Glass arch detection method and related device

Abstract

The present application provides a glass arch detection method and related devices. After being processed in the current process link of the tempering process, the present application obtains the real-time surface parameters of the target object by infrared detection; based on the real-time surface parameters, the surface morphology distribution data of the target object is generated; based on the surface morphology distribution data, the deviation between the target object and the standard surface morphology is obtained; if the deviation meets the deviation abnormality condition, the transmission process of the target object is stopped, and the target object is detected through the abnormality detection model to trigger the automatic processing mechanism for matching the candidate abnormal cause; if the deviation does not meet the deviation abnormality condition, the target object is transferred from the current process link to the next process link. The present application can dynamically predict and control the walking program parameters to control the thermal cycle speed of the target equipment and improve production efficiency and glass quality.

Description

Technical Field

[0001] The present application relates to the technical field of glass processing, and in particular to a glass arch detection method and related devices. Background Art

[0002] Glass tempering is a process that uses special treatment to make ordinary glass have higher strength and safety. During the heating process of the glass tempering process, the uneven temperature distribution inside the tempering furnace is a key factor affecting the flatness of the glass. If the glass is overheated locally, the expansion of this part of the glass will be greater than that of other parts, resulting in uneven thermal stress. When the glass enters the cooling stage, this uneven thermal stress is released, which can easily cause the glass to form an arch and increase the degree of arch. For example, in a large tempering furnace, if the heating element fails or the heating power is improperly adjusted, the temperature in the center area of ​​the glass is too high, and the center may arch upward after cooling, resulting in the waveform or arch not meeting the requirements.

[0003] In the related technologies, it is impossible to detect the arch or waveform immediately after the tempering furnace comes out of the furnace. Manual inspection takes a long time, which easily leads to unqualified products flowing into the subsequent links. During the tempering process, parameters such as temperature control, wind pressure, and placement cannot be dynamically adjusted based on real-time data, and the debugging cycle is long, which affects production efficiency. The detection data cannot be effectively integrated with the adjustment system, and a closed loop of continuous optimization cannot be formed.

[0004] Therefore, it is urgent to design a new technical solution to solve at least one of the above technical problems. Summary of the invention

[0005] In order to solve or improve the above problems, the embodiments of the present application provide a glass arch detection method and related devices.

[0006] In one embodiment of the present application, a method for detecting glass arch degree is provided, comprising:

[0007] After being processed in the current process link of the tempering process, the real-time surface parameters of the target object are obtained by infrared detection; the real-time surface parameters at least include: the local real-time temperature of each surface area of ​​the target object; the material of the target object is glass;

[0008] Generate surface morphology distribution data of the target object according to the surface real-time parameters; the data dimensions in the surface morphology distribution data at least include: local arch and waviness of each surface area in the target object;

[0009] Based on the surface morphology distribution data, obtaining a degree of deviation between the target object and a standard surface morphology;

[0010] If the deviation degree satisfies the deviation abnormality condition, the transmission process of the target object is stopped, and the target object is detected by an abnormality detection model to trigger an automatic processing mechanism for matching the candidate abnormality causes;

[0011] If the deviation degree does not satisfy the deviation abnormality condition, the target object is transferred from the current process link to the next process link.

[0012] In another embodiment of the present application, a glass dome detection device is provided, comprising at least the following units:

[0013] A detection unit is used to obtain the real-time surface parameters of the target object by infrared detection after the current process link of the tempering process; the real-time surface parameters at least include: the local real-time temperature of each surface area of ​​the target object; the material of the target object is glass;

[0014] A generating unit, configured to generate surface morphology distribution data of the target object according to the surface real-time parameters; the data dimensions of the surface morphology distribution data at least include: local camber and waviness of each surface area of ​​the target object;

[0015] An acquisition unit, configured to acquire a degree of deviation between the target object and a standard surface morphology based on the surface morphology distribution data;

[0016] An exception processing unit, configured to stop the transmission process of the target object if the deviation degree satisfies the deviation exception condition, and detect the target object through an exception detection model to trigger an automatic processing mechanism for matching the candidate exception causes;

[0017] A transmission unit is used to transmit the target object from the current process link to the next process link if the deviation degree does not meet the deviation abnormality condition.

[0018] In one embodiment of the present application, an electronic device includes:

[0019] at least one processor, memory, and input-output unit;

[0020] The memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the glass arch detection method in the first embodiment.

[0021] In one embodiment of the present application, a computer-readable storage medium is provided, which includes instructions. When the instructions are executed on a computer, the computer executes the glass dome detection method in the first embodiment.

[0022] In the technical solution provided by the embodiment of the present application, first, after being processed in the current process link of the tempering process, the surface real-time parameters of the target object are obtained by infrared detection; the surface real-time parameters at least include: the local real-time temperature of each surface area in the target object; the material of the target object is glass. Then, according to the surface real-time parameters, the surface morphology distribution data of the target object is generated; the data dimensions in the surface morphology distribution data at least include: the local arch and waviness of each surface area in the target object. Then, based on the surface morphology distribution data, the deviation between the target object and the standard surface morphology is obtained. Then, if the deviation meets the deviation abnormality condition, the transmission process of the target object is stopped, and the target object is detected by the abnormality detection model to trigger the automatic processing mechanism of matching the candidate abnormal cause. If the deviation does not meet the deviation abnormality condition, the target object is transferred from the current process link to the next process link. In the embodiment of the present application, the processing results of the tempering process are monitored in real time through infrared real-time detection and intelligent analysis through the abnormality detection model, so as to realize online monitoring and real-time regulation of glass flatness, and improve glass production efficiency and glass quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1 A schematic diagram of a process for detecting glass arch degree provided in an embodiment of the present application;

[0025] Figure 2 A schematic diagram of the structure of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0026] The present application provides the following embodiments to solve or improve at least some of the above problems. In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present application.

[0027] In some features described in the specification, claims and the above-mentioned drawings of this application, descriptions such as "first" and "second" are used to distinguish different components, parts, modules, devices, etc., and do not represent the order of precedence, nor do they limit "first" and "second" to be different types. In addition, the embodiments described below are only some embodiments of the present application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of this application.

[0028] In order to solve the above technical problems, a glass arch detection device provided in an embodiment of the present application includes the following units:

[0029] A detection unit is used to obtain the real-time surface parameters of the target object by infrared detection after the current process link of the tempering process; the real-time surface parameters at least include: the local real-time temperature of each surface area of ​​the target object; the material of the target object is glass;

[0030] A generating unit, configured to generate surface morphology distribution data of the target object according to the surface real-time parameters; the data dimensions of the surface morphology distribution data at least include: local camber and waviness of each surface area of ​​the target object;

[0031] An acquisition unit, configured to acquire a degree of deviation between the target object and a standard surface morphology based on the surface morphology distribution data;

[0032] An exception processing unit, configured to stop the transmission process of the target object if the deviation degree satisfies the deviation exception condition, and detect the target object through an exception detection model to trigger an automatic processing mechanism for matching the candidate exception causes;

[0033] A transmission unit is used to transmit the target object from the current process link to the next process link if the deviation degree does not meet the deviation abnormality condition.

[0034] The above-mentioned device has relevant functions or modules corresponding to the specific implementation methods involved in the method part of the embodiment of this application or other embodiments.

[0035] In the above device, the following is a detailed introduction to the functions and principles of each unit of the device:

[0036] In the glass tempering process, infrared detection is first used to obtain real-time parameters of the surface. When the current process is completed, the non-contact and fast characteristics of infrared detection are used to accurately scan the local real-time temperature of each surface area of ​​the glass. These temperature data are the cornerstone of subsequent analysis and can reflect the current thermal state of the glass.

[0037] Next, the surface morphology distribution data is generated. Based on the obtained temperature and other parameters, with the help of specific algorithms and models, combined with knowledge such as the thermal expansion characteristics of glass, the local arch and wave shape of each area are calculated, thereby integrating the complete surface morphology distribution data and intuitively presenting the details of the glass surface morphology.

[0038] Then, the deviation degree from the standard surface morphology is obtained. With the preset standard surface morphology as a reference, the degree of deviation of the glass surface morphology from the standard in terms of arch degree, wave degree and other dimensions is obtained through comparative analysis, and it is quantified as a deviation degree value or index system.

[0039] Finally, the subsequent process is determined based on the degree of deviation. If the degree of deviation meets the deviation abnormality condition, such as exceeding the threshold or multi-dimensional deviation exceeding the limit, it indicates that the surface morphology is abnormal. At this time, the glass transmission is stopped immediately, and the abnormality detection model and the automatic processing mechanism for matching candidate abnormal causes are started to check factors such as temperature, wind pressure, and placement method, and adjust the processing until the deviation is qualified or it is determined that it cannot be repaired; if the degree of deviation does not meet the deviation abnormality condition, it means that the surface morphology of the glass meets the standard and can be smoothly transferred to the next process link to continue processing, packaging and other processes. This series of processes effectively guarantees the surface morphology quality and production orderliness of glass tempered products through real-time detection, analysis and judgment.

[0040] It should be added here that the control logic of the relevant device will be introduced in detail in the method embodiment below.

[0041] An embodiment of the present application provides a method for detecting glass arch. The method of this embodiment can be performed by the glass arch detection device in the above embodiment, and is implemented on the basis of the structure of the glass arch detection device in the above embodiment. Figure 1 As shown, the method may include the following steps:

[0042] 101. After being processed in the current process link of the tempering process, the real-time surface parameters of the target object are obtained by infrared detection;

[0043] 102. Generate surface morphology distribution data of the target object according to the real-time surface parameters;

[0044] 103. Based on the surface morphology distribution data, obtain a deviation between the target object and a standard surface morphology;

[0045] 104. If the deviation degree satisfies the deviation abnormality condition, the transmission process of the target object is stopped, and the target object is detected by an abnormality detection model to trigger an automatic processing mechanism for matching the candidate abnormality causes;

[0046] 105. If the deviation degree does not meet the deviation abnormality condition, the target object is transferred from the current process link to the next process link.

[0047] In the embodiment of the present application, the surface real-time parameters at least include: local real-time temperature of each surface area of ​​the target object. The material of the target object is glass.

[0048] In the embodiment of the present application, glass is used as the target object for relevant process processing and parameter monitoring. Among them, the surface real-time parameter focuses on the local real-time temperature of each surface area of ​​the glass, which is of vital importance. During the tempering process, the glass undergoes heating and cooling, and the temperature changes in different surface areas directly affect its internal stress distribution and the final physical form. For example, during the heating stage, if the temperature of a certain area is too high or the heating rate is too fast, its thermal expansion degree will be greater than that of other areas, which may cause uneven contraction of the glass when cooling, thereby causing changes in the arch or waviness, affecting the flatness of the glass. In the cooling process, the same is true, and local temperature differences will cause different shrinkage stresses on the glass surface, which may cause deformation. By accurately measuring the local real-time temperature of each surface area, the thermal state changes of the glass in the tempering process can be understood in real time. This provides key basic data for the subsequent generation of surface morphology distribution data. For example, by utilizing the principles of heat conduction and mathematical models related to the thermal properties of glass materials, combined with these temperature data, it is possible to calculate the differences in thermal expansion in different areas of the glass, and further analyze the possible trends in changes in arch and wave shape, thereby providing a basis for accurately controlling tempering process parameters and ensuring the quality of glass products.

[0049] In the embodiment of the present application, the data dimensions in the surface morphology distribution data at least include: the local arch degree and waviness degree of each surface area in the target object.

[0050] Local camber refers to the degree to which a certain area of ​​the glass surface bulges upward or downward relative to the ideal plane. It is one of the key indicators for measuring the flatness of glass. In practical applications, such as building glass curtain walls, excessive camber will affect the installation accuracy and overall aesthetics of the glass, and may cause stress concentration and reduce the safety of the glass; in the field of optical instruments, precise local camber control is crucial for the refraction and reflection of light, and directly affects the optical performance of the instrument.

[0051] In the glass tempering process, the formation of local arch is closely related to the heating and cooling links. In the heating stage, if the temperature distribution in the tempering furnace is uneven, the glass is heated too much locally, and this part of the material expands greatly. When entering the cooling stage, this local area will be arched due to cooling and contraction. For example, if the temperature in the center of the glass is too high, it may arch upward after cooling, resulting in an increase in the local arch. In addition, the imbalance of the rapid cooling wind pressure will also affect the local arch. If the wind pressure on the upper surface of the glass is greater than the wind pressure on the lower surface, the upper surface cools faster and shrinks more, and the glass may arch downward, changing the local arch.

[0052] Through advanced detection methods, such as high-precision laser displacement sensors or detection equipment based on the principle of optical interference, the distance change of each local area of ​​the glass relative to the reference plane can be measured to obtain local arch data. These data can help operators or automatic control systems understand the state of the glass surface in real time. When the local arch exceeds the preset standard range, the process parameters can be adjusted in time, such as adjusting the heating temperature distribution or wind pressure balance, to correct the deformation of the glass.

[0053] Waviness describes the degree to which the glass surface presents a wavy shape. It reflects the change in the flatness of the glass surface in the horizontal direction. For application scenarios that require high-precision flatness, such as high-precision display screens, precision optical lenses, etc., the waviness must be controlled within a very small range. Poor waviness can cause problems such as light scattering and image distortion, affecting the use of the product.

[0054] In the glass tempering process, the wave shape is mainly caused by the uneven stress distribution of the glass during heating and cooling. For example, the failure of the heating element or the unreasonable layout in the tempering furnace may cause the glass to have a temperature gradient in the horizontal direction, making the thermal expansion and contraction of different positions of the glass inconsistent, thus forming a wave-like deformation. At the same time, the abnormal local wind pressure of the wind grid will also aggravate this situation. If the wind pressure distribution of the wind grid in the horizontal direction is uneven, the cooling speed of different areas on the glass surface is different, which is easy to cause wave deformation.

[0055] The waveform can be detected by using optical profilometers and other equipment. By scanning the glass surface, the surface height change data in the horizontal direction can be obtained, and then the waveform can be calculated. These data together with the local arch data constitute the surface morphology distribution data. When the waveform data shows abnormalities, it can provide a basis for troubleshooting, such as checking the uniformity of the heating furnace and the air grille in the horizontal direction, and adjusting the process parameters in a targeted manner to ensure that the flatness of the glass surface meets the requirements.

[0056] The following is a detailed introduction to the principles and effects of each step:

[0057] The following is a detailed introduction to the above content:

[0058] After the current process link of the glass tempering process is completed, infrared detection technology is used to collect the real-time surface parameters of the target object (i.e. glass). Infrared detection has the advantages of non-contact and rapid acquisition of temperature information, and can accurately detect the local real-time temperature of each surface area of ​​the glass. For example, after the glass comes out of the tempering furnace for heating and cooling, the infrared detection equipment will scan the entire surface of the glass to obtain the temperature values ​​corresponding to different position areas. These values ​​constitute the basic data for subsequent analysis and can reflect the thermal status of different parts of the glass surface at the current moment.

[0059] Based on the obtained local real-time temperature and other surface real-time parameters of each surface area, the surface morphology distribution data of the target object is further generated. The data dimensions here include key indicators such as the local arch and waviness of each surface area in the target object. Through specific algorithms and models, combined with temperature data and other relevant information (which may involve knowledge related to physical properties such as the thermal expansion characteristics of glass), the arch and waviness of different areas are calculated. For example, the temperature of a certain area is higher than the surrounding area, and the degree of thermal expansion is different. After calculation, the corresponding arch change value of the area and whether there is waveform deformation can be obtained. Finally, the complete surface morphology distribution data is integrated to intuitively present the specific situation of the glass surface morphology.

[0060] The generated surface morphology distribution data is compared with the preset standard surface morphology to calculate the deviation between the target object and the standard surface morphology. The standard surface morphology is the ideal glass surface morphology determined based on product quality requirements and application scenarios, with clear ranges of indicators such as arch and waviness. Through comparative analysis, the degree to which the current glass surface morphology deviates from the standard in various dimensions is obtained, such as how much the arch exceeds the standard range, whether the waviness meets the requirements, etc., and these deviations are quantified into a deviation value or a deviation indicator system.

[0061] If the calculated deviation reaches the set deviation abnormality condition (for example, the deviation value exceeds a certain threshold, or the deviations of multiple dimensions exceed their respective allowable ranges at the same time), it means that the surface morphology of the current glass has an abnormality that does not meet the requirements. At this time, the transmission process of the target object (glass) will be stopped immediately to prevent the glass with quality problems from entering the next link and causing more adverse effects in the future. Subsequently, the target object is detected through the anomaly detection model, and the automatic processing mechanism for matching candidate abnormal causes is started. This mechanism will follow the established process to check factors that may cause abnormalities, such as temperature distribution, wind pressure imbalance, glass placement, etc., and make corresponding adjustments and processing, such as the above-mentioned series of operations such as checking the source of abnormalities, partition adjustment parameters, closed-loop feedback, and alarm prompts, until the deviation of the glass surface morphology meets the requirements or is determined to be irreparable.

[0062] If the calculated deviation does not reach the abnormal deviation condition, it means that the current surface morphology of the glass meets the quality standards and can meet the subsequent production and application requirements. Then, the target object (glass) will be smoothly transferred from the current process link to the next process link, and the subsequent processing, packaging and other processes will continue to ensure the orderly progress of the entire glass tempering production process.

[0063] In summary, the above steps can accurately control the surface morphology quality of the glass after tempering through real-time detection, analysis and judgment, handle abnormal situations in a timely manner, and ensure the production of qualified tempered glass products.

[0064] As an optional embodiment, in 101, before obtaining the real-time surface parameters of the target object by infrared detection, multiple infrared sensors may be deployed at the exit of the current process link to form an infrared sensor array; the multiple infrared sensors are high-resolution sensors. Furthermore, in 101, obtaining the real-time surface parameters of the target object by infrared detection includes: real-time scanning of the glass surface temperature data of the target object; the glass surface temperature data is a two-dimensional data matrix obtained by real-time scanning using an infrared sensor array.

[0065] Careful layout is carried out at the exit of the current glass tempering process, and multiple high-resolution infrared sensors are deployed, which together constitute an infrared sensor array. The reason for choosing to deploy at the exit is that the glass here has just completed the current link processing, and timely detection can obtain its current status at the first time, providing the most immediate data support for subsequent analysis and adjustment. The high-resolution sensor is selected to accurately and meticulously capture the subtle changes in the temperature of the glass surface. For example, high resolution allows the sensor to accurately distinguish the temperature difference of every small area on the glass surface, and even small local temperature changes can be accurately detected, which is very critical for the subsequent accurate judgment of the glass surface morphology.

[0066] The deployed infrared sensor array is used to scan the surface of the target object (i.e., glass) in real time to obtain the temperature data of the glass surface. The data obtained here is presented in the form of a two-dimensional data matrix. In principle, the infrared sensor array is like a dense "monitoring network" covering the corresponding spatial position on the glass surface. Each sensor is responsible for detecting the temperature of its corresponding area. When many sensors work at the same time, it is equivalent to a comprehensive and detailed "scan" of the entire surface of the glass. For example, imagine the glass surface as a two-dimensional plane. Each sensor in the sensor array is like a coordinate point, each obtaining the temperature value of the corresponding position, and finally integrating them to form a complete two-dimensional data matrix, in which the rows and columns of the matrix can correspond to different coordinate positions on the glass surface, and the element value in the matrix is ​​the temperature value of the corresponding position.

[0067] Further optionally, the infrared sensors are arranged in a regular planar array above or on the side of the exit of the glass tempering process, perpendicular to the glass surface or at a certain angle to the glass surface to ensure that the plane area of ​​the glass can be completely covered. For example, the sensors can be arranged in a rectangular or square array, and the spacing between the sensors is determined according to the required temperature detection resolution. If you want to obtain a higher resolution, reduce the sensor spacing; if the resolution requirement is lower, you can increase the spacing appropriately. This layout can evenly cover the glass surface and fully obtain the temperature distribution information on the glass plane. It is very effective for detecting local temperature differences and overall temperature distribution on the glass surface, and can be well used to analyze the temperature gradient of the glass in the plane direction, and then determine whether the glass has waveform deformation and the like.

[0068] There may be some limitations in the temperature detection of the edge of the glass, especially in the corner area near the edge. Because the arrangement of sensors is mainly for the plane area, the detection accuracy of the edge part may be relatively low. It is necessary to increase the sensor density appropriately in the edge area or adopt a special edge detection strategy to compensate. Surround the glass transmission path with infrared sensors to form a cylindrical detection area, allowing the glass to pass through the middle. Sensors can be distributed at different heights and angles to ensure that the temperature of the glass surface can be obtained from multiple directions. For example, sensors are arranged in the upper and lower directions, left and right directions of the glass. By combining temperature data from different directions, all-round temperature information of the glass surface can be obtained. In this way, the temperature of each part of the glass surface can be fully detected, including the edge and corner areas. This deployment method is very beneficial for irregularly shaped glass or situations where high-precision edge temperature detection is required. It can provide more complete temperature data, which is conducive to accurately judging the thermal state changes of the glass in all directions, and then more accurately analyzing the reasons for the changes in the morphology of the glass surface.

[0069] Surround deployment systems are relatively complex and costly. Since data needs to be acquired from multiple directions, data processing and integration are also difficult. Moreover, sensors in different directions may be affected by environmental factors (such as reflected light, interference from other heat sources, etc.), and measures need to be taken to calibrate and compensate.

[0070] According to the key areas where temperature anomalies are prone to occur in the glass tempering process, infrared sensors are deployed around these areas in a targeted manner. For example, sensors are placed near the exit of the heating furnace and near the entrance of the wind grille, and a certain number of sensors are also set in the center and edge areas of the glass. This deployment method is based on the process characteristics and the areas where glass problems are prone to occur.

[0071] Resources can be concentrated to perform high-precision temperature detection on key areas, and abnormal temperature points in the process can be discovered in time. This is very helpful for optimizing glass tempering process parameters, especially for adjusting specific quality issues. Under limited costs, the role of sensors can be maximized, focusing on areas that have a greater impact on the surface morphology of the glass.

[0072] Since the sensors are mainly concentrated in key areas, the temperature detection of non-key areas may not be comprehensive. When the causes of abnormal glass surface morphology are complex or involve temperature changes in non-key areas, there may be detection blind spots, which require the combination of other detection methods or comprehensive judgment based on experience.

[0073] This glass surface temperature data, presented in the form of a two-dimensional data matrix, contains rich information and can intuitively reflect the temperature distribution of each area on the glass surface. Based on this data, combined with corresponding algorithms and models, the morphological distribution data of the glass surface can be further calculated, such as key indicators such as local arch and waveform, so as to accurately evaluate the quality of the glass, and timely discover possible abnormalities and make corresponding treatments to ensure that the glass tempering process produces high-quality products.

[0074] Further optionally, glass is a material with a relatively stable thermal expansion coefficient. When the glass is heated, different areas of the glass will expand to different degrees depending on the temperature. The expansion amount of different areas can be calculated based on the temperature data obtained by the infrared sensor array according to the thermal expansion formula. For example, if the sensor array detects that the temperature of the central area of ​​the glass is higher than that of the edge area, according to the principle of thermal expansion, the central area will expand more than the edge area. During the cooling process, this expansion difference will cause the surface morphology of the glass to change. If the cooling speed is fast, this uneven expansion and contraction may cause the glass to have an arch or wave shape. By analyzing the thermal expansion difference reflected by the temperature data, it can be preliminarily determined whether the glass surface may be deformed and the approximate direction of the deformation.

[0075] Temperature gradient refers to the rate of change of temperature in space. On the glass surface, the size and direction of the temperature gradient reflect the direction and intensity of heat transfer. By analyzing the two-dimensional temperature data matrix obtained by the infrared sensor array, the temperature gradient at different locations on the glass surface can be calculated. For example, the temperature difference between adjacent sensors in the horizontal and vertical directions is calculated to obtain the temperature gradient.

[0076] When the temperature gradient is large, it indicates that the heat transfer in this area is uneven, and the thermal expansion and contraction of the glass in this area will also be uneven. This unevenness often causes a wave-like change on the glass surface. If there is a large temperature gradient in the vertical direction, and the temperature gradients on the upper and lower surfaces are inconsistent during the cooling process, it may cause the glass to deform into an arch. By analyzing the temperature gradient, the possible deformation of the glass surface can be judged more carefully.

[0077] A large amount of experimental data and numerical simulation results can be used to establish a mathematical model between the temperature distribution and surface morphology of the glass surface. In this model, the input is the temperature data (two-dimensional data matrix) obtained by the infrared sensor array, and the output is the morphological parameters of the glass surface, such as local arch and waviness. This model can be constructed based on physical principles, such as thermoelastic equations, combined with the material properties of glass (such as elastic modulus, Poisson's ratio, etc.).

[0078] For example, the temperature data obtained in actual production is substituted into the model, and the predicted shape of the glass surface is calculated by the model. Then this predicted shape is compared with the standard glass surface shape to determine whether there is any abnormality on the glass surface. If the local arch or waviness predicted by the model exceeds the set standard range, it can be considered that the glass surface shape does not meet the requirements, and it is necessary to further analyze the cause of the abnormality and take corresponding measures.

[0079] During the glass tempering process, the infrared sensor array is continuously used to dynamically monitor the glass surface temperature and obtain temperature data at different time points. As time goes by, the trend of temperature data changes and the corresponding potential changes in the glass surface morphology can be observed.

[0080] Compare the temperature data at different time points with the corresponding glass surface morphology. For example, in the heating stage, observe the changes in the glass surface morphology during the uniform temperature rise; in the cooling stage, compare the differences in temperature data and glass surface morphology under fast cooling and slow cooling conditions. Through this comparative analysis, we can more accurately grasp the impact of temperature changes on the glass surface morphology, and thus better judge whether the glass surface morphology is normal.

[0081] Further optionally, in 102, generating the surface morphology distribution data of the target object according to the real-time surface parameters includes:

[0082] By using a thermal expansion formula, the difference between each temperature data in the glass surface temperature data and the surrounding temperature data is converted into a height difference between the corresponding area and the surrounding area, and the height difference distribution data of the target object is obtained; the thermal expansion formula is associated with the material type of the target object and the real-time temperature of the environment; wherein the thermal expansion formula is expressed as: ; is the glass thermal expansion coefficient corresponding to the target object, For surface area The temperature difference with the surrounding area, For surface area The location coordinates of For surface area Thickness, For surface area The glass surface deformation height is obtained; and the height difference distribution data of the target object is converted into the surface morphology distribution data.

[0083] In the analysis of glass tempering process, the first thing to focus on is how to derive height difference distribution data based on the acquired glass surface temperature data. For the glass surface temperature data acquired by the infrared sensor array, the focus is on analyzing the difference between each temperature data and its surrounding temperature data. Because glass material has the characteristics of thermal expansion and contraction, the temperature difference in different areas will cause it to expand or contract to different degrees, which in turn causes the surface area and the surrounding area to change in height.

[0084] For example, if the temperature detected in a certain local area of ​​the glass is higher than that in the adjacent areas, based on the thermal expansion coefficient (this is an inherent characteristic parameter of the glass material itself, different types of glass have different thermal expansion coefficients, and will be affected by the real-time temperature of the environment), combined with the thickness of the glass in this area, the thermal expansion formula can be used to calculate the height difference of this local area relative to the surrounding areas due to the temperature difference. On the entire glass surface, such temperature difference analysis and height difference calculation are performed on each area, and the height difference distribution data of the target object (glass) can be obtained. This data can reflect the changes in height of different positions on the glass surface relative to its surrounding positions, and intuitively present the ups and downs of the glass surface caused by temperature differences.

[0085] After obtaining the height difference distribution data of the glass, the next step is to convert it into surface morphology distribution data. The surface morphology distribution data mainly focuses on key indicators such as the local arch and waviness of each area on the glass surface, so as to fully describe the morphological characteristics of the glass surface.

[0086] The height difference distribution data is actually a basic data that reflects the local relative height change of the glass surface. These height difference data are integrated and analyzed through specific mathematical methods and models. For example, by analyzing the change rules of the height difference in the horizontal and vertical directions of the glass surface, it is determined which areas show a continuous, arch-like height change trend, and then the local arch degree is calculated; by observing the periodic changes of the height difference in the horizontal direction, etc., it is determined whether there is a waveform and the specific value of the waveform.

[0087] Through such a series of analysis and conversion operations, the height difference distribution data originally calculated based on the temperature difference can be converted into surface morphology distribution data containing key indicators such as local arch and waveform, thereby comprehensively and intuitively displaying the morphology of the glass surface, providing important basis for subsequent judgment on whether the glass meets the quality requirements and whether the process parameters need to be adjusted.

[0088] To summarize, this process first converts the temperature difference into height difference with the help of thermal expansion formula, and then further converts it into distribution data that can more accurately describe the surface morphology of the glass, thus realizing the effective conversion from temperature data to glass surface morphology information, which is of great significance for accurately controlling the quality of glass tempering process.

[0089] As an optional embodiment, converting the height difference distribution data of the target object into the surface morphology distribution data includes:

[0090] The height difference distribution data of the target object is converted into the local camber of each corresponding surface area; the formula for obtaining the local camber is expressed as: ; Expressed as the local camber of all surface areas, For surface area The glass surface deformation height, For surface area The location coordinates of For surface area The height distribution corresponding to the standard flatness of the glass, n is the total number of surface areas.

[0091] Local camber is one of the key indicators to measure the flatness of the glass surface. It reflects the degree of bulging or depression of a local area of ​​the glass surface relative to the standard flat state. In the glass tempering process, accurately obtaining the local camber of each surface area is crucial to controlling product quality. For example, in the application scenario of architectural glass curtain walls, glass with local camber that does not meet the requirements may be unevenly stressed after installation, increasing the risk of breakage and affecting the overall aesthetics. In the field of optical instruments, accurate local camber is a key factor in ensuring optical performance. If the camber deviation is too large, the refraction and reflection paths of light will be disturbed, which will affect the normal use of the instrument.

[0092] The local camber reflects the degree of bulging or depression of a certain area of ​​the glass surface relative to the standard flat state. When the local camber is zero, it means that the area is in an ideal flat state. If the local camber is not zero, it means that the glass surface is uneven in this area.

[0093] For example, on an ideal flat glass, the degree of camber in each local area should be zero. However, if a positive camber appears in a certain area, it means that the area is bulging upward relative to the surrounding area; if it is a negative camber, it is concave downward.

[0094] The arch threshold can be set according to the specific application scenario and quality standard of the glass. For example, for building glass curtain walls, the local arch is generally required to be within a very small range, such as pm0.2mm / m (millimeter per meter). If the local arch of a certain area exceeds this threshold during the detection process, it can be judged that the flatness of the glass surface in this area does not meet the requirements.

[0095] Assuming that a piece of glass used for curtain wall has a calculated curvature of 0.3mm / m in a certain local area, which exceeds the set threshold of pm0.2mm / m, it means that there is a problem with the flatness of this area, which may affect the installation and performance of the glass.

[0096] In addition to judging the threshold value of the camber of a single area, the distribution of the local camber on the entire glass surface also needs to be considered. If the camber of only a few small areas exceeds the threshold, while most other areas meet the requirements, then the impact on the overall flatness may be relatively small; but if the camber of multiple areas exceeds the threshold, or the local camber shows regular and large changes on the glass surface, such as the camber gradually increases from the center to the edge, then it can be judged that the overall flatness of the glass is poor.

[0097] For example, by scanning and analyzing the local arch of various areas on the glass surface, it was found that the arch of the central area of ​​the glass was 0.4mm / m, and the arch of the edge area was about 0.3mm / m, and the arch distribution of the entire glass surface showed a decreasing trend from the center to the edge. In this case, it can be judged that the overall flatness of the glass does not meet the requirements, and it may be caused by uneven heating or cooling during the tempering process.

[0098] The local camber is only an important indicator for judging the flatness of the glass surface, and it can also be combined with other surface morphology parameters such as waviness to make a comprehensive judgment. If the local camber meets the requirements, but the waviness is large, the glass surface may still be uneven.

[0099] For example, the local curvature of a piece of glass is within the threshold range, but optical inspection reveals obvious wavy fluctuations on its surface, that is, the waveform does not meet the requirements. Then the surface flatness of the glass still cannot meet high-quality application requirements.

[0100] As an optional embodiment, converting the height difference distribution data of the target object into the surface morphology distribution data includes:

[0101] The height difference distribution data of the target object is subjected to fast Fourier transform to obtain the corresponding surface periodic features; the formula for obtaining the surface periodic features is expressed as: ; For surface area The corresponding spectrum in the waveform distribution on the glass surface, For surface area The location coordinates of is the waveform frequency component, For surface area The deformation height of the glass surface; the surface periodic characteristics are used as the corresponding waveform of each surface area.

[0102] Fast Fourier Transform is an efficient algorithm that converts time domain (or spatial domain) signals into frequency domain signals. In this context, the height difference distribution data can be regarded as a signal at the spatial position of the glass surface. Through FFT, this spatial domain data can be converted into frequency domain data. From a mathematical point of view, it can decompose a complex, spatially distributed height difference signal into a combination of different frequency components. For analyzing the surface morphology of glass, FFT helps to reveal the periodic laws in the change of surface height. On the glass surface, the waviness reflects the degree to which the surface presents a wavy shape, and this wavy shape usually has a certain periodicity. FFT can help extract this periodic information in order to more accurately describe the waveform characteristics of the glass surface.

[0103] This formula is used to calculate the spectrum corresponding to the glass surface area in the waveform distribution. Among them, there is the element of the waveform frequency component and the element of the glass surface deformation height of the surface area. Its core is to calculate the spectrum by weighted summation of the deformation height of each surface area, and the weight here is a specific form.

[0104] For example, the glass surface can be divided into many small areas, each with its own corresponding deformation height. When calculating the spectrum at a specific frequency, add up the deformation heights of all areas according to the formula. The calculated spectrum value represents the intensity of the height difference change of the glass surface at this specific frequency. For example, if the spectrum value calculated at a certain frequency is relatively large, it means that the glass surface has a relatively obvious waveform change in the period corresponding to this frequency.

[0105] As an important indicator to describe the surface morphology of glass, waviness directly reflects the flatness of the glass surface. If the waviness is large, it means that there are obvious undulations on the glass surface, which is undesirable in many application scenarios. For example, in optical equipment, a large waviness will lead to uneven scattering and refraction of light, affecting optical performance; in architectural glass curtain walls, it will also affect the appearance and structural stability.

[0106] The surface periodic characteristics (spectrum) obtained by FFT can be used as a quantitative indicator of waveform degree. The frequency components and intensities in the spectrum can tell the period and amplitude information of the waveform on the glass surface. For example, high-frequency components correspond to short-period waveform changes, and low-frequency components correspond to long-period waveform changes. If the spectrum has a large value within a certain frequency range, it means that the waveform degree of the glass surface is high at the corresponding period, so the waveform degree of the glass surface can be comprehensively evaluated based on the spectrum information, and then it can be judged whether the flatness of the glass surface meets the requirements.

[0107] Further optionally, different frequency components in the spectrum correspond to different periods of the waveform on the glass surface. High-frequency components represent short-period waveform changes, while low-frequency components represent long-period waveform changes. For example, in the spectrum graph, if there is a high-frequency component with a higher amplitude, it means that the glass surface has many small-scale wave-like fluctuations with a shorter period; on the contrary, a low-frequency component with a higher amplitude indicates that the glass surface has large-scale, relatively gentle waveform changes with a longer period.

[0108] In the production of high-precision optical lenses, the glass surface is required to be as flat as possible, and any slight waveform change may affect the optical performance. Through spectrum analysis, if the high-frequency component amplitude is found to be high, it means that there is a small period of waveform on the glass surface, which indicates that the surface quality of the lens does not meet the high-precision requirements and the production process needs to be adjusted, such as optimizing the cooling link in the tempering process to reduce small-scale surface deformation.

[0109] The spectrum amplitude represents the intensity of the change in height difference of the glass surface at the corresponding frequency. The larger the amplitude, the larger the waveform amplitude of the glass surface at the period corresponding to the frequency. This is because the spectrum is obtained by weighted summation of the deformation height of the surface area, and the amplitude reflects the intensity of this deformation in the frequency domain.

[0110] In the production of architectural glass curtain walls, excessive waveform amplitude may affect the appearance and structural stability of the glass. If the spectrum analysis shows a large amplitude within a certain frequency range, it means that the glass surface has obvious wave-like fluctuations at the period corresponding to the frequency. For example, for large-scale architectural glass curtain walls, in order to ensure a smooth and safe appearance, the spectrum amplitude needs to be controlled within a certain range to ensure that the waveform amplitude of the glass surface meets the requirements and prevent stress concentration and increased risk of cracking due to excessive waveforms.

[0111] By observing the distribution of the spectrum in different frequency ranges, the uniformity of the glass surface quality can be evaluated. If the amplitude of the spectrum at each frequency is relatively uniform and low, it means that the waveform change of the glass surface is relatively uniform and the amplitude is small, and the surface quality is good. On the contrary, if the spectrum has sharp peaks at certain frequencies and the amplitude of other frequencies is low, it indicates that there are local and obvious waveform changes in the corresponding period area of ​​the glass surface, and the surface quality is uneven.

[0112] A standard spectrum can be established according to the specific application scenarios and quality standards of glass. The spectrum obtained by actual testing is compared with the standard spectrum to determine whether the surface quality of the glass meets the requirements. For example, in the production of curtain wall glass, there are specific spectrum standards. If the actual spectrum deviates greatly from the standard spectrum, such as the amplitude exceeds the standard or there are unnecessary frequency components at certain key frequencies, it means that the surface quality of the glass does not meet the requirements of curtain wall glass, and the production process needs to be adjusted or the product needs to be screened.

[0113] As an optional embodiment, the anomaly detection model includes at least the following structures: an input layer, a hybrid analysis layer, a decision layer, and a strategy layer; the detection of the target object by the anomaly detection model to trigger the automatic processing mechanism of matching the candidate anomaly causes includes:

[0114] Extracting multi-dimensional detection features from the surface real-time parameters and the surface morphology distribution data through the input layer;

[0115] Through the hybrid analysis layer, the multi-dimensional detection features are analyzed in real time through the multi-layer hybrid feature calculation layer to obtain multiple candidate anomaly types and their respective matching candidate anomaly change features;

[0116] Determine, through the decision layer, a target anomaly type matched by the target object according to a plurality of candidate anomaly types and the candidate anomaly change features matched respectively;

[0117] Through the strategy layer, an exception handling strategy is generated based on the target exception type; wherein the exception handling strategy at least includes: an exception area, an exception type, and an exception handling method.

[0118] Specifically, the role of the input layer is to extract multi-dimensional detection features from the acquired real-time surface parameters (such as the local real-time temperature of each area of ​​the glass) and surface morphology distribution data (including local arch and waviness, etc.). These features are the basis for subsequent analysis, and they can fully reflect the state of the glass surface. For example, by extracting the gradient characteristics of temperature, the trend characteristics of arch, the frequency and amplitude characteristics of waviness, etc., these complex data information are preliminarily sorted out to provide rich materials for subsequent reasoning analysis.

[0119] The hybrid analysis layer uses a multi-layer hybrid feature calculation layer to perform real-time reasoning and analysis on the extracted multi-dimensional detection features. This process is similar to a complex logical judgment process. It analyzes the association between various features based on pre-set rules and algorithms, combined with empirical knowledge in actual production. For example, when a certain combination of temperature gradient features and arch change trend features appears at the same time, it may be inferred that there is a candidate anomaly type that causes arch deformation due to uneven temperature; at the same time, the candidate abnormal change features that match them can be determined based on the specific values ​​of these features, such as the size of the temperature difference, the degree to which the arch exceeds the standard range, etc.

[0120] The decision layer determines the target anomaly type that matches the target object based on the multiple candidate anomaly types obtained by the hybrid analysis layer and the matching candidate anomaly change features. This is a comprehensive weighing and judgment process that takes into account factors such as the likelihood and severity of various candidate anomaly types. For example, if there are multiple candidate anomaly types, such as temperature anomaly, wind pressure anomaly, and placement anomaly, but according to the candidate anomaly change features, the surface morphology change caused by temperature anomaly is the most serious and most consistent with the actual state of the current glass, then temperature anomaly is determined as the target anomaly type.

[0121] The strategy layer generates an exception handling strategy based on the determined target exception type. It will formulate a complete strategy including the abnormal area, abnormal type and abnormal handling method according to different abnormal types and the characteristics and requirements of the production process. For example, for the temperature abnormal type, by analyzing the temperature data, it is determined that the abnormal area is the center of the glass, and the abnormal type is too high temperature. The abnormal handling method may be to adjust the heating power of the heating furnace in this area or optimize the wind pressure of the cooling air grille in this area to reduce the temperature and correct the surface morphology of the glass.

[0122] In this way, through this multi-layered anomaly detection model, the anomalies on the glass surface can be detected comprehensively and accurately. From the multi-dimensional feature extraction of real-time surface parameters and morphological data to complex reasoning analysis and the final determination of the target anomaly type, no factors that may cause glass surface quality problems are missed, which effectively improves the accuracy of anomaly detection. Targeted anomaly handling strategies can be automatically generated to improve processing efficiency. In actual production, once an anomaly is detected, the process parameters can be quickly adjusted according to the strategy, which reduces the time for manual investigation and adjustment, reduces the rate of unqualified products, and improves the production quality and efficiency of the entire glass tempering process. At the same time, clear abnormal areas, types and disposal methods also enable operators to more clearly understand the problems in the production process, which facilitates further optimization of the production process.

[0123] Optionally, for the input surface real-time parameters and surface morphology distribution data, more in-depth feature extraction and selection are performed. In addition to the existing temperature, arch and waviness related features, derived features such as temperature change rate, arch and waviness spatial gradient can also be mined. These new features can more comprehensively describe the state changes of the glass surface and provide more valuable information for anomaly detection. For example, the difference in the change of temperature in adjacent areas per unit time is calculated as the temperature change rate feature. In the tempering process, the abnormal temperature change rate may indicate the instability of the heating or cooling link, which is very important for early detection of potential anomalies. Ensure the quality of the input data, clean the collected data, and remove noise and outliers. Since the detection equipment or environmental factors may cause data deviation, these interference factors can be effectively removed by statistical methods (such as mean filtering, median filtering, etc.). At the same time, the data is normalized so that features of different magnitudes (such as large differences in the range of temperature values ​​and arch values) have the same weight to avoid excessive or insufficient attention to certain features during model training due to differences in data magnitude.

[0124] If there is enough annotated data (samples of known abnormal and normal glass surface conditions), you can consider increasing the depth or width of the model. For example, adding more hidden layers or neurons to the mixed analysis layer allows the model to learn more complex feature combinations and nonlinear relationships. This helps the model fit the data better and improves the ability to distinguish different types of anomalies. However, it should be noted that increasing the complexity of the model may bring the risk of overfitting, so it is necessary to cooperate with appropriate regularization methods, such as L1 and L2 regularization, Dropout, etc.

[0125] Consider combining unsupervised learning algorithms, such as cluster analysis or autoencoders. Cluster analysis can help find natural groupings in the data and identify abnormal data clusters that are different from normal state data clusters. Autoencoders can learn the potential representation of data and detect anomalies through reconstruction errors. For example, use autoencoders to train normal glass surface data so that the model can learn the encoding and decoding rules of data in normal state. During detection, if the reconstruction error of the input new data exceeds a certain threshold, it can be judged as abnormal. This method does not require a large amount of labeled data and is very useful for situations where the anomaly types are complex and it is difficult to obtain annotations. Choose a suitable model parameter optimization algorithm, such as stochastic gradient descent (SGD) and its variants Adagrad, Adadelta, Adam, etc. These algorithms can update model parameters more effectively according to the characteristics of the data and the gradient information of the loss function, accelerate the model convergence speed, and improve the model performance. For example, the Adam algorithm combines the advantages of the momentum method and the adaptive learning rate, and performs well in processing complex nonlinear models and large-scale data. By adjusting the hyperparameters of the Adam algorithm (such as learning rate, β1 and β2, etc.), the model can learn the optimal parameter combination faster. Establish an online learning mechanism so that the model can update parameters in real time or regularly. In the glass tempering production process, the distribution of data may change as the production environment changes, the process improves, or new abnormal situations arise. Through online learning, the model can continuously adapt to these changes. For example, at regular production batch intervals, newly collected data with accurate annotations (confirmed by manual inspection or other high-precision detection methods) is added to the training data, the model is retrained, and the model parameters are updated to maintain the model's adaptability to the latest production situation and detection accuracy.

[0126] Establish a comprehensive model evaluation index system. In addition to the commonly used classification indicators such as accuracy, recall, and F1-score, indicators such as model stability and robustness should also be considered. For example, by testing the model under different production environment conditions (such as different temperatures, humidity, equipment status, etc.), observe the fluctuation of model performance and evaluate the robustness of the model. For anomaly detection, false alarm rate and false negative rate are also very important indicators. A lower false alarm rate can avoid unnecessary production interruptions and waste of resources, while a lower false negative rate ensures that quality problems can be discovered in a timely manner. By adjusting the model threshold and optimizing the model structure, a suitable balance can be found between the false alarm rate and the false negative rate. Using cross-validation methods such as K-fold cross-validation, the data set is divided into multiple subsets, one of the subsets is used as the test set in turn, and the remaining subsets are used as the training set. The model is trained and tested multiple times to obtain more stable and reliable model performance evaluation results. In addition, simulation tests can also be conducted by constructing virtual abnormal scenarios (based on known abnormal causes and impact levels), inputting these simulation data into the model, and observing whether the model's detection capabilities and the generated processing strategies are reasonable, so as to optimize the model in a targeted manner.

[0127] As an optional embodiment, in 104, real-time reasoning analysis is performed on the multi-dimensional detection features through a multi-layer hybrid feature calculation layer to obtain multiple candidate anomaly types and respective matching candidate anomaly change features, including:

[0128] A fuzzy analysis layer is used to perform real-time reasoning and analysis on multi-dimensional detection features to obtain multiple corresponding candidate abnormality types; multiple fuzzy abnormality type identification rules are set in the fuzzy analysis layer; each fuzzy abnormality type identification rule is adaptively learned based on at least one historical deformation data and the corresponding deformation cause; each fuzzy abnormality type identification rule is dynamically configured based on material type and production environment information; and candidate abnormal change features in the multi-dimensional detection features that match the multiple candidate abnormality types are identified.

[0129] When performing real-time reasoning and analysis on multi-dimensional detection features, the fuzzy analysis layer is used to handle those uncertain and ambiguous situations more flexibly and effectively. The state of the glass surface is affected by many complex factors, and the correspondence between its detection features and abnormal types is often not absolutely clear, and there is a certain ambiguity, and the fuzzy analysis layer can just deal with this situation. The multi-dimensional detection features are analyzed and judged through multiple preset fuzzy abnormal type identification rules. For example, after the multi-dimensional detection features such as temperature distribution, arch, and waveform of the glass surface are input, the fuzzy analysis layer uses these rules to infer which feature combinations may point to specific abnormal situations, and then derives the corresponding multiple candidate abnormality types.

[0130] These fuzzy anomaly type recognition rules are not set arbitrarily, but are adaptively learned based on at least one historical deformation data and the corresponding deformation cause. In the past glass tempering production process, a large amount of actual data on various deformations on the glass surface has been accumulated, and the corresponding causes of deformation have been recorded. For example, the glass surface arch was too high once. After investigation, it was found that the temperature in a certain area of ​​the tempering furnace was too high and the cooling was uneven. Such actual case data is valuable learning material. Through in-depth analysis and summary of many similar historical deformation data and their causes, and using adaptive learning algorithms (such as some algorithms or mechanisms based on machine learning that can process fuzzy information), fuzzy anomaly type recognition rules have gradually been formed. These rules can summarize the association patterns between different detection features and possible anomaly types, so that when faced with new detection data, these rules can be used to infer possible anomaly types.

[0131] At the same time, each fuzzy anomaly type identification rule will also be dynamically configured based on the material type and production environment information. Glass materials vary, and different materials have different thermal expansion coefficients, mechanical properties and other characteristics. The response to the same process conditions during the tempering process will also be different. For example, ordinary soda-lime glass and high borosilicate glass have different deformation laws when heated and cooled. In addition, production environment information, such as the temperature range, wind pressure, humidity and other factors in the tempering furnace, will also affect the surface morphology of the glass. Therefore, the fuzzy anomaly type identification rules should adjust their own parameters, judgment thresholds and other key elements in a timely manner according to the specific material type of the glass currently being processed and the actual production environment information, to ensure that the rules can fit the actual situation and play their role accurately.

[0132] After inferring multiple candidate anomaly types based on rules through the fuzzy analysis layer, further identify candidate anomaly change features that match these candidate anomaly types in the multidimensional detection features. For example, when it is inferred that there may be a candidate anomaly type of "uneven temperature causing arch deformation", the multidimensional detection features are used to find out which specific temperature values, temperature gradients, arch values ​​and other features can reflect the specific situation of this abnormal change, such as determining which area has a higher temperature and the corresponding arch value that exceeds the standard range. These are the candidate anomaly change features that match it. They can describe the specific manifestations and extent of the anomaly in more detail, providing a more sufficient basis for the subsequent accurate judgment of the target anomaly type and the formulation of targeted anomaly handling strategies.

[0133] In short, through such a fuzzy analysis layer and its related rule application and feature recognition, it is possible to more accurately mine potential abnormal situation related information in the complex and ambiguous glass tempering production data, which will help the quality control and exception handling of the entire production process.

[0134] Further, optionally, a wider range of historical deformation data is collected, covering situations under different glass materials, various production environment conditions and multiple process parameter combinations. For example, not only common glass thicknesses and types are included, but also historical data of special-purpose glass should be considered. At the same time, the accuracy of these historical data is ensured, and data errors are reduced through precise detection equipment and multiple rounds of manual review. For noise and outliers in the data, data cleaning techniques can be used, such as outlier detection methods based on statistical distribution (such as the 3σ principle) to remove data points that are obviously inconsistent with the actual situation. In this way, the fuzzy abnormal type recognition rules learned based on richer and more accurate historical data can more comprehensively and truly reflect various abnormal situations in actual production. According to the physical principles and actual experience of glass production, the parameters and conditions in the fuzzy abnormal type recognition rules are refined. For example, for the rule that temperature causes arch deformation, not only the temperature difference is considered, but also the rate and duration of temperature change and the thermal conductivity characteristics of different areas of glass. By establishing more detailed rule conditions, such as "when the temperature in the center area of ​​​​the glass rises by more than 50°C within 10 seconds, and the temperature in the edge area does not rise by more than 30°C, and the thickness of the glass is 5mm, there is a higher probability of arch deformation", the abnormal type can be identified more accurately and the accuracy of the fuzzy analysis layer can be improved.

[0135] Furthermore, the expert experience library can be introduced to review and adjust the fuzzy anomaly type identification rules. Based on the practical experience of historical data, the possible unreasonableness in the identification rules or the key factors that need to be supplemented can be identified. For example, based on the tempering process characteristics of a special glass, it may be recommended to add a rule that the glass is more sensitive to wind pressure changes under a specific chemical composition, so as to make the rule system more complete and improve the accuracy of the fuzzy analysis layer in complex actual situations.

[0136] In-depth modeling of the physical and chemical properties of different glass materials. In addition to the thermal expansion coefficient, the influence of factors such as the elastic modulus, Poisson's ratio, and density of the glass on the surface morphology should also be considered. For example, by establishing a thermal-mechanical coupling model of glass materials, the deformation of glass under different temperature and stress conditions can be more accurately simulated. According to these detailed material property models, the relevant parameters in the fuzzy anomaly type recognition rules are adjusted so that the rules can perform more accurate reasoning and analysis based on the characteristics of different materials. For example, for glass with a low elastic modulus, when judging whether it has waveform deformation due to wind pressure, the judgment threshold of wind pressure anomaly should be appropriately lowered.

[0137] Real-time dynamic monitoring of production environment factors, such as real-time changes in temperature, humidity, wind pressure in the tempering furnace, and interference from the external environment. By setting up an input interface for environmental factors in the fuzzy analysis layer, these dynamically changing environmental information can be promptly integrated into the reasoning analysis process. For example, when the ambient humidity suddenly increases, the heat dissipation characteristics of the glass surface may change. The fuzzy analysis layer can adjust the judgment criteria for temperature-related anomaly types based on the pre-set humidity-heat dissipation relationship model, thereby improving the accuracy of analysis in complex and changing environments.

[0138] Establish an online learning mechanism so that the fuzzy analysis layer can update the fuzzy anomaly type recognition rules in real time or regularly according to new production data. During the production process, with the improvement of the process, the application of new materials or the emergence of new anomalies, the original rules may no longer apply. For example, when a new glass material is introduced or the tempering process is optimized, the system automatically collects new data samples and uses the incremental learning algorithm to update the fuzzy anomaly type recognition rules to ensure that the rules can adapt to the new production situation in a timely manner and maintain accuracy.

[0139] Model validation techniques, such as the holdout method and K-fold cross validation method, are used to evaluate the accuracy of the fuzzy analysis layer. The historical data is divided into a training set and a validation set. By comparing the predicted candidate anomaly types with the actual anomalies, the accuracy, recall rate, F1-score and other evaluation indicators are calculated. According to the validation results, the fuzzy anomaly type recognition rules are adjusted and optimized. For example, if the recall rate of a certain anomaly type is low, it means that there may be missed judgments. It is necessary to check the corresponding recognition rules to see if some conditions need to be relaxed or new rules need to be added to improve the recognition ability of this anomaly type.

[0140] As an optional embodiment, in 104, a fuzzy analysis layer is used to perform real-time reasoning analysis on the multi-dimensional detection features to obtain corresponding multiple candidate anomaly types, including:

[0141] The furnace feature network is reconstructed based on multi-dimensional detection features; each surface area in the target object in each node corresponds to the furnace area and the corresponding temperature data and wind pressure data; the temperature values ​​of each furnace area are compared to locate the abnormal temperature control area with the largest temperature deviation value; the wind pressure values ​​of each furnace area are compared to locate the abnormal wind grille area with the largest wind pressure deviation value; the abnormal type is identified based on the abnormal temperature control area and the abnormal wind grille area, and the identified candidate abnormal type is marked in the node corresponding to the furnace feature network.

[0142] In the whole process, the furnace feature network is first reconstructed based on multi-dimensional detection features. These multi-dimensional detection features contain a lot of key information corresponding to each surface area of ​​the target object (that is, glass), such as the furnace area and the corresponding temperature data, wind pressure data, etc. These data are integrated to build a network, and each node carries the relevant detailed information of a specific area, just like building a virtual model that can reflect the actual situation in the furnace. Each node is interconnected, showing the overall status of different areas inside the furnace.

[0143] Next, by comparing the temperature values ​​of each area in the furnace, we can accurately locate the abnormal temperature control area with the largest temperature deviation. Because in the glass tempering process, the uniformity of temperature has a great impact on the surface morphology of the glass. Once the temperature of a certain area deviates too much from that of other areas, it is likely to cause abnormal conditions such as deformation of the glass in that area that does not meet the requirements. By carefully comparing the temperature values ​​of each area, we can find the area with the most significant temperature difference, that is, the abnormal temperature control area.

[0144] Similarly, for wind pressure data, the wind pressure values ​​of each furnace area are compared to locate the abnormal wind grid area with the largest wind pressure deviation. The wind pressure blown out by the wind grid plays a key role in the glass cooling process. Unbalanced wind pressure will cause different cooling speeds on the upper and lower surfaces or different parts of the glass, thereby causing changes in the surface morphology. Finding the area with the largest wind pressure deviation, that is, the abnormal wind grid area, is very important for analyzing the cause of the abnormality.

[0145] Finally, the abnormal type is identified based on the abnormal temperature control area and abnormal wind grille area that have been determined. Because the abnormalities in these two key areas are often closely related to the abnormal performance of specific glass surfaces, for example, areas with excessively high temperatures may cause arch deformation in the corresponding parts of the glass, and areas with abnormal wind pressure may cause changes in waveform, etc. Based on these associations, accumulated experience and knowledge in the past, established analysis rules, etc., we identify the possible candidate abnormal types and mark them on the corresponding nodes of the furnace feature network. This is equivalent to clearly marking the possible problem areas and the corresponding problem types in this virtual model that reflects the state inside the furnace. In the future, we can further analyze and handle abnormal situations based on these marked information.

[0146] In terms of effect, this approach can make anomaly detection more accurate and intuitive. By reconstructing the furnace feature network and locating the abnormal temperature control area and abnormal wind grid area, the key influencing factors in the furnace are accurately controlled. It is no longer a general judgment of whether there is an abnormality, but it is known exactly in which specific temperature control area or wind grid area the problem may lie, which narrows the scope of investigation and improves the efficiency of subsequent analysis. The identified candidate abnormality types are marked on the corresponding nodes, making the entire abnormal situation more visual. The operator or the subsequent automatic processing mechanism can see at a glance what type of abnormality has occurred in which part of the furnace. This is very helpful for taking targeted measures to solve the problem in a timely manner. For example, the heating power of the abnormal temperature control area can be quickly adjusted, and the wind pressure in the abnormal wind grid area can be corrected, thereby effectively reducing the probability of abnormalities in the glass surface morphology, improving the product quality of the glass tempering process, and ensuring the smooth progress of the production process. At the same time, it also provides a clear basis for further analysis of possible systemic problems in the entire production process, which helps to continuously optimize process parameters and processes.

[0147] For example, in the glass tempering process, dynamic adjustment has multiple goals, including flatness control, which aims to achieve uniform temperature on the glass surface and prevent waveform or arch distortion; stress uniformity, ensuring matching of cooling rates in each area during the rapid cooling process; and efficiency and quality balance, that is, improving production efficiency on the basis of ensuring quality. For the algorithms and strategies of dynamic adjustment, in terms of dynamic adjustment of furnace temperature, since different thicknesses of glass have different requirements for heating time and temperature distribution, a partitioned PID control algorithm is adopted. The furnace body is divided into multiple temperature control areas, each of which is equipped with an independent PID controller, and the target temperature is set according to the glass thickness and material. Its dynamic adjustment strategy is to input real-time temperature data, glass thickness and material parameters, automatically adjust the heating wire power according to the regional temperature error, and increase the temperature control area for ultra-thick glass to improve temperature uniformity. In terms of dynamic adjustment of rapid cooling air pressure, due to excessive temperature difference, uneven cooling rate and stress concentration, the wind pressure gradient optimization model is used. Dynamically adjust the wind pressure in each area of ​​the wind grid according to the temperature difference distribution on the glass surface. Its dynamic adjustment strategy is to input the temperature difference data of the infrared detection system, increase the wind pressure in the high-temperature area, reduce the wind pressure in the low-temperature area, and keep the total wind pressure constant to avoid increasing the overall energy consumption. In terms of dynamic adjustment of tempering speed, adaptive speed control is adopted in view of the fact that the tempering speed affects the heating time and cooling uniformity. The conveying speed is adjusted in real time based on the heating time and cooling rate. Its dynamic adjustment strategy is to input the glass thickness and target temperature, and dynamically adjust the conveying speed according to the real-time temperature control data to prevent local overheating or uneven cooling. In addition, a multi-parameter joint optimization strategy is adopted, combining the furnace temperature, rapid cooling wind pressure and tempering speed, and realizing the dynamic joint adjustment of process parameters through genetic algorithm (GA) optimization. The operation steps include initializing process parameters, generating parameter populations based on the objective function, finding the global optimal parameter combination through selection, crossover and mutation, and outputting it for production.

[0148] In terms of environmental adaptability and scalability, environmental adaptability measures include interference suppression, such as using anti-glare filter materials to reduce the interference of ambient light on infrared detection, installing anti-vibration brackets to ensure sensor stability; and environmental control, setting up light shields and soundproof covers in the detection area to maintain a stable detection environment. In terms of scalability design, a modular architecture is adopted, and the furnace temperature control, wind pressure control and speed adjustment systems are independent of each other for easy expansion and upgrading; and it can adapt to multiple glass specifications, and the system automatically loads process parameter templates of different thicknesses and materials to support multi-variety switching.

[0149] In the embodiment of the present application, the equipment parameters of the target device are obtained; the equipment parameters at least include: furnace size, walking program parameters; the walking program parameters include a variety of process parameters associated with the process links and equipment structural components of the target device; the equipment parameters are predicted in real time by a thermal cycle control model to obtain the glass yield prediction information of the target device under different parameter settings; the thermal cycle control model is a lightweight model installed in the target device for learning the real-time situation of the production environment; based on the glass yield prediction information, an optimization strategy for the walking program parameters is generated; the optimization strategy includes the equipment parameters corresponding to the glass yield prediction value that reaches the preset glass finished product evaluation standard; the glass finished product evaluation standard is related to the equipment production demand information of the target device; the equipment production demand information at least includes: glass thickness, glass material composition; the walking program parameters of the target device are adjusted by using the optimization strategy to control the thermal cycle speed of the target device and improve the production efficiency of the target device. The present application can dynamically predict and control the walking program parameters, thereby achieving the control of the thermal cycle speed of the target device and improving the glass production efficiency and glass quality.

[0150] In the context of the embodiment, the architecture design follows the principle of modular layered design, including a data acquisition layer (obtaining flatness data from an infrared detection system and preprocessing), a data analysis and decision-making layer (using fuzzy logic and reinforcement learning algorithms to identify and analyze anomalies), a feedback and control layer (adjusting process parameters such as temperature, wind pressure, and speed based on the analysis results to achieve dynamic control), a data storage and optimization layer (storing historical data for long-term trend analysis and model optimization), and having real-time processing capabilities (using high-performance computing frameworks such as ApacheKafka or Flink to parallelly process real-time data streams to ensure low latency), fault tolerance and stability (setting up redundant sensors and backup servers). Key considerations include parallelization and Scalability (the data collection layer supports multi-channel parallel processing, and the data analysis layer adopts a distributed architecture to support dynamic expansion of computing nodes), low latency (using message queues such as Kafka and stream processing frameworks such as Flink to achieve millisecond-level latency data transmission and analysis), visualization and alarm (providing real-time data visualization interface and intelligent alarm system to help operators quickly locate problems); in terms of abnormality recognition algorithm selection and optimization, fuzzy logic is suitable for processing multi-variable and nonlinear process abnormality analysis, and real-time reasoning can be achieved by defining a rule base. It can be optimized by defining more precise rules and dynamically adjusting rule weights. Reinforcement learning learns the optimal parameter adjustment strategy through interaction with the actual production process, and uses deep reinforcement learning to utilize The neural network processes high-dimensional data and uses a policy gradient-based method to optimize the adjustment strategy. At the same time, it uses edge computing devices and model compression technology to ensure real-time performance. The design of the abnormality investigation and feedback mechanism includes grading by abnormality severity (temperature difference abnormality > wind pressure abnormality > support point abnormality) and performing logical investigation (temperature difference abnormality checks the area with the largest data positioning deviation in the temperature control area of ​​the furnace body, wind pressure abnormality analyzes the area with uneven wind pressure data positioning in the wind grid partition, and placement abnormality detects glass offset based on support point sensor data). It has a fast feedback mechanism (adjusting parameters for abnormal area partitions, retesting the flatness after adjustment to verify the effect, and automatically generating a problem report to push solution suggestions to the operator) and a high-frequency adjustment strategy (parameter partition independence , give priority to adjusting the most influential problem parameters to reduce interference with production); the system's scalability and adaptability design is reflected in the modular architecture (each functional module is independently deployed to support single-module upgrades and support dynamic addition of detection sensors or control nodes to meet more complex process requirements), parameter templates (preset process parameter templates for different glass types and sizes to quickly adapt to new products), intelligent learning mechanism (using historical data to optimize the rule base and reinforcement learning model to adapt to changes in the production environment), multi-process compatibility (supports switching between flat tempered glass and curved tempered glass production modes, and the parameter range supports different thicknesses and size specifications from 3mm to 19mm) and remote upgrade support (the system firmware can be remotely upgraded via the network to meet future process improvement needs).

[0151] As an optional embodiment, the system adopts a modular layered architecture design, covering the data collection layer, data analysis and decision layer, feedback and control layer, and data storage and optimization layer. It has real-time processing capabilities, fault tolerance and stability, and focuses on parallelization and scalability, low latency, visualization and alarm in key considerations. The abnormality recognition algorithm includes fuzzy logic (suitable for processing multi-variable, nonlinear process abnormality analysis, and can achieve real-time reasoning and optimize rules by defining a rule base) and reinforcement learning (learning the optimal parameter adjustment strategy through interaction with the actual production process, using deep reinforcement learning and optimization based on the policy gradient method), and uses edge computing devices and model compression technology to ensure real-time performance. The abnormality investigation and feedback mechanism is graded and logically investigated according to the severity of the abnormality, and has a fast feedback mechanism (partition adjustment parameters, closed-loop feedback, alarm prompts) and a high-frequency adjustment strategy (parameter partition independence, adjustment priority). The system's scalability and adaptability are reflected in the modular architecture, parameter templates, intelligent learning mechanism, multi-process compatibility and remote upgrade support.

[0152] In terms of dynamic parameter control, the system uses a multi-dimensional feedback mechanism to finely adjust the heating wire power in the furnace in different zones to improve heating uniformity, automatically optimize the wind grid operation mode of the upper and lower wind pressure difference, and adjust the glass placement method, such as the position and angle of the support point of the conveying system. At the same time, big data optimization is used to store the detection and control data in real time and compare it with historical data, and optimize future production parameters through big data analysis.

[0153] The specific process flow is as follows: when loading, the glass is automatically transferred to the tempering furnace, and the positioning support device automatically adjusts its placement angle and position; in the heating stage, the heating wire is controlled in the furnace to heat the glass to above 600°C and the temperature distribution is monitored in real time; in the rapid cooling stage, the upper and lower wind grids are used to quickly cool the glass, and the wind pressure is dynamically controlled to avoid uneven force; infrared detection uses a high-resolution infrared thermal imager as a core component, and multiple detection points are arranged along the glass transmission direction to fully cover the glass surface. After the glass is out of the furnace, it is scanned on the entire surface by an infrared sensor array. The detection system is equipped with professional data processing software to convert thermal radiation data into glass surface height information in real time, and generate high-precision arch and waveform distribution data; the abnormal feedback link sets a reasonable flatness threshold. If the detection result exceeds the standard range, the abnormal cause is automatically analyzed and fed back to the tempering furnace and wind grid control system. For example, the heating wire power is adjusted for uneven local heating, and the wind pressure parameters of the wind grid are adjusted for unbalanced wind pressure. Improper placement is corrected in real time through mechanical devices on the one hand, and prompts are issued to operators on the operation interface to guide manual assistance in adjusting the support points of the conveying device on the other hand; cycle optimization is to store data in the production database, and optimize the parameter model after regular analysis.

[0154] In terms of process conditions, the infrared detection system needs to be installed in an interference-free environment to ensure detection accuracy. The accuracy and stability of various sensors must meet the requirements and be calibrated and maintained regularly to ensure data accuracy. The intelligent analysis and control system must have strong data processing capabilities and rapid response capabilities.

[0155] In each process step, the infrared real-time detection system can promptly detect glass flatness problems, provide accurate data basis for subsequent adjustments, avoid further processing of unqualified products and reduce waste; the intelligent analysis and control system quickly determines the root cause of the problem through precise data analysis, improves the pertinence and effectiveness of adjustments, shortens process adjustment time and improves production efficiency; process parameter adjustment and glass placement correction can directly improve the glass stress state and molding conditions, effectively improve the glass flatness and ensure stable product quality.

[0156] The reasons for selecting each process parameter are as follows: the resolution and detection frequency of the infrared detection system are determined according to the glass production speed and accuracy requirements to capture tiny changes on the glass surface; the temperature parameters in the furnace are set according to factors such as glass material, thickness and target hardness to ensure the strength and flatness of the glass; the quenching wind pressure parameters take into account the thermal conductivity and size of the glass to allow the glass to be cooled evenly and quickly to reduce stress concentration; the tempering speed parameters are determined based on the glass type and production equipment performance to control the temperature change gradient during the glass tempering process to affect the flatness; the glass placement method is based on the glass shape, size and tempering process requirements to ensure uniform force during the heating and cooling process.

[0157] Based on the same implementation principle, in another embodiment of the present application, a glass arch detection device is provided, which at least includes the following units:

[0158] A detection unit is used to obtain the real-time surface parameters of the target object by infrared detection after the current process link of the tempering process; the real-time surface parameters at least include: the local real-time temperature of each surface area of ​​the target object; the material of the target object is glass;

[0159] A generating unit, configured to generate surface morphology distribution data of the target object according to the surface real-time parameters; the data dimensions of the surface morphology distribution data at least include: local camber and waviness of each surface area of ​​the target object;

[0160] An acquisition unit, configured to acquire a degree of deviation between the target object and a standard surface morphology based on the surface morphology distribution data;

[0161] An exception processing unit, configured to stop the transmission process of the target object if the deviation degree satisfies the deviation exception condition, and detect the target object through an exception detection model to trigger an automatic processing mechanism for matching the candidate exception causes;

[0162] A transmission unit is used to transmit the target object from the current process link to the next process link if the deviation degree does not meet the deviation abnormality condition.

[0163] Similar to the glass arch detection method, the specific functions of each unit in the above device will not be elaborated for the time being, and please refer to the relevant description in the context embodiment.

[0164] Based on the same implementation principle, the embodiment of the present application also provides an electronic device, Figure 2 is a structural block diagram of the electronic device 300, such as Figure 2 As shown, the electronic device 300 includes: a processor 301, a memory 302, a communication interface 303, a communication bus 304 and a controller 305; wherein the processor 301, the memory 302 and the communication interface 303 communicate with each other via the communication bus 304; the memory 302 is used to store computer programs; the processor 301 is used to execute the programs stored in the memory 302 to implement corresponding processing functions; the communication bus 304 is used for communication between the electronic device 300 and other devices; and the controller 305 is used to implement the glass arch detection method described in the above method embodiment.

[0165] In the embodiment of the present application, the communication bus 304 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus 304 may be divided into an address bus, a data bus, a control bus, etc., and the specific form is not limited. For ease of representation, Figure 2 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0166] The memory 302 may include a random access memory (RAM) or a non-volatile memory (non-volatile memory), such as at least one disk memory. Optionally, the memory may also be at least one storage device located far away from the processor 301 .

[0167] The processor 301 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other editable logic devices, discrete gate or transistor logic devices, discrete hardware components, and the specific form can be determined according to actual needs.

[0168] Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed, can implement each step that can be executed by an electronic device in the above method embodiment.

[0169] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for detecting glass arch, characterized in that: The method comprises: After being processed in the current process link of the tempering process, the real-time surface parameters of the target object are obtained by infrared detection; the real-time surface parameters at least include: the local real-time temperature of each surface area of ​​the target object; the material of the target object is glass; Generate surface morphology distribution data of the target object according to the surface real-time parameters; the data dimensions in the surface morphology distribution data at least include: local arch and waviness of each surface area in the target object; Based on the surface morphology distribution data, obtaining a degree of deviation between the target object and a standard surface morphology; If the deviation degree satisfies the deviation abnormality condition, the transmission process of the target object is stopped, and the target object is detected by an abnormality detection model to trigger an automatic processing mechanism for matching candidate abnormality causes; If the deviation does not satisfy the deviation abnormality condition, the target object is transferred from the current process link to the next process link; Before acquiring the real-time surface parameters of the target object by infrared detection, the method further includes: At the exit of the current process link, a plurality of infrared sensors are deployed to form an infrared sensor array; the plurality of infrared sensors are high-resolution sensors; The method of obtaining the real-time surface parameters of the target object by infrared detection includes: Scanning the glass surface temperature data of the target object in real time; the glass surface temperature data is a two-dimensional data matrix obtained by real-time scanning using an infrared sensor array; The step of generating the surface morphology distribution data of the target object according to the real-time surface parameters includes: By using a thermal expansion formula, the difference between each temperature data in the glass surface temperature data and the surrounding temperature data is converted into a height difference between the corresponding area and the surrounding area, and the height difference distribution data of the target object is obtained; the thermal expansion formula is associated with the material type of the target object and the real-time temperature of the environment; wherein the thermal expansion formula is expressed as: ; is the glass thermal expansion coefficient corresponding to the target object, For surface area The temperature difference with the surrounding area, For surface area The location coordinates of For surface area Thickness, For surface area The glass surface deformation height; Converting the height difference distribution data of the target object into the surface morphology distribution data; The converting the height difference distribution data of the target object into the surface morphology distribution data includes: converting the height difference distribution data of the target object into the local camber of each corresponding surface area; the formula for obtaining the local camber is expressed as: ; Expressed as the local camber of all surface areas, For surface area The glass surface deformation height, For surface area The location coordinates of For surface area The corresponding height distribution under the standard flatness of the glass, n is the total number of surface areas; The step of converting the height difference distribution data of the target object into the surface morphology distribution data further includes: performing fast Fourier transformation on the height difference distribution data of the target object to obtain corresponding surface periodic features; the formula for obtaining the surface periodic features is expressed as: ; For surface area The corresponding spectrum in the waveform distribution on the glass surface, is the waveform frequency component, For surface area The location coordinates of For surface area The glass surface deformation height; The surface periodicity characteristics are used as the corresponding waviness of each surface area.

2. The method according to claim 1, characterized in that The anomaly detection model at least includes the following structures: input layer, hybrid analysis layer, decision layer, and strategy layer; The detecting the target object by using the anomaly detection model to trigger the automatic processing mechanism for matching the candidate anomaly causes includes: Extracting multi-dimensional detection features from the surface real-time parameters and the surface morphology distribution data through the input layer; Through the hybrid analysis layer, the multi-dimensional detection features are analyzed in real time through the multi-layer hybrid feature calculation layer to obtain multiple candidate anomaly types and their respective matching candidate anomaly change features; Determine, through the decision layer, a target anomaly type matched by the target object according to a plurality of candidate anomaly types and the candidate anomaly change features matched respectively; Through the strategy layer, an exception handling strategy is generated based on the target exception type; wherein the exception handling strategy at least includes: an exception area, an exception type, and an exception handling method.

3. The method according to claim 2, characterized in that The multi-dimensional detection features are subjected to real-time reasoning analysis through a multi-layer hybrid feature calculation layer to obtain multiple candidate anomaly types and respective matching candidate anomaly change features, including: A fuzzy analysis layer is used to perform real-time reasoning and analysis on multi-dimensional detection features to obtain multiple corresponding candidate anomaly types; multiple fuzzy anomaly type identification rules are set in the fuzzy analysis layer; each fuzzy anomaly type identification rule is adaptively learned based on at least one historical deformation data and the corresponding deformation cause; each fuzzy anomaly type identification rule is dynamically configured based on material type and production environment information; Identify candidate abnormal change features that match the multiple candidate abnormality types in the multi-dimensional detection features.

4. The method according to claim 3, characterized in that The fuzzy analysis layer is used to perform real-time reasoning analysis on multi-dimensional detection features to obtain corresponding multiple candidate anomaly types, including: Reconstruct the furnace feature network based on multi-dimensional detection features; the furnace area corresponding to each surface area of ​​the target object in each node and the corresponding temperature data and wind pressure data; Compare the temperature values ​​of each furnace area and locate the abnormal temperature control area with the largest temperature deviation value; Compare the wind pressure values ​​of each furnace area and locate the abnormal wind grid area with the largest wind pressure deviation value; Abnormal types are identified based on the abnormal temperature control area and the abnormal air grille area, and the identified candidate abnormal types are marked in the nodes corresponding to the furnace feature network.

5. A glass dome detection device, characterized in that: For implementing the glass arch detection method according to any one of claims 1 to 4, the device comprises at least the following units, wherein: A detection unit is used to obtain the real-time surface parameters of the target object by infrared detection after the current process link of the tempering process; the real-time surface parameters at least include: the local real-time temperature of each surface area of ​​the target object; the material of the target object is glass; A generating unit, configured to generate surface morphology distribution data of the target object according to the surface real-time parameters; the data dimensions of the surface morphology distribution data at least include: local camber and waviness of each surface area of ​​the target object; An acquisition unit, configured to acquire a degree of deviation between the target object and a standard surface morphology based on the surface morphology distribution data; An exception processing unit, configured to stop the transmission process of the target object if the deviation degree satisfies the deviation exception condition, and detect the target object through an exception detection model to trigger an automated processing mechanism for matching candidate exception causes; A transmission unit is used to transmit the target object from the current process link to the next process link if the deviation degree does not meet the deviation abnormality condition.

6. An electronic device, characterized in that: include: Memory for storing computer software programs; A processor is used to read and execute the computer software program, thereby implementing the glass arch detection method according to any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium, characterized in that: The storage medium stores a computer software program, and when the computer software program is executed by the processor, the glass arch detection method according to any one of claims 1 to 4 is implemented.

Images