Temperature self-adaptive control method in glass annealing process

Abstract

The invention discloses a temperature self-adaptive control method in a glass annealing process, and relates to the related technical field of a glass annealing process, the method comprises the following steps: dividing an annealing kiln into N space areas, and respectively configuring independent temperature sensors in the N space areas; establishing a space temperature data set; acquiring information of currently annealed glass; establishing a weak classification model mapped with the N spatial regions; synchronizing the space temperature data set to the initialized corresponding weak classification model to generate a weak classification decision result; establishing a first important weight according to the N space regions; and establishing a global control strategy, and carrying out temperature self-adaptive control management through the global control strategy. The technical problems that in the prior art, annealing temperature control is not uniform and lagging, and glass types and various process requirements cannot be responded in real time, so that the glass annealing quality and efficiency are poor are solved, and the technical effects of improving the accuracy of glass annealing temperature control, the annealing quality and the production efficiency are achieved.

Description

Technical Field

[0001] The present application relates to the technical field related to glass annealing process, and in particular to a temperature adaptive control method in the glass annealing process. Background Art

[0002] Glass annealing is a crucial process in the glass production process. Glass is usually heated to a specific temperature and maintained for a certain period of time to eliminate the internal stress in the glass, thereby obtaining the required mechanical and optical properties. In the glass annealing process, temperature uniformity and precise control are crucial to the quality of the final glass. However, traditional glass annealing process control methods often rely on a single temperature sensor to detect the overall temperature changes in the annealing kiln, which cannot reflect the temperature distribution of different areas in the kiln in real time and accurately, resulting in hysteresis, volatility and unevenness of temperature control, thereby affecting the annealing effect of the glass. The current glass annealing temperature control is difficult to accurately formulate temperature control strategies according to different types of glass materials and different process requirements, and thus cannot cope with temperature fluctuations in the annealing kiln, etc., resulting in insufficient temperature control accuracy, thereby affecting the annealing quality and production efficiency of the product.

[0003] At present, relevant technologies have technical problems such as uneven annealing temperature control, hysteresis, and inability to respond in real time to diverse glass types and process requirements, resulting in poor glass annealing quality and efficiency. Summary of the invention

[0004] The present application solves the technical problems in the prior art of uneven annealing temperature control, hysteresis, and inability to respond in real time to diverse glass types and process requirements, resulting in poor glass annealing quality and efficiency by providing a temperature adaptive control method in the glass annealing process, thereby achieving the technical effect of improving the accuracy of glass annealing temperature control, annealing quality and production efficiency.

[0005] The present application provides a temperature adaptive control method in a glass annealing process, comprising: dividing an annealing furnace into N spatial regions, and configuring independent temperature sensors in the N spatial regions, respectively, wherein the N spatial regions are constructed based on a preset granularity division; using independent temperature sensors to detect the temperature of the annealing furnace and establish a spatial temperature data set; obtaining information of glass currently annealing, wherein the glass information includes glass type and process requirements, constructing a target annealing curve according to the glass information, and using the target annealing curve to allocate regional target temperatures of the N spatial regions; establishing a weak classification model mapped to the N spatial regions, and initializing a model of the corresponding weak classification model through the regional target temperature and the glass information; synchronizing the spatial temperature data set to the initialized corresponding weak classification model to generate a weak classification decision result; synchronizing the weak classification decision result to a strong classifier, and establishing a first important weight according to the N spatial regions, and establishing a second important weight according to the weak classification decision result; A global control strategy is established by the strong classifier according to the first important weight, the second important weight and the weak classification decision result, and temperature adaptive control management is performed through the global control strategy.

[0006] In a possible implementation, the temperature adaptive control method in the glass annealing process further performs the following processing: configuring a preset neighborhood capture space according to the preset granularity; establishing an input constraint, the input constraint is used to obtain input features, the input features include current region temperature features and neighborhood average temperature features generated based on the preset neighborhood capture space; configuring a loss function, as follows; ; in, Characterization Partition The loss function is Characterize the time range of consideration, is the time index, Characterization Partition In time The target temperature, The weight factor that characterizes the neighborhood temperature coupling term, Characterization Partition In time The regional temperature, Characterization Partition In time The average temperature of the neighborhood; and constructing a weak classification model according to the loss function and the input features.

[0007] In a possible implementation, the temperature adaptive control method during the glass annealing process also performs the following processing: establishing a neighborhood temperature difference trigger mechanism; if the neighborhood temperature difference triggers the neighborhood temperature difference trigger mechanism, generating a dynamic expansion instruction; dynamically adjusting the preset neighborhood capture space according to the dynamic expansion instruction, and using the dynamically adjusted preset neighborhood capture space to obtain the neighborhood average temperature characteristics.

[0008] In a possible implementation, the temperature adaptive control method in the glass annealing process further performs the following processing: temperature backtracking is performed according to the weak classification decision result, and the second important weight is calculated using the temperature backtracking result and the local weight calculation formula, as follows: ; in, Characterization Partition The second most important weight is Characterization Partition The real-time temperature error, To prevent the minimum value of 0, is the dynamic priority weight factor, Characterization Partition The rate of temperature change, Characterizes the neighborhood temperature difference weight factor, To adjust the parameters, For partition The zone temperature, For partition The average temperature of the neighborhood, is the historical feedback weight factor, Characterization Partition The historical control error variance is Characterizes the average control error variance of all partitions.

[0009] In a possible implementation, the temperature adaptive control method in the glass annealing process further performs the following processing: calling the weight fusion layer of the strong classifier to fuse the first important weight and the second important weight to establish a global weight; and calculating the global adjustment amount based on the global weight, as follows: ; in, Representation Moment The global control adjustment of Characterization Partition The global weight of Characterization Partition In time The weak classifier decision result; establish the global control strategy with the global adjustment amount.

[0010] In a possible implementation, the temperature adaptive control method in the glass annealing process further performs the following processing: establishing a neighborhood temperature coordination constraint, as follows: ; in, Characterizes the neighborhood temperature difference smoothing term, Characterizing a smoothing factor; optimizing the global adjustment amount by using the neighborhood temperature collaborative constraint, and establishing the global control strategy.

[0011] In a possible implementation, the temperature adaptive control method in the glass annealing process further performs the following processing: acquiring regional importance data of N spatial regions, and establishing the first importance weight according to the regional importance data.

[0012] In a possible implementation, the temperature adaptive control method in the glass annealing process also performs the following processing: establishing feedback monitoring, which is a verification monitoring strategy generated based on the global control strategy; performing actual response analysis based on the feedback monitoring and the global control strategy to establish a response optimization coefficient; and using the response optimization coefficient to perform strong classifier self-optimization management.

[0013] The temperature adaptive control method in the glass annealing process proposed in this application is to divide the annealing furnace into N spatial regions, and configure independent temperature sensors in each of the N spatial regions; establish a spatial temperature data set; obtain the current annealing glass information; establish a weak classification model mapped to the N spatial regions; synchronize the spatial temperature data set to the corresponding weak classification model after initialization to generate a weak classification decision result; establish the first important weight according to the N spatial regions; establish a global control strategy, and perform temperature adaptive control management through the global control strategy. The technical problems of uneven annealing temperature control, hysteresis, and inability to respond to the diverse requirements of glass types and processes in real time, resulting in poor glass annealing quality and efficiency, are solved in the prior art, and the technical effect of improving the accuracy of glass annealing temperature control, annealing quality and production efficiency is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0015] Figure 1A schematic flow chart of a method for adaptively controlling temperature during glass annealing provided in an embodiment of the present application.

[0016] Figure 2 A schematic diagram of a flow chart for configuring a preset neighborhood capture space in a temperature adaptive control method during a glass annealing process provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0018] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0019] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of the objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, product, or server comprising a series of steps is not necessarily limited to those steps clearly listed, but may include other steps that are not clearly listed or inherent to these processes, methods, products, or devices, and unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field to which this application belongs. The terms used herein are for the purpose of describing the embodiments of this application only.

[0020] The present application provides a temperature adaptive control method during glass annealing, such as Figure 1 As shown, the method includes: Step S100, dividing the annealing furnace into N spatial regions, and configuring independent temperature sensors in the N spatial regions respectively, wherein the N spatial regions are constructed based on a preset granularity division.

[0021] Preferably, the preset granularity division refers to dividing the internal space of the annealing furnace into several smaller areas (i.e., N spatial areas, N is a positive integer greater than 1, indicating the total number of spatial areas) according to certain standards or parameters. The division of these areas is based on different needs (such as the distribution of temperature changes, the characteristics of the glass processing process, the spatial structure in the kiln, etc.). Specifically, the annealing furnace is a larger device, and its internal temperature is not uniformly distributed. There may be different temperature gradients. The interior of the annealing furnace is divided into multiple spatial areas to more accurately monitor and adjust the temperature of different areas; wherein, the granularity refers to the fineness or meticulousness of the division. For example, the space of the annealing furnace may determine the size and number of areas according to the temperature distribution, process requirements or the structure in the kiln according to the change of heat flow or the different needs of glass materials. If the granularity If the setting is finer, more areas may be divided, and the temperature of each area is easier to control accurately. If the granularity is set coarser, fewer areas may be divided, and the control accuracy is relatively low. An independent temperature sensor is installed in each divided space area, and each sensor is responsible for monitoring the temperature of the area. Assume that the interior of the annealing furnace is divided into 5 areas, and the division of each area is determined according to the temperature change characteristics of the glass melt during the annealing process. For an area with a large temperature change, a finer division (i.e., a smaller area or a finer granularity) may be required, while for an area with a small temperature change, it can be divided into a larger area (i.e., a coarser granularity). The basis for the division may include: the heat flow model in the kiln, the thermal properties of the glass, the temperature control requirements of the annealing process, etc., so as to ensure that the temperature can be accurately adjusted according to the specific needs of different areas.

[0022] Step S200: Detect the temperature of the annealing furnace using an independent temperature sensor to establish a spatial temperature data set.

[0023] Preferably, an independent temperature sensor is used to detect the temperature of the annealing furnace, wherein the temperature sensor may be based on thermocouples, infrared detection or other detection technologies suitable for high-temperature environments, and can provide stable and high-precision temperature measurement results. The temperature data of each spatial area is recorded through the collection of independent sensors, which may include the temperature value at each time point, the temperature change trend over time, and the temperature difference between each area. The collected temperature data is organized into a spatial temperature data set, including the temperature value of each area and its time series data and the correlation data between different areas (such as the temperature gradient between areas). Through the analysis and modeling of these temperature data, the law of temperature distribution in the annealing furnace can be identified, and it can be judged whether the temperature of each area meets the target requirements, thereby improving the accuracy and adaptability of temperature control.

[0024] Step S300, obtaining glass information currently being annealed, the glass information including glass type and process requirements, constructing a target annealing curve according to the glass information, and using the target annealing curve to allocate regional target temperatures of N spatial regions.

[0025] Preferably, the information of the glass currently being annealed (glass type and process requirements) is obtained, and a suitable annealing temperature curve is determined according to the obtained glass type and process requirements, and target temperatures are allocated to different spatial areas in the annealing furnace according to the curve. Specifically, different types of glass (such as float glass, tempered glass, optical glass, etc.) have different temperature requirements during the annealing process. Each type of glass has specific thermal expansion coefficients, melting temperatures, softening temperatures and other characteristics, which affect the temperature requirements during the annealing process. For example, optical glass may require more precise temperature control to avoid changes in transmittance due to excessive temperature differences, and tempered glass may require higher temperatures and faster heating rates to ensure that a compressive stress layer is formed on its surface. Different production processes will also have different requirements for the annealing process. For example, large-scale production of glass may require a faster annealing rate, but the temperature control must be kept within a certain range. Some glass products require specific physical or chemical properties, such as higher heat resistance and impact resistance, which require the annealing curve to be specially designed.

[0026] Preferably, a target annealing curve suitable for the current glass is designed according to the acquired glass type and process requirements. The annealing curve generally includes information such as temperature change, heating rate, holding temperature and cooling rate during the glass heating process. Specifically, the construction of the target annealing curve takes into account the temperature range of the glass (the upper and lower limits of temperature control are set according to the softening temperature, solidification temperature and other characteristics of the glass), the heating rate (according to the process requirements, the heating speed is controlled to avoid thermal stress caused by excessive temperature changes on the glass), the holding temperature (the glass needs to be kept at a certain temperature for a period of time to eliminate internal stress, and this temperature is the key part of the annealing curve), and the cooling rate (the cooling rate during the annealing process is controlled to ensure the structural stability of the glass). The target annealing curve is the annealing requirement for a specific type of glass, reflecting the temperature control of the entire process from heating to cooling; finally, the target annealing curve is used to determine the temperature of the glass. The target annealing curve is used to allocate regional target temperatures to N spatial regions. The temperature in the annealing furnace is not evenly distributed, and there are different temperature gradients. Different target temperatures need to be allocated to different regions in the furnace according to the target annealing curve. Specifically, after the annealing furnace is divided into N spatial regions, the temperature requirements of each region may be different. The target temperature of each spatial region can be determined by numerical simulation, experimental data, etc. The specific temperature allocation needs to be adjusted according to the target annealing curve, that is, these regional temperature values ​​will be adjusted with different glass types and process requirements. For example, higher temperatures may be required in some areas to promote the softening of the glass, while lower temperatures may be required in other areas to avoid excessive temperature causing glass deformation or unnecessary stress, thereby achieving precise temperature control and ensuring the quality and performance of the glass.

[0027] Step S400: establishing a weak classification model mapped to N spatial regions, and initializing the model of the corresponding weak classification model through the regional target temperature and the glass information.

[0028] Preferably, a weak classification model in machine learning is used to control the temperature of different spatial areas of the annealing furnace, and the model is initialized by the relevant information of the glass and the target temperature of the area. Specifically, in each spatial area of ​​the annealing furnace (assuming that there are N areas), due to the different temperature targets of each area and different processing requirements of the glass, a weak classification model needs to be established separately for each area to perform temperature control management, wherein the weak classification model refers to a simple classifier (such as linear regression, decision tree, etc.), which can provide valuable prediction information for the final strong classifier, and is used to deal with the decision-making problem of regional temperature control during the annealing process. The weak classification model maps the temperature target, glass type, process requirements and other information corresponding to each spatial area to achieve auxiliary decision-making on the temperature control of each area; the weak classification model is initialized by converting the glass information ( The initial state of the model is set by taking information such as glass type, process requirements) and regional target temperature as input, and the model of each spatial area starts training on this basis. Specifically, the glass information includes glass type (such as tempered glass, float glass, etc.) and process requirements (such as annealing speed, holding temperature, etc.), which determine the initial setting of the temperature control strategy. The regional target temperature is the target temperature set for each area according to the target annealing curve, which reflects the temperature control requirements of each area during the annealing process. By passing this information to the weak classification model, the model learns how to make preliminary temperature adjustment decisions according to different situations based on the input regional target temperature and glass information, and then can more finely customize the temperature control strategy of each area according to the specific glass type and production process requirements, thereby realizing adaptive adjustment and control of the temperature in the glass annealing furnace and ensuring accuracy.

[0029] Further, step S400 also includes step S410, configuring a preset neighborhood capture space according to the preset granularity; step S420, establishing an input constraint, the input constraint is used to obtain input features, the input features include current regional temperature features and neighborhood average temperature features generated based on the preset neighborhood capture space; step S430, configuring a loss function, as follows; ; in, Characterization Partition The loss function is Characterize the time range of consideration, is the time index, Characterization Partition In time The target temperature, The weight factor that characterizes the neighborhood temperature coupling term, Characterization Partition In time The regional temperature, Characterization Partition In time The average temperature of the neighborhood; Step S440, constructing a weak classification model according to the loss function and the input features.

[0030] Preferably, configuring a preset neighborhood capture space according to a preset granularity means that, according to the preset granularity of the division (area size and number) of the specific needs of the glass annealing process, a neighborhood range is defined around each partition (such as the temperature data of several surrounding adjacent partitions) to capture the neighborhood temperature characteristics of the partition to take into account the coupling relationship between the temperatures between partitions (that is, the temperature change of a partition may be affected by the temperature of the surrounding area); the input constraints are used to obtain input features, which may include the temperature characteristics of the current partition (for example, the actual temperature at the current time point), the neighborhood average temperature generated by the neighborhood capture space (calculated from the temperature data of the preset neighborhood), and then a weak classification model is constructed according to the configured loss function and input features. Specifically, the current temperature of the partition and the neighborhood capture temperature are obtained from real-time sensors, and by continuously optimizing the loss function and using historical data or real-time data for supervised learning, the model can predict the adjustment direction and amplitude of the partition temperature (for example, heating, cooling or maintaining), and then learn the optimal temperature control adjustment strategy to improve the accuracy of the partition temperature control of the glass annealing furnace.

[0031] Further, such as Figure 2 As shown, step S410 also includes step S411, establishing a neighborhood temperature difference trigger mechanism; step S412, if the neighborhood temperature difference triggers the neighborhood temperature difference trigger mechanism, generating a dynamic expansion instruction; step S413, dynamically adjusting the preset neighborhood capture space according to the dynamic expansion instruction, and obtaining the neighborhood average temperature characteristics using the dynamically adjusted preset neighborhood capture space.

[0032] Preferably, the neighborhood temperature difference refers to the degree of difference between the temperature of a partition and the temperature of its neighborhood area, and a neighborhood temperature difference trigger mechanism is established, that is, a judgment condition is constructed, and when the neighborhood temperature difference exceeds a preset threshold, a subsequent action is triggered to identify the incoordination between the partition temperature and its neighborhood temperature; if the neighborhood temperature difference is too large, the neighborhood temperature difference trigger mechanism is triggered, and a dynamic expansion instruction is generated, that is, an operation instruction after detecting that the neighborhood temperature difference exceeds the threshold, which is used to expand or adjust the range of the neighborhood capture space to obtain more comprehensive neighborhood temperature characteristics, such as adjusting the parameters of the neighborhood capture range (such as capture radius, number of capture partitions), and the expanded neighborhood structure (such as a new neighborhood relationship diagram or range definition); dynamic adjustment refers to real-time updating of the range or structure of the capture space according to the dynamic expansion instruction, including expanding the neighborhood range and redefining the neighborhood relationship, so that the captured neighborhood data is more in line with the current temperature control requirements, which helps to improve the coordination of the neighborhood temperature difference, and finally generates a new neighborhood average temperature feature based on the adjusted neighborhood capture space, reflecting the adjusted neighborhood temperature control state, improving the flexibility of temperature control, and realizing temperature adaptive control.

[0033] Step S500: synchronizing the spatial temperature data set to the initialized corresponding weak classification model to generate a weak classification decision result.

[0034] Preferably, the spatial temperature data collected in real time are input into the initialized weak classification model of each area according to the spatial area, so that the weak classification model can generate a classification result that matches the current actual situation based on the latest spatial temperature data. Specifically, the weak classification model compares and processes these data with the target annealing curve, glass information, etc. according to the input actual spatial temperature data, glass information at the time of initialization, and target temperature, to determine whether the current temperature control meets the target requirements, analyze whether the temperature of the current area deviates from the target temperature, and the degree and trend of the deviation, and generate a weak classification decision result. The weak classification decision result is a preliminary judgment made by the model based on the input data and initialization conditions, which may include whether the temperature of the current area reaches the target value; whether the temperature needs to be increased or decreased, and the adjustment range; the priority of temperature control adjustment, for example, some areas may need to be adjusted more quickly; the decision result is expressed in the form of classification labels, such as "need to heat up", "need to cool down" or "maintain current temperature", thereby improving the accuracy and dynamic adaptability of temperature control, and ensuring that the temperature of different areas in the annealing furnace always meets the process requirements of glass annealing.

[0035] Step S600, synchronizing the weak classification decision result to the strong classifier, and establishing a first important weight according to the N spatial regions, and establishing a second important weight according to the weak classification decision result.

[0036] Preferably, the preliminary decision results obtained from each weak classification model are further processed by a strong classifier, and the weights of each region are adjusted on this basis to optimize the temperature control strategy. Specifically, the weak classification decision results are input into a strong classifier for analysis and processing, wherein the strong classifier is a more complex and more powerful decision module, which can integrate information from multiple weak classifiers to make a more accurate global temperature control strategy, that is, combining the weak classifier results from each region, further analyzing and integrating them, and making a global temperature control decision based on the importance, temperature control requirements and priorities of different regions, thereby determining the final temperature control adjustment strategy; the first important weight refers to assigning a weight value to each region based on the importance of each spatial region, reflecting the importance of different regions to the global temperature control during the temperature control process. Specifically, the temperature of different regions has different effects on the glass annealing process. The weight of each area is different. For example, the area close to the heating source may require more precise control, and temperature fluctuations may have a greater impact on the entire annealing process; while the area close to the cooling end may be relatively stable and have less impact on the global control; according to the temperature control accuracy requirements, importance and sensitivity of each area to temperature changes in each area during the glass annealing process; the second most important weight refers to the weight assigned to different areas based on the weak classification decision results. This weight is adjusted according to the output results of each weak classifier (such as "heating up", "cooling down" and "maintaining"). Different temperature control suggestions (heating up, cooling down or maintaining) reflect the urgency or priority of temperature control. If the weak classifier outputs "maintaining the current temperature" in a certain area, the weight of the area may be relatively low; thereby improving the accuracy and flexibility of temperature control and ensuring the production efficiency and quality of glass.

[0037] Furthermore, step S600 also includes performing temperature backtracking according to the weak classification decision result, and calculating the second important weight using the temperature backtracking result and the local weight calculation formula, as follows: ; in, Characterization Partition The second most important weight is Characterization Partition The real-time temperature error, To prevent the minimum value of 0, is the dynamic priority weight factor, Characterization Partition The rate of temperature change, Characterizes the neighborhood temperature difference weight factor, To adjust the parameters, For partition The zone temperature, For partition The average temperature of the neighborhood, is the historical feedback weight factor, Characterization Partition The historical control error variance is Characterizes the average control error variance of all partitions.

[0038] Preferably, temperature backtracking is to analyze the historical temperature data of the partition to understand the temperature change trend and control error of each partition in the past time period. By backtracking the historical data, the control status of the current partition temperature can be evaluated to determine whether it deviates from the target temperature or whether there is a historical control deviation accumulation. The weak classification decision result includes the current temperature control suggestion (such as heating, cooling or maintaining) and temperature control deviation of the partition, and the historical temperature data is the input of temperature backtracking. The output result and the local weight calculation formula are used to calculate the second important weight, where, is the real-time temperature error term, The smaller it is (i.e. the current temperature is close to the target temperature), the greater the weight is, indicating that when the current temperature control effect is good, the importance of this partition is higher; is the dynamic temperature change rate term, the temperature change rate When it is larger, the weight increases, reflecting the priority of dynamic changes, that is, if the temperature of the partition changes rapidly, a higher priority is required; is the neighborhood temperature difference term. The larger the neighborhood temperature difference (i.e. The larger the value is), the greater the weight is, emphasizing the coordination of the neighborhood temperature difference. That is, if the temperature difference between the partition and the neighborhood is too large, a higher weight needs to be given to adjust the temperature difference. is the historical control error term, and the historical control error variance The larger it is (i.e., the greater the fluctuation of the zone temperature control error), the greater the weight, which reflects the problem of historical accumulation. That is, if a zone has a large error in the past temperature control process, it needs to be given a higher weight for adjustment.

[0039] Furthermore, step S600 also includes step S610, obtaining regional importance data of N spatial regions, and establishing the first importance weight according to the regional importance data.

[0040] Preferably, the relative weight of each area is evaluated according to the importance of each spatial area in the annealing furnace to the glass annealing quality in the whole temperature control process, and the first important weight is generated accordingly. The regional importance is to measure the influence of a certain spatial area (such as a part of the annealing furnace) on key indicators such as glass quality and temperature control accuracy in the whole annealing process. The area with high importance is usually the area where the temperature fluctuation is sensitive to the process results, and the area with low importance may be the area with relatively stable temperature and low temperature control accuracy requirements. For example, the area close to the heating source may determine the overall efficiency of the temperature control; the area close to the glass discharge port may directly affect the final quality of the glass; Specifically, the importance of each area is quantified according to the annealing furnace design and glass process requirements, or based on sensor data, historical annealing quality data, etc., to obtain regional importance data of N spatial areas, and then the importance value of each area is normalized to ensure that the sum of the weights is 1, so as to allocate the proportion in the global strategy. If the real-time data in the temperature control process shows that the control demand of a certain area has changed (for example, the temperature fluctuation of a certain area has become larger), the weight can be recalculated according to the real-time importance data to finally obtain the first important weight. Through the first important weight, those areas that have a greater impact on the annealing quality can be given priority, thereby improving the temperature control efficiency.

[0041] Step S700: establishing a global control strategy with the strong classifier according to the first important weight, the second important weight and the weak classification decision result, and performing temperature adaptive control management through the global control strategy.

[0042] Preferably, the strong classifier calculates the global control strategy according to the first important weight, the second important weight and the weak classification decision result, that is, comprehensively considering the first weight (regional importance) and the second weight (priority adjusted according to the weak classifier decision), adjusting the temperature control strategy of each region, ensuring that the global temperature control is more accurate, and finally generating a global temperature control strategy. Specifically, the first important weight represents the importance of each spatial region in the annealing furnace (for example, the temperature control of some regions has a greater impact on the quality of the glass), and the second important weight represents the importance of the weak classification decision result. After the strong classifier comprehensively considers the temperature control decisions of each region (not only adjusting the control priority based on the importance of the region, but also dynamically adjusting the temperature control strategy according to the output of the weak classifier), a global temperature control strategy is generated. A global control strategy is developed to determine how to dynamically adjust the temperature of each area during the entire annealing process to ensure that the temperature of each area meets the annealing process requirements, thereby improving the quality and production efficiency of the glass; finally, the temperature of different areas in the annealing furnace is adjusted in real time according to the temperature data, glass characteristics and process requirements in the current annealing process. The annealing process is dynamic, and temperature changes may be affected by many factors (such as changes in glass, changes in the external environment, etc.). The global control strategy can monitor the temperature changes of each area in real time, adjust the heating and cooling processes in time, ensure the uniformity of temperature in the annealing furnace, reduce defects or quality problems of glass caused by uneven temperature, thereby improving production efficiency, reducing scrap rate and ensuring product quality.

[0043] Further, step S700 further includes step S710, calling the weight fusion layer of the strong classifier to fuse the first important weight and the second important weight to establish a global weight; step S720, calculating a global adjustment amount based on the global weight, as follows: ; in, Representation Moment The global control adjustment of Characterization Partition The global weight of Characterization Partition In time weak classifier decision result; step S730, establishing the global control strategy with the global adjustment amount.

[0044] Preferably, the weight fusion layer of the strong classifier is called to fuse the first important weight and the second important weight. For example, the two types of weights (first weight and second weight) are comprehensively processed using weighted average to generate the final global weight, which represents the influence of the partition on the global control strategy at the current moment. A large weight indicates that the adjustment demand of the partition is more urgent, or its temperature is more important to the overall control. Then the global weight is used according to the formula: Calculate the global adjustment amount, that is, the decision result of each partition is weighted according to the global weight, which represents the contribution of the partition to the global adjustment amount. The weighted sum is normalized (divided by the total weight) to ensure that the proportion of the global adjustment amount is reasonable. The global adjustment amount represents the comprehensive adjustment of the temperature control of the entire annealing furnace to ensure that the direction and amplitude of the global temperature control adjustment are reasonable. Finally, establish a global control strategy based on the global adjustment amount, including generating global control instructions for adjusting the heating, cooling and other equipment of the entire annealing furnace. The global strategy will be assigned to each partition, and the adjustment amplitude will be further refined according to the specific needs of each partition. It can more intelligently integrate the partition temperature control information and provide a high-precision temperature adaptive management solution, thereby improving the quality and production efficiency of glass annealing.

[0045] Furthermore, step S730 also includes step S731, establishing a neighborhood temperature coordination constraint, as follows: ; in, Characterizes the neighborhood temperature difference smoothing term, Characterizing the smoothing factor; Step S732, optimizing the global adjustment amount by using the neighborhood temperature collaborative constraint, and establishing the global control strategy.

[0046] Preferably, the neighborhood temperature synergy constraint is used to optimize the global adjustment amount, so that the temperature control strategy not only meets the overall goal, but also takes into account the coordination of the temperature of adjacent areas to avoid the problem of excessive local temperature difference. The synergy constraint balances the temperature difference between each partition and its neighbors by introducing a neighborhood temperature difference smoothing term. If the temperature of some partitions is significantly different from that of their neighbors, it may cause local stress concentration and affect the uniformity and quality of glass annealing. Among them, the smoothing factor It is used to control the influence of the neighborhood temperature difference smoothing term on the global adjustment amount. It refers to the absolute difference between the temperature of partition i and the average temperature of its neighborhood, representing the neighborhood temperature difference. Characterize the average neighborhood temperature difference of the entire annealing kiln; use the neighborhood temperature collaborative constraint to optimize the global adjustment amount. Specifically, based on the weight fusion layer and the partition decision results, calculate the initial global adjustment amount, calculate the difference between the current temperature and the neighborhood average temperature of all partitions and calculate the average value. According to the formula, the neighborhood temperature difference smoothing term is multiplied by the smoothing factor , subtract this value from the initial global adjustment to obtain the optimized global adjustment. Finally, specific temperature control instructions are generated based on the optimized global adjustment, which may include adjusting the power of the heating / cooling equipment, reallocating temperature control resources, etc. This is used as a global control strategy to dynamically adjust the temperature control amplitude of each partition to ensure the temperature coordination between neighborhoods, thereby significantly improving the temperature control accuracy and quality stability of glass annealing.

[0047] Furthermore, step S732 also includes step A, establishing feedback monitoring, wherein the feedback monitoring is a verification monitoring strategy generated based on the global control strategy; step B, performing actual response analysis using the feedback monitoring and the global control strategy to establish a response optimization coefficient; and step C, using the response optimization coefficient to perform self-optimization management of the strong classifier.

[0048] Preferably, an optimization mechanism based on closed-loop feedback is established to improve the adaptive ability of the strong classifier and the accuracy of the global control strategy. Specifically, feedback monitoring is a verification monitoring strategy generated based on the global control strategy, that is, feedback monitoring is a real-time monitoring mechanism based on the execution result of the global control strategy, which is used to evaluate whether the actual execution effect is consistent with the expected goal. By monitoring and analyzing the actual temperature control data (such as partition temperature changes, neighborhood temperature balance, global adjustment amount, etc.), the effectiveness of the global control strategy is verified, the deviation in the control process is identified, and whether the global control strategy needs to be adjusted is evaluated. For example, whether the deviation between the partition temperature and the target temperature is less than the threshold, whether the global neighborhood temperature difference is reduced to a specified range, and whether the global adjustment amount significantly affects the temperature of certain areas; feedback monitoring and the global control strategy are then used to perform actual response analysis, including analyzing the execution effect of the global control strategy and feedback monitoring. The relationship between the results is used to judge whether the execution effect of the strategy meets the expectations, that is, to calculate and quantify the effectiveness of the control strategy, discover the deviations or problems in the control, and thus obtain the response optimization coefficient, which is used to evaluate the control strategy effect of the strong classifier and reflect the accuracy and effectiveness of the global control strategy, mainly including the global deviation coefficient, which measures the average deviation between the target temperature and the actual temperature; the neighborhood smoothing coefficient, which measures the amplitude of change of the neighborhood temperature difference; the dynamic adjustment coefficient, which evaluates the impact of the global adjustment on the actual temperature; finally, the response optimization coefficient is used to perform self-optimization management of the strong classifier, that is, the response optimization coefficient generated by the actual response analysis is input into the strong classifier. The strong classifier adjusts its own parameters through real-time feedback, so that it can adapt to the complex dynamic environment of the annealing kiln, and gradually improve the accuracy of the control strategy, thereby continuously improving the accuracy and adaptability of control decisions, thereby achieving more accurate and efficient annealing kiln temperature control management.

[0049] The above specific implementation manner does not constitute a limitation to the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A temperature adaptive control method in a glass annealing process, characterized in that: The method comprises: The annealing furnace is divided into N spatial regions, and independent temperature sensors are respectively configured in the N spatial regions, wherein the N spatial regions are constructed based on a preset granularity division; Use independent temperature sensors to detect the temperature of the annealing kiln and establish a spatial temperature data set; Acquire glass information currently being annealed, the glass information including glass type and process requirements, construct a target annealing curve according to the glass information, and use the target annealing curve to allocate regional target temperatures of N spatial regions; Establishing a weak classification model mapped to N spatial regions, and initializing the model of the corresponding weak classification model according to the regional target temperature and the glass information; Synchronizing the spatial temperature data set to the corresponding weak classification model after initialization to generate a weak classification decision result; Synchronize the weak classification decision result to the strong classifier, and establish a first important weight according to the N spatial regions, and establish a second important weight according to the weak classification decision result; A global control strategy is established by the strong classifier according to the first important weight, the second important weight and the weak classification decision result, and temperature adaptive control management is performed through the global control strategy.

2. The temperature adaptive control method in the glass annealing process according to claim 1, characterized in that: The step of establishing a weak classification model mapped to N spatial regions and initializing the corresponding weak classification model by using the regional target temperature and the glass information includes: Configuring a preset neighborhood capture space according to the preset granularity; Establishing input constraints, where the input constraints are used to obtain input features, where the input features include current regional temperature features and neighborhood average temperature features generated based on a preset neighborhood capture space; Configure the loss function as follows; ； in, Characterization Partition The loss function is Characterize the time range of consideration, is the time index, Characterization Partition In time The target temperature, The weight factor that characterizes the neighborhood temperature coupling term, Characterization Partition In time The regional temperature, Characterization Partition In time The average temperature of the neighborhood; A weak classification model is constructed according to the loss function and the input features.

3. The temperature adaptive control method in the glass annealing process according to claim 2, characterized in that: The configuring a preset neighborhood capture space according to the preset granularity includes: Establish a neighborhood temperature difference trigger mechanism; If the neighborhood temperature difference triggers the neighborhood temperature difference triggering mechanism, a dynamic expansion instruction is generated; The preset neighborhood capture space is dynamically adjusted according to the dynamic expansion instruction, and the neighborhood average temperature feature is obtained by using the dynamically adjusted preset neighborhood capture space.

4. The temperature adaptive control method in the glass annealing process according to claim 1, characterized in that: The establishing of the second important weight according to the weak classification decision result comprises: Temperature backtracking is performed according to the weak classification decision result, and the second most important weight is calculated using the temperature backtracking result and the local weight calculation formula, as follows: ； in, Characterization Partition The second most important weight is Characterization Partition The real-time temperature error, To prevent the minimum value of 0, is the dynamic priority weight factor, Characterization Partition The rate of temperature change, Characterizes the neighborhood temperature difference weight factor, To adjust the parameters, For partition The zone temperature, For partition The average temperature of the neighborhood, is the historical feedback weight factor, Characterization Partition The historical control error variance is Characterizes the average control error variance of all partitions.

5. The temperature adaptive control method in the glass annealing process according to claim 1, characterized in that: The establishing of a global control strategy using the strong classifier according to the first important weight, the second important weight and the weak classification decision result includes: Calling the weight fusion layer of the strong classifier to fuse the first important weight and the second important weight to establish a global weight; The global adjustment amount is calculated based on the global weight as follows: ； in, Representation Moment The global control adjustment of Characterization Partition The global weight of Characterization Partition In time The decision result of the weak classifier; The global control strategy is established using the global adjustment amount.

6. The temperature adaptive control method in the glass annealing process according to claim 5, characterized in that: The establishing of the global control strategy using the global adjustment amount comprises: Establish the neighborhood temperature coordination constraint as follows: ； in, Characterizes the neighborhood temperature difference smoothing term, Characterize the smoothing factor; After optimizing the global adjustment amount by using the neighborhood temperature cooperative constraint, the global control strategy is established.

7. The temperature adaptive control method in the glass annealing process according to claim 1, characterized in that: The establishing of the first important weight according to the N spatial regions comprises: The regional importance data of N spatial regions are obtained, and the first importance weight is established according to the regional importance data.

8. The temperature adaptive control method in the glass annealing process according to claim 1, characterized in that: The method further comprises: Establishing feedback monitoring, wherein the feedback monitoring is a verification monitoring strategy generated based on the global control strategy; Performing actual response analysis using the feedback monitoring and the global control strategy to establish a response optimization coefficient; The response optimization coefficient is used to perform self-optimization management of the strong classifier.

Images