Liquid crystal glass heating control method based on up-down symmetrical temperature prediction

Abstract

The invention relates to the technical field of liquid crystal glass manufacturing temperature control, and discloses a liquid crystal glass heating control method based on longitudinal symmetry temperature prediction. The method comprises the following steps: establishing an in-furnace three-dimensional temperature field real-time monitoring system, synchronously capturing temperature distribution image sequences of upper and lower surfaces of glass, and finishing time alignment; gridding segmentation is carried out on the image sequence, the instantaneous symmetry degree of the upper surface temperature and the lower surface temperature in each grid unit is calculated, and according to the instantaneous symmetry degree, a temperature conduction abnormal area is identified and a primary mark graph is generated; and performing dynamic threshold correction on the primary mark graph by fusing historical cycle data and the current conduction rate to generate a more accurate secondary abnormal region mark graph. And positioning a heat source interference center point based on the second-level marker graph, calculating a heat source compensation vector by combining heater power distribution, and adjusting working parameters of the heater according to the heat source compensation vector. According to the invention, fine monitoring and self-adaptive regulation and control of the symmetry of the temperature field are realized, and the heating uniformity and the product quality of the liquid crystal glass are improved.

Description

Technical Field

[0001] This invention relates to the field of temperature control technology in liquid crystal glass manufacturing, specifically a heating control method for liquid crystal glass based on top-bottom symmetrical temperature prediction. Background Technology

[0002] In the production of LCD glass substrates, the heating process is crucial. The glass needs to undergo a uniform heating process within the furnace; any temperature unevenness between the upper and lower surfaces or between different areas can lead to internal stress, warping, or microscopic defects, severely impacting product quality. Currently, monitoring the temperature field within the furnace generally relies on a limited number of thermocouples or single-sided infrared thermography. The temperature information obtained by these methods is discrete or localized, making it difficult to comprehensively and intuitively reflect the overall temperature distribution across the entire glass substrate, especially the upper and lower surfaces.

[0003] Most existing control strategies are based on feedback of the absolute temperature values ​​of the measuring points, or use preset, fixed temperature curves and thresholds for regulation. This control method is slow to respond to dynamic and localized temperature anomalies caused by factors such as heat source fluctuations, subtle changes in glass loading position, or equipment aging during the heating process. Due to the lack of real-time quantitative assessment of the temperature symmetry between the upper and lower surfaces, the system can only react when the temperature difference accumulates to a certain level and is captured by the local measuring point, by which time thermal damage to the glass may have already occurred. In addition, anomaly judgment rules based on fixed thresholds cannot adapt to the natural changes in temperature conduction characteristics under different heating stages or different process formulations, which easily leads to misjudgments or missed judgments, thus restricting further improvement in control accuracy. Summary of the Invention

[0004] The purpose of this invention is to provide a liquid crystal glass heating control method based on top and bottom symmetrical temperature prediction, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for controlling the heating of liquid crystal glass based on top-bottom symmetrical temperature prediction, the method comprising: Establish a real-time monitoring system for the three-dimensional temperature field inside the liquid crystal glass heating furnace, and simultaneously capture temperature distribution image sequences of the upper and lower surface regions of the glass; A time-aligned model for the temperature distribution image sequence is constructed to map the upper and lower surface temperature distribution images to a unified time coordinate system. The time-aligned temperature distribution image sequence is divided into regional grids, and the instantaneous symmetry between the upper and lower surface temperature values ​​in each grid cell is calculated. Identify anomalous temperature conduction regions based on the spatial distribution pattern of instantaneous symmetry, and generate a primary anomalous region marker map; By integrating historical heating cycle data with the temperature conduction rate of the current heating stage, dynamic threshold correction is applied to the primary anomaly region marker map to generate a secondary anomaly region marker map. The heat source interference center point is located based on the secondary anomaly region marking map, and the heat source compensation vector is calculated by combining the heater power distribution map. The operating parameters of the corresponding heater are adjusted according to the heat source compensation vector, heating control operations are performed, and temperature response data is recorded.

[0006] Preferably, establishing a real-time monitoring system for the three-dimensional temperature field inside the liquid crystal glass heating furnace includes: Infrared thermal imager arrays are deployed at the top and bottom of the heating furnace, respectively, to synchronously acquire temperature distribution images of the upper and lower surfaces of the glass at a fixed sampling frequency; Add timestamps, heater power status, and ambient pressure parameters to each temperature distribution image to form a multimodal data frame; Time reference calibration is performed on multimodal data frames to eliminate timing errors caused by acquisition time delays from different thermal imagers; The calibrated multimodal data frames are stored in a distributed cache database to construct a real-time temperature monitoring data stream.

[0007] Preferably, the time alignment model for constructing the temperature distribution image sequence includes: Extract the upper surface temperature distribution image sequence and the lower surface temperature distribution image sequence within a continuous time window from the real-time temperature monitoring data stream; A feature point tracking algorithm is used to identify corresponding thermal feature points in the upper and lower surface temperature distribution images; A time warp function is constructed based on the motion trajectory of thermal feature points to map the lower surface temperature distribution image sequence to the time axis of the upper surface temperature distribution image sequence; Spatial resampling is performed on the image sequence after time axis mapping to ensure that the upper surface temperature distribution image and the lower surface temperature distribution image at each time point have the same spatial resolution.

[0008] Preferably, performing region gridding segmentation on the time-aligned temperature distribution image sequence includes: The single-frame temperature distribution image is divided into uniform square grid cells, and the grid size is dynamically adjusted according to the glass thermal expansion coefficient. Within each grid cell, the variance of the temperature distribution on the upper surface and the variance of the temperature distribution on the lower surface are statistically analyzed. Calculate the absolute difference between the average temperature value of the upper surface and the average temperature value of the lower surface within the grid cell, and divide this absolute difference by the spatial area of ​​the grid cell to obtain the instantaneous symmetry. The instantaneous symmetry of all grid cells is combined to form an instantaneous symmetry distribution matrix.

[0009] Preferably, identifying temperature conduction anomaly regions based on the spatial distribution pattern of instantaneous symmetry includes: Perform morphological opening operations on the instantaneous symmetry distribution matrix to eliminate isolated noise points; A region growing algorithm is used to connect adjacent high instantaneous symmetry grid cells to form the outline of potential anomaly regions; Calculate the geometric center coordinates and area of ​​each potential anomaly region's contour; When the area of ​​a region exceeds the area threshold and the geometric center is located within the influence range of the heater, the outline of the potential abnormal region is marked as a primary abnormal region.

[0010] Preferably, dynamic threshold correction of the primary anomaly region marker map includes: Search the historical heating cycle database for records of temperature conduction anomalies during similar heating phases; Extract the duration and temperature deviation of abnormal areas from historical records as reference features; Compare the similarity between the anomalous regions in the current primary anomaly region labeling map and historical reference features; When the similarity is lower than the similarity threshold, the area threshold is increased and abnormal region markers are regenerated; When the similarity is higher than the similarity threshold, the boundary of the abnormal region is finely adjusted based on the temperature conduction rate of the current heating stage to form a secondary abnormal region marker map.

[0011] Preferably, the location of the heat source interference center point includes: Extract the centroid coordinates of each anomaly region from the secondary anomaly region marker map; A circular search area is constructed with the centroid coordinates as the center and the heater's influence radius as the search radius; Analyze the gradient direction of the heater power distribution map within the circular search area; Heaters whose gradient direction points to the centroid of the abnormal region are identified as candidate interference sources; Calculate the distance weight and power weight from each candidate interference source to the centroid of the abnormal region, and select the candidate interference source with the largest comprehensive weight as the center point of the heat source interference.

[0012] Preferably, calculating the heat source compensation vector includes: Obtain the current power setting and temperature setting at the center point of the heat source interference; Measure the actual temperature values ​​of preset monitoring points around the center point of the heat source interference; Calculate the deviation vector between the actual temperature value and the temperature setpoint; The power compensation direction is determined by the projection of the deviation vector into the heater power space. The power compensation amplitude is determined by combining the dynamic response characteristic curve of the heater, and the power compensation direction and power compensation amplitude are combined into a heat source compensation vector.

[0013] Preferably, determining the power compensation amplitude based on the dynamic response characteristic curve of the heater, and synthesizing the power compensation direction and power compensation amplitude into a heat source compensation vector includes: The dynamic response characteristic curve of the target heater is extracted from the pre-stored heater characteristic library. The dynamic response characteristic curve is obtained by fitting historical experimental data and characterizes the functional relationship between power change and temperature change. Read the magnitude of the deviation vector, which represents the overall magnitude of the deviation between the actual temperature value and the temperature setpoint; Based on the dynamic response characteristic curve, the local slope corresponding to the current power setting value is queried and used as the power-temperature conversion coefficient. Divide the magnitude of the deviation vector by the power-temperature conversion coefficient to obtain the basic power compensation amplitude; An ambient temperature compensation factor is introduced. This factor is calculated by monitoring the difference between the ambient temperature inside the heating furnace and the standard operating conditions. The base power compensation range is scaled and corrected to obtain the final power compensation range. The power compensation direction is vector normalized to obtain a unit direction vector; The heat source compensation vector is generated by performing a scalar multiplication operation between the unit direction vector and the final power compensation amplitude.

[0014] Preferably, performing heating control operations and recording temperature response data includes: The heat source compensation vector is decomposed into a power adjustment command sequence for each heater; The heater power is adjusted step by step according to the time sequence of the power adjustment command sequence; After each power adjustment, a sequence of temperature distribution images is acquired, and the instantaneous symmetry distribution matrix after adjustment is calculated. Compare the differences in the instantaneous symmetry distribution matrix before and after adjustment to generate control effect evaluation indicators; The power adjustment command sequence, temperature distribution image sequence, and control effect evaluation index are associated and stored in the historical heating cycle database.

[0015] Compared with the prior art, the beneficial effects of the present invention are: By segmenting time-aligned temperature distribution images of the upper and lower surfaces into regional grids and calculating the instantaneous symmetry of each grid cell, a microscopic and quantitative perception of temperature field uniformity is achieved. This elevates traditional point or line temperature measurement to a full-field scanning analysis level, enabling the precise detection of minute temperature conduction asymmetries occurring at any spatial location. This symmetry-based anomaly identification mechanism is more sensitive than methods relying on absolute temperature thresholds, accurately pinpointing the location and extent of anomalies in their early stages—before the absolute temperature value deviates significantly but symmetry is disrupted—thus creating conditions for timely intervention.

[0016] By integrating historical heating cycle data with the current real-time temperature conduction rate to dynamically correct the initial anomaly markers, the criteria for judging anomaly areas possess self-learning and adaptive capabilities. Historical data provides a reference benchmark for normal temperature conduction patterns during the process, while the real-time conduction rate reflects the dynamic characteristics of the current operating conditions. The combination of these two factors distinguishes between symmetry fluctuations caused by normal process adjustments and genuine equipment failures or process anomalies, reducing the false alarm rate. This dynamic correction mechanism avoids the high risk of missed or false alarms that may arise from fixed thresholds at different stages of heating, ensuring that the reliability of anomaly detection remains stable under various production conditions and improving the level of control intelligence. Attached Figure Description

[0017] Figure 1 This is a schematic diagram illustrating the working principle of the liquid crystal glass heating control method based on top-bottom symmetrical temperature prediction as described in this invention. Figure 2 A flowchart for establishing a real-time monitoring system; Figure 3 A flowchart for region gridding and symmetry calculation; Figure 4 Thermographic diagram showing the instantaneous symmetry distribution of temperature on the upper and lower surfaces of the LCD glass heating furnace; Figure 5 A comparison chart of the comprehensive weights of candidate interference sources in the heating control of liquid crystal glass. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1This invention provides a method for heating and controlling liquid crystal glass based on top-bottom symmetry temperature prediction. The method includes: establishing a real-time monitoring system for the three-dimensional temperature field inside the liquid crystal glass heating furnace, and simultaneously capturing temperature distribution image sequences of the upper and lower surface regions of the glass; constructing a time alignment model for the temperature distribution image sequences, mapping the upper and lower surface temperature distribution images to a unified time coordinate system; performing regional grid segmentation on the time-aligned temperature distribution image sequences, and calculating the instantaneous symmetry between the upper and lower surface temperature values ​​within each grid cell; identifying abnormal temperature conduction regions based on the spatial distribution pattern of the instantaneous symmetry, and generating a primary abnormal region marker map; fusing historical heating cycle data with the temperature conduction rate of the current heating stage, and dynamically thresholding the primary abnormal region marker map to generate a secondary abnormal region marker map; locating the heat source interference center point based on the secondary abnormal region marker map, and calculating the heat source compensation vector by combining it with the heater power distribution map; adjusting the operating parameters of the corresponding heater according to the heat source compensation vector, executing heating control operations, and recording temperature response data.

[0020] Example 1: See Figure 2 In practical implementation, establishing a real-time monitoring system for the three-dimensional temperature field within the LCD glass heating furnace involves deploying infrared thermal imager arrays at the top and bottom of the furnace. These arrays synchronously acquire temperature distribution images of the upper and lower surfaces of the glass at a fixed sampling frequency of 10 frames per second, set according to the heating process requirements to ensure continuous capture of temperature changes. Each temperature distribution image is appended with a timestamp, heater power status, and environmental pressure parameters, forming a multimodal data frame. The timestamp is generated by a high-precision clock source, the heater power status is read in real-time from the controller, and the environmental pressure parameters are obtained from an in-furnace pressure sensor. The multimodal data frames undergo time reference calibration, using a network time protocol to synchronize the internal clocks of each infrared thermal imager, eliminating timing errors caused by acquisition delays between different thermal imagers. These timing errors are typically controlled within milliseconds. The calibrated multimodal data frames are stored in a distributed cache database, which employs a memory storage architecture for high-speed read and write operations, constructing a real-time temperature monitoring data stream. This real-time temperature monitoring data stream is pushed to subsequent processing modules via a message queue. In some embodiments, the deployment location of the infrared thermal imager array is optimized to cover the entire surface area of ​​the glass, avoiding blind spots in monitoring.

[0021] The construction of a time-aligned model for the temperature distribution image sequence involves extracting upper and lower surface temperature distribution image sequences within a continuous time window from the real-time temperature monitoring data stream. The length of the continuous time window is determined based on the heating cycle, typically one-tenth of the complete heating process. A feature point tracking algorithm is used to identify corresponding thermal feature points in the upper and lower surface temperature distribution images. This algorithm is based on optical flow, and the center points of regions with significant temperature gradients are selected as thermal feature points. A time warp function is constructed based on the motion trajectories of the thermal feature points. This time warp function uses a polynomial fitting method to map the lower surface temperature distribution image sequence to the time axis of the upper surface temperature distribution image sequence. The mapping process uses an interpolation algorithm to ensure temporal continuity. The time warp function models the motion trajectory of the thermal feature points using a polynomial fitting method. Polynomial fitting, based on the least squares principle, fits discrete location points into a continuous function curve to describe the motion law of the thermal feature points changing over time. Mapping the lower surface temperature distribution image sequence to the time axis of the upper surface temperature distribution image sequence involves calculating the time offset corresponding to each lower surface image using the time warp function and adjusting its timestamp to synchronize it with the upper surface image sequence. Spatial resampling is performed on the time-mapped image sequence using bilinear interpolation to ensure that the upper and lower surface temperature distribution images at each time point have the same spatial resolution, uniformly set to 512 pixels × 512 pixels. It is understood that the accuracy of the time warp function directly affects the alignment effect; therefore, the polynomial order must be selected based on the motion complexity. In some embodiments, the feature point tracking algorithm can be replaced with a template matching method to improve robustness in low-contrast regions. Optionally, an anti-aliasing filter can be introduced during the spatial resampling process to reduce image distortion.

[0022] Example 2: See Figure 3In specific implementation, the time-aligned temperature distribution image sequence is divided into uniform square grid cells by dividing a single frame of the temperature distribution image. The grid size is dynamically adjusted according to the glass's thermal expansion coefficient. The adjustment strategy is to establish an inverse relationship between the grid size and the glass's thermal expansion coefficient to prevent excessive differences in the physical scale of the grid cells due to glass thermal deformation. Within each grid cell, the distribution variances of the upper and lower surface temperature values ​​are statistically analyzed. These variances are used to initially assess the temperature uniformity within the grid cell. The absolute difference between the average upper and lower surface temperature values ​​within the grid cell is calculated, and this absolute difference is divided by the spatial area of ​​the grid cell to obtain the instantaneous symmetry. The instantaneous symmetry characterizes the magnitude of the temperature difference between the upper and lower surfaces of the glass per unit area. The instantaneous symmetry of all grid cells is combined to form an instantaneous symmetry distribution matrix, which serves as the basic data structure for subsequent anomaly region identification. In some embodiments, the shape of the grid cells can be adapted to rectangles to match the geometric features of non-square heating regions.

[0023] Identifying anomalous temperature conduction regions based on the spatial distribution pattern of instantaneous symmetry involves performing a morphological opening operation on the instantaneous symmetry distribution matrix. This morphological opening operation uses a rectangular kernel with a 3×3 pixel structuring element to eliminate isolated noise points within the matrix. A region growing algorithm is then used to connect adjacent high instantaneous symmetry mesh cells. The growth threshold of this algorithm is automatically set based on the overall statistical characteristics of the instantaneous symmetry distribution matrix, forming the contours of potential anomalous regions. The geometric center coordinates and area of ​​each potential anomalous region contour are calculated. The geometric center coordinates are obtained by calculating the centroid of the contour polygon, and the area is obtained by multiplying the total number of mesh cells enclosed by the contour by the area of ​​a single mesh cell. When the area exceeds the area threshold and the geometric center is within the heater's influence range, the area threshold is set to one percent of the furnace bottom plate area based on historical normal production data, and the potential anomalous region contour is marked as a primary anomalous region. It is understood that the effectiveness of the morphological opening operation depends on the size and shape of the structuring element. In some embodiments, the region growing algorithm can be replaced by a watershed algorithm based on edge detection to handle anomalous regions with complex shapes. Optionally, the area threshold setting can be dynamically fine-tuned in conjunction with the current heating stage, for example, using a larger threshold during the heating stage.

[0024] Instantaneous symmetry The calculation method is as follows: in: This represents the instantaneous symmetry of the grid cell located in the i-th row and j-th column. This represents the average temperature value of all upper surfaces within the grid cell. This represents the average temperature value of all lower surfaces within the grid cell. This represents the spatial area of ​​the grid cell in the i-th row and j-th column.

[0025] See Figure 4 This image is the core visualization result of time-aligned temperature distribution image meshing segmentation. It presents the instantaneous symmetry distribution of the 10×10 grid cells on the glass surface in the form of a heatmap: the horizontal axis represents the number of grid columns, the vertical axis represents the number of grid rows, and the color brightness is positively correlated with the instantaneous symmetry. This image serves as a preliminary data carrier for identifying primary anomalous regions: it transforms the abstract symmetry distribution matrix into an intuitive thermal field distribution, and also provides a visual reference for subsequent morphological opening operations for noise reduction and region growing algorithms for connecting anomalous grids. It demonstrates the technical advantage of this heating control method, upgrading from point temperature measurement to full-field scanning analysis, and can quickly locate local areas with abnormal temperature differences.

[0026] Example 3: In specific implementation, dynamic threshold correction of the primary anomaly region marker map involves querying temperature conduction anomaly records of similar heating stages in the historical heating cycle database. The determination of similar heating stages is based on the Euclidean distance between the currently set temperature curve and the historical temperature curve. The duration and temperature deviation amplitude of the anomaly region in the historical records are extracted as reference features. The duration refers to the number of time frames spanned from the appearance to the disappearance of the anomaly region, and the temperature deviation amplitude refers to the difference between the maximum instantaneous symmetry value within the anomaly region and the background average value. The similarity between the anomaly region in the current primary anomaly region marker map and the historical reference features is compared. The similarity calculation uses a cosine similarity method based on feature vectors. When the similarity is lower than the similarity threshold, the area threshold is increased and the anomaly region marker is regenerated. The similarity threshold is set to 0.7, and the area threshold is adjusted by 20% of the initial value. When the similarity is higher than the similarity threshold, the boundary of the anomaly region is fine-tuned in conjunction with the temperature conduction rate of the current heating stage. The temperature conduction rate of the current heating stage is obtained by calculating the rate of change of the area of ​​the anomaly region per unit time, forming a secondary anomaly region marker map. In some embodiments, environmental pressure parameters can be added as a filter item to the query conditions of the historical heating cycle database. Optionally, similarity calculation can use a weighted Euclidean distance method, assigning different weights to the duration and temperature deviation magnitude.

[0027] Locating the heat source interference center point involves extracting the centroid coordinates of each anomalous region from the secondary anomalous region marker map. The centroid coordinates are obtained by calculating the average of the pixel coordinates of the anomalous region. A circular search area is constructed with the centroid coordinates as the center and the heater's influence radius as the search radius. The heater's influence radius is a characteristic parameter of the heater, determined experimentally based on the thermal field distribution. Within the circular search area, the gradient direction of the heater power distribution map is analyzed. The gradient direction is obtained by calculating the convolution result of the power distribution map using the Sobel operator. Heaters whose gradient direction points towards the centroid of the anomalous region are identified as candidate interference sources. The directional judgment criterion is that the angle between the gradient direction vector and the vector pointing from the centroid to the heater coordinates is less than 45 degrees. The distance weight and power weight from each candidate interference source to the centroid of the anomalous region are calculated. The distance weight is inversely proportional to the distance, and the power weight is directly proportional to the current power setting of the heater. The candidate interference source with the largest comprehensive weight is selected as the heat source interference center point. It is understood that the radius of the circular search area needs to cover the heaters that may cause influence. In some embodiments, the gradient direction analysis can be combined with historical trends in power changes for auxiliary judgment. Optionally, the calculation of the comprehensive weight can be corrected by incorporating the heater aging coefficient.

[0028] Overall weight The calculation formula is as follows: in: This represents the overall weight of the candidate interference sources. This indicates the current power setting value of the candidate interference source. This indicates the maximum power setting value of the heater in the system. This represents the Euclidean distance from the candidate interference source to the centroid of the anomaly region. This represents the gradient magnitude of the power distribution map at the location of the candidate interference source. This represents the maximum gradient magnitude within the circular search region. and It is the weighting coefficient for balancing the power factor and the distance factor. It is the distance decay index.

[0029] See Figure 5 This bar chart is a core data visualization result of the heat source interference center point localization process, intuitively presenting the comprehensive weight distribution of the five candidate interference sources. The horizontal axis represents the candidate interference source identifiers, and the vertical axis represents the comprehensive weights. The weight values ​​are calculated using a weighted formula of distance weight + power weight. The value of this chart lies in concretizing the abstract weight calculation results, helping to quickly identify high-risk interference sources. This provides a direct decision-making basis for subsequent calculation of heat source compensation vectors and precise adjustment of heater parameters, demonstrating the technical characteristics of anomaly tracing and quantification in this heating control method.

[0030] Example 4: In specific implementation, calculating the heat source compensation vector includes obtaining the current power setting value and temperature setting value of the heat source interference center point. The current power setting value is directly read from the heater controller, and the temperature setting value is obtained from the process recipe database. The actual temperature values ​​of preset monitoring points around the heat source interference center point are measured. These preset monitoring points are arranged in a ring around the heat source interference center point, and the actual temperature values ​​are collected by a thermocouple array arranged on the ring. The deviation vector between the actual temperature value and the temperature setting value is calculated. The deviation vector is a multi-dimensional vector, with its dimension matching the number of preset monitoring points. Each component corresponds to the temperature deviation of one monitoring point. The power compensation direction is determined based on the projection of the deviation vector into the heater power space. The power compensation direction is a unit vector pointing in the direction where power needs to be increased. The power compensation amplitude is determined by combining the heater's dynamic response characteristic curve. The power compensation direction and power compensation amplitude are combined to form the heat source compensation vector, which is ultimately output in the form of a power adjustment command. In some embodiments, the ring radius of the preset monitoring points can be differentiated according to the thermal field characteristics of different heaters.

[0031] The power compensation amplitude is determined by combining the dynamic response characteristic curve of the heater. The power compensation direction and amplitude are combined to form a heat source compensation vector. The dynamic response characteristic curve of the target heater is extracted from a pre-stored heater characteristic library. The dynamic response characteristic curve is obtained by fitting historical experimental data and represents the functional relationship between power change and temperature change. The magnitude of the deviation vector is read. The magnitude of the deviation vector represents the overall deviation between the actual temperature value and the temperature setpoint. It is obtained by calculating the square root of the sum of the squares of each component of the deviation vector. According to the dynamic response characteristic curve, the power-temperature conversion coefficient corresponding to the current power setpoint is looked up. The power-temperature conversion coefficient is the local slope of the dynamic response characteristic curve at the current power setpoint. An ambient temperature compensation factor is introduced. The ambient temperature compensation factor is calculated by monitoring the difference between the ambient temperature inside the heating furnace and the standard operating conditions. The magnitude of the deviation vector, the power-temperature conversion coefficient, and the ambient temperature compensation factor are substituted into the compensation formula to calculate the final power compensation amplitude. The power compensation direction is vector normalized to obtain a unit direction vector. The unit direction vector and the final power compensation amplitude are multiplied by a scalar to generate the heat source compensation vector. It is understood that the power-temperature conversion coefficient reflects the thermal efficiency of the heater at a specific operating point. In some embodiments, the calculation of the ambient temperature compensation factor can comprehensively consider ambient temperature readings at multiple locations within the furnace.

[0032] Heat source compensation vector The synthesis formula is: in: Represents the heat source compensation vector. This represents the ambient temperature compensation factor. Indicates the current power setting value The power-temperature conversion coefficient is obtained from the dynamic response characteristic curve. This represents the vector representing the deviation between the actual temperature value and the temperature setpoint. The magnitude of the deviation vector is shown in Table 1.

[0033] Table 1: Ambient Temperature Compensation Factors Reference table for values Example 5: In specific implementation, executing heating control operations and recording temperature response data involves decomposing the heat source compensation vector into a power adjustment command sequence for each heater. The decomposition process, based on the communication protocol of the heater control system, converts the vector components into specific power setpoint commands. The heater power is adjusted step-by-step according to the time sequence of the power adjustment command sequence, with the adjustment interval set to a fixed time period. Only one power adjustment command is executed within each period to prevent sudden power fluctuations from impacting the heating furnace's thermal field. After each power adjustment, a temperature distribution image sequence is acquired. The acquisition action is triggered by the power adjustment completion signal, and the temperature distribution image sequence contains data from at least three complete sampling periods after adjustment. The adjusted instantaneous symmetry distribution matrix is ​​calculated using the same method as after regional gridding, employing the same grid division rules and instantaneous symmetry calculation formula. The difference between the instantaneous symmetry distribution matrices before and after adjustment is compared to generate a control effect evaluation index, which quantifies the effectiveness of the heating control operation. The power adjustment command sequence, temperature distribution image sequence, and control effect evaluation index are associated and stored in a historical heating cycle database. This association is achieved through a unified timestamp and batch number. In some embodiments, the timing order of the power adjustment command sequence can be rearranged according to the command priority, which is set based on the comprehensive weight of the heat source interference center point. Optionally, the acquisition of the temperature distribution image sequence can be delayed to allow the thermal field to reach a stable state.

[0034] Control effectiveness evaluation indicators The calculation is performed using the following formula: in: Indicates the indicators for evaluating the effectiveness of control. This represents the value of the grid cell in the i-th row and j-th column of the adjusted instantaneous symmetry distribution matrix. This represents the value of the grid cell in the i-th row and j-th column of the instantaneous symmetry distribution matrix before adjustment. and These represent the number of rows and columns of the instantaneous symmetry distribution matrix, respectively. This can be understood as a control effectiveness evaluation index. The closer the value is to 1, the more significant the effect of the control operation on reducing the temperature difference between the upper and lower surfaces. The power adjustment command sequence includes the command number, target heater identifier, power setpoint, planned execution timestamp, and actual execution timestamp. The temperature distribution image sequence data block includes the start timestamp, end timestamp, image data index, and a snapshot of the corresponding instantaneous symmetry distribution matrix. The control effect evaluation index record includes the calculated... The data includes the value, the calculation timestamp, and the associated power adjustment instruction sequence identifier. In some embodiments, the historical heating cycle database employs a time-series database architecture to optimize time-range-based query efficiency.

[0035] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A liquid crystal glass heating control method based on upper and lower symmetry temperature prediction, characterized by, The implementation steps are as follows: A real-time monitoring system for a three-dimensional temperature field in a liquid crystal glass heating furnace is established, and temperature distribution image sequences of the upper surface region and the lower surface region of the glass are synchronously captured; A time alignment model for the temperature distribution image sequences is constructed, and the upper surface temperature distribution image and the lower surface temperature distribution image are mapped to a unified time coordinate system; The regionally gridded segmentation of the time-aligned temperature distribution image sequences is performed, and the instantaneous symmetry degree of the upper surface temperature value and the lower surface temperature value in each grid cell is calculated; According to the spatial distribution pattern of the instantaneous symmetry degree, an abnormal temperature conduction region is identified, and a primary abnormal region marker map is generated; The primary abnormal region marker map is dynamically threshold corrected by fusing historical heating cycle data and the temperature conduction rate in the current heating stage, and a secondary abnormal region marker map is generated; Based on the secondary abnormal region marker map, a heat source disturbance center point is located, and a heat source compensation vector is calculated in combination with a heater power distribution map; According to the heat source compensation vector, the working parameters of the corresponding heater are adjusted, the heating control operation is performed, and the temperature response data is recorded.

2. The liquid crystal glass heating control method based on the up-down symmetric temperature prediction according to claim 1, characterized in that, The establishment of the real-time monitoring system for the three-dimensional temperature field in the liquid crystal glass heating furnace includes: An array of infrared thermal imagers is arranged on the top and bottom of the heating furnace respectively to synchronously collect temperature distribution images of the upper surface and the lower surface of the glass at a fixed sampling frequency; A time stamp, a heater power state and an environmental pressure parameter are attached to each temperature distribution image to form a multi-modal data frame; The multi-modal data frame is time reference calibrated to eliminate the timing error caused by the time delay of different thermal imagers; The calibrated multi-modal data frame is stored in a distributed cache database to construct a real-time temperature monitoring data stream.

3. The liquid crystal glass heating control method based on the up-down symmetric temperature prediction of claim 2, wherein, The construction of the time alignment model for the temperature distribution image sequences includes: The upper surface temperature distribution image sequence and the lower surface temperature distribution image sequence in a continuous time window are extracted from the real-time temperature monitoring data stream; A feature point tracking algorithm is used to identify corresponding thermal feature points in the upper surface temperature distribution image and the lower surface temperature distribution image; A time warping function is constructed according to the motion trajectory of the thermal feature points to map the lower surface temperature distribution image sequence to the time axis of the upper surface temperature distribution image sequence; The image sequence mapped to the time axis is spatially resampled to ensure that the upper surface temperature distribution image and the lower surface temperature distribution image at each time point have the same spatial resolution.

4. The liquid crystal glass heating control method based on the up-down symmetric temperature prediction of claim 1, wherein, The regionally gridded segmentation of the time-aligned temperature distribution image sequences includes: A single temperature distribution image is divided into uniform square grid cells, and the grid size is dynamically adjusted according to the thermal expansion coefficient of the glass; The distribution variance of the upper surface temperature value and the distribution variance of the lower surface temperature value are counted in each grid cell; The absolute difference between the average value of the upper surface temperature value and the average value of the lower surface temperature value in the grid cell is calculated, and the instantaneous symmetry degree is obtained by dividing the absolute difference by the spatial area of the grid cell; All the instantaneous symmetry degrees of the grid cells are combined to form an instantaneous symmetry degree distribution matrix.

5. The liquid crystal glass heating control method based on the up-down symmetric temperature prediction according to claim 4, characterized in that, The identification of the abnormal temperature conduction region according to the spatial distribution pattern of the instantaneous symmetry degree includes: The instantaneous symmetry degree distribution matrix is subjected to morphological opening operation to eliminate isolated noise points; The region growing algorithm is used to connect adjacent high instantaneous symmetry grid cells to form a potential anomaly area profile; The geometric center coordinates and area of each potential anomaly area profile are calculated; When the area exceeds the area threshold and the geometric center is located within the influence range of the heater, the potential anomaly area profile is marked as a primary anomaly area.

6. The upper and lower symmetry temperature prediction based liquid glass heating control method according to claim 1, wherein, The dynamic threshold correction of the primary anomaly area marking map includes: Querying the temperature conduction anomaly records of similar heating stages in the historical heating cycle database; Extracting the duration length and temperature deviation amplitude of the anomaly area in the historical records as reference features; Comparing the similarity of the anomaly area in the current primary anomaly area marking map with the historical reference features; When the similarity is lower than the similarity threshold, the area threshold is increased and the anomaly area marking is regenerated; When the similarity is higher than the similarity threshold, the anomaly area boundary is fine-tuned in combination with the temperature conduction rate of the current heating stage to form a secondary anomaly area marking map.

7. The upper-lower symmetric temperature prediction based liquid crystal glass heating control method according to claim 1, wherein, The positioning of the heat source disturbance center point includes: Extracting the centroid coordinates of each anomaly area in the secondary anomaly area marking map; Building a circular search area with the centroid coordinates as the center and the heater influence radius as the search radius; Analyzing the gradient direction of the heater power distribution map within the circular search area; Identifying the heater with the gradient direction pointing to the anomaly area centroid as the candidate disturbance source; Calculating the distance weight and power weight of each candidate disturbance source to the anomaly area centroid, and selecting the candidate disturbance source with the largest comprehensive weight as the heat source disturbance center point.

8. The upper-lower symmetric temperature prediction based liquid glass heating control method according to claim 7, wherein, The calculation of the heat source compensation vector includes: Obtaining the current power setting value and temperature setting value of the heat source disturbance center point; Measuring the actual temperature values of the preset monitoring points around the heat source disturbance center point; Calculating the deviation vector of the actual temperature values and the temperature setting value; Determining the power compensation direction according to the projection of the deviation vector in the heater power space; Combining the dynamic response characteristic curve of the heater to determine the power compensation amplitude, and synthesizing the power compensation direction and the power compensation amplitude into the heat source compensation vector.

9. The upper-lower symmetric temperature prediction based liquid glass heating control method according to claim 8, wherein, The combination of the dynamic response characteristic curve of the heater to determine the power compensation amplitude, and the synthesis of the power compensation direction and the power compensation amplitude into the heat source compensation vector includes: Extracting the dynamic response characteristic curve of the target heater from the pre-stored heater characteristic library, which is fitted by historical experimental data and represents the functional relationship between the power change and the temperature change; Reading the module length of the deviation vector, which represents the overall deviation size of the actual temperature values and the temperature setting value; According to the dynamic response characteristic curve, the local slope corresponding to the current power setting value is queried as the power-temperature conversion coefficient; Divide the module length of the deviation vector by the power-temperature conversion coefficient to obtain the basic power compensation amplitude; Introducing an environmental temperature compensation factor, which is calculated by monitoring the difference between the environmental temperature in the heating furnace and the standard working condition, to scale and correct the basic power compensation amplitude to obtain the final power compensation amplitude; Performing vector normalization processing on the power compensation direction to obtain a unit direction vector; Performing scalar multiplication operation on the unit direction vector and the final power compensation amplitude to generate the heat source compensation vector.

10. The upper-lower symmetric temperature prediction based liquid crystal glass heating control method according to claim 1, wherein, The heating control operation is performed and temperature response data is recorded, which comprises: decomposing the heat source compensation vector into power adjustment instruction sequences of each heater; adjusting the heater power step by step according to the time sequence of the power adjustment instruction sequences; collecting a temperature distribution image sequence after each power adjustment, and calculating an adjusted instantaneous symmetry distribution matrix; comparing the difference between the instantaneous symmetry distribution matrices before and after the adjustment to generate a control effect evaluation index; storing the power adjustment instruction sequences, the temperature distribution image sequences and the control effect evaluation index into a historical heating cycle database in association.

Images