Fireproof glass processing system and control method thereof

Abstract

The application provides a fireproof glass processing system and a control method thereof, and belongs to the technical field of automatic control of glass processing, and is used for solving the problems of low temperature control precision of adhesive tape laying, and inconsistent bonding quality caused by fixed pressing time in the prior art. The method establishes a double-body thermal contact model by acquiring glass geometric parameters and physical parameters, performs rolling time domain state-parameter joint estimation in each control cycle, solves the full trajectory optimal control problem from the current time to the end of laying based on the estimated value as a starting point, applies the first control amount of the optimal control sequence, and repeats until the end of laying; after the laying is completed, the natural cooling temperature curve of the adhesive tape is monitored, and the hot pressing is triggered when falling into the pressing temperature interval. The application realizes accurate online regulation and control of the adhesive tape laying temperature and dynamic pressing, and significantly improves the consistency of the bonding quality and the process robustness.

Description

Technical Field

[0001] This application relates to the technical field of automated control in glass processing, and in particular to a fireproof glass processing system and its control method. Background Technology

[0002] During the processing of fire-resistant glass, sealing strips need to be laid along the edges of the glass and then hot-pressed to ensure fire resistance. The bonding quality of the sealing strips is closely related to the heating temperature during installation, the installation speed, environmental conditions, and the timing of subsequent pressing.

[0003] Currently, most common methods for laying adhesive strips involve manual assistance and semi-automatic equipment. Operators set the heating temperature based on experience and lay the strips at a constant speed. Some automated systems have introduced temperature sensors and speed control, but they still use open-loop or simple closed-loop methods, such as fixing the temperature curve with constant-speed laying, or simply compensating for ambient temperature. In addition, existing systems usually treat laying heating and subsequent pressing as independent steps, lacking dynamic monitoring and linkage control of the adhesive strip cooling process.

[0004] The shortcomings of the existing technology are as follows: First, it does not consider the spatial lag effect caused by the physical installation position of the heating module, resulting in the temperature control phase being ahead or behind, affecting the actual heating uniformity of the adhesive strip; second, it lacks dynamic modeling of the thermal contact process between the adhesive strip and the glass, making it unable to adapt to different glass thicknesses, ambient temperature and humidity, and batch differences of adhesive strips; third, the laying speed and heating temperature are planned independently, making it difficult to achieve precise temperature gradient control from the beginning to the end, often resulting in insufficient bonding at the beginning or overheating and aging at the end; finally, the pressing timing is fixed and cannot be adaptively adjusted according to the actual cooling process, resulting in poor consistency of bonding strength. Summary of the Invention

[0005] This application provides a fireproof glass processing system and its control method, which can achieve temperature and speed optimization throughout the entire process of adhesive strip laying through optimal control and joint estimation in the rolling time domain, ensuring that the temperature of the adhesive strip falls precisely into the optimal bonding range before lamination.

[0006] In a first aspect, this application provides a control method for a fireproof glass processing system. The method includes: acquiring the geometric and physical property parameters of the glass to be processed and establishing a system dynamic model; within each control cycle, performing rolling time-domain state-parameter joint estimation based on measurement data within a past time window to obtain estimates of the current system state and unknown thermal parameters; starting from the estimated values, solving the optimal control problem for the entire trajectory from the current moment to the end of adhesive strip laying to obtain an optimal control sequence; applying the first control variable in the optimal control sequence, which includes a heating power command and a laying speed command; and repeating the above estimation, solving, and application steps until the laying is completed.

[0007] By adopting the above technical solution, this method integrates state estimation, parameter identification and full trajectory optimal control into a rolling time domain framework. Each control decision is based on the current system state and future full-process optimization, which can adaptively handle spatial lag, environmental changes and model uncertainties, and achieve accurate temperature trajectory tracking from start to finish.

[0008] Furthermore, the rolling time-domain state-parameter joint estimation adopts the rolling time-domain estimation method, which obtains the current state estimate and parameter estimate by solving an optimization problem. This optimization problem takes the rubber strip temperature measurement value, heating power command and laying speed data of a preset number of sampling times in the past as input, and the system dynamic model as constraint, to minimize the weighted sum of squares of initial state deviation, process noise, parameter prior deviation and measurement residual.

[0009] By adopting the above technical solution, rolling time-domain estimation can update the system status and unknown thermal parameters in real time using historical data in each control cycle, reducing the dependence on the accuracy of the initial calibration and improving the adaptability to batch differences of rubber strips and changes in operating conditions.

[0010] Furthermore, the objective function of the full-trajectory optimal control problem includes terminal cost and operational cost; the terminal cost penalizes the deviation between the adhesive strip temperature and the target final pressure temperature at the end of the laying and the deviation between the glass temperature and the target final pressure temperature; the operational cost includes penalties for the adhesive strip temperature deviating from the preset bonding temperature range, heating energy consumption penalties, and laying speed change rate penalties.

[0011] By adopting the above technical solution, the objective function simultaneously considers terminal accuracy, process interval constraints, and energy efficiency, so that the optimization result can not only guarantee the final bonding quality but also take into account economy, and the speed change rate penalty avoids the impact of drastic acceleration and deceleration on the mechanical system.

[0012] Furthermore, the full-trajectory optimal control problem also satisfies the following constraints: upper and lower limits of heating power, upper and lower limits of laying speed, and laying speed change rate constraint.

[0013] By adopting the above technical solution, the clear constraints ensure that the obtained control quantity is within the capability range of the actual actuator, thus avoiding command saturation or exceeding physical limits.

[0014] Furthermore, the full trajectory optimal control problem is discretized into a nonlinear programming problem using the direct collocation method and solved using sequential quadratic programming. The gradient of the objective function with respect to the control quantity is calculated through adjoint sensitivity analysis, which is based on the inverse integration of the adjoint equation of the system dynamic model.

[0015] By adopting the above technical solution, the accompanying sensitivity analysis significantly improves the gradient calculation efficiency, enabling the full trajectory optimization to be solved online in milliseconds, thus meeting the requirements of real-time control.

[0016] Furthermore, the system dynamic model is a two-body thermal contact model, which regards the adhesive strip and the glass as two nodes with finite heat capacity, and considers the spatial lag effect of the heating module. The spatial lag effect is expressed as follows: there is a delay in the moment when the heating power acts on the adhesive strip-glass contact point, and the delay time is equal to the fixed arc length distance between the heating center of the heating module and the contact point divided by the current laying speed.

[0017] By adopting the above technical solution, the dual-body thermal contact model accurately reflects the heat exchange process between the adhesive strip and the glass. Spatial lag compensation eliminates the phase error caused by the physical installation position of the heater, thus improving the accuracy of temperature prediction.

[0018] Furthermore, the dual-body thermal contact model includes a dynamic equation for the adhesive strip temperature and a dynamic equation for the glass temperature. The rate of change of the adhesive strip temperature with respect to time is equal to the lag term of the heating power minus the convective heat dissipation term of the adhesive strip to the environment minus the contact heat transfer term between the adhesive strip and the glass, and then divided by the heat capacity of the adhesive strip. The rate of change of the glass temperature with respect to time is equal to the contact heat transfer term minus the heat dissipation term of the glass to the environment, and then divided by the heat capacity of the glass. The lag term of the heating power represents the heating power at the current moment acting on the adhesive strip at a future moment, and the time difference between the future moment and the current moment is equal to the fixed arc length distance divided by the current laying speed.

[0019] By adopting the above technical solution, the dynamic model in the form of differential equations, based on physical mechanisms, can adapt to different glass sizes, environmental temperature and humidity, and changes in the material of the sealing strip, providing an accurate prediction basis for optimal control.

[0020] Furthermore, after the installation is completed, the natural cooling temperature decay curve of the adhesive strip is monitored. When the temperature of the adhesive strip falls into the preset pressing temperature range, the hot pressing unit is triggered to perform the pressing operation.

[0021] By adopting the above technical solution, the pressing trigger is dynamically linked to the actual cooling process of the adhesive strip, avoiding the problem of pressing too early or too late caused by a fixed waiting time, and significantly improving the consistency of bonding quality.

[0022] Furthermore, the pressing temperature range is 80°C to 90°C; when the temperature of the adhesive strip is higher than 90°C, auxiliary cooling is activated; when the temperature of the adhesive strip is lower than 80°C, an alarm is issued and pressing is suspended.

[0023] By adopting the above technical solutions, the optimal pressing process window was identified and an anomaly handling mechanism was set up, further ensuring production safety and product yield.

[0024] Secondly, this application provides a fire-resistant glass processing system. The system includes: an image acquisition unit for acquiring contour images of the glass to be processed; an environmental sensing unit for acquiring ambient temperature and humidity; an adhesive strip laying unit, including a robot, an infrared temperature sensor, and a miniature heating module, for performing adhesive strip laying and providing real-time feedback on the adhesive strip temperature; a hot-pressing unit for hot-pressing the laid adhesive strip; a conveying and positioning unit for feeding the glass and providing position feedback; and a central controller connected to each of the above units, wherein the central controller is configured to execute the control method of the fire-resistant glass processing system described in the first aspect.

[0025] By adopting the above technical solutions, the system integrates image perception, environmental monitoring, temperature feedback and high-speed computing capabilities, and can fully realize the aforementioned control methods to form a closed-loop automated processing equipment from data acquisition to execution control.

[0026] In summary, this application has at least the following beneficial effects:

[0027] A fireproof glass processing control scheme that deeply integrates rolling time-domain state-parameter joint estimation with full trajectory optimal control is provided, which significantly improves the temperature consistency and bonding quality of adhesive strip laying;

[0028] By using a two-body thermal contact model and spatial lag compensation, the control phase error caused by the physical installation position of the heater was solved, thus improving the temperature control accuracy.

[0029] By utilizing natural cooling and dynamic monitoring to trigger pressing, the temperature of the adhesive strip is always within the optimal bonding window during the bonding process.

[0030] It should be understood that the description in the Summary Section is not intended to limit the key or essential features of the embodiments of this application, nor is it intended to restrict the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0031] The above and other features, advantages, and aspects of the embodiments of this application will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0032] Figure 1 A schematic diagram of a fireproof glass processing system according to an embodiment of this application is shown.

[0033] Figure 2 A flowchart of a control method for a fireproof glass processing system according to an embodiment of this application is shown. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0035] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0036] This application provides a fireproof glass processing system and its control method, which achieves precise online control of the adhesive strip laying temperature and natural cooling dynamic pressing through the deep integration of rolling time-domain state-parameter joint estimation and full trajectory optimal control, significantly improving the consistency of bonding quality and process robustness.

[0037] In a first aspect, embodiments of this application disclose a fireproof glass processing system. Figure 1 A schematic diagram of a fireproof glass processing system according to an embodiment of this application is shown.

[0038] Reference Figure 1 The system includes a central controller and image acquisition unit, environmental sensing unit, adhesive strip laying unit, hot pressing unit, and conveying and positioning unit, all connected to the central controller. The entire system is deployed in a clean processing workshop where the ambient temperature is typically maintained between 15 and 30 degrees Celsius and the relative humidity is between 30% and 70%.

[0039] The image acquisition unit is fixed above the glass conveyor line, approximately 500 to 800 millimeters above the glass surface. This unit includes an industrial line scan camera and a ring-shaped LED supplementary light source. The line scan camera has a resolution of at least 4096 pixels by 4096 pixels, and the acquisition frame rate can be set to 50 frames per second. The supplementary light source has a color temperature of 5000 Kelvin and adjustable illuminance, typically set to 20,000 lux. This unit is used to acquire a complete outline image of the glass to be processed and transmit the image data to the central controller via gigabit Ethernet. Before each laying task begins, the glass moves with the conveyor belt to a position directly below the camera, and a photoelectric switch triggers the camera to acquire a complete image.

[0040] The environmental sensing unit is installed in the center of the processing area, approximately 1.5 meters above the ground, and away from air conditioning vents and heating equipment. This unit includes an integrated temperature and humidity sensor with a temperature measurement accuracy of ±0.3 degrees Celsius and a humidity measurement accuracy of ±2% relative humidity, sampling at a frequency of 1 Hz. This sensor collects ambient temperature and humidity data in real time and transmits it to the central controller via an RS485 bus. Typical ambient temperature values ​​are 20 degrees Celsius and typical humidity values ​​are 50%. The system will issue an alert when the ambient temperature is below 15 degrees Celsius or above 30 degrees Celsius.

[0041] The adhesive strip laying unit includes a six-DOF collaborative robot, an infrared temperature sensor mounted on the robot's end effector, and a miniature heating module. The collaborative robot has a working radius of 1200 mm, a repeatability of ±0.05 mm, and an end-effector load capacity of 5 kg. The robot's base is fixed to one side of the conveyor belt, and its range of motion covers the entire edge of the glass. The infrared temperature sensor is non-contact, with a measurement range of 0 to 200 degrees Celsius, an accuracy of ±0.5 degrees Celsius, a response time of 10 milliseconds, and the sensor probe is approximately 15 mm from the adhesive strip surface, with a measurement spot diameter of 8 mm. The miniature heating module is an electrothermal ceramic heating element with a power adjustment range of 0 to 500 watts. The fixed arc length distance along the laying path between the heating center and the adhesive strip-glass contact point is 25 mm, calibrated through a step response test before shipment. The temperature rise time constant of the heating module is approximately 0.3 seconds. This adhesive strip laying unit performs the adhesive strip laying action along the glass edge according to the laying speed and heating power commands sent by the central controller, while simultaneously feeding back the adhesive strip temperature to the central controller in real time.

[0042] The conveying and positioning unit includes a conveyor belt driven by a servo motor and an incremental encoder mounted on the conveyor belt drive shaft. The conveyor belt is 1,500 mm wide, and its maximum conveying speed after carrying glass is 500 mm per second. The encoder outputs 2,500 pulses per revolution, which are connected to the counter channel of the central controller via differential signal lines for real-time feedback of the glass's current position. A set of through-beam photoelectric switches is installed at both the beginning and end of the conveyor belt to trigger glass feeding and discharging signals.

[0043] The hot-pressing unit includes a pneumatically operated lifting hot-pressing plate, measuring 1000 mm x 600 mm and 20 mm thick. Four heating wires and six thermocouples are evenly embedded inside. The total power of the heating wires is 2 kW, and the thermocouples have a measurement accuracy of ±0.2 degrees Celsius. The hot-pressing plate is driven to rise and fall by a double-rod cylinder. The cylinder's working air pressure is 0.4 to 0.6 MPa, and the pressing pressure is adjustable from 0.2 to 0.5 MPa, preferably 0.3 MPa. Under the control of a central controller, this unit rapidly lowers the hot-pressing plate to press the adhesive strip when its temperature naturally cools and falls within a preset range. The pressing time is typically set to three seconds.

[0044] The central controller is an industrial PC with an Intel Core i7-12700 processor, 32 gigabytes of memory, and an NVIDIA RTX 3060 graphics processor to accelerate matrix operations. The PC has a real-time operating system with a task cycle set to ten milliseconds. The controller connects to the robot driver, heating module driver, conveyor belt servo driver, and various sensor acquisition modules via an EtherCAT bus. The controller internally stores pre-calibrated camera parameters, adhesive strip property parameters, and initial parameters of the two-body thermal contact model. The controller receives image data, environmental data, temperature feedback, and position feedback. Based on the control method of the fireproof glass processing system provided in this application embodiment, it performs rolling time-domain state-parameter joint estimation and full-trajectory optimal control calculation in each control cycle. Then, it sends heating power commands and laying speed commands to the adhesive strip laying unit and a pressing trigger signal to the hot-pressing unit. Through the coordinated operation of the aforementioned hardware units, the entire system provides a complete physical execution and sensing environment for the automated control of fireproof glass adhesive strip laying. The system is capable of performing all the control steps described in the following method embodiments, thereby achieving intelligent control of the entire process from glass contour recognition to adhesive strip laying to hot pressing.

[0045] Secondly, embodiments of this application disclose a control method for a fireproof glass processing system.

[0046] Figure 2 A flowchart illustrating a control method for a fire-resistant glass processing system according to an embodiment of this application is shown. (Refer to...) Figure 2 The method specifically includes the following steps:

[0047] S1: Obtain the geometric and physical property parameters of the glass to be processed and establish a dynamic model of the system.

[0048] The specific steps of this method include: obtaining the geometric and physical property parameters of the glass to be processed and establishing a system dynamic model; the system dynamic model is a two-body thermal contact model, which considers the adhesive strip and the glass as two nodes with finite heat capacity, and considers the spatial lag effect of the heating module; the spatial lag effect is expressed as: there is a delay in the moment when the heating power acts on the adhesive strip-glass contact point at the current moment, and the delay time is equal to the fixed arc length distance between the heating center of the heating module and the contact point divided by the current laying speed; the two-body thermal contact model includes the dynamic equation of adhesive strip temperature and the dynamic equation of glass temperature; the rate of change of adhesive strip temperature with respect to time is equal to the lag term of heating power minus the convective heat dissipation term of adhesive strip to the environment minus the contact heat transfer term between adhesive strip and glass, and then divided by the heat capacity of adhesive strip; the rate of change of glass temperature with respect to time is equal to the contact heat transfer term minus the heat dissipation term of glass to the environment, and then divided by the heat capacity of glass; wherein the lag term of heating power represents the heating power at the current moment acting on the adhesive strip at a future moment, and the time difference between the future moment and the current moment is equal to the fixed arc length distance divided by the current laying speed.

[0049] Specifically, before each installation task begins, the central controller first triggers the image acquisition unit to photograph the glass stationary on the conveyor belt, obtaining a high-resolution grayscale image. The controller then performs Gaussian filtering for noise reduction and Canny edge detection on the image, extracting the pixel set of the glass edges. Subsequently, the Douglas-Pockmark polygon fitting algorithm is used to extract the precise polygon of the glass's outer contour from this pixel set, thereby calculating the glass's perimeter, area, and minimum bounding rectangle size. The perimeter, in millimeters, is used to determine the total length of the adhesive strip installation; the area is used for subsequent calculations of the glass's heat capacity. Simultaneously, the controller reads a pre-stored library of adhesive strip properties from its internal non-volatile memory, obtaining the adhesive strip width, thickness, density, and specific heat capacity based on the current production batch, and similarly reads the glass's thickness, density, and specific heat capacity. If the glass thickness is not pre-stored, the system allows the operator to manually input it via a human-machine interface or automatically measure it using an external laser thickness gauge. All of the above geometric and property parameters are stored in the controller's memory as fixed inputs for the system's dynamic model.

[0050] After obtaining the above parameters, the controller establishes the mathematical framework of the two-body thermal contact model. This model treats the adhesive strip as a concentrated heat capacity node, whose heat capacity is the product of the mass per unit length of the adhesive strip and its specific heat capacity. The mass per unit length of the adhesive strip is calculated from its width, thickness, and density. The glass is treated as another concentrated heat capacity node, whose heat capacity is the product of the total mass of the glass and its specific heat capacity. The total mass of the glass is calculated from its area, thickness, and density. The adhesive strip-air convection heat transfer coefficient, the glass-air convection heat transfer coefficient, and the contact thermal resistance between the adhesive strip and the glass in the model are initially set to typical values ​​calibrated experimentally. These typical values ​​are stored in the controller and will be corrected online in subsequent steps through rolling time-domain estimation. The spatial lag distance in the model was obtained through a step response experiment before shipment: with the adhesive strip laying unit stationary, a step power command was applied to the micro heating module, and an infrared temperature sensor was used to monitor the temperature response of the adhesive strip at different arc length positions away from the heating center of the heating module. Correlation analysis was used to find the spatial offset corresponding to the peak temperature response; this offset is the fixed arc length distance, which is stored in the controller and directly referenced in this step. The convective heat transfer area and the adhesive strip heat dissipation surface area in the model were pre-calculated as constants based on the adhesive strip geometry and the glass surface area. At this point, the system dynamic model was fully established, providing a physical prediction basis for subsequent rolling time-domain estimation and optimal control.

[0051] S2: Within each control cycle, perform rolling time-domain state-parameter joint estimation based on measurement data from the past time window to obtain estimates of the current system state and unknown thermal parameters.

[0052] The specific steps of this method include: within each control cycle, performing rolling time-domain state-parameter joint estimation based on measurement data within a past time window to obtain estimates of the current system state and unknown thermal parameters; the rolling time-domain state-parameter joint estimation adopts a rolling time-domain estimation method, which obtains the current state estimate and parameter estimate by solving an optimization problem. This optimization problem takes the rubber strip temperature measurement value, heating power command and laying speed data at a preset number of sampling times in the past as input, and the system dynamic model as constraint, minimizing the weighted sum of squares of initial state deviation, process noise, parameter prior deviation and measurement residual.

[0053] Specifically, at the beginning of each control cycle, the central controller first reads recorded data from the historical buffer for a preset number of consecutive sampling moments. This preset number is typically twenty, corresponding to a 0.2-second time window. The recorded data for each sampling moment includes the actual temperature measurement of the adhesive strip acquired by the infrared temperature sensor, the heating power command value sent by the central controller from the previous cycle, and the laying speed value converted from encoder feedback. These measurements are all timestamped and arranged in chronological order to form a fixed-length sliding window. The oldest data in the window is discarded as new data arrives, ensuring that the estimation is always based on the most recent system behavior.

[0054] The controller embeds the system dynamic model as a constraint into the rolling time-domain estimation problem. This dynamic model, established in step S1, includes the temperature states of the two thermal capacity nodes—the adhesive strip and the glass—as well as the unknown thermal parameters to be estimated. The unknown thermal parameters mainly include the contact thermal resistance between the adhesive strip and the glass, and the convective heat transfer coefficient between the adhesive strip and the air. These two parameters drift slowly in actual production due to variations in adhesive strip batches, glass surface conditions, and environmental changes; therefore, they need to be tracked in real-time through online estimation.

[0055] The objective function of the rolling time-domain estimation problem consists of a weighted sum of four parts. The first part is the initial state bias penalty, which measures the difference between the estimated state at the beginning of the window and the prior predicted state, obtained recursively from the previous estimation result through the model. The second part is the process noise penalty, which measures unmodeled random disturbances in the system's dynamic model, allowing for a reasonable degree of fluctuation in the state based on the model prediction. The third part is the parameter prior bias penalty, which encourages the estimated unknown thermal parameters not to deviate too far from the factory-calibrated typical values, avoiding parameter estimation divergence. The fourth part is the measurement residual penalty, which measures the difference between the model-predicted rubber strip temperature and the actual sensor measurement, and is the main source of information for the estimation problem. The controller assigns weighting coefficients to each of these four components. These weighting coefficients are pre-tuned through simulation experiments to achieve a balance between fast response and noise resistance in the estimation results.

[0056] Solving this rolling time-domain estimation problem is equivalent to solving a constrained nonlinear least squares problem. The controller employs the Gauss-Newton iterative method, with each iteration requiring the calculation of the Jacobian matrix using the linearized form of the system's dynamic model. Due to the fixed and relatively small window length, this optimization problem typically converges within three to five iterations, with a computation time of less than five milliseconds, meeting the real-time requirement of a ten-millisecond control cycle. After solving, the controller obtains the system state estimate at the end of the window, i.e., the current moment, including the current adhesive strip temperature and the current glass temperature. Simultaneously, it obtains updated estimates of the unknown thermal parameters, namely the contact thermal resistance and convective heat transfer coefficient that best match the measured data. These estimates are stored in the controller's memory for use in the next step, S3, for the full-trajectory optimal control problem.

[0057] S3: Starting from the estimated value, solve the optimal control problem of the entire trajectory from the current time to the end of the rubber strip laying time to obtain the optimal control sequence.

[0058] The specific steps of this method include: starting from the estimated value, solving the optimal control problem of the entire trajectory from the current time to the end time of adhesive strip laying to obtain the optimal control sequence; the objective function of the optimal control problem of the entire trajectory includes terminal cost and operating cost; the terminal cost penalizes the deviation between the adhesive strip temperature and the target final pressure temperature at the end time of laying, as well as the deviation between the glass temperature and the target final pressure temperature; the operating cost includes penalties for the adhesive strip temperature deviating from the preset bonding temperature range, heating energy consumption penalties, and laying speed change rate penalties; the optimal control problem of the entire trajectory also satisfies the following constraints: upper and lower limits of heating power constraints, upper and lower limits of laying speed constraints, and laying speed change rate constraints; the optimal control problem of the entire trajectory is discretized into a nonlinear programming problem by the direct collocation method, and solved by sequential quadratic programming, wherein the gradient of the objective function with respect to the control quantity is calculated by adjoint sensitivity analysis, which is realized by inverse integration of the adjoint equation of the system dynamic model.

[0059] Specifically, after obtaining the estimated temperature of the adhesive strip, the estimated temperature of the glass, and the updated contact thermal resistance and convective heat transfer coefficient given in step S2, the central controller begins to construct the optimal control problem for the entire trajectory from the current moment to the end of the laying process. The reference laying speed adopts a preset baseline speed value, which is 50 millimeters per second in this embodiment. The laid arc length is obtained by real-time feedback accumulation from the encoder. The end of the laying process is estimated by dividing the remaining arc length by the reference laying speed. The remaining arc length is equal to the total circumference of the glass minus the laid arc length. The controller first divides the continuous time interval into multiple discrete time nodes, and the step size between adjacent nodes is determined adaptively: a smaller step size is used in the region where changes are rapid in the initial stage, typically 0.01 seconds; in the subsequent relatively stable region, the step size can be appropriately widened to 0.1 seconds. The system state at each discrete time node is recursively derived through the two-body thermal contact model established in step S1, where the control quantities heating power and laying speed are assumed to be piecewise constants within each step size. This transforms the original continuous-time optimal control problem into a finite-dimensional nonlinear programming problem, with the decision variables being the heating power value and laying speed value at each discrete time node.

[0060] The objective function of this nonlinear programming problem consists of the sum of two parts: the terminal cost and the operational cost. The terminal cost focuses on the state at the end of the installation, namely the deviation between the adhesive strip temperature and the glass temperature and the target final pressure temperature. The target final pressure temperature is preset to 85 degrees Celsius, which is the median of the optimal bonding window of 80 to 90 degrees Celsius for the adhesive strip, ensuring that the adhesive strip has sufficient time to fall into the window during subsequent natural cooling. The terminal cost is in the form of a sum of squares, so that the penalty increases sharply as the deviation increases, guiding the optimization result to hit the target temperature as accurately as possible. The operational cost accumulates the instantaneous costs throughout the entire laying process, comprising three sub-items: The first sub-item is the penalty for the adhesive strip temperature deviating from the preset bonding temperature range of 80-90 degrees Celsius. When the adhesive strip temperature is below 80 degrees or above 90 degrees, a penalty proportional to the square of the deviation is incurred. This ensures the adhesive strip remains within the optimal bonding temperature range as much as possible throughout the laying process. The second sub-item is the heating energy consumption penalty, which is the integral of the square of the heating power over time, used to avoid unnecessary energy waste. The third sub-item is the integral of the square of the rate of change of laying speed, used to suppress drastic speed fluctuations, protect the robot's mechanical structure, and ensure laying uniformity. The controller assigns weighting coefficients to each of these three sub-items. These weighting coefficients are determined through offline simulation and on-site debugging. Typical values ​​prioritize temperature tracking accuracy, followed by energy consumption, and then speed smoothing.

[0061] In addition to the objective function, this nonlinear programming problem must also satisfy a series of hard constraints. The heating power command cannot exceed the rated power limit of the micro-heating module (500 watts) nor fall below zero watts; the laying speed cannot exceed the robot's maximum safe speed of 100 millimeters per second, nor fall below the minimum speed of 20 millimeters per second to ensure proper adhesion of the adhesive strip; simultaneously, the rate of change of the laying speed is limited by the robot's acceleration and deceleration capabilities, with a speed change not exceeding 50 millimeters per second. These constraints are added to the optimization problem in the form of inequalities to ensure that the solved control commands are within the capabilities of the actual actuator.

[0062] The core challenge in solving the aforementioned nonlinear programming problem lies in the highly nonlinear objective function and constraints, and the large number of decision variables, typically reaching hundreds. To improve efficiency, the controller employs a sequential quadratic programming method. This method approximates the original problem as a quadratic programming subproblem in each iteration, obtains the search direction by solving this subproblem, and then performs a one-dimensional search along this direction to update the decision variables, repeating this process until convergence. The success of sequential quadratic programming depends on the accuracy of the objective function gradient, but directly performing analytical differentiation on the complex two-body thermal contact model is extremely cumbersome. Therefore, the controller uses adjoint sensitivity analysis to calculate the gradient of the objective function with respect to each control variable. The specific approach is as follows: First, starting from the current moment, the system dynamic model is integrated forward, recording the state trajectory at each discrete time node. Then, the adjoint equation is constructed, which is a linearized form of the original dynamic model, progressing backward along time. Its right-hand side consists of the partial derivatives of the objective function with respect to the state. Starting from the final moment, the adjoint equation is integrated backward to obtain the adjoint state at each time node. Finally, the gradient of the objective function with respect to a certain control variable is equal to the integral over time of the product of the adjoint state and the partial derivative of the system dynamic model with respect to that control variable within the time period. All gradient values ​​can be efficiently obtained through one forward integration and one backward integration. The computational cost is approximately independent of the number of decision variables, significantly reducing the solution time. The entire sequence quadratic programming iteration typically converges within five to ten iterations, with the total computation time controlled within twenty milliseconds, meeting the real-time scheduling requirements of a ten-millisecond control cycle. After the solution is completed, the controller obtains the optimal control sequence from the current moment to the end of the laying process. This sequence includes the heating power command value and laying speed command value at each discrete time node. These optimal control variables are temporarily stored in memory for use in the next step, S4.

[0063] S4: Apply the first control quantity in the optimal control sequence, which includes heating power command and laying speed command.

[0064] The method of this step specifically includes: applying the first control quantity in the optimal control sequence, which includes a heating power command and a laying speed command.

[0065] Specifically, the central controller extracts the first control variable from the optimal control sequence obtained in step S3, which corresponds to the heating power command value and the laying speed command value at the start of the current control cycle. The heating power command value is a continuous value between zero watts and 500 watts, its magnitude determined by the comprehensive optimization of the full-trajectory optimal control problem based on the current system state and the future target trajectory. The laying speed command value is a value between 20 millimeters per second and 100 millimeters per second, also determined by the optimization process. The controller sends the heating power command value to the driver of the micro-heating module via the EtherCAT bus. The driver converts it into the duty cycle of a pulse-width modulation signal, thereby controlling the actual heating power of the heating element. Simultaneously, the controller sends the laying speed command value to the motion controller of the six-DOF robot. The robot adjusts the movement speed of the end effector along the tangential direction of the glass contour in real time according to this speed command, while maintaining the adhesive strip clamping mechanism always aligned with the glass edge. It is worth noting that the controller only applies the first control variable in the optimal control sequence, discarding the remaining control variables for future moments in the sequence. This is because the rolling time-domain control strategy assumes that after applying the optimal first action in the current cycle, the system state will change when the next control cycle arrives. At that time, it is necessary to re-estimate the state and optimize the entire trajectory to obtain an updated optimal control sequence. This strategy of "only executing the optimal first step at the current moment" allows the control system to continuously use the latest feedback information to correct subsequent decisions, possessing inherent robustness and adaptability. While applying the control input, the controller continues to collect real-time adhesive strip temperature from the infrared temperature sensor, the actual laying speed fed back by the encoder, and the current laid arc length position every ten milliseconds. This real-time data will serve as the input measurement data for step S2 in the next control cycle, forming a complete "estimation-optimization-execution-feedback" closed loop. In this way, the system calculates the globally optimal control action in each cycle using the most accurate model and state information, and executes it precisely, thereby ensuring that the adhesive strip temperature always approaches the target trajectory throughout the laying process and creating optimal conditions for subsequent pressing at the end of the laying.

[0066] S5: Repeat the above estimation, solution and application steps until the laying is completed.

[0067] The specific method for this step includes: repeatedly performing the above estimation, solution, and application steps until the laying is completed.

[0068] Specifically, after completing the estimation, solution, and action application for the first control cycle, the central controller determines whether the currently laid arc length has reached or exceeded the total circumference of the glass. The laid arc length is accumulated in real time by the encoder in the conveying and positioning unit. Each encoder pulse corresponds to a fixed distance the glass moves, which is predetermined through wheel diameter calibration. If the currently laid arc length is less than the total circumference of the glass, the system automatically enters the next control cycle. At the beginning of each new control cycle, the controller repeats the rolling time-domain state-parameter joint estimation in step S2. The measurement data from the previous time window has been updated, including the newly acquired adhesive strip temperature, actual laying speed, and encoder position after the control input was applied in the previous cycle. Since the system state and unknown thermal parameters may change with the laying process and the external environment, the rolling time-domain estimation can recalibrate the state and parameters using the latest measurements to ensure that the model always fits the current actual working conditions. The optimal control problem for the entire trajectory in step S3 is then repeated. Since the estimated current state may deviate from the predicted state in the previous cycle, the re-optimized optimal control sequence from the current moment to the end of the laying process is adjusted accordingly, allowing subsequent control actions to compensate for earlier model errors or external disturbances. Finally, step S4 is repeated, applying the first control variable from the newly obtained optimal control sequence. This process repeats continuously, with each control cycle outputting a control action based on the latest state estimation and global optimization results, forming a recursive, adaptive closed-loop control process. This process continues until the laid arc length fed back by the encoder reaches or exceeds the total circumference of the glass. At this point, the controller determines that the laying task is complete and exits the loop. Throughout the entire loop, the control cycle is strictly fixed at ten milliseconds, ensuring the system's real-time performance and stability. Through this rolling time-domain repetition mechanism, the system can achieve high-precision temperature and speed coordinated control without relying on a precise initial model and offline trajectory, relying on online estimation and online optimization. This effectively addresses practical engineering problems such as batch differences in adhesive strips, fluctuations in ambient temperature and humidity, and sensor noise.

[0069] S6: After the installation is completed, monitor the natural cooling temperature decay curve of the adhesive strip. When the temperature of the adhesive strip falls into the preset pressing temperature range, trigger the hot pressing unit to perform the pressing operation.

[0070] The specific steps of this process include: after the installation is completed, monitoring the natural cooling temperature decay curve of the adhesive strip; when the temperature of the adhesive strip falls into the preset pressing temperature range, triggering the hot pressing unit to perform the pressing operation; the pressing temperature range is 80°C to 90°C; when the temperature of the adhesive strip is higher than 90°C, auxiliary cooling is started; when the temperature of the adhesive strip is lower than 80°C, an alarm is issued and pressing is paused.

[0071] Specifically, when step S5 determines that the laid arc length has reached the total circumference of the glass, the central controller immediately stops the movement of the adhesive strip laying unit and the power output of the heating module, and records the final temperature of the adhesive strip measured by the infrared temperature sensor at the end of the laying. After laying, the adhesive strip no longer receives external heating, and its temperature begins to naturally cool down. The controller enters the pressing waiting stage. The core task of this stage is to monitor the natural cooling decay curve of the adhesive strip temperature in real time and accurately capture the moment when the temperature falls into the pressing temperature window of 80 to 90 degrees Celsius. To this end, the controller continues to collect the readings of the infrared temperature sensor at a period of 10 milliseconds, while pausing all control outputs and running only one temperature monitoring task. This task compares the current temperature of the adhesive strip collected each time with the preset pressing temperature range. If the temperature of the adhesive strip is higher than 90 degrees Celsius, it indicates that the adhesive strip is still overheated and not suitable for immediate pressing. At this time, the controller starts auxiliary cooling measures: turns on the fan or compressed air nozzle installed near the hot pressing unit to perform forced convection cooling on the surface of the adhesive strip to accelerate the temperature drop process, while continuing to monitor the temperature. The intensity of auxiliary cooling is automatically adjusted based on the temperature deviation; the greater the deviation, the higher the fan speed or airflow, but localized overcooling must be avoided. If the adhesive strip temperature is below 80 degrees Celsius, it indicates excessive natural cooling, and the adhesive strip may have lost sufficient tack. In this case, the controller determines that the pressing conditions are not met, immediately issues an audible and visual alarm signal, and pauses the operation of the hot pressing unit, while recording the abnormal event for operator handling. When the adhesive strip temperature falls precisely within the 80-90 degree Celsius range, the controller considers the optimal pressing process window reached and immediately triggers the hot pressing unit to perform the pressing operation. Upon receiving the trigger signal, the hot pressing unit drives the pneumatic lifting hot pressing plate to descend rapidly, pressing the adhesive strip against the glass edge with a pressure of 0.3 MPa, maintaining the pressure for three seconds to ensure full adhesion between the adhesive strip and the glass. After pressing is complete, the hot pressing plate automatically lifts, ending the entire fireproof glass processing process. Through the above dynamic monitoring and adaptive triggering mechanism, this method effectively solves the problem that fixed waiting times cannot adapt to different environmental conditions and differences in the initial temperature of the adhesive strip, ensuring that each pressing is performed at the moment when the adhesive strip has optimal tack, significantly improving the consistency of bonding quality.

[0072] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0073] To more clearly illustrate the specific algorithm calculation process involved in the above method, an exemplary mathematical implementation is given below. This implementation is completely consistent with the description of the system dynamic model, rolling time-domain estimation, and full trajectory optimal control in the foregoing embodiments, and each parameter, function, and symbol used therein is clearly defined.

[0074] Define the sampling period as Seconds, the current control cycle number is At that moment The system is aware of the following data: the laid arc length obtained from encoder feedback. (Unit: mm) Measured temperature of the adhesive strip obtained from an infrared temperature sensor. (Unit: degrees Celsius) Heating power command applied in the previous cycle and laying speed instructions And the system state estimated at the end of the previous cycle. and parameter estimation .in Indicates the temperature of the rubber strip. Indicates the glass temperature. The convective heat transfer coefficient between the rubber strip and the air (unit: watts per square millimeter Kelvin). The contact thermal resistance between the sealing strip and the glass (unit: Kelvin per watt), symbol This represents the transpose of a vector.

[0075] First, perform rolling time-domain state-parameter joint estimation. Select the estimation window length. This means using data from the past 20 sampling points. Define the state sequence within the window. ,in Define the vector of unknown parameters to be estimated. Solve the following optimization problem:

[0076]

[0077] in It is a priori state estimate, which is obtained by recursion from the estimation result of the previous step through the model; These are the prior values ​​of the parameters (i.e., the factory calibration values); These are the covariance matrices of initial state deviation, process noise, measurement noise, and prior parameter deviation, respectively, all of which are constant matrices pre-tuned through simulation experiments; Representing a quadratic form ; This is the discretized form of the two-body thermal contact model, specifically expressed as:

[0078]

[0079] In the above formula, The mass of the adhesive strip per unit length (grams per millimeter). The specific heat capacity of the adhesive strip (joules per gram Kelvin). The total mass of the glass (grams). Specific heat capacity of glass (joules per gram Kelvin). The convective heat transfer area per unit length of the rubber strip between the rubber strip and the air (square millimeters). The convective heat transfer coefficient between glass and air (watts per square millimeter Kelvin). The area of ​​the glass for convective heat transfer between the glass and the air (square millimeters). The ambient temperature (in degrees Celsius) is collected in real time by the environmental sensing unit. Lag steps. ,in The fixed arc length distance (in millimeters) between the heating center of the heating module and the contact point between the adhesive strip and the glass. For the first Laying speed at any given moment (millimeters per second). Indicates rounding up; Represents the heating power command (watts) at a historical moment. Measurement function. This means only the temperature of the adhesive strip is measured. The above optimization problem is solved using the Gauss-Newton method, and the state estimate at the current moment is obtained after three iterations. and parameter estimation .

[0080] Next, full-trajectory optimal control is executed. Define the starting point from the current time step. By the end of the laying Remaining time ,in , This is the total circumference of the glass (in millimeters). Millimeters per second is used as the reference velocity. The remaining time is discretized into... Intervals, step size An adaptive approach is used to determine the initial stage when the temperature change rate is large. Seconds, taken during the subsequent stable phase Seconds. The decision variable is the heating power in each interval. and laying speed subscript The corresponding discrete interval index. The objective function is:

[0081]

[0082] Terminal cost ,in The target final pressure temperature is used as the weighting coefficient. , Operating costs ,in and These are the lower and upper limits (in degrees Celsius) of the preset bonding temperature range, with weighting coefficients set to... , , ; This indicates taking the larger of zero and the independent variable. The constraint is:

[0083]

[0084] The power unit is watts, the speed unit is millimeters per second, and the rate of change of speed is constrained to no more than 50 millimeters per second. The state recursion uses the same discretized model as the estimation step. ,in Let the vector of the right-hand side of the above differential equation be the initial state. To achieve efficient solution, the adjoint method is used to calculate the gradient of the objective function with respect to the decision variables. The adjoint variables are defined. Its reverse recursive relationship is:

[0085]

[0086] in Let Jacobian matrix be the system model. and These are the partial derivatives of the running cost and the terminal cost with respect to the state, respectively. The objective function with respect to the control quantity... The gradient is:

[0087]

[0088] Similarly, we can obtain the correct answer. gradient After obtaining all gradients, a sequential quadratic programming iteration is used to solve the problem, which typically converges in 5 iterations. The first set of optimal control sequences obtained is selected. As the output instruction for the current cycle.

[0089] Finally, the central controller will Send to the miniature heating module driver, The data is sent to the robot's motion controller, and the actual execution data for the current cycle (including actual power, actual speed, and measured rubber strip temperature) is recorded as input for the rolling time-domain estimation of the next cycle. This calculation process is repeated until the laid arc length is reached. Reaching or exceeding the total circumference of the glass Through the algorithm examples described by the above mathematical formulas, those skilled in the art can clearly understand the specific calculation process of rolling time-domain estimation and full trajectory optimal control in this application, and can implement it in programming without creative effort.

[0090] In summary, the fire-resistant glass processing system and its control method provided in this application, through the close integration of system hardware and method, achieve intelligent control of the entire process from glass contour recognition and adhesive strip laying to hot pressing. The system hardware provides the necessary sensing and execution capabilities, while the method, through the deep fusion of rolling time-domain state-parameter joint estimation and full-trajectory optimal control, adapts online to environmental changes and model uncertainties, ensuring precise control of the adhesive strip laying temperature; simultaneously, through natural cooling monitoring and dynamic pressing triggering, it ensures the utilization of the optimal bonding window. The entire technical solution has significant inventiveness, practicality, and industrial promotion value.

[0091] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A control method for a fireproof glass processing system, characterized in that, include: Obtain the geometric and physical property parameters of the glass to be processed and establish a system dynamic model; within each control cycle, perform rolling time-domain state-parameter joint estimation based on the measurement data in the past time window to obtain the estimated values ​​of the current system state and unknown thermal parameters; Starting with the estimated value, solve the optimal control problem of the entire trajectory from the current time to the end of the tape laying to obtain the optimal control sequence; apply the first control variable in the optimal control sequence, which includes the heating power command and the laying speed command; repeat the above estimation, solving and application steps until the laying is completed.

2. The method according to claim 1, characterized in that, The rolling time-domain state-parameter joint estimation adopts the rolling time-domain estimation method, which obtains the current state estimate and parameter estimate by solving an optimization problem. The optimization problem takes the rubber strip temperature measurement value, heating power command and laying speed data of a preset number of sampling times as input, and the system dynamic model as constraint, and minimizes the weighted sum of squares of initial state deviation, process noise, parameter prior deviation and measurement residual.

3. The method according to claim 1, characterized in that, The objective function of the full-trajectory optimal control problem includes terminal cost and operational cost; the terminal cost penalizes the deviation between the adhesive strip temperature and the target final pressure temperature at the end of the laying and the deviation between the glass temperature and the target final pressure temperature; the operational cost includes penalties for the adhesive strip temperature deviating from the preset bonding temperature range, heating energy consumption penalties, and laying speed change rate penalties.

4. The method according to claim 1, characterized in that, The full trajectory optimal control problem also satisfies the following constraints: upper and lower limits of heating power, upper and lower limits of laying speed, and laying speed change rate constraint.

5. The method according to claim 1, characterized in that, The full trajectory optimal control problem is discretized into a nonlinear programming problem using the direct collocation method and solved by sequential quadratic programming. The gradient of the objective function with respect to the control quantity is calculated through adjoint sensitivity analysis, which is based on the inverse integration of the adjoint equation of the system dynamic model.

6. The method according to claim 1, characterized in that, The system dynamic model is a two-body thermal contact model, which considers the adhesive strip and the glass as two nodes with finite heat capacity, and takes into account the spatial lag effect of the heating module. The spatial lag effect is expressed as follows: there is a delay in the moment when the heating power acts on the adhesive strip-glass contact point. The delay time is equal to the fixed arc length distance between the heating center of the heating module and the contact point divided by the current laying speed.

7. The method according to claim 6, characterized in that, The dual-body thermal contact model includes dynamic equations for the adhesive strip temperature and glass temperature. The rate of change of the adhesive strip temperature with respect to time is equal to the lag term of the heating power minus the convective heat dissipation term of the adhesive strip to the environment minus the contact heat transfer term between the adhesive strip and the glass, divided by the heat capacity of the adhesive strip. The rate of change of the glass temperature with respect to time is equal to the contact heat transfer term minus the heat dissipation term of the glass to the environment, divided by the heat capacity of the glass. The lag term of the heating power represents the heating power at the current moment acting on the adhesive strip at a future moment. The time difference between the future moment and the current moment is equal to the fixed arc length distance divided by the current laying speed.

8. The method according to claim 1, characterized in that, After the installation is completed, monitor the natural cooling temperature decay curve of the adhesive strip. When the temperature of the adhesive strip falls into the preset pressing temperature range, trigger the hot pressing unit to perform the pressing operation.

9. The method according to claim 8, characterized in that, The pressing temperature range is 80°C to 90°C; when the temperature of the adhesive strip is higher than 90°C, auxiliary cooling is activated; when the temperature of the adhesive strip is lower than 80°C, an alarm is issued and pressing is paused.

10. A fire-resistant glass processing system, characterized in that, include: An image acquisition unit is used to acquire the contour image of the glass to be processed; an environmental sensing unit is used to acquire the ambient temperature and humidity; an adhesive strip laying unit, including a robot, an infrared temperature sensor and a micro heating module, is used to perform adhesive strip laying and provide real-time feedback on the adhesive strip temperature; a hot pressing unit is used to hot press the laid adhesive strip; a conveying and positioning unit is used for glass feeding and position feedback; and a central controller is connected to each of the above units, wherein the central controller is configured to execute the control method of the fireproof glass processing system according to any one of claims 1 to 9.

Images