dividing the absolute value of the difference between the reference baking temperature between the main production order and the filling batch by the preset temperature adjustment rate of the baking station. The length of the virtual physical gap is the product of the linear velocity of the mesh transmission line and the effective response time, wherein the effective response time is the time difference between the temperature regulation response time and the normal physical gap time, and when the time difference is less than zero, the effective response time is zero.

4. The intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles according to claim 2, characterized in that, The step of calculating the surface thermal radiation absorption conversion factor of the batch of products based on the predicted concentration of the solution and the material characteristic parameters of the caulking batch, and thereby determining the temperature feedforward bias, specifically includes: The surface thermal radiation absorption conversion factor is obtained by subtracting the attenuation product from the preset value 1. The attenuation product is the product of the material coupling attenuation factor of the filling batch and the difference between the preset calibration concentration and the predicted concentration of the solution. The temperature feedforward bias is the product of the thermodynamic proportional compensation coefficient of the baking station, the conversion deviation rate, and the reference baking temperature of the filling batch, wherein the conversion deviation rate is obtained by subtracting the surface thermal radiation absorption conversion coefficient from a preset value of 1.

5. The intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles according to claim 4, characterized in that, When the computer program is executed by the processor, the following closed-loop feedback steps are also implemented: Obtain the actual surface temperature of the product after the grouting batch leaves the baking station; Based on the actual surface temperature of the product, the temperature feedforward bias, and the material characteristic parameters of the filling batch, the observed value of the material coupling attenuation factor under the processing state of the corresponding batch is calculated in reverse. Using the Kalman filter algorithm, the observed value of the material coupling attenuation factor is used as the observation input to perform a posteriori iterative update on the original material coupling attenuation factor in the memory.

6. The intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles according to claim 5, characterized in that, The reverse calculation of the observed material coupling attenuation factor under the corresponding batch processing state specifically includes: The observed value of the material coupling attenuation factor is the quotient of the residual temperature deviation and the theoretical maximum compensation. The residual temperature deviation is the sum of the temperature feedforward offset and the reference baking temperature of the caulking batch, minus the actual surface temperature of the product. The theoretical maximum compensation amount is the product of the thermodynamic proportional compensation coefficient of the baking station, the difference between the preset calibration concentration and the predicted concentration of the solution, and the reference baking temperature of the filling batch.

7. The intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles according to claim 6, characterized in that, The step of using the Kalman filter algorithm to perform a posteriori iterative update of the original material coupling attenuation factor in the memory specifically includes: Calculate the Kalman gain, which is the ratio of the error covariance of the previous batch to the sum of the error covariance of the previous batch and the preset measurement noise variance. The compensation adjustment term is obtained by multiplying the difference between the observed material coupling attenuation factor and the prior material coupling attenuation factor by the Kalman gain; The compensation adjustment term is superimposed on the prior material coupling attenuation factor to obtain the posterior updated material coupling attenuation factor, and then overwritten into the memory; The gain margin is obtained by subtracting the Kalman gain from the preset value of 1. The gain margin is then multiplied by the error covariance of the previous batch to obtain the synchronously reduced and updated error covariance.

8. The intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles according to claim 2, characterized in that, The step of predicting the solution concentration during processing based on the initial concentration of the frosting solution in the initial physicochemical state variables and a preset concentration decay model specifically includes: For each batch of work orders, the theoretical solution consumption of the corresponding batch of work orders is obtained by multiplying the empirical coefficient of unit area consumption, the surface area of ​​a single item, and the number of work orders by the physical volume of the frosting tank. The predicted solution concentration at any production node is obtained by successively subtracting the cumulative theoretical solution consumption of the scheduled production orders from the initial concentration of the frosting solution.

9. The intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles according to claim 1, characterized in that, When the computer program is executed by the processor, the following exception handling steps are also implemented: When the predicted concentration of the solution is lower than the minimum allowable concentration corresponding to the main production work order, and no candidate work order with a minimum allowable concentration not higher than the current predicted concentration is found when traversing the work orders to be scheduled, a physical replenishment or liquid replacement shutdown command is issued to the field edge control layer.

10. The intelligent scheduling and collaborative system for a flexible production line for deep processing of glass bottles according to claim 1, characterized in that, The control of the mesh belt transmission line of the flexible production line to operate according to the virtual physical idle time specifically includes: After the last workpiece of the main production schedule enters the mesh belt conveyor line, the linear speed of the mesh belt conveyor line is kept constant, and the entry of subsequent workpieces is suspended. After a preset delay time, the suspended state is released, and the first workpiece of the filling batch is scheduled to enter the mesh belt conveyor line. The preset delay time is the quotient obtained by dividing the length of the virtual physical gap by the linear speed of the network transmission line.

Images