Method for coupling deep processing, cutting, sorting and tempering matching of building glass

Abstract

The invention provides a coupling method for building glass deep processing cutting, sheet arranging and tempering sheet matching, and relates to the field of glass deep processing, and the method comprises the following steps: receiving production order data, generating a cutting instruction set according to the maximum loading size and process characteristics of a tempering furnace, transferring cut glass sheets to a sheet arranging cage, synchronously updating the dynamic excess material database; calculating equipment production data uploaded by the nodes to form a global production view; predicting the thickness distribution of glass batches which are about to enter a tempering link in the next 24 hours through a long short-term memory (LSTM) network, and generating a thickness distribution report; carrying out batch grouping optimization through a clustering algorithm, and generating a batch grouping scheme for efficient loading of the toughening furnace and a corresponding sheet sorting cage sorting instruction; and running the loading simulation algorithm, and generating the optimal placement layout of the corresponding batch of glass in the toughening furnace. The technical problems that all links are mutually separated and have no closed loop, production data form islands, the production rhythm cannot be adjusted, and the efficiency is low are solved.

Description

Technical Field

[0001] This application relates to the field of deep glass processing, and in particular to a coupling method for deep processing of architectural glass, including cutting, shearing, and tempering. Background Technology

[0002] Cutting, glass preparation, and tempering / fitting are the core processes in transforming raw glass into finished products, and their collaborative efficiency directly determines overall production capacity and cost control. In existing technologies, the cutting, glass preparation, and tempering / fitting processes in architectural glass deep processing generally operate independently, with each process performing operations based on only local information, without forming a closed-loop collaborative mechanism. The cutting process relies solely on the glass size and quantity requirements of the production order for cutting the original glass sheets. It fails to optimize the layout based on process characteristics such as the maximum loading size of the tempering furnace and the minimum gap between glass sheets. As a result, the cut glass sheets often cannot be efficiently loaded into the tempering furnace due to size mismatch and poor combination compatibility. This requires manual readjustment or waiting for a suitable batch, causing stagnation when switching between the cutting and tempering processes. The sheet sorting process is used only as a temporary storage device and does not sort the glass sheets in accordance with the real-time operating conditions of the tempering furnace (such as the current load and the required thickness of the glass to be processed). This results in the glass sheets being transferred to the tempering furnace needing to be sorted again, further increasing the waiting time between processes. The tempering process passively receives the glass sheets transferred from the previous process and cannot obtain key information such as the glass thickness distribution in advance. It requires frequent shutdowns to adjust process parameters such as temperature, creating a "breakpoint" that separates the various processes. Meanwhile, production data from the cutting, sorting, and tempering equipment are also scattered and stored locally on each piece of equipment, forming "data silos." The central control system cannot summarize the overall production view in real time, nor can it dynamically adjust the production rhythm according to changes in equipment load. This often results in situations where cutting is too fast, causing congestion in the sorting cage, or an imbalance where the tempering furnace is idle while waiting for glass sheets. Furthermore, there is no mechanism for predicting glass thickness distribution. The thickness can only be detected and the process adjusted after the glass sheets arrive at the tempering furnace, leading to frequent start-ups and shutdowns of the tempering furnace and severely disrupting the production rhythm. Summary of the Invention

[0003] This application provides a coupling method for deep processing of architectural glass, including cutting, glass arrangement, and tempering and matching. It solves the technical problems of existing technologies where the cutting, glass arrangement, and tempering and matching processes are isolated and lack closed-loop coordination, resulting in isolated production data that makes it impossible to dynamically adjust the production rhythm. At the same time, it lacks optimization of cutting layout, batch grouping, loading layout, and precise matching and closed-loop updating mechanism for process parameters, which leads to production stagnation and low efficiency.

[0004] To achieve the above objectives, this application adopts the following technical solution: Firstly, a coupling method for deep processing of architectural glass, including cutting, glass arrangement, and tempering, comprises: receiving production order data and, based on the maximum loading size and process characteristics of the tempering furnace, cutting the raw glass sheets required for the order with the goal of maximizing material utilization and tempering furnace loading compatibility, generating a cutting instruction set, and synchronously updating a dynamic surplus material database; controlling the cutting equipment to execute the cutting instruction set and transferring the cut glass sheets to the glass arrangement cage; and aggregating the equipment production data uploaded by each edge computing node through a central scheduling platform to form a global production view, wherein the equipment production data includes at least the real-time inventory status, size, and location information of the glass sheets in the glass arrangement cage, as well as the current operating status of the tempering furnace; the central scheduling platform is based on the global... In the production view, a Long Short-Term Memory (LSTM) network is used to predict the thickness distribution of glass batches entering the tempering process within the next 24 hours, generating a thickness distribution report. Based on the thickness distribution report and real-time information of the glass sheets in the sorting cage, the central scheduling platform optimizes batch grouping through a clustering algorithm, generating a batch grouping scheme for efficient loading into the tempering furnace and corresponding sorting cage instructions. Based on the batch grouping scheme and sorting instructions, the central scheduling platform runs a loading simulation algorithm to generate the optimal placement layout of the corresponding batch of glass in the tempering furnace. Combining the optimal placement layout with the physical characteristics of the batch of glass, the central scheduling platform automatically matches and sends the optimal tempering process parameters from the process parameter mapping library to the tempering furnace execution unit.

[0005] Based on the above technical solution, in the coupling method of deep processing of architectural glass cutting, glass arrangement and tempering and matching provided in this application, a closed-loop intelligent system can be formed by connecting the central scheduling platform to the edge perception layer (equipment nodes) and the execution layer (cutting, transfer and tempering equipment). During the operation of the intelligent system, the central scheduling platform first optimizes the cutting layout based on the tempering furnace constraints and generates an instruction set, and then combines it with the loading simulation program to perform cutting and transfer, thereby avoiding the problem of process interruption leading to stagnation. At the same time, during its operation, the edge computing nodes will also collect data in real time and summarize it to the platform to form a global production view. Then, combined with LSTM prediction and clustering algorithms, batch grouping and sorting instructions are dynamically generated. While avoiding the formation of data silos, the coupling efficiency of cutting, glass arrangement and tempering and matching is improved, thereby systematically solving the problems of process interruption, inability to adjust production rhythm and low efficiency.

[0006] In conjunction with the first aspect mentioned above, one possible implementation involves optimizing the cutting and layout of the raw glass sheets required for the order to generate a cutting instruction set, with the goal of maximizing material utilization and tempering furnace loading compatibility. Specifically, this includes: parsing production order data to obtain the area A of a single glass sheet to be processed. glass The constraint data of the tempering furnace is obtained from the tempering furnace parameter database. The constraint data includes the maximum loading area A. furnaceMinimum gap requirements between glass plates; based on maximizing material utilization. Maximizing compatibility with tempering furnaces Establish a multi-objective optimization model, where A raw Let A be the area of ​​the original glass plate. pattern The projected area of ​​the cutting pattern within the tempering furnace is defined. Based on a multi-objective optimization model, a mixed-integer programming solver is used to process the production order data and constraint data to generate a Pareto optimal solution set. Each solution in the Pareto optimal solution set corresponds to a cutting layout pattern. The cutting layout pattern with the highest comprehensive score is selected from the Pareto optimal solution set, and the selected pattern is parsed to generate a corresponding cutting instruction set. The cutting instruction set includes the cutting coordinates, cutting path sequence, and process parameters for each glass sheet. Information on leftover material with an area greater than or equal to a preset threshold generated by cutting is recorded and updated to the dynamic leftover material database.

[0007] In conjunction with the first aspect mentioned above, one possible implementation involves using a Long Short-Term Memory (LSTM) network to predict the thickness distribution of glass batches that will enter the tempering process within the next 24 hours and generating a thickness distribution report. This process specifically includes: Historical glass thickness sequences, order intervals, and equipment status information are extracted from the global production view. The mean, variance, and trend slope of the glass thickness are calculated to form a time-series feature vector. This time-series feature vector is then input into an LSTM prediction model to obtain the thickness distribution probability P for the next 24 hours. thickness(i,t） The prediction results, where i represents the thickness level and t represents the time point; through Calculate the time-series rate of change of thickness distribution P identifies the key time points for switching thickness distribution and the corresponding tempering furnace temperature adjustment time points, and generates a thickness distribution report. The thickness distribution report includes the dominant thickness change trend, the suggested batch grouping scheme, and the temperature adjustment time points. Based on the temperature adjustment time points in the thickness distribution report, the central scheduling platform controls the tempering furnace to complete the pre-adjustment of process parameters before executing the tempering of the corresponding batch of glass.

[0008] In conjunction with the first aspect mentioned above, in one possible implementation, the construction process of the LSTM prediction model specifically includes: constructing a hierarchical LSTM prediction model comprising an input layer, an LSTM hidden layer, and a prediction output layer; the input layer receives feature vectors, the number of elements in the feature vectors is denoted as the input layer dimension, and the length of the input sequence is set to 24 hours; the LSTM hidden layer uses three stacked LSTM units, with the hidden state passed between layers via a fully connected network using a Dropout mechanism; the prediction output layer uses a fully connected layer to obtain the glass thickness distribution trend for the next 24 hours, and outputs a probability distribution vector P. thickness , used to represent the predicted proportion of different thicknesses.

[0009] In conjunction with the first aspect mentioned above, one possible implementation involves optimizing batch grouping using a clustering algorithm to generate a batch grouping scheme and corresponding sorting instructions for efficient loading into the tempering furnace. Specifically, this includes extracting the size and thickness data of the glass sheets within the sorting cage from the global production view and thickness distribution report; to maximize the uniformity of glass thickness within the same group. To achieve the goal of maximizing the space utilization of the tunnel, parameters and distance metrics for the clustering algorithm were set, where h... k Let δ be the set of thicknesses of all glass sheets in the k-th group, and let δ be an indicator function, which is 1 when the condition is met and 0 otherwise. A density-based clustering algorithm is used to group the glass sheets to generate an initial batch grouping scheme. The clustering algorithm uses the thickness difference between glass sheets as the core distance metric dimension. Based on the initial batch grouping scheme, a sorting instruction sequence for glass sheets to enter the glass sheet handling cage aisle is generated. The sorting instruction sequence ensures that the thickness difference of adjacent glass sheets entering the aisle is ≤2mm as much as possible.

[0010] In conjunction with the first aspect mentioned above, one possible implementation involves using a density-based clustering algorithm to group glass sheets and generate an initial batch grouping scheme. Specifically, this includes: obtaining the maximum allowable thickness deviation of the tempering furnace as the cluster radius, and obtaining the minimum number of glass sheets loaded into the tempering furnace in a single operation as the minimum cluster size; constructing a three-layer clustering architecture comprising a physical layer, a spatiotemporal layer, and a process layer; the physical layer performing initial grouping based on glass thickness characteristics, dividing glass sheets with different thickness difference cluster radii into candidate clusters, and feeding these candidate clusters back to the spatiotemporal layer; the spatiotemporal layer receiving the candidate clusters and, combining area similarity and time continuity constraints, merging and optimizing the candidate clusters to obtain spatiotemporally optimized clusters; and the process layer receiving the spatiotemporally optimized clusters and, in conjunction with the tempering process... The furnace loading process requires final adjustments to the optimized clusters to obtain target clusters. The thickness uniformity within the target cluster is less than the cluster radius, and the number of glass sheets in the target cluster is between the minimum cluster size and the maximum loading capacity of the tempering furnace. All glass sheets in the sheet handling cage are traversed, and unvisited glass sheets are marked as unprocessed. For each unprocessed glass sheet, all glass sheet sets within the cluster radius's neighborhood are searched. If the glass sheet set is smaller than the minimum cluster size, a new cluster is created, and the unprocessed glass sheet is added to the new cluster. A comprehensive score for the target cluster is calculated based on the glass thickness variance and the space utilization rate of the sheet handling cage aisle. Clusters with comprehensive scores below a threshold are re-clustered or split and merged to generate an initial batch grouping scheme that meets the requirements for efficient tempering furnace loading.

[0011] In conjunction with the first aspect mentioned above, in one possible implementation, the process of running a loading simulation algorithm to generate the optimal placement layout of the corresponding batch of glass in the tempering furnace specifically includes: receiving the batch grouping scheme, sorting instructions, and constraint data of the tempering furnace; sorting the glass sheets in the sorting cage according to the sorting instructions to obtain the sorted glass sheet data; arranging the sorted glass sheet data in descending order of area to form a queue to be placed; initializing the effective loading area of ​​the tempering furnace as a set of free rectangles; and based on the maximum rectangle algorithm, placing the glass in the queue into the free rectangles that can accommodate it and have the smallest wasted area, while updating the list of free rectangles and recording the layout coordinate set; calculating the space utilization rate, heating uniformity index, and layout stability index of the current layout based on the layout coordinate set, and weighting them to obtain a layout evaluation score; traversing different glass placement orders to obtain all candidate layout schemes and their evaluation scores, and selecting the candidate layout scheme with the highest evaluation score as the optimal placement layout.

[0012] In conjunction with the first aspect mentioned above, in one possible implementation, the process of automatically matching and distributing the optimal tempering process parameters from the process parameter mapping library to the tempering furnace execution unit, based on the optimal placement layout and the physical characteristics of the batch of glass, specifically includes: extracting the physical characteristics of the current batch of glass from the batch grouping scheme, and receiving the optimal placement layout and its layout coordinate set generated by the loading simulation algorithm; matching the glass physical characteristics and layout coordinate set through the process parameter mapping library constructed based on the attention mechanism to obtain the final optimal tempering process parameter set; and distributing the optimal tempering process parameter set to the corresponding tempering furnace execution unit through the central scheduling platform.

[0013] In conjunction with the first aspect mentioned above, one possible implementation also includes, after tempering is completed, the central dispatch platform collects the quality inspection data of the final product and performs correlation analysis with the actual cutting parameters and tempering process parameters to update the process parameter mapping library: receiving quality inspection data from the quality inspection equipment, including glass flatness and stress distribution uniformity indicators; comparing the quality inspection data with the actual cutting parameters and the optimal tempering process parameters issued and executed in the central dispatch platform; and optimizing and correcting the mapping relationships stored in the process parameter mapping library based on the correlation comparison results, so that for the same or similar glass physical characteristics and optimal placement layout, the process parameter mapping library can match better tempering process parameters.

[0014] Secondly, a coupling device for deep processing of architectural glass, including cutting, glass arrangement, and tempering, is provided. The device comprises a communication unit and a processing unit. The communication unit receives production order data and, based on the maximum loading size and process characteristics of the tempering furnace, cuts the raw glass sheets required for the order, generates a cutting instruction set, and synchronously updates a dynamic surplus material database, aiming to maximize material utilization and tempering furnace loading compatibility. It controls the cutting equipment to execute the cutting instruction set and transfers the cut glass sheets to the glass arrangement cage. A central scheduling platform aggregates the equipment production data uploaded by each edge computing node to form a global production view. The equipment production data includes at least the real-time inventory status, size, and location information of the glass sheets in the glass arrangement cage, as well as the tempering furnace's... Current operating conditions: Based on the global production view, the central scheduling platform uses a Long Short-Term Memory (LSTM) network to predict the thickness distribution of glass batches entering the tempering process within the next 24 hours, generating a thickness distribution report. Based on the thickness distribution report and real-time information of the glass sheets in the sorting cage, the central scheduling platform optimizes batch grouping through a clustering algorithm, generating a batch grouping scheme for efficient loading into the tempering furnace and corresponding sorting cage instructions. Based on the batch grouping scheme and sorting instructions, the central scheduling platform runs a loading simulation algorithm to generate the optimal placement layout of the corresponding batch of glass in the tempering furnace. Combining the optimal placement layout with the physical characteristics of the batch of glass, the central scheduling platform automatically matches and sends the optimal tempering process parameters from the process parameter mapping library to the tempering furnace execution unit.

[0015] Thirdly, this application provides a coupling device for deep processing of architectural glass, including cutting, shearing, and tempering, comprising: a processor and a storage medium; the storage medium includes instructions, and the processor is used to execute the instructions to implement the method described in the first aspect and any possible implementation thereof. This coupling device for deep processing of architectural glass, including cutting, shearing, and tempering, can be an electronic device or a chip within an electronic device.

[0016] Fourthly, this application provides a computer-readable storage medium storing instructions that, when executed on a coupling device for deep processing, cutting, shearing, and tempering of architectural glass, cause the coupling device to perform the method described in the first aspect and any possible implementation thereof.

[0017] Fifthly, this application provides a computer program product containing instructions that, when run on a coupling device for deep processing, cutting, shearing, and tempering of architectural glass, causes the coupling device to perform the methods described in the first aspect and any possible implementation thereof.

[0018] This application provides a coupling method for deep processing of architectural glass, including cutting, glass arrangement, and tempering. By connecting a central scheduling platform to the edge perception layer (equipment nodes) and the execution layer (cutting, transfer, and tempering equipment), a closed-loop intelligent system can be formed. During operation, the central scheduling platform first optimizes the cutting layout based on the tempering furnace constraints and generates an instruction set. This is then combined with a loading simulation program for cutting and transfer, thus avoiding the problem of process interruption leading to stagnation. Simultaneously, during operation, edge computing nodes collect data in real time and aggregate it to the platform to form a global production view. Combined with LSTM prediction and clustering algorithms, batch grouping and sorting instructions are dynamically generated. This avoids the formation of data silos and improves the coupling efficiency of cutting, glass arrangement, and tempering, thereby systematically solving the problems of process interruption, inability to adjust production rhythm, and low efficiency. Attached Figure Description

[0019] Figure 1 A system architecture diagram of a coupling method for deep processing, cutting, shearing, and tempering of architectural glass provided in this application embodiment; Figure 2 A schematic flowchart illustrating a coupling method for deep processing of architectural glass, including cutting, shearing, and tempering, provided in an embodiment of this application. Figure 3 A schematic flowchart illustrating another coupling method for deep processing of architectural glass, including cutting, shearing, and tempering, provided in an embodiment of this application. Figure 4 A schematic flowchart illustrating another coupling method for deep processing of architectural glass, including cutting, shearing, and tempering, provided in an embodiment of this application. Figure 5 A schematic flowchart illustrating another coupling method for deep processing of architectural glass, including cutting, shearing, and tempering, provided in an embodiment of this application. Figure 6 A schematic flowchart illustrating another coupling method for deep processing of architectural glass, including cutting, shearing, and tempering, provided in an embodiment of this application. Figure 7 This is a schematic diagram of the coupling device for deep processing, cutting, shearing, and tempering of architectural glass, provided in an embodiment of this application. Detailed Implementation

[0020] In the description of this application, unless otherwise stated, "" means "or," for example, A / B can mean A or B. "And / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The words "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0021] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0022] To address the disconnect and lack of coordination between the cutting, glass sorting, and tempering processes in existing technologies, the following issues arise: The cutting process relies solely on order requirements for cutting and layout, neglecting the maximum loading size and process characteristics of the tempering furnace. This leads to potential manual adjustments or waiting for compatible batches due to poor furnace compatibility, causing inter-process stagnation. Similarly, the glass sorting process only serves as temporary storage, failing to consider real-time furnace conditions. After being transferred to the furnace, glass requires secondary sorting. The tempering process passively receives glass, unable to access information from previous stages, necessitating frequent shutdowns to adjust process parameters, further exacerbating stagnation. Furthermore, the production rhythm cannot be dynamically adjusted, lacking a mechanism to predict the thickness distribution of future glass batches entering the tempering process. Tempering furnace process parameters are only adjusted after glass arrives, disrupting the production rhythm. Additionally, production data from the cutting, sorting, and tempering equipment is fragmented. Data silos are created in the storage, preventing the central control system from aggregating global information and making it difficult to dynamically adjust the production process based on equipment load changes. The lack of optimized batch grouping logic also leads to an imbalanced production rhythm. Furthermore, the cutting process lacks a multi-objective optimization model that maximizes material utilization and ensures compatibility with the tempering furnace loading, potentially resulting in material waste or low furnace space utilization. Due to the lack of pre-adjustment of process parameters and reasonable batch grouping, the tempering furnace's capacity is not fully utilized, and adaptation and parameter adjustment at each stage rely on manual labor, increasing costs and reducing efficiency. Finally, the low accuracy of process parameter matching and the existence of repeated trial-and-error issues are problems. The lack of simulation algorithms to optimize the placement of glass sheets within the tempering furnace during tempering leads to uneven heating. The absence of a process parameter mapping library that can be updated and corrected based on quality inspection data means that the same or similar glass batches require repeated adjustments to process parameters, increasing trial-and-error costs.

[0023] like Figure 1As shown in the embodiment of this application, a coupling method for deep processing of architectural glass, including cutting, shearing, and tempering, comprises: Step 101: Receive production order data and, based on the maximum loading size and process characteristics of the tempering furnace, cut the original glass required for the order with the goal of maximizing material utilization and tempering furnace loading compatibility, generate a cutting instruction set, and synchronously update the dynamic surplus material database. Production order data refers to the glass processing requirements submitted by customers, typically including key parameters such as glass size, thickness, quantity, and delivery date. The maximum loading size of the tempering furnace refers to the limit in length and width of the glass sheets that the furnace can accommodate in a single operation; this is crucial spatial constraint data for cutting and layout. Process characteristics refer to the technical requirements of the tempering furnace in terms of heating temperature, cooling rate, and loading gaps. Material utilization rate is the ratio of the effective glass area after cutting to the total area of ​​the original glass sheet; maximizing this ratio reduces waste. Tempering furnace loading compatibility measures the fit of the cut and arranged glass assembly within the tempering furnace, minimizing wasted furnace space. The cutting instruction set is the set of commands that control the cutting machine to perform specific cutting operations, including cutting paths, coordinates, and sequence. The dynamic scrap database is a real-time updated database used to record and manage reusable glass scraps generated after cutting that reach a certain size threshold (e.g., 500mm x 500mm).

[0024] In some implementations, production order data from an Enterprise Resource Planning (ERP) system is received via an interface, and the specific specifications of the glass to be processed are parsed out. Simultaneously, the maximum loading size of the currently used tempering furnace (e.g., 6000mm x 3800mm) and the minimum gap between glass sheets are retrieved from a pre-set tempering furnace parameter database. This allows for a dual optimization objective of maximizing material utilization and tempering furnace loading compatibility. A multi-objective optimization mathematical model is established to ensure the matching degree between the area of ​​the original glass sheet, the total area of ​​the glass sheets required for the order, and the projected area of ​​the cutting pattern within the tempering furnace with the furnace's maximum loading area. A mixed-integer programming algorithm is then used to calculate and select the cutting layout scheme with the highest comprehensive score. This scheme is then parsed into a specific set of cutting instructions that can be executed by the cutting machine. This instruction set includes the cutting coordinates, path sequence, and process parameters such as cutting speed and pressure for each glass sheet. Furthermore, while generating the cutting instructions, the size and location of any excess material with an area greater than or equal to a pre-set threshold generated during the cutting process are automatically recorded and updated to a dynamic excess material database.

[0025] It should be noted that the objectives of maximizing material utilization and maximizing the loading of the tempering furnace are weighed during operation using a multi-objective optimization algorithm. This algorithm combines the specific business and production data of the cutting and processing enterprise to seek the optimal balance point and avoid potential conflicts and incompatibilities.

[0026] Step 102: Control the cutting equipment to execute the cutting instruction set and transfer the cut glass sheets to the glass handling cage; Cutting equipment refers to automated machinery that performs glass cutting operations, such as CNC cutting machines. Slab handling cages are automated storage and warehousing equipment used for the temporary storage and sorting of cut glass slabs; they typically contain multiple aisles for categorized storage.

[0027] In some implementations, the central dispatch platform sends the optimized cutting instruction set to the designated CNC cutting machine via the network. The cutting machine then receives the instruction and automatically performs positioning and cutting operations. After cutting the original glass sheet, the cut glass sheet is picked up from the cutting worktable and transferred to the designated aisle entrance of the glass handling cage by an automated robotic arm or conveyor system.

[0028] During the transfer process, based on information such as the size and thickness of the glass sheets, and in conjunction with pre-generated sorting instructions for the sorting cage, the glass sheets are placed into specific aisles of the sorting cage for temporary storage and sorting.

[0029] Step 103: The central scheduling platform aggregates the equipment production data uploaded by each edge computing node to form a global production view. The equipment production data includes at least the real-time inventory status, size and location information of the glass sheets in the glass handling cage, as well as the current operating status of the tempering furnace. The central dispatch platform refers to the software system responsible for the unified coordination and command of the entire production process. Edge computing nodes are computing units deployed near field equipment such as cutting machines, glass handling cages, and tempering furnaces, possessing data acquisition and preliminary processing capabilities. Equipment production data refers to the data set collected by edge computing nodes, reflecting the equipment's operating status and production process. The global production view is an overall picture reflecting the real-time status of the entire production line, formed by the central dispatch platform integrating all data. Size and location information refers to the specific length, width, and height dimensions of each piece of glass and its specific storage coordinates in the glass handling cage aisles. The current operating status of the tempering furnace refers to parameters such as the tempering furnace's real-time temperature, air pressure, and operating status.

[0030] In some implementations, edge computing nodes deployed near key equipment such as cutting machines, glass racks, and tempering furnaces continuously collect local equipment production data (e.g., real-time inventory status of glass sheets in each aisle, precise dimensions of each sheet, and specific stacking location information obtained through sensors and recognition systems on the glass racks; and current operating parameters such as temperature and pressure obtained through the tempering furnace control system). The collected data is then uploaded to a central dispatch platform via an industrial network. The central dispatch platform then performs data fusion processing, removes redundant and erroneous information, and integrates the inventory, size, and location information of the glass racks with the operating conditions of the tempering furnace. Finally, a unified and visualized global production view is formed on the platform interface, which can comprehensively and in real-time display the overall production situation from the inventory of the glass racks to the status of the tempering furnace.

[0031] Step 104: Based on the global production view, the central scheduling platform uses a Long Short-Term Memory (LSTM) network to predict the thickness distribution of glass batches that will enter the tempering process in the next 24 hours and generates a thickness distribution report. Among them, the Long Short-Term Memory (LSTM) network is a special recurrent neural network capable of learning and predicting time-series data, and can handle data with long-term dependencies. The glass batches to be tempered within the next 24 hours refer to the collection of glass expected to be sent to the tempering furnace for processing within the next day, based on the current production schedule and order plan. Thickness distribution refers to the quantity or proportion of glass of different thicknesses (e.g., 6mm, 8mm, 12mm, etc.) among these glass pieces to be tempered. The thickness distribution report includes the predicted trend of glass thickness over a specific future time period, key time points, and other information from the LSTM network's prediction results.

[0032] In some implementations, the central scheduling platform extracts historical glass thickness sequences, order intervals, and related equipment status information from the established global production view as basic data for preprocessing. It then calculates features such as the mean, variance, and trend slope of the glass thickness to construct a time-series feature vector. This time-series feature vector is then input into a pre-trained Long Short-Term Memory (LSTM) network prediction model to calculate the probability of different thickness grades of glass appearing at each time point within the next 24 hours. Based on the predicted probability distribution, the platform calculates its rate of change over time, identifying key time points where the thickness distribution changes significantly and pre-adjustment time points where the tempering furnace needs to adjust process parameters (such as temperature). Finally, all these prediction results, trend analyses, and key time points are integrated into a structured thickness distribution report.

[0033] Step 105: Based on the thickness distribution report and the real-time information of the glass sheets in the glass rack, the central dispatch platform optimizes batch grouping through clustering algorithm to generate a batch grouping scheme for efficient loading of the tempering furnace and the corresponding glass rack sorting instructions. Clustering algorithms are unsupervised machine learning methods that automatically group data points (glass sheets) with similar characteristics (such as similar thickness). Batch grouping schemes refer to the planning of dividing glass sheets in a sorting cage into several glass sets according to specific rules (such as thickness uniformity) for loading into the tempering furnace. Sorting cage sorting instructions are the sequential instructions generated to control the entry and exit of glass sheets in the sorting cage aisles in order to implement the grouping scheme, which can optimize the stacking order of glass in the aisles.

[0034] In some implementations, the central dispatch platform first reads the thickness distribution report and the real-time information of the glass sheets in the sorting cage (including the size, thickness, and position of each glass sheet). Then, based on the preset optimization goals of maximizing the thickness uniformity of the glass sheets in the same group (e.g., the maximum thickness difference within the group does not exceed 2 mm) and the space utilization of the sorting cage aisles, it uses a density-based clustering algorithm (such as the improved DBSCAN algorithm) to automatically group all the glass sheets in the sorting cage using the thickness difference between the glass sheets as the core distance metric dimension. This generates an initial batch grouping scheme, ensuring that the glass thickness in each group is as similar as possible and that the quantity meets the loading requirements of the tempering furnace. Furthermore, it generates corresponding sorting cage sorting instructions. At this point, the sorting cage sorting instruction sequence will plan the order in which the glass sheets enter the sorting cage aisles, ensuring that the thickness difference of adjacent glass sheets placed in the aisles is as small as possible.

[0035] Step 106: Based on the batch grouping scheme and sorting instructions, the central scheduling platform runs the loading simulation algorithm to generate the optimal placement layout of the corresponding batch of glass in the tempering furnace. Among them, the loading simulation algorithm is a computer simulation program used to simulate the placement process of glass sheets in a tempering furnace in a virtual environment to find the optimal solution. The optimal placement layout refers to the arrangement of glass sheets obtained through simulation calculations that maximizes the space utilization of the tempering furnace, achieves the best heating uniformity, and provides the most stable layout.

[0036] In some implementations, upon receiving a predetermined batch grouping scheme and corresponding sorting instructions for the glass sheets in the sorting cage, the algorithm logically organizes the glass sheets of a specified batch within the cage according to the sorting instructions. It obtains the accurate size data of all glass sheets in the batch and sorts them by area from largest to smallest, forming a queue of glass sheets to be placed. Then, a loading simulation algorithm is run, sequentially attempting to place the glass sheets from the queue into the rectangle in the current set of available rectangles that can accommodate them while minimizing wasted area. The list of available rectangles is updated after each glass sheet is placed, and the layout coordinates of that glass sheet are recorded. Simultaneously, after generating a candidate layout, the simulation algorithm calculates the space utilization rate, estimated heating uniformity, and stability index of the layout based on the layout coordinate set, and obtains a comprehensive layout evaluation score through weighted calculation. It then iterates through different glass sheet placement orders, generating multiple candidate layout schemes and calculating their respective evaluation scores. Finally, the candidate layout scheme with the highest evaluation score is selected as the optimal placement layout for that batch of glass in the tempering furnace.

[0037] For example, consider a pre-sorted batch containing 20 pieces of glass of varying sizes. The simulation algorithm will first select the piece with the largest area (e.g., 2400 mm²). Take a 1200mm glass piece and place it in a corner of the tempering furnace. Then take the second largest glass piece and try placing it in different positions in the remaining space, choosing the position that makes the remaining space after cutting the most regular and the waste the least. Repeat this process until all 20 pieces of glass have found a place, and finally form a compact layout pattern with uniform gaps. This pattern is the optimal placement layout, and the precise coordinates of each piece of glass in the furnace are recorded.

[0038] Step 107: The central dispatch platform, combining the optimal placement layout and the physical characteristics of this batch of glass, automatically matches and sends the optimal tempering process parameters from the process parameter mapping library to the tempering furnace execution unit.

[0039] The process of automatically matching and issuing the optimal tempering process parameters from the process parameter mapping library to the tempering furnace execution unit, based on the optimal placement layout and the physical characteristics of this batch of glass, specifically includes: Extract the physical characteristics of the glass in the current batch from the batch grouping scheme, and receive the optimal placement layout and its layout coordinate set generated by the loading simulation algorithm; Based on the physical characteristics of glass and the layout coordinate set, the optimal tempering process parameter set is obtained by matching through a process parameter mapping library built based on an attention mechanism. The optimal set of tempering process parameters is distributed to the corresponding tempering furnace execution unit through the central dispatch platform.

[0040] The process parameter mapping library is a knowledge base that stores the correspondence between glass physical characteristics, placement layout, and corresponding optimal tempering process parameters. Optimal tempering process parameters refer to a set of parameters, such as heating temperature, heating time, and cooling rate, verified through historical data to achieve the best tempering quality (e.g., stress, flatness) for a specific glass under a specific layout. The tempering furnace execution unit refers to the controller on the tempering furnace that receives instructions and specifically controls the furnace temperature, fans, and other equipment.

[0041] Based on the above technical solution, a closed-loop intelligent system can be formed by connecting the central scheduling platform to the edge perception layer (equipment nodes) and the execution layer (cutting, transfer, and tempering equipment). During operation, the central scheduling platform first optimizes the cutting layout based on the tempering furnace constraints and generates an instruction set. Combined with the loading simulation program, cutting and transfer are carried out, thus avoiding the problem of process fragmentation leading to stagnation. At the same time, during its operation, the edge computing nodes will also collect data in real time and summarize it to the platform to form a global production view. Then, combined with LSTM prediction and clustering algorithms, batch grouping and sorting instructions are dynamically generated. While avoiding the formation of data silos, the coupling efficiency of cutting, sheet handling, and tempering sheet matching is improved, thus systematically solving the problems of process fragmentation, inability to adjust production rhythm, and low efficiency.

[0042] In one possible implementation of the embodiments of this application, combined with Figure 1 ,like Figure 2 As shown, with the goal of maximizing material utilization and tempering furnace loading compatibility, the cutting and layout of the raw glass required for the order is optimized, and the cutting instruction set is generated through the following steps 201 to 205, which are explained in detail below: Step 201: Parse the production order data to obtain the area A of a single glass sheet to be processed. glass The constraint data of the tempering furnace is obtained from the tempering furnace parameter database. The constraint data includes the maximum loading area A. furnace Minimum clearance requirement between the glass plate and the glass plate; Among them, production order data is structured data generated by planning based on customer order information, including glass specifications (length, width), thickness, quantity, and delivery time, which can be parsed and calculated to determine the area A of a single glass sheet. glass The tempering furnace parameter database is a system that stores key constraint data for tempering equipment, including the maximum loading area A of the tempering furnace. furnace (Refers to the maximum total area of ​​glass that a tempering furnace can process at one time) and the minimum gap requirement between glass sheets (the minimum distance that must be reserved between adjacent glass sheets to ensure uniform heating and cooling and to prevent glass from colliding in the furnace).

[0043] In some implementations, production order data is first extracted from the acquisition module to obtain the specifications (length and width) of each piece of glass to be processed; the length and width values ​​of the glass are then parsed, allowing the area A of a single glass sheet to be automatically calculated using the basic area formula (length × width). glass The system then accesses a tempering furnace parameter database containing pre-stored process constraint parameters for a specific model of tempering furnace to obtain the maximum loading area A. furnace After determining the minimum gap requirement between the glass plates, compare it with A obtained from the order. glass The data, when combined, serve as key inputs for the pre-production scheduling process. Based on the actual dimensions of the loading platform and the minimum clearance requirements, the total number of glass pieces that the current production batch (furnace) can accommodate, the specific arrangement of the glass pieces in the furnace, and the total area of ​​these glass pieces are determined, thereby completing the optimized arrangement of the furnace.

[0044] For example, suppose the dimensions of a glass sheet to be tempered are 1000mm (length) × 800mm (width) as determined from a production order. Then, the area A of a single sheet is... glass The calculated value is 0.8 square meters. If the maximum loading area A is set in the target tempering furnace parameter database... furnace The area is 8 square meters, and the minimum gap between glass sheets is required to be 50 mm. This ensures that the total area of ​​all glass sheets (including the glass in this example and other glass that may be processed in the same furnace) within a single batch does not exceed 8 square meters. Secondly, when arranging the position of each glass sheet on the loading platform, an algorithm automatically arranges them to ensure that the closest distance between the edges of any two glass sheets is not less than 50 mm. If 8 sheets of this specification of glass are arranged in one furnace, theoretically the total net area occupied by the glass is 6.4 square meters. However, the layout must include the area occupied by the gaps in the overall space planning to ensure that the overall layout (including gaps) is within the acceptable range. furnace The distribution should be reasonable within a limited area of ​​8 square meters.

[0045] Step 202: Maximizing material utilization Maximizing compatibility with tempering furnaces Establish a multi-objective optimization model, where A raw Let A be the area of ​​the original glass plate. pattern The projected area of ​​the cut pattern inside the tempering furnace; Among them, the multi-objective optimization model is a mathematical framework for solving problems that simultaneously optimize multiple conflicting objectives. Its solution is usually a set of trade-off solutions (Pareto solution set). Maximizing material utilization means optimizing the cutting layout to maximize the total effective glass sheet area cut from the original glass plate (area Araw), thereby reducing material waste. Maximizing tempering furnace compatibility focuses on the efficiency of the arrangement of the cut glass sheets within the tempering furnace. By optimizing the arrangement of the projected area (Apattern) of the cutting pattern within the tempering furnace, the total area of ​​glass placed into the furnace at a time is made as close as possible to the furnace's maximum loading area, while also meeting the minimum gap requirement between glass sheets, thereby improving the furnace's single-pass processing capacity.

[0046] In some implementations, the effective total area of ​​the glass sheet and the original glass plate area A are first quantified by maximizing the material utilization objective function. raw The ratio of the two glass plates is such that the closer the ratio is to 1, the higher the material utilization rate. Simultaneously, the total projected area A occupied by all glass plates and their necessary gaps in a single loading of the tempering furnace is quantified using the objective function of maximizing furnace compatibility. pattern With the maximum loading area A of the tempering furnace furnace The ratio of the two objectives is such that the closer the ratio is to 1, the more fully the furnace space is utilized. When these two objectives conflict (e.g., improving material utilization may require producing more irregularly shaped or small-sized glass sheets, affecting efficient layout within the furnace), a multi-objective optimization model is constructed with the objectives of maximizing material utilization and maximizing furnace compatibility. The decision variables of the multi-objective optimization model include integer and continuous variables such as the cutting position, rotation angle, and layout of the glass sheets, and are subject to multiple constraints from production orders and furnace parameters. This allows for the effective search of the Pareto optimal solution set, balancing the structures of the two objective functions. Furthermore, by determining the decision variables (the cutting pattern layout on the original glass sheet, which determines the utilization of Araw) and the arrangement of these cut glass sheets on the furnace's carrying area (corresponding to Apattern), and setting the maximum loading area A of the furnace as the maximum loading area of ​​the furnace, the Pareto optimal solution set can be effectively found, balancing the structures of the two objective functions. furnace The upper limit, along with the minimum gap requirement between glass plates that must be met to ensure uniform heating and cooling, serves as constraints. Therefore, specialized multi-objective optimization algorithms (such as Non-Dominated Sorting Genetic Algorithm (NSGA-II) or Multi-Objective Particle Swarm Optimization (MOPSO)) are used to search for and output a uniformly distributed set of Pareto optimal solutions. Production decision-makers can then select the most suitable compromise from the Pareto solution set to guide actual production, based on current production needs and preferences (such as prioritizing material cost savings or improving equipment throughput efficiency).

[0047] Step 203: Based on the multi-objective optimization model, a mixed integer programming solver is used to process the production order data and constraint data to generate a Pareto optimal solution set. Each solution in the Pareto optimal solution set corresponds to a cutting and layout pattern. Among them, the mixed-integer programming solver is a computational tool used to solve mathematical programming problems containing both continuous and integer variables, capable of handling complex constraints. The cutting and layout pattern is determined based on a solution in the Pareto optimal solution set, specifying the cutting positions and arrangement of each glass piece on the original glass plate. Its goal is typically to maximize material utilization and meet the loading requirements of the tempering furnace.

[0048] In some implementations, The solver uses its internal algorithms (such as branch and bound, and cutting plane method) to search the feasible region of decision variables, analyzes production order data, and obtains the specific specifications and quantity requirements of the glass sheets to be processed. The constraint data ensures that the generated layout scheme meets the actual production capacity and process requirements of the tempering furnace (such as minimum gap). Each solution in the Pareto optimal solution set is transformed into a specific, executable cutting layout pattern. This cutting layout pattern specifies in detail the precise cutting position and direction of each glass sheet on the raw material plate. Production planners can select the most suitable scheme from this solution set for production based on the core optimization objectives at the time (such as prioritizing material saving or equipment throughput efficiency).

[0049] For example, suppose there is a large original glass plate with an area Araw of 10 square meters, which needs to be cut into several small glass pieces of different sizes. The maximum loading area A of the tempering furnace is then... furnace The area to be used is 8 square meters, and the minimum gap between the glass panes is required to be 5 centimeters. After running the multi-objective optimization model, a set of Pareto optimal solutions will be generated: for example, solution A has a high material utilization rate, reaching 95% (i.e., effectively utilizing 9.5 square meters of the original glass panes), but its layout projection area A... pattern It might only be 7.5 square meters, in which case the utilization rate of the tempering furnace loading space is relatively low; the material utilization rate of scheme B is 90% (using 9 square meters of raw materials), but its layout projection area A pattern The area could reach 7.9 square meters, very close to the furnace's maximum loading area, allowing a single furnace to process more glass and improving equipment efficiency. At this point, the production planner can make a final decision based on whether the current objective is to prioritize reducing material waste (Option A) or increasing production capacity (Option B).

[0050] Step 204: Select the cutting layout pattern with the highest comprehensive score from the Pareto optimal solution set, and parse the selected pattern to generate the corresponding cutting instruction set. The cutting instruction set includes the cutting coordinates, cutting path sequence and process parameters of each glass piece. The highest overall score refers to the highest score obtained after quantifying the objective function values ​​(such as material utilization and tempering furnace compatibility) of each solution in the Pareto optimal solution set using a weighted model. The weights can be dynamically allocated based on actual needs. The cutting instruction set is a collection of programs that guide the CNC cutting machine to perform specific operations, including the cutting coordinates (defining the start and end points of the contour), the cutting path sequence (optimized tool movement sequence), and process parameters (such as cutting speed and laser power) for each glass sheet.

[0051] In some implementations, a comprehensive score for each cutting pattern in the Pareto optimal solution set can be calculated using a weighted sum model by setting weighting factors based on the coefficient of variation method or production requirements (such as material cost priority and equipment throughput requirements). The pattern with the highest comprehensive score is then selected. Geometric data containing the vertex coordinates, boundary relationships, and relative positions of each glass sheet on the original glass plate is analyzed. A path optimization algorithm (such as the nearest neighbor method) is used to plan the cutting sequence to minimize idle travel time. The analytical results are converted into an instruction set recognizable by the CNC cutting machine. The cutting coordinates are mapped from the design coordinate system to the machine coordinate system through geometric transformation. The cutting path sequence is generated based on the contour topology to reduce the number of tool lifts. Process parameters are matched from a preset database based on glass thickness and material.

[0052] Step 205: Record and update the information of leftover material with an area greater than or equal to the preset threshold generated by cutting to the dynamic leftover material database.

[0053] The preset threshold is a manually set minimum area, length, or width limit value used to determine whether surplus material has secondary utilization value. It is usually set based on the size of commonly used small workpieces or the processing capacity of the equipment. The dynamic surplus material database is a structured data system that records and manages surplus material information in real time. Its core fields include surplus material number, dimensions (length, width, thickness), area, shape outline, inventory location, status (available / occupied), etc., and it can support multi-user collaborative updates and queries.

[0054] In some implementations, after the cutting equipment completes the cutting instructions for the current order, it automatically scans the remaining area of ​​the sheet metal. Using geometric algorithms (such as calculating the minimum bounding rectangle or actual outline polygon of the remaining area), it calculates the area of ​​the leftover material in real time and compares it with preset thresholds (such as minimum area, minimum width, and minimum length). If the area of ​​the leftover material is greater than or equal to the threshold, a leftover material warehousing process is triggered to generate a unique number for the leftover material and record its size, shape coordinates, material, and other attributes. Simultaneously, the "available leftover material list" in the dynamic leftover material database is updated. If the area of ​​the leftover material is less than the threshold, it is marked as waste and excluded from management. The dynamic leftover material database also employs a real-time update mechanism. When new leftover material is added or existing leftover material is called, the database status is updated synchronously. For example, during the material layout optimization stage, the system can prioritize retrieving matching leftover material from the database for layout. If the leftover material is successfully selected, its status automatically changes to "occupied." Furthermore, the database supports leftover material lifecycle tracking, including warehousing time and call records, and integrates with production management systems (such as MES and ERP) to ensure data flow and production process coordination.

[0055] Based on the above technical solution, by establishing a multi-objective optimization model with the goals of maximizing material utilization and tempering furnace compatibility, a mixed-integer programming solver can be used to generate a Pareto optimal solution set containing multiple non-dominated solutions, with each solution corresponding to a specific cutting and layout pattern. This automatically generates a series of feasible solutions with different trade-offs between material utilization and equipment efficiency, providing decision-makers with a wealth of optimization options. This avoids the problem of traditional cutting and layout methods, which often rely on manual experience or simple single-objective optimization algorithms, making it difficult to simultaneously consider material cost and equipment efficiency. This allows the layout scheme to fully utilize the tempering furnace's loading capacity or meet material utilization requirements in actual production. Simultaneously, by parsing production order data and querying the tempering furnace parameter database, the actual order requirements (such as glass sheet size and quantity) and physical equipment constraints (such as maximum loading area) can be considered. AfurnaceUsing minimum clearance requirements as core input parameters and constraints of the optimization model, and in conjunction with a dynamic scrap database, the system can automatically identify and record information on reusable scrap (such as size, shape, and inventory location) generated after cutting by setting an area threshold, and update the database status in real time. This ensures that the generated cutting layout scheme is not only theoretically optimized but can also be directly applied to the production line, reducing data silos. Furthermore, it prioritizes the use of these scraps in order layout optimization, promoting the recycling of scraps within the factory and reducing raw material waste. Finally, by automatically selecting the cutting layout pattern with the highest comprehensive score, multi-objective decisions are effectively transformed into single-objective decisions, achieving automatic optimization of the scheme. The selected pattern is then analyzed and a cutting instruction set containing cutting coordinates, path sequence, and process parameters is automatically generated, directly driving the CNC cutting machine for high-precision operation. This avoids the need for manual conversion or cumbersome secondary programming from the layout diagram to executable machine tool instructions (such as G-code), reducing error rates and improving overall production efficiency.

[0056] In one possible implementation of this application embodiment, the thickness distribution of glass batches that will enter the tempering process within the next 24 hours is predicted using a Long Short-Term Memory (LSTM) network, and a thickness distribution report is generated. This can be achieved through the following steps 301 to 304, which are described in detail below: Step 301: Extract historical glass thickness sequence, order interval time and equipment status information from the global production view, and calculate the mean, variance and trend slope of glass thickness to form a time series feature vector; The global production view refers to a real-time panoramic view of production status formed by the central scheduling platform after integrating equipment production data uploaded from various edge computing nodes. It covers multi-dimensional information such as glass handling cage inventory and tempering furnace operating conditions. The historical glass thickness sequence is a collection of processed glass thickness data recorded in chronological order, used to analyze thickness variation patterns. The order interval refers to the time difference between the arrival or processing of consecutive orders, reflecting fluctuations in production rhythm. Equipment status information includes real-time operating parameters such as tempering furnace temperature, air pressure, and glass handling cage aisle occupancy rate.

[0057] In some implementations, the central scheduling platform extracts historical glass thickness sequences from the global production view, sorts them by timestamp, and removes outliers. Simultaneously, it collects order interval data (such as the difference between the completion time of the previous order and the start time of the current order) and equipment status information (such as the current temperature of the tempering furnace and the number of idle sections in the glass handling cage aisles). The platform then sums the thickness data within a specified time window and divides by the number of data points to obtain the average thickness level. Furthermore, it averages the squared differences between each thickness value and the mean to quantify the thickness fluctuation range. Simultaneously, it uses the least squares method to fit a linear relationship between thickness and time, using the slope value to indicate the trend of thickness increase or decrease. The mean, variance, trend slope, and key equipment status indicators (such as the tempering furnace load rate) are then concatenated into a multi-dimensional vector to form a standardized time-series feature vector, which serves as the input to the LSTM model.

[0058] Step 302: Input the time series feature vector into the LSTM prediction model to obtain the thickness distribution probability P for the next 24 hours. thickness(i,t） The prediction results are given, where i represents the thickness level and t represents the time point. The LSTM prediction model is a recurrent neural network that learns long-term dependencies in time series data through a gating mechanism and possesses long short-term memory capabilities. The thickness distribution probability P for the next 24 hours... thickness(i,t） It is the probability matrix output by the model.

[0059] The construction process of the LSTM prediction model specifically includes: A hierarchical LSTM prediction model is constructed, consisting of an input layer, an LSTM hidden layer, and a prediction output layer. The input layer receives feature vectors, and the number of elements in the feature vectors is denoted as the input layer dimension. The length of the input sequence is set to 24 hours. The LSTM hidden layer uses three stacked LSTM cells, and the hidden state is passed between layers through a fully connected network using the Dropout mechanism. The prediction output layer uses a fully connected layer to obtain the glass thickness distribution trend for the next 24 hours, and outputs a probability distribution vector P. thickness , used to represent the predicted proportion of different thicknesses.

[0060] In some implementations, the preprocessed time-series feature vectors are segmented into 24-hour segments and input into a trained three-layer LSTM hidden layer structure. In this case, the input layer dimension of the LSTM prediction model is consistent with the feature vector dimension (e.g., if it includes six features such as mean, variance, and slope, the dimension is 6). The LSTM hidden layers process the sequence data cyclically through forget gates, input gates, and output gates to capture periodic patterns of thickness changes. A dropout mechanism is used between layers to randomly disable some neurons to prevent overfitting. Finally, the final state of the hidden layers is passed to the prediction output layer through a fully connected layer, and the output is converted into a probability distribution using the Softmax activation function, generating the probability value P for each thickness level i at each time point t (1-hour intervals) within the next 24 hours. thickness(i,t） This forms a 24×i-dimensional probability matrix. This allows the probability matrix to be packaged into a thickness distribution report.

[0061] Step 303, through Calculate the time-series rate of change of thickness distribution P identifies the key time points for the switching of thickness distribution and the corresponding time points for adjusting the tempering furnace temperature, and generates a thickness distribution report. The thickness distribution report includes the dominant thickness change trend, the suggested batch grouping scheme, and the temperature adjustment time points. Among them, the time-series change rate of thickness distribution P is a quantitative indicator obtained by calculating the maximum absolute value of the difference in thickness probability distributions between adjacent time points, used to measure the intensity of fluctuations in the thickness distribution. The key time point for the switching of the thickness distribution refers to... The time point at which the probability distribution changes significantly when P exceeds the preset threshold. The tempering furnace temperature adjustment time point is a pre-adjustment time for process parameters reserved by shifting the process parameters forward from the critical switching time point.

[0062] In some implementations, the thickness distribution probability P for the next 24 hours is based on the output of the LSTM model. thickness(i,t） Calculate the rate of change over time points. P, and set The threshold for P is 0.3 (i.e., the probability change exceeds 30%), when two consecutive time points... When all values ​​(P) exceed the threshold, this time point is marked as the critical switching point for thickness distribution. Based on the lead time required for tempering furnace temperature adjustment (e.g., 15 minutes), the corresponding tempering furnace temperature adjustment time point is derived by subtracting the adjustment time from the critical switching time point. At this point, a thickness distribution report can be automatically generated using the dominant thickness change trend identified by the trend slope (e.g., "the proportion of 8mm glass continues to rise"), suggested batch grouping schemes generated based on thickness clustering results (e.g., "process 6mm and 8mm glass in separate furnaces"), and a list of temperature adjustment time points accurate to the minute.

[0063] Step 304: Based on the temperature adjustment time points in the thickness distribution report, the central dispatch platform controls the tempering furnace to complete the pre-adjustment of process parameters before performing the tempering of the corresponding batch of glass.

[0064] The central dispatch platform is the intelligent control center responsible for coordinating the entire process of cutting, glass handling, and tempering. Process parameter pre-adjustment refers to setting key parameters such as heating temperature, heating time, and cooling rate of the tempering furnace in advance based on the thickness characteristics of the glass to be processed. The corresponding batch refers to the set of glass that is about to enter the tempering stage, grouped by thickness clustering algorithm.

[0065] In some implementations, the central dispatch platform first parses the thickness distribution report, extracts the temperature adjustment time point sequence marked in it (e.g., t1=10:15, t2=14:30), and matches it with the batch entry time into the tempering furnace in the production plan. Then, based on the dominant thickness change trend in the thickness distribution report (e.g., "8mm glass accounts for more than 70%)" and the batch grouping scheme, it retrieves the corresponding optimal tempering process parameter set from the process parameter mapping library (e.g., the heating temperature of 8mm glass is set to 695℃). Thus, before each temperature adjustment time point arrives, it issues parameter pre-adjustment instructions (including specific values ​​such as heating zone temperature curve and fan speed) to the target tempering furnace execution unit through the industrial network. This allows the tempering furnace to complete the parameter switching before the actual entry of the glass batch and feed back the ready status to the central dispatch platform, ensuring seamless connection of subsequent tempering operations.

[0066] It should be noted that the pre-adjustment instruction must be issued at least 15 minutes before the batch enters the tempering furnace to allow sufficient time for the furnace temperature to stabilize. Furthermore, if multiple thickness switching requirements exist within the same time period, the platform will merge the adjustment nodes according to the principle of "minimizing thickness differences."

[0067] Based on the above technical solution, by integrating fragmented thickness sequences, order intervals, equipment status, and other multi-source information into a time-series feature vector, a comprehensive quantification of production status can be achieved, laying a data foundation for accurate prediction. This avoids the problems of incomplete information and biased decision-making caused by data silos. Simultaneously, leveraging the advantages of neural networks in processing time-series data, the input feature vector is transformed into a probability distribution of different thickness levels for the next 24 hours. This effectively bridges the gap between qualitative experience-based judgment and quantitative probability prediction, making production planning more forward-looking. This avoids the problems of existing technologies where the lack of effective prediction tools leads to an inability to predict thickness changes and passive adaptation to production rhythm. Furthermore, differential analysis of the predicted probability distribution automatically captures significant changes in the distribution, transforming thickness switching timing from fuzzy estimation to precise identification, providing clear time anchors for process adjustments. Finally, the central scheduling platform directly drives the tempering furnace execution unit based on the conclusions (trends, grouping schemes, adjustment time points) in the thickness distribution report, enabling seamless and early switching of process parameters and minimizing production preparation time. This avoids the problems of long equipment waiting times and severe production stoppages caused by process fragmentation, information transmission delays, and manual parameter settings.

[0068] In one possible implementation of this application embodiment, batch grouping optimization using a clustering algorithm to generate a batch grouping scheme for efficient loading of the tempering furnace and the corresponding rack sorting instructions can be achieved through the following steps 401 to 404, which are described in detail below: Step 401: Extract the size and thickness data of the glass sheets in the glass rack from the global production view and thickness distribution report; Step 402: To maximize the uniformity of glass thickness within the same group. To achieve the goal of maximizing the space utilization of the alleyway, the parameters and distance metric of the clustering algorithm are set, where h... k Let δ be the set of thicknesses of all glass plates in the k-th group, and let δ be an indicator function, which is 1 when the condition is met and 0 otherwise. Among these, glass thickness uniformity within a group refers to minimizing the thickness difference between glass sheets entering the tempering furnace in the same batch, typically with an optimization objective of no more than 2 mm for the maximum thickness difference within the group. Sheet handling cage aisle space utilization rate refers to the proportion of aisle space within the sheet handling cage that is effectively used for storing glass, characterizing the maximization of the number of glass sheets accommodated and the compactness of their arrangement. Clustering algorithms are unsupervised machine learning methods used to automatically group data objects with similar characteristics. Distance metrics are mathematical methods used in clustering algorithms to measure the similarity or difference between data points, such as Euclidean distance or custom distance functions.

[0069] In some implementations, the goal is to maximize the uniformity of glass thickness within the same group and the space utilization of the glass handling cage aisle. The parameters of the clustering algorithm are set (including the cluster radius representing the maximum allowable thickness deviation of the tempering furnace and the minimum cluster size representing the minimum number of glass sheets loaded per tempering furnace load) and the distance metric (using the thickness difference between glass sheets as the core distance metric dimension). A density-based clustering algorithm can then be used to group the glass sheets, generating an initial batch grouping scheme. Based on this initial batch grouping scheme, a sequence of sorting instructions for the glass sheets entering the glass handling cage aisle is generated. This sorting instruction sequence ensures that the thickness difference between adjacent glass sheets entering the aisle is ≤2mm, thereby optimizing thickness uniformity and space utilization.

[0070] Step 403: Use a density-based clustering algorithm to group the glass slides and generate an initial batch grouping scheme. The clustering algorithm uses the thickness difference between the glass slides as the core distance metric dimension. Density-based clustering is a machine learning method that groups data points according to their density distribution in the feature space. It can identify clusters of arbitrary shapes and effectively handle noisy points. The process of grouping glass slides using density-based clustering to generate an initial batch grouping scheme specifically includes: The maximum allowable thickness deviation of the tempering furnace is recorded as the cluster radius, and the minimum number of glass sheets loaded into the tempering furnace in a single operation is recorded as the minimum cluster size. A three-layer clustering architecture consisting of a physical layer, a spatiotemporal layer, and a process layer is constructed. The physical layer performs initial grouping based on glass thickness characteristics, divides glass sheets with different glass thickness clustering radii into candidate clusters, and feeds the candidate clusters back to the spatiotemporal layer. The spatiotemporal layer receives candidate clusters and, combining area similarity and temporal continuity constraints, merges and optimizes the candidate clusters to obtain the spatiotemporally optimized clusters. The process layer receives the spatiotemporally optimized clusters and, in conjunction with the tempering furnace loading process requirements, makes final adjustments to the optimized clusters to obtain the target clusters. The thickness uniformity within the target clusters is less than the cluster radius, and the number of glass sheets in the target clusters is between the minimum cluster size and the maximum loading capacity of the tempering furnace. Traverse all glass slides in the processing cage, mark unvisited glass slides as unprocessed, for each unprocessed glass slide, find all glass slide sets within the cluster radius neighborhood, if the glass slide set is smaller than the minimum cluster size, create a new cluster and add the unprocessed glass slide to the new cluster; The comprehensive score of the target cluster is calculated based on the glass thickness variance and the space utilization rate of the glass processing cage aisle. Clusters with comprehensive scores below the threshold are re-clustered or split and merged to generate an initial batch grouping scheme that meets the requirements of efficient loading of the tempering furnace.

[0071] In this context, intra-cluster thickness uniformity refers to the maximum difference in thickness among all glass sheets within the same cluster. Glass thickness variance is a statistical indicator that measures the dispersion of a set of glass thicknesses.

[0072] Step 404: Based on the initial batch grouping scheme, generate a sorting instruction sequence for glass sheets to enter the glass sheet handling cage aisle. The sorting instruction sequence ensures that the thickness difference between adjacent glass sheets entering the aisle is ≤2mm as much as possible.

[0073] The initial batch grouping scheme refers to a method of initially grouping glass slides using a clustering algorithm to maximize the uniformity of glass thickness within the same group and the space utilization of the glass handling cage aisles. A glass handling cage aisle refers to a specific channel or compartment inside the glass handling cage used for storing and organizing glass slides.

[0074] In some implementations, based on the generated initial batch grouping scheme, the specific order in which the glass sheets enter the sorting cage aisle is planned. That is, the core optimization objective can be "to ensure that the thickness difference between adjacent glass sheets entering the aisle is ≤2mm as much as possible". Combined with the thickness distribution of all glass sheets in the current batch, a specific sorting instruction sequence is generated. Thus, through algorithm optimization, the thickness of two consecutive glass sheets is made as close as possible during the process of the glass sheets being sent into the sorting cage aisle, thereby minimizing thickness fluctuations.

[0075] Based on the above technical solution, by combining real-time global production views with predictive thickness distribution reports as input data for clustering, grouping decisions can be made not only based on current inventory status but also incorporating predictions of future production trends, thus achieving a shift from passive response to proactive planning. Simultaneously, with the dual objectives of "maximizing the uniformity of glass thickness within the same group" and "maximizing the space utilization rate of the sheet cage aisles," the parameters of the clustering algorithm (such as cluster radius and minimum cluster size) and the core distance metric (glass thickness difference) are directly set. This effectively ensures that the clustering algorithm serves the tempering process from its inception, not only improving the processing efficiency and quality stability of the tempering furnace but also effectively optimizing the space management and material flow efficiency of the sheet handling cage. This makes the storage of glass sheets in the handling cage more orderly and compact, reducing space waste and making the subsequent feeding process to the tempering furnace smoother and more predictable.

[0076] In one possible implementation of this application embodiment, running the loading simulation algorithm to generate the optimal placement layout of the corresponding batch of glass in the tempering furnace can be achieved through the following steps 501 to 504, which are described in detail below: Step 501: Receive the batch grouping scheme, sorting instructions and tempering furnace constraint data, sort the glass sheets in the sorting cage according to the sorting instructions, obtain the sorted glass sheet data, and arrange the sorted glass sheet data in descending order of area to form a queue to be placed. The queue to be placed is a sequence formed by sorting the glass sheets from largest to smallest area, which is used for subsequent tempering furnace loading simulation.

[0077] In some implementations, the central dispatch platform receives the determined batch grouping scheme, the corresponding sorting cage instructions, and the constraint data of the tempering furnace. It then uses sensors and recognition systems to acquire the precise dimensions, thickness, and aisle location information of each glass pane. Based on the sorting instructions, it automatically organizes the glass panes of a specified batch within the sorting cage, forming structured, organized glass pane data. Based on this data, the glass panes are arranged by area, with the largest panes at the front of the queue and the smallest at the back, creating an ordered queue for placement. This queue then serves as input to the loading simulation algorithm, ensuring that large-area glass panes are prioritized during subsequent tempering furnace layout optimization, thereby improving space utilization.

[0078] Step 502: Initialize the effective loading area of ​​the tempering furnace as a set of free rectangles, and based on the maximum rectangle algorithm, place the glass pieces in the queue into the free rectangles that can accommodate them and have the smallest wasted area in turn, while updating the list of free rectangles and recording the layout coordinate set. The set of free rectangles initializes the effective loading area as a collection of empty rectangles, each representing a contiguous available space. The maximum rectangle algorithm optimizes spatial layout by always selecting the largest available empty rectangle for placement. Minimizing wasted area means minimizing the unused area remaining after cutting when a glass sheet is placed in a free rectangle. The layout coordinate set is a set of precise coordinate data recording the final placement position of each glass sheet within the tempering furnace.

[0079] In some implementations, the effective loading area of ​​the tempering furnace is initialized as a set of free rectangles containing a single maximum rectangle (the rectangle completely covers the available space of the loading area). Then, using the maximum rectangle algorithm, glass sheets are sequentially retrieved from the queue of sheets to be placed, sorted by area in descending order. The current set of free rectangles is traversed, and the remaining area generated after each rectangle accommodates the glass sheet is calculated. The free rectangle that minimizes wasted area is then selected as the optimal placement position, and the glass sheet is placed at the lower left corner of that rectangle or at a specified starting position. Immediately after placement, the set of free rectangles is updated. A rectangle segmentation algorithm is used to remove used rectangles from the set, and the remaining space is divided into new, smaller rectangles and added to the set. Simultaneously, the layout coordinates of the current glass sheet within the tempering furnace, including X-axis, Y-axis coordinates and rotation angle, are recorded to form a continuously updated set of layout coordinates. This process is repeated until all glass sheets in the queue have completed their placement calculations.

[0080] Step 503: Based on the layout coordinate set, calculate the space utilization rate, heating uniformity index and layout stability index of the current layout, and obtain the layout evaluation score by weighting. Among them, the heating uniformity index is a parameter that evaluates the uniformity of heating of the glass sheets in the furnace, and is calculated based on the glass sheet spacing distribution and edge distance. The layout stability index is a quantitative parameter that analyzes the glass sheets' resistance to displacement during the tempering process, and is evaluated through the center of gravity distribution and support structure.

[0081] In some implementations, based on the generated layout coordinate set, the geometric position data of each glass pane and the boundary information of the effective loading area of ​​the tempering furnace are extracted. The sum of the projected areas of each glass pane can be divided by the total furnace area to obtain a percentage utilization value. Then, the heating uniformity is quantified by measuring the uniformity of the minimum gap distribution between the glass panes and the variance of the distance from the edge of each glass pane to the furnace boundary. At the same time, the offset distance between the overall center of gravity of the glass pane group and the geometric center of the furnace is calculated, and the presence of any suspended or insufficiently supported glass panes is detected. Finally, these three indicators are multiplied by preset weighting coefficients (such as space utilization weight 0.5, heating uniformity weight 0.3, and layout stability weight 0.2), and the weighted sum is used to obtain the final layout evaluation score, which serves as the basis for selecting the optimal placement layout.

[0082] Step 504: Traverse different glass placement orders to obtain all candidate layout schemes and their evaluation scores, and select the candidate layout scheme with the highest evaluation score as the optimal placement layout.

[0083] In some implementations, a permutation and combination algorithm is used to traverse different glass placement orders to generate multiple candidate layout schemes. For each candidate scheme, an evaluation score is calculated based on its layout coordinate set. Then, the evaluation scores of all candidate schemes are sorted in descending order, and the scheme with the highest score is automatically selected as the final optimal placement layout, which can be directly used to guide the actual loading operation of the tempering furnace.

[0084] Based on the above technical solution, by receiving batch grouping schemes and sorting instructions from upstream, the input data is ensured to be synchronized with the overall production rhythm. A closed-loop algorithm process of "initialization-placement-update-evaluation" can then be adopted. The effective loading area of ​​the tempering furnace is abstracted into a dynamically changing set of free rectangles. A greedy placement method using the maximum rectangle algorithm is employed, and the spatial status is updated in real time. Through multi-index weighted scoring and global traversal comparison, quantitative optimization of the layout scheme is achieved. This transforms the layout process, which originally relied on manual experience, into a precise control process driven by data and automatically optimized by algorithms. This effectively improves the furnace loading density through area descending queues and the principle of minimum waste, while ensuring thermal uniformity and physical safety during the tempering process through heating uniformity and layout stability indicators. Finally, the optimal scheme is automatically locked through a traversal evaluation mechanism, avoiding manual trial and error.

[0085] In one possible implementation of this application embodiment, after tempering is completed, the central scheduling platform collects the quality inspection data of the final product and performs correlation analysis with the actual cutting parameters and tempering process parameters. Updating the process parameter mapping library can be achieved through the following steps 601 to 603, which are described in detail below: Step 601: Receive quality inspection data from the quality inspection equipment. The quality inspection data includes indicators such as the flatness of the glass and the uniformity of stress distribution. Among them, the quality inspection data refers to the set of quality index data of the final product (tempered glass) collected by quality inspection equipment. This includes the flatness of the glass (i.e. the degree of deviation of the glass surface from the ideal plane, usually measured by the difference between the maximum peak and trough heights per unit area) and the stress distribution uniformity index (referring to the distribution of residual stress inside the glass after tempering, which can reflect the uniformity and quality stability of the tempering process).

[0086] Step 602: Compare and correlate the quality inspection data with the actual cutting parameters and the optimal tempering process parameters issued and executed in the central dispatch platform; In some implementations, the central dispatch platform first receives quality inspection data uploaded from the quality inspection equipment. Then, it compares and correlates this data with the actual cutting parameters stored internally (i.e., the real parameter records generated when the cutting equipment executes the cutting instruction set) and the optimal tempering process parameters issued for execution (i.e., the parameter set previously matched from the process parameter mapping library and issued to the tempering furnace execution unit). The analysis yields a quantitative relationship between the current process parameter settings and product quality, and establishes a correspondence between quality results and process inputs.

[0087] Step 603: Based on the correlation comparison results, optimize and correct the mapping relationship stored in the process parameter mapping library so that the process parameter mapping library can match better tempering process parameters for the same or similar glass physical characteristics and optimal placement layout.

[0088] In some implementations, the mapping relationships stored in the process parameter mapping library are optimized and corrected based on the results of the correlation comparison. Thus, when encountering the same or similar glass physical characteristics (such as size and thickness) and optimal placement layout, the process parameter mapping library can match better tempering process parameters based on the corrected mapping relationships.

[0089] For example, suppose the correlation comparison results show that for a batch of glass with physical characteristics of "8mm thickness, 2440mm x 3660mm" and a specific optimal placement layout, the original process parameter mapping library recommends a heating temperature of 690°C. However, actual quality inspection data shows that its stress distribution uniformity is only 88%. Data analysis reveals that when the heating temperature is finely adjusted to 695°C, the stress distribution uniformity of similar glass can stably reach over 92%. In this case, optimization correction can be directly performed, updating the recommended heating temperature in the mapping library for this specific condition (i.e., the same glass physical characteristics and optimal placement layout) from 690°C to 695°C. Subsequently, when production tasks again require processing glass with the same or highly similar characteristics, the latest optimized and superior tempering process parameters will be matched from the mapping library.

[0090] Based on the above technical solution, by systematically correlating and comparing the final product quality data (flatness, stress distribution) with the source parameters (actual cutting parameters) and core process parameters (optimal tempering parameters) in the production process, rather than viewing quality problems in isolation, the data streams from quality inspection equipment, cutting equipment, and tempering equipment can be aggregated and cross-analyzed on a central scheduling platform. This breaks down the data silos between the "cutting-slab handling-tempering-quality inspection" stages, allowing the attribution analysis of quality problems to trace back from the final product to the initial cutting and key tempering process parameters, providing a data foundation for precise optimization. By transforming the process parameter mapping library from a static reference database into a dynamic knowledge base capable of self-correction and continuous optimization based on the correlation comparison results, its internal mapping relationships (i.e., the correspondence rules from "glass features + layout" to "process parameters") are continuously optimized and improved according to specific production conditions.

[0091] The above primarily describes the solutions of the embodiments of this application from the perspective of device implementation. It is understood that each device, for example, a coupling device for deep processing of architectural glass—cutting, arranging, and tempering glass—in order to achieve the above functions, includes at least one of the hardware structures and software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0092] When using integrated units, Figure 7 The diagram shows a possible structural schematic of a coupling device for deep processing, cutting, and tempering of architectural glass (referred to as a coupling device 70 for deep processing, cutting, and tempering of architectural glass) involved in the above embodiments. The coupling device 70 for deep processing, cutting, and tempering of architectural glass includes a processing unit 701 and a communication unit 702, and may also include a storage unit 703. Figure 7 The structural diagram shown can be used to illustrate the structure of the coupling device for deep processing, cutting, shearing, and tempering of architectural glass involved in the above embodiments.

[0093] when Figure 7 The schematic diagram shown illustrates the structure of the coupling device for deep processing, cutting, shearing, and tempering of architectural glass in the above embodiments. The processing unit 701 is used to control and manage the operation of the coupling device for deep processing, cutting, shearing, and tempering of architectural glass. The communication unit 702 is used for the coupling device to communicate with other devices. The storage unit 703 is used to store the program code and data of the coupling device for deep processing, cutting, shearing, and tempering of architectural glass.

[0094] For example, communication unit 702 is used to receive production order data and, based on the maximum loading size and process characteristics of the tempering furnace; Processing unit 701 is used to cut the raw glass sheets required for the order with the goal of maximizing material utilization and tempering furnace loading compatibility, generate a cutting instruction set, and synchronously update the dynamic surplus material database; control the cutting equipment to execute the cutting instruction set, and transfer the cut glass sheets to the sorting cage; and aggregate the equipment production data uploaded by each edge computing node through the central scheduling platform to form a global production view. The equipment production data includes at least the real-time inventory status, size and location information of the glass sheets in the sorting cage, as well as the current operating status of the tempering furnace; based on the global production view, the central scheduling platform predicts the next 24 hours using a Long Short-Term Memory (LSTM) network. Within a short time, the thickness distribution of glass batches entering the tempering process is recorded, generating a thickness distribution report. Based on the thickness distribution report and real-time information of the glass sheets in the sorting cage, the central scheduling platform optimizes batch grouping through a clustering algorithm, generating a batch grouping scheme for efficient loading into the tempering furnace and corresponding sorting cage sorting instructions. Based on the batch grouping scheme and sorting instructions, the central scheduling platform runs a loading simulation algorithm to generate the optimal placement layout of the corresponding batch of glass in the tempering furnace. Combining the optimal placement layout with the physical characteristics of the batch of glass, the central scheduling platform automatically matches and sends the optimal tempering process parameters from the process parameter mapping library to the tempering furnace execution unit.

[0095] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, the disclosure, and the appended claims, can understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple instances. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0096] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely illustrative descriptions of the application as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from the spirit and scope of this application. Thus, if such modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is also intended to include such modifications and variations.

Claims

1. A coupling method for deep processing of architectural glass, including cutting, shearing, and tempering, characterized in that... include: Receive production order data and, based on the maximum loading size and process characteristics of the tempering furnace, aim to maximize material utilization and tempering furnace loading compatibility, cut the raw glass required for the order, generate a cutting instruction set, and synchronously update the dynamic surplus material database; The cutting equipment is controlled to execute the cutting instruction set, and the cut glass sheets are transferred to the glass handling cage. The central scheduling platform aggregates the equipment production data uploaded by each edge computing node to form a global production view. The equipment production data includes at least the real-time inventory status, size and location information of the glass sheets in the glass handling cage, as well as the current operating status of the tempering furnace. Based on the global production view, the central scheduling platform uses a Long Short-Term Memory (LSTM) network to predict the thickness distribution of glass batches that will enter the tempering process within the next 24 hours and generates a thickness distribution report. Based on the thickness distribution report and the real-time information of the glass sheets in the glass rack, the central scheduling platform optimizes batch grouping through a clustering algorithm to generate a batch grouping scheme for efficient loading of the tempering furnace and corresponding glass rack sorting instructions. Based on the batch grouping scheme and sorting instructions, the central scheduling platform runs a loading simulation algorithm to generate the optimal placement layout of the corresponding batch of glass in the tempering furnace. The central dispatch platform, combining the optimal placement layout with the physical characteristics of the batch of glass, automatically matches and sends the optimal tempering process parameters from the process parameter mapping library to the tempering furnace execution unit.

2. The coupling method for deep processing, cutting, shearing, and tempering of architectural glass according to claim 1, characterized in that, The process of optimizing the cutting and layout of the raw glass sheets required for the order, with the goal of maximizing material utilization and tempering furnace loading compatibility, and generating a cutting instruction set, specifically includes: Parse production order data to obtain the area A of a single glass sheet to be processed. glass The constraint data of the tempering furnace is obtained from the tempering furnace parameter database, and the constraint data includes the maximum loading area A. furnace Minimum clearance requirement between the glass plate and the glass plate; Based on maximizing material utilization Maximize compatibility with tempering furnace Establish a multi-objective optimization model, where A raw Let A be the area of ​​the original glass plate. pattern The projected area of ​​the cut pattern inside the tempering furnace; Based on the multi-objective optimization model, a mixed integer programming solver is used to process the production order data and constraint data to generate a Pareto optimal solution set, where each solution in the Pareto optimal solution set corresponds to a cutting and layout pattern. The cutting layout pattern with the highest comprehensive score is selected from the Pareto optimal solution set, and the selected pattern is parsed to generate the corresponding cutting instruction set, which includes the cutting coordinates, cutting path sequence and process parameters of each glass piece. Information on leftover material with an area greater than or equal to a preset threshold generated by cutting is recorded and updated to the dynamic leftover material database.

3. The coupling method for deep processing, cutting, shearing, and tempering of architectural glass according to claim 2, characterized in that, The process of predicting the thickness distribution of glass batches entering the tempering process within the next 24 hours using a Long Short-Term Memory (LSTM) network and generating a thickness distribution report specifically includes: Historical glass thickness sequence, order interval time and equipment status information are extracted from the global production view, and the mean, variance and trend slope of the glass thickness are calculated to form a time series feature vector. The time series feature vector is input into the LSTM prediction model to obtain the thickness distribution probability P for the next 24 hours. thickness(i,t） The prediction results are given, where i represents the thickness level and t represents the time point. pass Calculate the time-series rate of change of thickness distribution P identifies the key time points for the switching of thickness distribution and the corresponding time points for adjusting the tempering furnace temperature, and generates a thickness distribution report. The thickness distribution report includes the dominant thickness change trend, the suggested batch grouping scheme, and the temperature adjustment time points. Based on the temperature adjustment time points in the thickness distribution report, the central dispatch platform controls the tempering furnace to pre-adjust the process parameters before performing the tempering of the corresponding batch of glass.

4. The coupling method for deep processing, cutting, shearing, and tempering of architectural glass according to claim 3, characterized in that, The construction process of the LSTM prediction model specifically includes: A hierarchical LSTM prediction model is constructed, comprising an input layer, an LSTM hidden layer, and a prediction output layer. The input layer receives a feature vector, and the number of elements in the feature vector is denoted as the input layer dimension. The length of the input sequence is set to 24 hours. The LSTM hidden layer uses three stacked LSTM units, and the hidden state is passed between layers through a fully connected network using the Dropout mechanism. The prediction output layer uses a fully connected layer to obtain the glass thickness distribution trend for the next 24 hours, and outputs a probability distribution vector P. thickness , used to represent the predicted proportion of different thicknesses.

5. The coupling method for deep processing, cutting, shearing, and tempering of architectural glass according to claim 4, characterized in that, The batch grouping optimization using a clustering algorithm to generate a batch grouping scheme and corresponding rack sorting instructions for efficient loading of the tempering furnace specifically includes: Extract the size and thickness data of the glass sheets inside the glass rack from the global production view and the thickness distribution report; To maximize the uniformity of glass thickness within the same group To achieve the goal of maximizing the space utilization of the tunnel, parameters and distance metrics for the clustering algorithm were set, where h... k Let δ be the set of thicknesses of all glass plates in the k-th group, and let δ be an indicator function, which is 1 when the condition is met and 0 otherwise. A density-based clustering algorithm is used to group the glass slides to generate an initial batch grouping scheme. The clustering algorithm uses the thickness difference between the glass slides as the core distance metric dimension. Based on the initial batch grouping scheme, a sorting instruction sequence is generated for glass sheets to enter the glass sheet handling cage aisle. The sorting instruction sequence ensures that the thickness difference between adjacent glass sheets entering the aisle is ≤2mm as much as possible.

6. The coupling method for deep processing, cutting, shearing, and tempering of architectural glass according to claim 5, characterized in that, The process of grouping glass slides using a density-based clustering algorithm to generate an initial batch grouping scheme specifically includes: The maximum allowable thickness deviation of the tempering furnace is recorded as the cluster radius, and the minimum number of glass sheets loaded into the tempering furnace in a single operation is recorded as the minimum cluster size. A three-layer clustering architecture consisting of a physical layer, a spatiotemporal layer, and a process layer is constructed. The physical layer performs initial grouping based on glass thickness characteristics, divides glass sheets with different glass thickness clustering radii into candidate clusters, and feeds the candidate clusters back to the spatiotemporal layer. The spatiotemporal layer receives candidate clusters and, combining area similarity and temporal continuity constraints, merges and optimizes the candidate clusters to obtain spatiotemporally optimized clusters. The process layer receives the spatiotemporally optimized clusters and, in conjunction with the tempering furnace loading process requirements, makes a final adjustment to the optimized clusters to obtain the target clusters. The uniformity of the thickness within the target clusters is less than the cluster radius, and the number of glass sheets in the target clusters is between the minimum cluster size and the maximum loading capacity of the tempering furnace. Traverse all glass slides in the processing cage, mark unvisited glass slides as unprocessed, for each unprocessed glass slide, find all glass slide sets within the cluster radius neighborhood, if the glass slide set is smaller than the minimum cluster size, create a new cluster and add the unprocessed glass slide to the new cluster; The comprehensive score of the target cluster is calculated based on the glass thickness variance and the space utilization rate of the glass processing cage aisle. Clusters with comprehensive scores below the threshold are re-clustered or split and merged to generate an initial batch grouping scheme that meets the requirements of efficient loading of the tempering furnace.

7. The coupling method for deep processing, cutting, shearing, and tempering of architectural glass according to claim 6, characterized in that, The process of running the loading simulation algorithm to generate the optimal placement layout of the corresponding batch of glass in the tempering furnace specifically includes: Receive the batch grouping scheme, sorting instructions and tempering furnace constraint data, sort the glass sheets in the sorting cage according to the sorting instructions, obtain sorted glass sheet data, and arrange the sorted glass sheet data in descending order of area to form a queue to be placed. The effective loading area of ​​the tempering furnace is initialized as a set of free rectangles. Based on the maximum rectangle algorithm, the glass sheets in the queue are placed sequentially into the free rectangles that can accommodate them and have the smallest wasted area. At the same time, the list of free rectangles is updated and the layout coordinate set is recorded. Based on the layout coordinate set, the space utilization rate, heating uniformity index and layout stability index of the current layout are calculated, and the layout evaluation score is obtained by weighting them. By iterating through different glass placement orders, all candidate layout schemes and their evaluation scores are obtained, and the candidate layout scheme with the highest evaluation score is selected as the optimal placement layout.

8. The coupling method for deep processing, cutting, shearing, and tempering of architectural glass according to claim 7, characterized in that, The process of automatically matching and issuing the optimal tempering process parameters from the process parameter mapping library to the tempering furnace execution unit, based on the optimal placement layout and the physical characteristics of the batch of glass, specifically includes: Extract the physical characteristics of the glass in the current batch from the batch grouping scheme, and receive the optimal placement layout and its layout coordinate set generated by the loading simulation algorithm; Based on the physical characteristics of the glass and the set of layout coordinates, the final optimal set of tempering process parameters is obtained by matching through a process parameter mapping library built based on an attention mechanism. The central scheduling platform distributes the optimal set of tempering process parameters to the corresponding tempering furnace execution unit.

9. The coupling method for deep processing, cutting, shearing, and tempering of architectural glass according to claim 8, characterized in that, This also includes, after tempering is completed, the central scheduling platform collecting quality inspection data of the final product, and performing correlation analysis between this quality data and the actual cutting parameters and tempering process parameters to update the process parameter mapping library: Receive quality inspection data from quality inspection equipment, the quality inspection data including glass flatness and stress distribution uniformity indicators; The quality inspection data is compared and correlated with the actual cutting parameters and the optimal tempering process parameters issued and executed in the central scheduling platform. Based on the correlation comparison results, the mapping relationships stored in the process parameter mapping library are optimized and corrected so that the process parameter mapping library can match better tempering process parameters for the same or similar glass physical characteristics and the optimal placement layout.

10. A coupling device for deep processing of architectural glass, including cutting, shearing, and tempering, characterized in that... The device includes: a communication unit and a processing unit; The communication unit is used to receive production order data and, based on the maximum loading size and process characteristics of the tempering furnace; The processing unit is used to cut the original glass required for the order with the goal of maximizing material utilization and tempering furnace loading compatibility, generate a cutting instruction set, and synchronously update the dynamic surplus material database. The cutting equipment is controlled to execute the cutting instruction set, and the cut glass sheets are transferred to the glass handling cage. The central scheduling platform aggregates the equipment production data uploaded by each edge computing node to form a global production view. The equipment production data includes at least the real-time inventory status, size and location information of the glass sheets in the glass handling cage, as well as the current operating status of the tempering furnace. Based on the global production view, the central scheduling platform uses a Long Short-Term Memory (LSTM) network to predict the thickness distribution of glass batches that will enter the tempering process within the next 24 hours and generates a thickness distribution report. Based on the thickness distribution report and the real-time information of the glass sheets in the glass rack, the central scheduling platform optimizes batch grouping through a clustering algorithm to generate a batch grouping scheme for efficient loading of the tempering furnace and corresponding glass rack sorting instructions. Based on the batch grouping scheme and sorting instructions, the central scheduling platform runs a loading simulation algorithm to generate the optimal placement layout of the corresponding batch of glass in the tempering furnace. The central dispatch platform, combining the optimal placement layout with the physical characteristics of the batch of glass, automatically matches and sends the optimal tempering process parameters from the process parameter mapping library to the tempering furnace execution unit.

Images