Glass processing parameter optimization method and system based on edge computing

Abstract

The application relates to the technical field of glass processing, and provides a glass processing parameter optimization method and system based on edge computing, which comprises the following steps: acquiring physical state data generated in a glass processing process; performing feature analysis and processing on the physical state data to obtain processing area feature data; generating a physical state diagram for reflecting local heat distribution and internal stress distribution left on a glass plate by a previous process according to the processing area feature data and physical characteristics of the glass material; acquiring a processing position of a next process, and determining corresponding local heat values and internal stress values from the physical state diagram according to the processing position; adjusting initial processing parameters of the next process according to the local heat values and the internal stress values to obtain adjusted processing parameters; and controlling a processing device to execute the next process according to the adjusted processing parameters. The application has the effect of improving the overall quality and efficiency of glass processing.

Description

Technical Field

[0001] This invention relates to the technical field of glass processing, and specifically to a method and system for optimizing glass processing parameters based on edge computing. Background Technology

[0002] In modern industrial production, especially in the field of glass deep processing, automated processing equipment uses precise CNC systems to perform complex processes such as cutting and drilling to meet the ever-increasing demands for quality and efficiency. To ensure processing quality and efficiency, intelligent parameter optimization systems are typically introduced. These systems deploy edge computing units on the production floor to collect information from various sensors in real time, quickly analyze this information, and then dynamically adjust processing parameters. This real-time adjustment mechanism can effectively improve product quality and production efficiency when handling single, continuous processing tasks.

[0003] However, in actual production scheduling, a large sheet of glass often requires a complex processing sequence rather than a single task. A typical work order might require first cutting several rectangular areas from the glass sheet, then drilling several mounting holes within each rectangular area, and finally performing fine edge grinding on the rectangles. This means that the processing equipment needs to continuously switch between different operating modes in a short period of time, from high-speed linear cutting to low-speed, high-torque drilling, and then to smooth grinding. Each switch in operating mode is accompanied by drastic changes in the mechanical state of the equipment and the physical characteristics of the processing. For example, at the moment of switching from cutting to drilling, the vibration frequency and amplitude of the cutting tool will change abruptly, and the temperature in the processing area will change from a gentle linear increase to a sharp, point-like increase.

[0004] This drastic shift in state between operations presents a severe challenge to existing edge computing optimization methods. These methods are trained on data collected under relatively stable processing conditions, excelling at fine-tuning within specific scenarios. When the equipment switches from cutting to drilling mode, the sensor data stream changes abruptly. The system may misinterpret this normal, expected mode switch as a serious processing anomaly. For example, it might interpret a sudden increase in spindle vibration at the start of drilling as a malfunction, outputting incorrect correction commands, such as excessively reducing the drill bit speed, which could cause the glass to chip or crack due to uneven stress at the moment of drilling. This "optimization error" occurring at the initial stage of task switching not only fails to improve quality but also becomes a new source of defects.

[0005] To address this issue, a straightforward approach is to temporarily disable the parameter adjustment function of the edge computing system during the initial phase after each task switch. This allows the device to run for a short period using a fixed, conservative set of "safe parameters," and then re-enable real-time optimization once the processing status stabilizes. While this method avoids misjudgments during the transition period, it introduces new problems. Glass processing is a highly continuous physical process; the final state of the previous operation directly affects the initial conditions of the next. For example, a previous long-distance cutting process leaves residual stress fields and heat distributions inside the glass sheet. When drilling occurs immediately near the cutting line, these residual stresses and heat significantly impact the drilling quality. If the system ignores this inherited state information from the previous process because the optimization function is disabled at the start of drilling, it cannot set the most suitable initial drilling parameters immediately. This is equivalent to forfeiting the opportunity to optimize during the most critical drilling stage, significantly diminishing the value of the entire real-time optimization system.

[0006] Therefore, the core technical challenge we face is how to enable edge computing systems to identify and understand the sequence context of processing tasks, and use the state data at the end of the previous process to predictively generate and apply the optimal initial processing parameters for the next process, thereby achieving a seamless and high-quality transition between different processes and avoiding optimization failures or accuracy losses caused by process switching.

[0007] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0008] This application discloses a glass processing parameter optimization method and system based on edge computing, which aims to solve the problem that existing edge computing systems cannot effectively utilize the end state data of the previous process to predictively generate and apply the best initial processing parameters for the next process when switching glass processing steps, resulting in optimization failure or loss of accuracy.

[0009] The technical solution of this application is as follows: In a first aspect, this application discloses a glass processing parameter optimization method based on edge computing, comprising the following steps: Acquire physical state data generated during glass processing; Feature analysis is performed on the physical state data to obtain the feature data of the processing area; Based on the characteristic data of the processing area and the physical properties of the glass material, a physical state diagram is generated to reflect the local heat distribution and internal stress distribution left on the glass sheet by the previous process. Obtain the processing position of the next process, and determine the corresponding local heat value and internal stress value from the physical state diagram based on the processing position; Based on the local heat value and internal stress value, adjust the initial processing parameters of the next process to obtain the adjusted processing parameters; Control the processing equipment to execute the next process according to the adjusted processing parameters.

[0010] This technical solution effectively addresses the problem of optimization failure or precision loss caused by process switching in existing technologies. By proactively utilizing the physical state information of the previous process, it provides optimized initial processing parameters for the next process, thereby achieving a seamless and high-quality transition between different processes and significantly improving the overall quality and efficiency of glass processing.

[0011] Secondly, this application also discloses a glass processing parameter optimization system based on edge computing, used to perform glass processing parameter optimization based on edge computing, including: The physical state acquisition module is used to acquire physical state data generated during glass processing; The feature data analysis module is used to perform feature analysis on physical state data to obtain feature data of the processing area; The physical state diagram generation module is used to generate a physical state diagram that reflects the local heat distribution and internal stress distribution left on the glass sheet by the previous process, based on the characteristic data of the processing area and the physical properties of the glass material. The physical state diagram module is used to obtain the processing position of the next process and, based on the processing position, to determine the corresponding local heat value and internal stress value from the physical state diagram. The processing parameter adjustment module is used to adjust the initial processing parameters of the next process based on the local heat value and local stress value, so as to obtain the adjusted processing parameters. The processing parameter execution module is used to control the processing equipment to execute the next process according to the adjusted processing parameters.

[0012] This application provides a system that can optimize glass processing parameters through this technical solution. The system, through modular design, can efficiently acquire and analyze physical state data, generate and utilize physical state diagrams, and ultimately achieve intelligent adjustment and execution of processing parameters, thereby effectively improving the quality and efficiency of glass processing.

[0013] Beneficial Effects: This application discloses a glass processing parameter optimization method based on edge computing. This method acquires physical state data generated during glass processing and performs feature analysis to obtain processing area feature data. Based on this, and combined with the physical properties of the glass material, a physical state map is generated to reflect the local heat distribution and internal stress distribution left on the glass sheet by the previous process. Subsequently, the method obtains the processing position of the next process and determines the corresponding local heat value and internal stress value from the physical state map. Finally, based on these local heat values ​​and internal stress values, the initial processing parameters of the next process are adjusted to obtain the adjusted processing parameters, and the processing equipment is controlled to execute the next process according to the adjusted processing parameters.

[0014] Through the above technical solution, this application effectively solves the technical problem in the prior art where edge computing systems, when switching glass processing steps, cannot effectively utilize the end-state data of the previous step to predictively generate and apply the optimal initial processing parameters for the next step, leading to optimization failure or accuracy loss. Specifically, this application constructs a physical state diagram to quantify and visualize the local heat distribution and internal stress distribution left on the glass sheet by the previous step, enabling the system to "memorize" and "understand" the sequence context of the processing task. Before the next step begins, the system can obtain the actual impact of the previous step on the glass sheet based on the physical state diagram and predictively adjust the initial processing parameters for the next step accordingly. This forward-looking parameter optimization mechanism avoids "optimization errors" caused by drastic state changes at the beginning of task switching and overcomes the loss of optimization opportunities caused by temporarily disabling optimization functions in traditional methods. Therefore, this application can achieve seamless, high-quality transitions between different processes, significantly improving the overall quality and efficiency of glass processing, and has significant progressiveness and practicality. Attached Figure Description

[0015] Figure 1 This is a flowchart of a glass processing parameter optimization method based on edge computing in one embodiment of the present invention; Figure 2 This is a flowchart of a glass processing parameter optimization method based on edge computing in another embodiment of the present invention; Figure 3 This is a system block diagram of a glass processing parameter optimization system based on edge computing in another embodiment of the present invention; Explanation of reference numerals in the attached figures: 1. Edge computing-based glass processing parameter optimization system; 11. Physical state acquisition module; 12. Feature data analysis module; 13. Physical state diagram generation module; 14. Physical state diagram usage module; 15. Processing parameter adjustment module; 16. Processing parameter execution module. Detailed Implementation

[0016] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0017] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0018] This application proposes a glass processing parameter optimization method based on edge computing, combined with... Figure 1 As shown, it includes: S1, acquire physical state data generated during glass processing; S2, perform feature analysis on the physical state data to obtain the feature data of the processing area; S3, Based on the processing area feature data and the physical properties of the glass material, generate a physical state diagram that reflects the local heat distribution and internal stress distribution left on the glass sheet by the previous process. S4, obtain the processing position of the next process, and determine the corresponding local heat value and internal stress value from the physical state diagram based on the processing position; S5. Based on the local heat value and internal stress value, adjust the initial processing parameters of the next process to obtain the adjusted processing parameters; S6 controls the processing equipment to execute the next process according to the adjusted processing parameters.

[0019] To better understand the technical solution proposed in this application, it is necessary to explain some key terms involved. "Physical state data" refers to data collected by various sensors during glass processing that reflects the real-time physical state of the glass sheet and its processing area, such as temperature, stress, vibration, and acoustic signals. This data reflects the thermal effects, mechanical disturbances, and structural responses experienced by the glass sheet during processing, and forms the basis for subsequent processing area state identification and processing parameter optimization. "Processing area characteristic data" refers to key information extracted from the original physical state data that characterizes specific attributes or potential defect risks in the processing area, such as the area of ​​a sudden temperature rise, the rate of temperature change, the high-frequency oscillation components of the cutting force, force abrupt changes, and the energy distribution and instantaneous peak value of the acoustic signal. Compared to the original physical state data, processing area characteristic data more comprehensively reflects whether there is localized heat accumulation, stress concentration, abnormal vibration, or crack initiation risk in the processing area.

[0020] A "physical state diagram" is a data structure or visualization that integrates the characteristic data of the processing area and the physical properties of the glass material. It reflects the localized heat distribution and internal stress distribution left on the glass sheet by the previous process. Because glass materials are susceptible to thermal shock, residual stress, and localized mechanical disturbances during continuous processing, the localized heat and internal stress generated in the previous process do not immediately dissipate completely, affecting the processing stability of subsequent processes. Therefore, the physical state diagram records the continuity of the processing state of the glass sheet at different locations, providing a locational basis for adjusting parameters in the next process. "Local heat values ​​and internal stress values" refer to the heat and stress information extracted from the physical state diagram that precisely corresponds to the processing location of the next process. Local heat values ​​reflect the degree of heat accumulation at the corresponding processing location, while internal stress values ​​reflect the degree of residual stress or stress concentration at the corresponding processing location. These values ​​directly reflect the current physical state of the processing location in the next process and are crucial inputs for adjusting the processing parameters of the next process. "Initial machining parameters" refer to the machining parameters set based on standard processes, historical experience, or default machining templates before optimization and adjustment, such as tool spindle speed, feed rate, depth of cut, and coolant flow rate. "Adjusted machining parameters" refer to the machining parameters that are more suitable for the current machining conditions, obtained through rule-based judgment, optimization calculation, or parameter matching after considering local heat values ​​and internal stress values.

[0021] This application provides a glass processing parameter optimization method based on edge computing. Its core lies in sensing, extracting, and modeling the physical state left over from the previous process. Before executing the next process, it obtains the local heat and internal stress values ​​corresponding to the processing location and adjusts the initial processing parameters accordingly. This allows the processing control of the next process to adapt to the current actual physical state of the glass sheet. This method does not simply perform fixed processing based on standard process parameters; instead, it uses the heat and internal stress distribution generated by the previous process as the basis for adjusting subsequent processing parameters. This reduces the risk of edge chipping, crack propagation, local breakage, or decreased processing accuracy caused by local heat accumulation or stress concentration during continuous processing.

[0022] In practical implementation, the first step is to acquire physical state data generated during glass processing. This data can be obtained in various ways. For example, infrared thermal sensors, force sensors, vibration sensors, and acoustic sensors can be deployed in the glass sheet processing area, and these sensors can simultaneously collect temperature, force, vibration, and acoustic signals during processing to form multi-source physical state data. Alternatively, data such as spindle load, cutting force, cooling status, equipment vibration, or processing temperature can be directly read from the sensor system built into the processing equipment via the existing production line data interface. For critical local areas, key physical state data can also be collected periodically using handheld measuring devices through manual inspection. The physical state data acquired through these methods can cover thermal, mechanical, and dynamic disturbance states, providing a raw data foundation for subsequent feature data extraction from the processing area.

[0023] After acquiring the physical state data, feature analysis is performed on the acquired physical state data to obtain the processing area feature data. Specifically, statistical analysis can be performed on the collected temperature data to calculate the mean, variance, maximum, and minimum values ​​of the temperature data, and further extract the area of ​​local temperature surge regions and the rate of temperature change; Fourier transform or time-frequency analysis can be performed on the mechanical data to extract the high-frequency oscillation components and force value mutation characteristics of the cutting force; energy distribution analysis and peak detection can be performed on the acoustic signals to extract the energy distribution and instantaneous peak values ​​of the sound wave signals. As another implementation method, machine learning models can be used to perform pattern recognition on the original physical state data, such as training a classifier to identify whether there are abnormal patterns in the data, and using the recognition results as processing area feature data. Alternatively, an expert system can be used to extract specific feature values ​​from the physical state data according to preset rules and thresholds. For example, when the temperature exceeds a preset threshold, the corresponding area is marked as a high-temperature area feature; when the cutting force undergoes an instantaneous mutation, the corresponding time period or area is marked as a high-stress risk feature. Through this step, the original physical state data is transformed into processing area feature data that can characterize the processing area state and potential defect risks.

[0024] Subsequently, based on the characteristic data of the processing area and the physical properties of the glass material, a physical state diagram (PSD) is generated to reflect the local heat distribution and internal stress distribution left on the glass sheet by the previous process. The physical properties of the glass material can include thermal conductivity, specific heat capacity, coefficient of thermal expansion, elastic modulus, fracture toughness, and stress release characteristics. When generating the PSD, finite element analysis software can be used, taking the characteristic data of the processing area as input and combining it with the thermodynamic and mechanical parameters of the glass material to simulate and calculate the heat and stress distribution inside the glass sheet, and then generating the PSD from the calculation results. Alternatively, a simulation system based on a physical model can be established, allowing the system to calculate the local heat distribution and internal stress distribution at different locations on the glass sheet in real time based on the input processing area characteristic data and material physical properties, and generate the corresponding PSD. A data-driven approach can also be used, such as training a deep learning model to learn and predict heat and stress distribution based on the processing area characteristic data, thereby generating the PSD. Through the PSD, the thermal and stress effects of the previous process on the glass sheet can be transformed into location-specific, queryable state data.

[0025] Next, the processing position of the next step is obtained, and based on this position, the corresponding local heat value and internal stress value are determined from the physical state diagram. The processing position of the next step can be obtained by analyzing the CNC program. For example, the starting processing coordinates, path point coordinates, and processing trajectory range of the next step can be automatically extracted from the CNC program, and these coordinates can be mapped onto the physical state diagram to read the corresponding local heat value and internal stress value. Alternatively, the operator can manually input or select the processing position of the next step on the human-machine interface, and the system can query the corresponding local heat value and internal stress value in the physical state diagram based on the input position information. Image recognition technology can also be used to identify marker points or feature areas on the glass sheet to determine the processing position of the next step, and further extract the corresponding local heat value and internal stress value from the physical state diagram. Through this step, the next step is no longer executed solely based on the theoretical processing position and default process parameters, but rather obtains the current thermal and stress states corresponding to the actual processing position.

[0026] After determining the local heat value and internal stress value, the initial machining parameters for the next process are adjusted based on these values ​​to obtain the adjusted machining parameters. One approach is to pre-set parameter adjustment rules, such as reducing the feed rate when the local heat value is higher than a preset threshold, reducing the depth of cut when the internal stress value is higher than a preset threshold, and simultaneously reducing the tool speed and increasing the coolant flow rate when both local heat value and internal stress value are at a high level. Another approach is to utilize optimization algorithms, such as genetic algorithms or particle swarm optimization algorithms, using machining quality or efficiency as the objective function and local heat value and internal stress value as constraints, to calculate the optimal combination of machining parameters suitable for the current physical state of the machining position. Alternatively, a lookup table method can be used to find the corresponding parameter adjustment amount from a pre-established parameter adjustment table based on different combinations of local heat value and internal stress value, thus obtaining the adjusted machining parameters. Through this step, the initial machining parameters can be specifically corrected based on the current local heat value and internal stress value of the glass sheet.

[0027] Finally, the processing equipment is controlled to execute the next process according to the adjusted processing parameters. Specifically, the adjusted processing parameters can be sent to the CNC system of the processing equipment. After receiving the adjusted processing parameters, the CNC system adjusts the control quantities such as tool movement trajectory, tool speed, feed rate, depth of cut, and coolant flow rate accordingly, so that the next process is executed according to parameters that match the current physical state of the glass sheet. In this way, this application can introduce the local heat distribution and internal stress distribution generated in the previous process into the parameter control process of the next process, enabling the glass processing process to have the state continuity between the previous and subsequent processes, reducing the risk of processing defects caused by the failure to consider the thermal stress left over from the previous process, and improving the stability of the glass processing process and the quality of the finished product.

[0028] Optional, combined Figure 2 As shown, the steps for generating a physical state diagram reflecting the local heat distribution and internal stress distribution left on the glass sheet by the previous process, based on the characteristic data of the processing area and the physical properties of the glass material, include: A1, acquire the image sequence of surface temperature distribution in the glass plate processing area; A2, extract the curves of temperature values ​​at specific points on the glass plate and the corresponding preset areas over time from the surface temperature distribution image sequence; the preset area refers to the neighborhood area determined based on the preset neighborhood radius; A3. Based on the curve of temperature change over time, calculate the rate at which heat diffuses outward from a specific point to obtain the heat diffusion rate. A4. Identify the heat conduction characteristics of the local area to which a specific point belongs based on the heat diffusion rate; A5. Based on the heat conduction characteristics, adjust the thermal diffusivity used to estimate the local heat distribution to obtain the adjusted thermal diffusivity. A6. Based on the adjusted thermal diffusivity, processing area characteristic data, and physical properties of the glass material, calculate the local heat distribution and internal stress distribution. A7. Based on the local heat distribution and internal stress distribution, generate a physical state diagram that reflects the local heat distribution and internal stress distribution left on the glass plate by the previous process.

[0029] Specifically, when generating the physical state diagram, it is first necessary to obtain a sequence of images showing the surface temperature distribution of the glass sheet's processing area. This sequence of images provides dynamic information about how the glass surface temperature changes over time, rather than a static snapshot at a single moment, which is crucial for understanding the conduction and diffusion of heat within the glass.

[0030] Furthermore, from the surface temperature distribution image sequence, curves showing the temperature changes over time at specific points on the glass plate and their corresponding preset regions can be extracted. The preset region refers to a neighborhood area determined based on a preset neighborhood radius. By focusing on the temperature changes at specific points and their neighborhoods, local thermal effects can be analyzed more precisely. For example, key points or areas of concentrated heat along the processing path can be selected as specific points, and a preset neighborhood radius can be set around these points to define a local observation area.

[0031] By analyzing the temperature-time curve, the rate at which heat diffuses outward from a specific point can be calculated, thus yielding the heat diffusion rate. This diffusion rate is a key parameter for measuring the speed at which heat propagates within a glass material, and its calculation can be based on Fourier's law of heat conduction or achieved through numerical simulation methods.

[0032] Based on the heat diffusion rate, the heat conduction characteristics of a local area at a specific point can be identified. The thermal conductivity of glass materials may vary locally due to their microstructure, impurity distribution, or processing history. By analyzing the actual heat diffusion rate, the actual heat conduction behavior of that local area can be understood more accurately.

[0033] Based on the heat conduction characteristics, the thermal diffusivity coefficient used to estimate local heat distribution can be adjusted to obtain a modified thermal diffusivity coefficient. Traditional thermal diffusivity coefficients are usually based on the macroscopic average value of the material, but in actual processing, the thermal diffusivity coefficient in local areas may deviate. By dynamically adjusting the thermal diffusivity coefficient in conjunction with actually observed heat conduction characteristics, the thermal model can be made closer to reality.

[0034] Based on this, and using the adjusted thermal diffusivity, processing area characteristic data, and the physical properties of the glass material, the local heat distribution and internal stress distribution can be calculated. The adjusted thermal diffusivity makes the calculation of heat distribution more accurate, and thus a more accurate internal stress distribution can be derived through a thermo-mechanical coupling model.

[0035] Finally, based on the local heat distribution and internal stress distribution, a physical state diagram is generated to reflect the local heat distribution and internal stress distribution left on the glass sheet by the previous process. This physical state diagram is presented in a visual or data model format, providing a precise physical basis for parameter optimization in subsequent processes.

[0036] In some preferred embodiments, it is assumed that fine grinding is required after the glass sheet undergoes a laser cutting process. Laser cutting generates localized high temperatures and thermal stress near the cutting path. To optimize the parameters of the subsequent grinding process and avoid edge chipping or cracking due to residual thermal stress, this application specifically implements the following: First, a high-frame-rate infrared thermal imaging sensor continuously acquires a sequence of images showing the surface temperature distribution in the laser-cut area. This sequence records the entire process of heat diffusion and cooling on the glass surface after cutting.

[0037] Next, using image processing algorithms, multiple specific points on the cutting path are identified, and a preset neighborhood radius of 5 mm is set for each point. The curves of temperature changes over time for these points and their neighborhoods are then extracted.

[0038] Then, based on these temperature curves, the instantaneous rate of heat diffusion outward at each specific point is calculated using a heat conduction model. For example, in some regions, the heat diffusion rate may be slower, indicating that the heat conduction characteristics in that region may differ from those of the overall material.

[0039] Based on the calculated heat diffusion rate, the system identifies the heat conduction characteristics of different local areas. For example, it finds that the heat conduction characteristics of the cut edge area differ from those of the center area of ​​the board.

[0040] Based on these identified local heat conduction characteristics, the system dynamically adjusts the thermal diffusivity used to calculate local heat distribution. For example, for areas where heat diffuses slowly, the thermal diffusivity is appropriately lowered to more accurately simulate actual heat retention.

[0041] Finally, using the adjusted thermal diffusivity, processing area characteristic data of the laser cutting process (such as cutting power and speed), and the known physical properties of the glass material, the local heat distribution and internal stress distribution of the entire cutting area are calculated. Based on these calculation results, a high-resolution physical state map is generated, clearly showing the heat concentration areas and stress distribution patterns near the cutting path. This physical state map is then used to guide the precise adjustment of parameters such as the feed rate, grinding pressure, and coolant flow rate of the grinding equipment to minimize processing defects.

[0042] Optionally, the steps of adjusting the initial processing parameters for the next process based on local heat values ​​and internal stress values ​​to obtain the adjusted processing parameters include: Obtain and determine the processing type for the next step; Select a parameter adjustment strategy based on the processing type, local heat value, and internal stress value; Based on the parameter adjustment strategy, calculate the adjustment amount of multiple related processing parameters; Based on the adjustment amount, the initial processing parameters for the next process are adjusted to generate the adjusted processing parameters.

[0043] In this context, acquiring and determining the processing type of the next step means that before adjusting parameters, the system needs to clearly define the specific category of the processing operation to be performed, such as cutting, grinding, polishing, drilling, or chamfering. This processing type can be determined by reading processing task instructions, analyzing CAD / CAM data, or through operator input.

[0044] Furthermore, selecting parameter adjustment strategies based on processing type, local heat value, and internal stress value means that once the processing type is determined, the system combines the current local heat value and internal stress value of the glass sheet and selects the most suitable parameter adjustment strategy from a preset strategy library. For example, for the cutting process, if the local heat value is too high, a strategy of reducing the cutting speed or increasing the coolant flow rate may be selected; if the internal stress value is too high, a strategy of step-by-step cutting or optimizing the tool path may be selected. These strategies can be constructed based on expert experience rules, lookup tables, machine learning models, or optimization algorithms.

[0045] Specifically, calculating the adjustment amount of multiple related machining parameters based on the parameter adjustment strategy means that the selected parameter adjustment strategy guides the system to calculate specific adjustment values ​​for multiple related machining parameters in the next process. These related machining parameters may include, but are not limited to, feed rate, spindle speed, tool pressure, coolant flow rate, machining depth, or machining path offset. The calculation of the adjustment amount can be based on mathematical models, empirical formulas, or through iterative optimization.

[0046] Therefore, adjusting the initial processing parameters of the next process based on the adjustment amount to generate adjusted processing parameters means applying the calculated adjustment amount to the initial processing parameters of the next process, thereby generating a new set of optimized processing parameters. These adjusted parameters will be more adapted to the current physical state of the glass sheet, in order to avoid processing defects or improve processing quality.

[0047] Optionally, the steps for acquiring physical state data generated during glass processing include: Collect initial physical state data of the glass processing area; Interference identification is performed on the initial physical state data to obtain the interference identification results; Based on the interference identification results, the filtering parameters are adjusted or signal reconstruction is initiated to remove the interference components in the initial physical state data and obtain the interference-free physical state data. Among them, signal reconstruction is a process of recovering the interfered data segments based on adjacent time window data, adjacent spatial location data, or multi-sensor homogeneous data when there are missing segments, local saturation, or abnormal peak interference in the initial physical state data. Environmental compensation calibration is performed on the interference-free physical state data to obtain calibrated physical state data. The calibrated physical state data is used as the physical state data generated during the glass processing.

[0048] Specifically, collecting initial physical state data in the glass processing area refers to acquiring real-time information on the raw physical quantities of the glass sheet and its processing through various sensors deployed in the glass sheet processing area, such as high-frame-rate infrared thermal imaging sensors, high-sampling-rate force sensors, and high-sensitivity acoustic sensors. This data may include surface temperature distribution, cutting force, vibration, and acoustic signals.

[0049] Interference identification in the initial physical state data aims to detect and locate anomalies caused by non-real physical processes within the data. This can be achieved through various signal processing techniques, such as detecting outliers using statistical methods, identifying periodic noise based on frequency domain analysis, or using machine learning models to identify specific interference patterns. The interference identification results can indicate the type, intensity, and location of the interference.

[0050] In practical applications, based on the interference identification results, filter parameters are adjusted or signal reconstruction is initiated to remove interference components from the initial physical state data. When a specific type of noise interference is identified, the parameters of the digital filter, such as the cutoff frequency and filter order, can be dynamically adjusted to effectively filter out the noise. When the initial physical state data contains missing segments, local saturation, or abnormal spike interference, signal reconstruction is initiated. Signal reconstruction specifically refers to utilizing the inherent correlation of the data itself, such as data trends within adjacent time windows, data distribution at adjacent spatial locations, or homogeneous data from different sensors measuring the same physical phenomenon, to recover and fill in the interfered data segments, thereby reconstructing data that more closely approximates the true state.

[0051] Furthermore, environmental compensation calibration is performed on the interference-free physical state data to eliminate the influence of environmental factors on the sensor measurement results. For example, changes in ambient temperature, humidity, and air pressure may cause sensor output drift or measurement errors. Environmental compensation calibration can be achieved through a preset calibration model, real-time environmental parameter measurement and correction algorithms, or a compensation model trained based on historical data, ensuring that the data reflects the physical state of the glass processing itself, rather than environmental changes.

[0052] Therefore, after the above series of processes, the calibrated physical state data is used as the physical state data generated during glass processing. This data has higher accuracy, reliability and consistency, providing high-quality input for subsequent processing parameter optimization.

[0053] Optionally, the step of performing feature analysis on the physical state data to obtain the feature data of the processing area may include the following: Obtain an image of the surface temperature distribution in the glass sheet processing area; Calculate the area and rate of temperature change of regions with sudden temperature increases in a surface temperature distribution image; Obtain cutting force data when the tool contacts the glass; Analyze the high-frequency oscillation components and abrupt force changes in cutting force data; Acquire acoustic signals generated during the processing; Analyze the energy distribution and instantaneous peak value of acoustic signals in a specific frequency band; By integrating the area of ​​the local temperature rise region, the temperature change rate of the local temperature rise region, the high-frequency oscillation components, the force value change, the energy distribution, and the instantaneous peak value, characteristic data reflecting the defect risk of the processing area are obtained. Feature data reflecting the risk of defects in the processing area will be used as feature data of the processing area.

[0054] Acquiring an image of the surface temperature distribution in the glass processing area refers to capturing the surface temperature distribution of the glass sheet in real time or near real time during processing using devices such as infrared thermal imaging sensors. This image can intuitively reflect the heat accumulation and distribution in the processing area and is an important basis for assessing potential defects such as thermal stress and microcracks.

[0055] Furthermore, the area and rate of temperature change of localized temperature spikes in the surface temperature distribution image are calculated to quantify potential localized overheating during processing. The area of ​​the localized temperature spike indicates the extent of heat concentration, while the rate of temperature change reflects the severity of heat accumulation. These parameters are crucial for identifying potential thermal damage or material phase transition risks.

[0056] In addition, cutting force data when the tool contacts the glass is typically monitored in real time using force sensors. Cutting force data reflects the interaction between the tool and the workpiece and is a key indicator for evaluating machining stability, tool wear, and material removal efficiency.

[0057] Specifically, analyzing high-frequency oscillations and force abrupt changes in cutting force data is crucial for identifying abnormal dynamic behaviors during machining. High-frequency oscillations may be related to phenomena such as tool chatter and resonance, while force abrupt changes may indicate sudden material fracture, chipping, or abnormal stress on the tool. These abnormal signals are important indicators for assessing the risk of machining defects.

[0058] Meanwhile, the acoustic signals generated during processing are typically acquired using high-sensitivity acoustic sensors. These signals can capture minute vibrations and friction sounds that are difficult for the human ear to detect, providing supplementary information for diagnosing the processing status.

[0059] The analysis of the energy distribution and instantaneous peak values ​​of acoustic signals in specific frequency bands aims to extract features related to specific processing defect patterns from complex acoustic signals. For example, abnormal increases in energy or the appearance of instantaneous peak values ​​in certain frequency bands may be closely related to phenomena such as microcrack propagation, material spalling, or abnormal friction.

[0060] Therefore, by integrating the area of ​​the localized temperature surge region, the rate of temperature change in the localized temperature surge region, high-frequency oscillation components, force abrupt changes, energy distribution, and instantaneous peak values, characteristic data reflecting the defect risk of the processing area is obtained. This integration process aims to fuse characteristic information from different physical modes to form a multi-dimensional and more comprehensive description of the processing area's state. The characteristic data reflecting the defect risk of the processing area can be understood as a comprehensive set of indicators used to quantify the probability and severity of defects occurring in the current processing area.

[0061] Optionally, the step of integrating the area of ​​the local temperature surge region, the temperature change rate of the local temperature surge region, the high-frequency oscillation components, the force abrupt change, the energy distribution, and the instantaneous peak value to obtain characteristic data reflecting the defect risk of the processing area includes: Read surface temperature distribution images, cutting force data, and acoustic signals; Add timestamps and spatial location markers to the surface temperature distribution image, cutting force data, and acoustic signal; Based on timestamps and spatial location markers, and using a unified time-space grid, surface temperature distribution images, cutting force data, and acoustic signals are synchronized. Spatial interpolation processing is performed on the synchronized surface temperature distribution image, cutting force data, and acoustic signal to ensure that the data at the same time point corresponds to the same local area of ​​the glass plate. Based on the surface temperature distribution image after spatial interpolation, cutting force data, and acoustic signal, the area, temperature change rate, high-frequency oscillation component, force value mutation, energy distribution, and instantaneous peak value of the extracted local temperature rise region are aligned and calibrated at the same time point and in the same local region to obtain the updated area, temperature change rate, high-frequency oscillation component, force value mutation, energy distribution, and instantaneous peak value of the local temperature rise region. By fusing the updated data on the area of ​​the local temperature surge region, the rate of temperature change, the high-frequency oscillation components, the force abrupt change, the energy distribution, and the instantaneous peak value, characteristic data reflecting the defect risk of the processing area are obtained.

[0062] Specifically, reading surface temperature distribution images, cutting force data, and acoustic signals refers to acquiring raw or pre-processed multimodal data from corresponding sensors or data storage units. Surface temperature distribution images are typically acquired by high-frame-rate infrared thermal imaging sensors, cutting force data by high-sampling-rate force sensors, and acoustic signals by high-sensitivity acoustic sensors.

[0063] The above-mentioned surface temperature distribution images, cutting force data, and acoustic signals are supplemented with timestamps and spatial location markers to provide a unified time and space reference for data from different sources. The timestamps can be high-precision time synchronization signals to ensure accurate alignment of each data point on the time axis; the spatial location markers can be physical location information based on the coordinate system of the processing equipment or the local coordinate system of the glass plate to ensure accurate correspondence of each data point in the spatial dimension.

[0064] Based on the aforementioned timestamps and spatial location markers, synchronizing the aforementioned surface temperature distribution image, cutting force data, and acoustic signal using a unified time-space grid refers to mapping data with different sampling frequencies and spatial resolutions to a common time-space reference frame through algorithms. For example, techniques such as resampling, interpolation, or extrapolation can be employed to ensure that all data are comparable at each time step and at each spatial grid point.

[0065] Spatial interpolation processing is performed on the synchronized surface temperature distribution image, cutting force data, and acoustic signal to ensure that the data at the same time point corresponds to the same local area of ​​the glass plate. This means that when there are slight spatial deviations in the data from different sensors, the data are adjusted using interpolation algorithms (such as bilinear interpolation, nearest neighbor interpolation, etc.) to ensure that when analyzing a specific local area, all relevant data accurately point to that area.

[0066] Based on the spatially interpolated surface temperature distribution image, cutting force data, and acoustic signals, the area, temperature change rate, high-frequency oscillation components, force abrupt changes, energy distribution, and instantaneous peak values ​​of the extracted local temperature surge regions are aligned and calibrated at the same time point and within the same local region. This yields updated data on the area, temperature change rate, high-frequency oscillation components, force abrupt changes, energy distribution, and instantaneous peak values. This process involves further refining the alignment of high-level features extracted from the data after temporal and spatial synchronization and interpolation. For example, if a temperature surge event and a cutting force abrupt change event are slightly offset in time or slightly misaligned in space, the calibration algorithm corrects these discrepancies to ensure that these features truly reflect the same physical event.

[0067] The updated data, including the area of ​​the local temperature surge region, the rate of temperature change, the high-frequency oscillation components, the force abrupt change, the energy distribution, and the instantaneous peak value, are fused to obtain characteristic data reflecting the defect risk of the processing area. This means integrating various characteristic data that have been precisely aligned and calibrated into a unified and more representative feature vector or index through multimodal data fusion technology (such as weighted averaging, machine learning model input, etc.) for the final defect risk assessment.

[0068] Optionally, the step of integrating the area of ​​the local temperature surge region, the temperature change rate of the local temperature surge region, the high-frequency oscillation components, the force abrupt change, the energy distribution, and the instantaneous peak value to obtain characteristic data reflecting the defect risk of the processing area includes: Obtain and determine the type of processing operation currently being performed on the glass sheet; Based on the processing procedure type, the defect pattern definition corresponding to the processing procedure type is obtained from the preset process defect pattern library; Based on the defect mode definition, a feature weight configuration is generated; the feature weight configuration is used to define the area of ​​the local temperature rise region, the rate of temperature change, the high-frequency oscillation component, the force value change, the energy distribution, and the relative importance of the instantaneous peak value under the current processing procedure type. Based on the feature weight configuration, the area, temperature change rate, high-frequency oscillation components, force value mutation, energy distribution, and instantaneous peak value of the local temperature rise region are weighted and fused to obtain the weighted fusion result; Based on the weighted fusion results, characteristic data reflecting the defect risk in the processing area are obtained.

[0069] Specifically, acquiring and determining the current processing step type of the glass sheet refers to the system identifying the processing operation currently being performed on the glass sheet by reading the processing equipment's program instructions, sensor feedback, or operator input, such as rough grinding, fine grinding, polishing, drilling, and chamfering. This process type is the basis for subsequent defect pattern matching and feature weight configuration.

[0070] Specifically, based on the type of processing step, the system retrieves the corresponding defect pattern definition from a pre-defined process defect pattern library. This can be understood as the system maintaining a database containing various processing steps and their corresponding typical defect patterns. For example, for the "rough grinding" step, possible defect patterns include "edge chipping," "scratches," and "excessive surface roughness"; for the "drilling" step, possible defect patterns include "inaccurate hole diameter" and "inner wall cracking." Each defect pattern has its specific physical characteristics.

[0071] In practical applications, generating feature weight configurations based on defect mode definitions refers to the system dynamically assigning different weight values ​​to various physical features (such as the area of ​​a localized temperature surge region, the rate of temperature change, high-frequency oscillation components, force abrupt changes, energy distribution, and instantaneous peak values) according to the defect mode definition corresponding to the current process type. Feature weight configurations define the relative importance of these features to defect risk assessment under the current machining process type. For example, in rough grinding, the high-frequency oscillation components and force abrupt changes of the cutting force may be more indicative of edge chipping defects and therefore will be assigned higher weights; while in polishing, the area of ​​a localized temperature surge region and the rate of temperature change may be more indicative of surface burns or microcracks and will therefore be assigned higher weights. These weight values ​​can be trained and optimized based on historical data analysis, expert experience, or machine learning models.

[0072] Furthermore, based on the feature weight configuration, the area of ​​the local temperature surge region, the rate of temperature change, high-frequency oscillation components, force abrupt changes, energy distribution, and instantaneous peak values ​​are weighted and fused to obtain a weighted fusion result. This is typically achieved by multiplying each feature value by its corresponding weight and then summing all weighted feature values, or by using other fusion algorithms (such as weighted average, weighted summation, weighted decision tree, etc.). Thus, a comprehensive numerical value is obtained, which more accurately reflects the defect risk under the current processing procedure.

[0073] Finally, based on the weighted fusion results, characteristic data reflecting the defect risk in the processing area is obtained. This characteristic data is an optimized and contextualized indicator that can more accurately guide subsequent adjustments to processing parameters.

[0074] Optionally, the steps for acquiring physical state data generated during glass processing include: Multiple sensors are deployed in the glass sheet processing area, and a hardware-level time synchronization triggering mechanism is adopted to enable the sensors to collect data synchronously. The multiple sensors include high frame rate infrared thermal imaging sensors, high sampling rate force sensors, and high sensitivity acoustic sensors. Each sensor is equipped with an edge preprocessing unit. After the sensor completes data acquisition, the edge preprocessing unit performs preliminary noise reduction and formatting on the acquired data to obtain preprocessed data. A data buffer is set up in the edge computing unit to receive and temporarily store preprocessed data from various sensors; Based on the current high-speed or high-precision mode of the processing equipment, the sampling frequency and data transmission bandwidth of each sensor are dynamically adjusted. The data collected after adjusting the sampling frequency and data transmission bandwidth of each sensor is processed by the edge preprocessing unit and then written into the data buffer. The high-speed mode is the operating mode in which the feed speed or spindle speed of the processing equipment is higher than the preset speed threshold and the processing cycle priority is higher than the precision priority. The high-precision mode is the operating mode in which the processing tolerance requirements, positioning accuracy requirements, or surface quality requirements of the processing equipment are higher than the preset precision threshold and the precision priority is higher than the processing cycle priority. In the edge computing unit, when the processing equipment is detected to have entered the micro-feature processing area, a high-frequency data acquisition mode is activated, and time series analysis is performed on the preprocessed data written to the data buffer to identify the transient physical state changes. The micro-feature processing area is the processing area on the glass plate where the processing width, hole diameter, groove width, chamfer size, or local radius of curvature is smaller than the corresponding preset size threshold. Transient features are extracted based on the identified transient physical state changes, and the extracted transient features are compared with a preset transient feature library to determine whether there are any missed or false alarms, thus obtaining the data acquisition reliability judgment result. When the reliability assessment result indicates that there are missed or false alarms, adjust the sensor gain or focal length and reacquire the data to obtain the reacquired data; When the reliability assessment result indicates that there are no missed or false alarms, the data corresponding to the transient characteristics will be marked as data that has not been judged as missed or false alarms. Based on the re-collected data or data that was not identified as a missed or false alarm, physical state data generated during the glass processing are obtained.

[0075] Specifically, multiple sensors are deployed in the glass processing area to comprehensively capture multi-dimensional physical information during the processing. These sensors include high-frame-rate infrared thermal imaging sensors, high-sampling-rate force sensors, and high-sensitivity acoustic sensors. The high-frame-rate infrared thermal imaging sensors are used to monitor the temperature distribution on the glass surface in real time, capturing heat changes; the high-sampling-rate force sensors are used to measure the cutting force between the tool and the glass during processing, reflecting mechanical stress; and the high-sensitivity acoustic sensors are used to listen to the sound wave signals generated during processing, identifying abnormal vibrations or noise. A hardware-level time synchronization triggering mechanism ensures that data collected by different types of sensors at the same time are accurately aligned, laying the foundation for subsequent multimodal data fusion and analysis.

[0076] Furthermore, an edge preprocessing unit is configured for each sensor to perform preliminary denoising and formatting on the raw data immediately after data acquisition. This edge-side processing effectively reduces the burden on the central computing unit, lowers data transmission latency, and improves the real-time availability of data. Denoising removes interference introduced by the sensor itself or the environment, while formatting unifies the data from different sensors into a standard format, facilitating subsequent integration processing.

[0077] In addition, a data buffer is set up in the edge computing unit to receive and temporarily store preprocessed data from various sensors. This buffer can effectively address the mismatch between data acquisition rate and processing rate, ensuring the continuity and integrity of the data stream and preventing data loss.

[0078] As a preferred implementation, the sampling frequency and data transmission bandwidth of each sensor are dynamically adjusted based on whether the machining equipment is currently operating in high-speed or high-precision mode. High-speed mode typically means that the feed rate or spindle speed of the machining equipment exceeds a preset speed threshold, and the machining cycle time priority is higher than the precision priority. In this case, it may be necessary to reduce the sampling frequency to decrease the amount of data and ensure timely data transmission and processing. High-precision mode, on the other hand, means that the machining tolerance requirements, positioning accuracy requirements, or surface quality requirements of the machining equipment exceed a preset precision threshold, and the precision priority is higher than the machining cycle time priority. In this case, it is necessary to increase the sampling frequency and data transmission bandwidth to capture finer changes in physical state and ensure machining quality. This dynamic adjustment mechanism allows data acquisition to better adapt to different machining needs and optimize resource utilization.

[0079] In the edge computing unit, a high-frequency data acquisition mode is activated when the processing equipment is detected to have entered a micro-feature processing area. The micro-feature processing area refers to regions on the glass sheet where the processing width, aperture, groove width, chamfer size, or local radius of curvature is smaller than a corresponding preset size threshold. These areas have extremely high requirements for processing accuracy and surface quality and are highly susceptible to transient physical state changes. The high-frequency data acquisition mode can capture the physical state data of these areas more densely and, through time-series analysis, identify transient physical state changes, such as sudden increases in local temperature, abrupt changes in cutting force, or abnormal acoustic signals.

[0080] Based on the identified transient physical state changes, transient features are extracted and compared with a pre-set transient feature library to determine whether there are any missed or false alarms, thus obtaining a reliability assessment result. The transient feature library stores multimodal collaborative change patterns corresponding to different defect types, and comparison can verify the validity and accuracy of the collected data. When the reliability assessment result indicates the presence of missed or false alarms, the system adjusts the sensor gain or focal length and re-acquires data to correct data acquisition deviations and ensure data authenticity. When the assessment result indicates no missed or false alarms, the data corresponding to the transient features is marked as reliable data. Finally, based on the re-acquired data or the data not judged as missed or false alarms, the physical state data generated during glass processing is obtained, which has higher accuracy and reliability.

[0081] Optionally, the steps of extracting transient features based on the identified transient physical state changes, comparing the extracted transient features with a preset transient feature library, determining whether there are missed or false alarms, and obtaining the data acquisition reliability judgment result include: Transient features are extracted based on the identified transient physical state changes; wherein, the transient physical state change is a processing state change event characterized by a sudden change or coordinated change in at least one of thermal data, mechanical data and acoustic data within a preset short time window; the transient feature extraction includes extracting the local temperature gradient change rate from the thermal data corresponding to the transient physical state change, extracting the cutting force fluctuation amplitude and frequency from the corresponding mechanical data, and extracting the instantaneous energy increment of a specific frequency band from the corresponding acoustic data; Calculate the correlation between the rate of change of local temperature gradient, the amplitude and frequency of cutting force fluctuation, and the instantaneous energy increment in a specific frequency band; Based on the correlation, the cooperative change patterns among the local temperature gradient change rate, cutting force fluctuation amplitude and frequency, and instantaneous energy increment in a specific frequency band were identified. The identified collaborative change patterns are compared with a preset transient feature library to obtain the comparison results; the transient feature library contains multimodal collaborative change patterns corresponding to different defect types. Based on the comparison results, it is determined whether there are any missed or false alarms, and the reliability of the data collection is judged.

[0082] Specifically, transient physical state changes can be understood as significant, nonlinear changes in the thermal, mechanical, or acoustic data of the processing area within an extremely short time window (e.g., milliseconds or microseconds) due to factors such as the interaction between the tool and the glass, the release of internal material stress, or the accumulation of local heat during glass processing. These changes may manifest as abrupt changes in a single data type, or as synergistic changes occurring synchronously and in relation to multiple data types. For example, when a tool experiences a minor chipping, it may simultaneously cause high-frequency oscillations in the cutting force, a rapid increase in local temperature, and a transient enhancement of sound wave energy in a specific frequency band.

[0083] Transient feature extraction refers to extracting quantitative indicators that characterize the essential properties of these transient physical state changes from raw data of different modes. Specifically, extracting the rate of change of local temperature gradient from thermal data can reflect the speed and intensity of local heat accumulation or dissipation; extracting the amplitude and frequency of cutting force fluctuations from mechanical data can reveal the stability of the tool-workpiece contact state and the presence of abnormal vibrations; and extracting the instantaneous energy increment of specific frequency bands from acoustic data can indicate the generation of microcracks or changes in friction states within the material. The aim of these feature extractions is to transform raw, complex time-series data into more discriminative numerical features.

[0084] In practical applications, calculating the correlation between the rate of change of local temperature gradient, the amplitude and frequency of cutting force fluctuations, and the instantaneous energy increment in a specific frequency band aims to quantify the interdependence and synchronicity of different physical phenomena during transient changes. For example, methods such as Pearson correlation coefficient, mutual information, or dynamic time warping can be used to assess the linear or nonlinear correlations between these characteristic sequences. Correlation analysis can identify cooperative change patterns among these characteristics; that is, when a certain transient event occurs, different physical quantities exhibit specific and repeatable linkage patterns. For example, a certain type of microcrack may always be accompanied by a specific combination of temperature gradient and cutting force fluctuations.

[0085] Furthermore, the identified cooperative change patterns are compared with a pre-built transient feature library. This library is a pre-established knowledge base that stores multimodal cooperative change patterns corresponding to different defect types (e.g., microcracks, chipping, sintering, stress concentration, etc.). These patterns can be trained and constructed using extensive experimental data, simulations, or expert experience. By matching the currently identified cooperative change pattern with known patterns in the library, it can be determined whether the current processing state matches a known anomaly or defect pattern, thus obtaining a comparison result. Ultimately, based on the comparison result, the system can determine whether there are missed reports (i.e., an anomaly actually occurred but the system failed to identify it) or false alarms (i.e., the system identified an anomaly but it did not actually occur), thereby obtaining a more accurate and reliable assessment of data acquisition reliability.

[0086] This application proposes a glass processing parameter optimization system based on edge computing, used to perform glass processing parameter optimization based on edge computing, combined with... Figure 3 As shown, the edge computing-based glass processing parameter optimization system 1 includes: Physical state acquisition module 11 is used to acquire physical state data generated during glass processing; Feature data analysis module 12 is used to perform feature analysis on physical state data to obtain feature data of the processing area; The physical state diagram generation module 13 is used to generate a physical state diagram that reflects the local heat distribution and internal stress distribution left on the glass sheet by the previous process, based on the processing area feature data and the physical properties of the glass material. The physical state diagram uses module 14 to obtain the processing position of the next process, and determines the corresponding local heat value and internal stress value from the physical state diagram based on the processing position. The processing parameter adjustment module 15 is used to adjust the initial processing parameters of the next process based on the local heat value and internal stress value, so as to obtain the adjusted processing parameters. The processing parameter execution module 16 is used to control the processing equipment to execute the next process according to the adjusted processing parameters.

[0087] In order to better understand the technical solutions proposed in this application, it is necessary to explain some key terms involved therein.

[0088] "Physical state data" refers to data collected by various sensors during the glass processing process that reflects the real-time physical state of the glass sheet and its processing area, such as temperature, stress, vibration, and acoustic signals. This data forms the basis for subsequent analysis and optimization.

[0089] "Machining area characteristic data" refers to key information extracted from the original physical state data that can characterize specific attributes or potential defect risks of the machining area, such as the area of ​​the local temperature rise zone, the rate of temperature change, the high-frequency oscillation component of the cutting force, the sudden change in force value, the energy distribution and instantaneous peak value of the acoustic signal, etc.

[0090] A "physical state diagram" is a data structure or visualization that integrates the characteristic data of the processing area and the physical properties of the glass material. It graphically reflects the local heat distribution and internal stress distribution left on the glass sheet by the previous process. This diagram can provide the "historical memory" of the glass sheet at a specific location, providing a basis for adjusting parameters in subsequent processes.

[0091] "Local heat value and internal stress value" refers to the heat and stress information extracted from the physical state diagram that precisely corresponds to the processing position of the next process. These values ​​directly reflect the current physical state of the processing position and are key inputs for adjusting the processing parameters of the next process.

[0092] "Initial machining parameters" refer to machining parameters set according to standard processes or experience before optimization and adjustment, such as tool speed, feed rate, depth of cut, coolant flow rate, etc.

[0093] "Adjusted processing parameters" refers to processing parameters that are more suitable for the current processing conditions, calculated by an optimization algorithm after taking into account local heat values ​​and internal stress values.

[0094] In some embodiments of this application, the various modules of the above system can be implemented in a variety of ways.

[0095] Specifically, the physical state acquisition module is used to acquire physical state data generated during glass processing. Various methods for acquiring physical state data have been described in the above embodiments and will not be repeated here. It is important to emphasize that this physical state acquisition module can be configured as a standalone hardware unit, such as a dedicated data acquisition card integrated into an edge computing unit. This card connects directly to various sensors via a physical interface and acquires data using a fixed sampling frequency and data transmission protocol. Alternatively, the physical state acquisition module can also be implemented as a software service running on the edge computing unit. This service communicates with the sensor management system via a network interface or API and periodically requests and receives sensor data according to a preset scheduling strategy.

[0096] The feature data analysis module is used to perform feature analysis on the physical state data to obtain the processing area feature data. Various methods for performing feature analysis on physical state data have been described in the above embodiments and will not be repeated here. It is important to emphasize that this feature data analysis module can be implemented as a software library containing a series of predefined algorithms, such as a set of functions for calculating statistical features, performing spectral analysis, or performing simple pattern matching. These algorithms, upon receiving the physical state data, perform calculations according to a predetermined process and output the processing area feature data. Alternatively, the feature data analysis module can also be implemented as a lightweight machine learning model, such as a support vector machine or decision tree model, which runs on an edge computing unit and performs feature extraction and classification on the input physical state data based on the trained model, thereby obtaining the processing area feature data.

[0097] The physical state diagram generation module generates a physical state diagram reflecting the local heat distribution and internal stress distribution left on the glass sheet by the previous process, based on the processing area feature data and the physical properties of the glass material. Various methods for generating physical state diagrams have been described in the above embodiments and will not be repeated here. It is important to emphasize that this physical state diagram generation module can be implemented as a physical model-based simulator. This simulator receives the processing area feature data and glass material parameters as input, performs calculations using preset heat transfer and mechanical equations, and generates a two-dimensional or three-dimensional mesh data structure to represent the heat and stress distribution. Alternatively, the physical state diagram generation module can also be implemented as a data mapping service. This service matches the processing area feature data with pre-stored physical state templates generated based on historical data or offline simulation, and generates the corresponding physical state diagram based on the matching results.

[0098] The physical state diagram module is used to obtain the processing position of the next process step and, based on the processing position, determine the corresponding local heat value and internal stress value from the physical state diagram. The above embodiments have already described various methods for obtaining the processing position and determining the corresponding heat and stress values, which will not be repeated here. It is important to emphasize that this physical state diagram module can be implemented as a coordinate transformation and query interface. This interface receives the coordinates of the processing position of the next process step, maps them to the coordinate system of the physical state diagram, and then extracts the corresponding local heat and internal stress values ​​from the physical state diagram through direct indexing or interpolation algorithms. Alternatively, the physical state diagram module can also be implemented as a graphical user interface component, allowing the operator to select the processing position on a visual interface and obtain the heat and stress information of the corresponding area through clicking or box selection.

[0099] The machining parameter adjustment module is used to adjust the initial machining parameters for the next process based on the local heat value and internal stress value, thus obtaining the adjusted machining parameters. Various methods for adjusting machining parameters have been described in the above embodiments and will not be repeated here. It is important to emphasize that this machining parameter adjustment module can be implemented as a rule-based expert system. This system adjusts the initial machining parameters according to a preset set of condition-action rules, such as "if the local heat value exceeds X, then reduce the feed rate Y%". Alternatively, the machining parameter adjustment module can also be implemented as an optimization algorithm engine. This engine aims at machining quality or efficiency, using the local heat value and internal stress value as constraints, and determines the optimal combination of machining parameters through iterative calculation or optimization algorithms.

[0100] The machining parameter execution module is used to control the machining equipment to execute the next process according to the adjusted machining parameters. Various methods for controlling the machining equipment to execute processes have been described in the above embodiments and will not be repeated here. It is important to emphasize that this machining parameter execution module can be implemented as an industrial communication interface. This interface exchanges data with the CNC system of the machining equipment through standard industrial Ethernet protocols (such as Profinet, EtherCAT) or serial communication protocols (such as RS-232 / 485), sending the adjusted machining parameter instructions to the CNC system. Alternatively, the machining parameter execution module can also be implemented as an analog output unit. This unit converts the digitized adjustment parameters into analog voltage or current signals, directly driving the actuators in the machining equipment, such as frequency converters or servo controllers.

[0101] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for optimizing glass processing parameters based on edge computing, characterized in that, include: Acquire physical state data generated during glass processing; The physical state data is subjected to feature analysis to obtain the processing area feature data; Based on the processing area feature data and the physical properties of the glass material, a physical state diagram is generated to reflect the local heat distribution and internal stress distribution left on the glass sheet by the previous process. Obtain the processing position of the next process, and determine the corresponding local heat value and internal stress value from the physical state diagram based on the processing position; Based on the local heat value and the internal stress value, the initial processing parameters for the next process are adjusted to obtain the adjusted processing parameters; Control the processing equipment to execute the next process according to the adjusted processing parameters.

2. The glass processing parameter optimization method based on edge computing according to claim 1, characterized in that, The step of generating a physical state diagram reflecting the local heat distribution and internal stress distribution left on the glass sheet by the previous process, based on the characteristic data of the processing area and the physical properties of the glass material, includes: Obtain a sequence of surface temperature distribution images of the glass sheet processing area; From the surface temperature distribution image sequence, extract the curves of temperature values ​​at specific points on the glass plate and the corresponding preset areas as a function of time; the preset area refers to a neighborhood area determined based on a preset neighborhood radius. Based on the curve of temperature change over time, the rate at which heat diffuses outward from a specific point is calculated, and the heat diffusion rate is obtained. Based on the heat diffusion rate, identify the heat conduction characteristics of the local area to which a specific point belongs; Based on the heat conduction characteristics, the thermal diffusivity used to estimate the local heat distribution is adjusted to obtain the adjusted thermal diffusivity. Based on the adjusted thermal diffusivity, the characteristic data of the processing area, and the physical properties of the glass material, the local heat distribution and internal stress distribution are calculated. Based on the local heat distribution and the internal stress distribution, a physical state diagram is generated to reflect the local heat distribution and internal stress distribution left on the glass plate by the previous process.

3. The glass processing parameter optimization method based on edge computing according to claim 1, characterized in that, The step of adjusting the initial processing parameters of the next process based on the local heat value and the internal stress value to obtain the adjusted processing parameters includes: Obtain and determine the processing type of the next process; Based on the processing type, the local heat value, and the internal stress value, a parameter adjustment strategy is selected; Based on the parameter adjustment strategy, calculate the adjustment amount of multiple related processing parameters; Based on the adjustment amount, the initial processing parameters for the next process are adjusted to generate the adjusted processing parameters.

4. The glass processing parameter optimization method based on edge computing according to claim 1, characterized in that, The steps for acquiring physical state data generated during glass processing include: Collect initial physical state data of the glass processing area; The initial physical state data is subjected to interference identification to obtain interference identification results; Based on the interference identification results, the filtering parameters are adjusted or signal reconstruction is initiated to remove the interference components in the initial physical state data and obtain the interference-free physical state data. The signal reconstruction is a process of recovering the interfered data segments based on adjacent time window data, adjacent spatial location data, or multi-sensor homogeneous data when the initial physical state data has missing segments, local saturation, or abnormal peak interference. Environmental compensation calibration is performed on the interference-free physical state data to obtain calibrated physical state data. The calibrated physical state data is used as the physical state data generated during the glass processing.

5. The glass processing parameter optimization method based on edge computing according to claim 1, characterized in that, The step of performing feature analysis on the physical state data to obtain the processing area feature data includes: Obtain an image of the surface temperature distribution in the glass sheet processing area; Calculate the area and rate of temperature change of the local temperature surge region in the surface temperature distribution image; Obtain cutting force data when the tool contacts the glass; Analyze the high-frequency oscillation components and abrupt force changes in the cutting force data; Acquire acoustic signals generated during the processing; Analyze the energy distribution and instantaneous peak value of the acoustic signal in a specific frequency band; By integrating the area of ​​the local temperature surge region, the temperature change rate of the local temperature surge region, the high-frequency oscillation component, the force value mutation, the energy distribution, and the instantaneous peak value, characteristic data reflecting the defect risk of the processing area are obtained. Feature data reflecting the risk of defects in the processing area will be used as the feature data of the processing area.

6. The glass processing parameter optimization method based on edge computing according to claim 5, characterized in that, The step of integrating the area of ​​the local temperature surge region, the temperature change rate of the local temperature surge region, the high-frequency oscillation component, the force value mutation, the energy distribution, and the instantaneous peak value to obtain characteristic data reflecting the defect risk of the processing area includes: Read the surface temperature distribution image, the cutting force data, and the acoustic signal; Add timestamps and spatial location markers to the surface temperature distribution image, the cutting force data, and the acoustic signal; Based on the timestamp and the spatial location marker, and using a unified time-space grid, the surface temperature distribution image, the cutting force data, and the acoustic signal are synchronized. Spatial interpolation processing is performed on the synchronized surface temperature distribution image, cutting force data, and acoustic signal to ensure that the data at the same time point corresponds to the same local area of ​​the glass plate. Based on the surface temperature distribution image after spatial interpolation, cutting force data, and acoustic signal, the area, temperature change rate, high-frequency oscillation component, force change, energy distribution, and instantaneous peak value of the extracted local temperature rise region are aligned and calibrated at the same time point and in the same local region to obtain the updated area, temperature change rate, high-frequency oscillation component, force change, energy distribution, and instantaneous peak value of the local temperature rise region. The updated area of ​​the local temperature surge region, the rate of temperature change, the high-frequency oscillation component, the force value mutation, the energy distribution, and the instantaneous peak value are fused to obtain characteristic data reflecting the defect risk of the processing area.

7. The glass processing parameter optimization method based on edge computing according to claim 5, characterized in that, The step of integrating the area of ​​the local temperature surge region, the temperature change rate of the local temperature surge region, the high-frequency oscillation component, the force value mutation, the energy distribution, and the instantaneous peak value to obtain characteristic data reflecting the defect risk of the processing area includes: Obtain and determine the type of processing operation currently being performed on the glass sheet; Based on the processing procedure type, obtain the defect pattern definition corresponding to the processing procedure type from the preset process defect pattern library; Based on the defect mode definition, a feature weight configuration is generated; the feature weight configuration is used to define the area of ​​the local temperature rise region, the temperature change rate, the high-frequency oscillation component, the force value mutation, the energy distribution, and the relative importance of the instantaneous peak value under the current processing procedure type; Based on the aforementioned feature weight configuration, the area of ​​the local temperature surge region, the rate of temperature change, the high-frequency oscillation component, the force value mutation, the energy distribution, and the instantaneous peak value are weighted and fused to obtain a weighted fusion result; Based on the weighted fusion results, characteristic data reflecting the defect risk of the processing area are obtained.

8. The glass processing parameter optimization method based on edge computing according to claim 1, characterized in that, The steps for acquiring physical state data generated during glass processing include: Multiple sensors are deployed in the glass sheet processing area, and a hardware-level time synchronization triggering mechanism is adopted to enable the sensors to collect data synchronously. The multiple sensors include high frame rate infrared thermal imaging sensors, high sampling rate force sensors, and high sensitivity acoustic sensors. Each sensor is equipped with an edge preprocessing unit. After the sensor completes data acquisition, the edge preprocessing unit performs preliminary noise reduction and formatting on the acquired data to obtain preprocessed data. A data buffer is set up in the edge computing unit to receive and temporarily store preprocessed data from various sensors; Based on the current high-speed or high-precision mode of the processing equipment, the sampling frequency and data transmission bandwidth of each sensor are dynamically adjusted. The data collected after adjusting the sampling frequency and data transmission bandwidth of each sensor is processed by the edge preprocessing unit and then written into the data buffer. The high-speed mode is the operating mode in which the feed speed or spindle speed of the processing equipment is higher than a preset speed threshold and the processing cycle priority is higher than the precision priority. The high-precision mode is the operating mode in which the processing tolerance requirements, positioning accuracy requirements, or surface quality requirements of the processing equipment are higher than a preset precision threshold and the precision priority is higher than the processing cycle priority. In the edge computing unit, when the processing equipment is detected to enter the micro-feature processing area, a high-frequency data acquisition mode is activated, and time series analysis is performed on the preprocessed data written to the data buffer to identify transient physical state changes; wherein, the micro-feature processing area is a processing area on the glass plate whose processing width, hole diameter, groove width, chamfer size, or local radius of curvature is smaller than the corresponding preset size threshold. Transient features are extracted based on the identified transient physical state changes, and the extracted transient features are compared with a preset transient feature library to determine whether there are any missed or false alarms, thus obtaining the data acquisition reliability judgment result. When the reliability assessment result indicates that there are missed or false alarms, adjust the sensor gain or focal length and re-acquire the data to obtain the re-acquired data; When the data acquisition reliability judgment result indicates that there are no missed or false alarms, the data corresponding to the transient characteristics will be marked as data that has not been judged as missed or false alarms. Based on the re-collected data or data that was not identified as a missed or false alarm, physical state data generated during the glass processing are obtained.

9. The glass processing parameter optimization method based on edge computing according to claim 8, characterized in that, The steps of extracting transient features based on the identified transient physical state changes, comparing the extracted transient features with a preset transient feature library, determining whether there are missed or false alarms, and obtaining the data acquisition reliability judgment result include: Transient feature extraction is performed based on the identified transient physical state changes; wherein, the transient physical state change is a processing state change event characterized by a sudden change or coordinated change in at least one of thermal data, mechanical data, and acoustic data within a preset short time window; the transient feature extraction includes extracting the local temperature gradient change rate from the thermal data corresponding to the transient physical state change, extracting the cutting force fluctuation amplitude and frequency from the corresponding mechanical data, and extracting the instantaneous energy increment of a specific frequency band from the corresponding acoustic data; Calculate the correlation between the local temperature gradient change rate, the cutting force fluctuation amplitude and frequency, and the instantaneous energy increment in the specific frequency band; Based on the correlation, a cooperative change pattern is identified among the local temperature gradient change rate, the cutting force fluctuation amplitude and frequency, and the instantaneous energy increment in the specific frequency band; The identified collaborative change patterns are compared with a preset transient feature library to obtain the comparison results; the transient feature library contains multimodal collaborative change patterns corresponding to different defect types. Based on the comparison results, it is determined whether there are any missed or false alarms, and the reliability of the data collection is judged.

10. A glass processing parameter optimization system based on edge computing, used to perform glass processing parameter optimization based on edge computing, characterized in that, include: The physical state acquisition module is used to acquire physical state data generated during glass processing; The feature data analysis module is used to perform feature analysis processing on the physical state data to obtain the processing area feature data; The physical state diagram generation module is used to generate a physical state diagram that reflects the local heat distribution and internal stress distribution left on the glass sheet by the previous process, based on the processing area feature data and the physical properties of the glass material. The physical state diagram module is used to obtain the processing position of the next process and, based on the processing position, to determine the corresponding local heat value and internal stress value from the physical state diagram. The processing parameter adjustment module is used to adjust the initial processing parameters of the next process according to the local heat value and the internal stress value, so as to obtain the adjusted processing parameters. The processing parameter execution module is used to control the processing equipment to execute the next process according to the adjusted processing parameters.

Images