Inorganic panel production line management system and method based on digital twin

Abstract

The present application provides an inorganic panel production line management system and method based on digital twin, which relates to the field of inorganic panel production. It first constructs a digital twin of the inorganic panel production line, then receives the mixed slurry state image collected by the industrial camera and the time queue of the scraper speed value collected by the speed sensor, and maps these collected data to the digital twin, processes these images and time queue data in the digital twin to obtain a speed management result for indicating an increase, decrease or maintenance of the speed gear, and finally controls the scraper speed value based on the result. In this way, intelligent management and optimized control of the inorganic panel production line are realized based on digital twin, which significantly improves the real-time and accuracy of the production process.

Description

Technical Field

[0001] The present application relates to the field of inorganic panel production, and more specifically, to an inorganic panel production line management system and method based on digital twins. Background Art

[0002] Inorganic board is a board with inorganic materials as the main component, which has excellent fire resistance, moisture resistance, sound insulation and environmental protection performance. This type of board is widely used in the fields of construction, decoration and fire protection. Common types of inorganic boards include calcium silicate board, fiber reinforced cement board and gypsum board, etc. These boards can not only meet the high standards of modern buildings for safety and environmental protection, but also provide a variety of functional options. They are an important part of modern building materials.

[0003] In the production process of inorganic boards, the slurry after the raw materials are mixed needs to be stirred to ensure that the various raw materials are evenly distributed in the slurry, to prevent the generation of bubbles and voids, and to improve the density and strength of the inorganic board.

[0004] However, traditional mixing control methods mainly rely on the operator's experience and preset process parameters. Although they can meet basic production needs in some simple scenarios, they have significant defects and disadvantages when faced with complex production environments and high-precision requirements. Specifically, traditional mixing control relies on the operator's experience and visual inspection to judge the state of the mixed slurry. However, in the actual production process, the state of the mixed slurry may change rapidly. Traditional methods cannot capture these changes in time and respond quickly, which may lead to instability in the production process. In addition, the state changes of the mixed slurry are complex, including consistency, uniformity, bubbles and other factors. It is difficult to evaluate these complex states comprehensively and accurately manually, so it is difficult to determine the optimal scraper speed adjustment point in a short time, resulting in inconsistent uniformity and consistency of the mixed slurry, which in turn affects the performance of the final product.

[0005] Therefore, an optimized inorganic panel production line management solution is desired. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide an inorganic panel production line management system and method based on digital twin.

[0007] According to one aspect of the present application, a digital twin-based inorganic panel production line management system is provided, which includes:

[0008] A digital twin construction module, used to construct a digital twin of the inorganic panel production line;

[0009] A sensor data receiving module is used to receive a mixed slurry state image collected by an industrial camera and a time queue of a speed value of a scraper collected by a speed sensor;

[0010] A sensor data mapping module, used for mapping the mixed slurry state image and the time queue of the rotation speed value of the scraper to the digital twin of the inorganic plate production line;

[0011] A twin intelligent center module, used for processing the time queue of the mixed slurry state image and the speed value of the scraper to obtain a speed management result, and the speed management result is used to indicate increasing the speed gear, reducing the speed gear or maintaining the speed gear, wherein the twin intelligent center module includes: a state scraper speed analysis unit, used for encoding the time queue of the mixed slurry state image and the speed value of the scraper to obtain the semantic coding feature of the mixed slurry state and the time series correlation feature of the scraper speed; a slurry scraper cross-modal joint unit, used for cross-modal continuous prompt joint encoding of the semantic coding feature of the mixed slurry state and the time series correlation feature of the scraper speed to obtain the slurry state-scraper speed cross-modal joint feature; a speed management result generation unit, used for obtaining the speed management result based on the slurry state-scraper speed cross-modal joint feature;

[0012] The remote control module is commanded to control the rotation speed value of the scraper based on the rotation speed management result.

[0013] In the above-mentioned inorganic panel production line management system based on digital twin, the state scraper speed analysis unit includes: a mixed slurry state semantic coding feature extraction unit, which is used to input the mixed slurry state image into a slurry state feature extractor based on a deep separable convolutional neural network model to obtain a mixed slurry state semantic coding feature map as the mixed slurry state semantic coding feature; a scraper speed timing association feature extraction unit, which is used to input the time queue of the scraper speed value into a scraper speed timing encoder based on a forward LSTM model to obtain a scraper speed timing association feature vector as the scraper speed timing association feature.

[0014] In the above-mentioned inorganic panel production line management system based on digital twin, the slurry scraper cross-modal joint unit includes: a mixed slurry state semantic coding local feature decomposition sub-unit, which is used to perform local feature decomposition of the mixed slurry state semantic coding feature map along the channel dimension to obtain a set of mixed slurry state local feature matrices; a slurry state-scraper speed prompt perception sub-unit, which is used to perform prompt perception cross-modal joint on the set of the scraper speed time-series correlation feature vector and the mixed slurry state local feature matrix to obtain the slurry state-scraper speed cross-modal joint feature.

[0015] In the above-mentioned inorganic panel production line management system based on digital twin, the slurry state-scraper speed prompt perception subunit includes: a slurry state-scraper speed query prompt secondary subunit, which is used to use the scraper speed time series association feature vector as the query vector and each mixed slurry state local feature matrix in the set of mixed slurry state local feature matrices as the key matrix, and input the query vector and the key matrix into the prompt learning network based on the converter structure to obtain a set of slurry state-scraper speed local query prompt semantic coding vectors; a slurry state-scraper speed feature mask secondary subunit, which is used to input the scraper speed time series association feature vector and each slurry state-scraper speed local query prompt semantic coding vector in the set of slurry state-scraper speed local query prompt semantic coding vector into a cross-modal mask weaving network based on prompt information to obtain a base. A set of slurry state-scraper speed cross-modal local feature mask weight matrices based on prompt information; a slurry state-scraper speed feature interaction secondary sub-unit, used to calculate the set of slurry state-scraper speed cross-modal local feature mask weight matrices based on prompt information and the set of mixed slurry state local feature matrices, and to obtain a set of slurry state-scraper speed cross-modal local granularity significant interaction matrices by multiplying the position points between each corresponding slurry state-scraper speed cross-modal local feature mask weight matrix based on prompt information and the mixed slurry state local feature matrix; a slurry state-scraper speed cross-modal joint feature generation secondary sub-unit, used to perform feature aggregation on the set of slurry state-scraper speed cross-modal local granularity significant interaction matrices to obtain a slurry state-scraper speed cross-modal joint feature map as the slurry state-scraper speed cross-modal joint feature.

[0016] In the above-mentioned inorganic panel production line management system based on digital twin, the slurry state-scraper speed query prompt secondary sub-unit includes: a slurry scraper similarity score calculation tertiary sub-unit, which is used to calculate the similarity score between the scraper speed time series correlation feature vector and each row vector in the mixed slurry state local feature matrix to obtain a set of slurry state-scraper speed similarity scores; a slurry scraper weight calculation tertiary sub-unit, which is used to standardize each slurry state-scraper speed similarity score in the set of slurry state-scraper speed similarity scores to obtain a set of slurry state-scraper speed weight coefficients; a slurry state-scraper speed local weighted sum coding tertiary sub-unit, which is used to calculate the weighted sum of each row vector in the mixed slurry state local feature matrix based on the set of slurry state-scraper speed weight coefficients to obtain the slurry state-scraper speed local query prompt semantic coding vector.

[0017] In the above-mentioned inorganic panel production line management system based on digital twin, the slurry scraper similarity score calculation three-level sub-unit is used to: calculate the norm of each row vector in the local feature matrix of the mixed slurry state to obtain a set of local norm values ​​of the mixed slurry state; calculate the norm of the scraper speed timing-related feature vector to obtain the scraper speed timing-related feature vector norm value; multiply the scraper speed timing-related feature vector norm value and the set of local norm values ​​of the mixed slurry state by position to obtain a set of slurry scraper joint norm values; calculate the product between the scraper speed timing-related feature vector and the transposed vector of each row vector in the local feature matrix of the mixed slurry state to obtain a set of slurry scraper related values; divide the set of the slurry scraper related values ​​and the set of the slurry scraper joint norm values ​​by the slurry scraper related values ​​and the slurry scraper joint norm value for each group of corresponding slurry scraper related values ​​to obtain the set of slurry state-scraper speed similarity scores.

[0018] In the above-mentioned inorganic panel production line management system based on digital twin, the slurry scraper weight calculation tertiary sub-unit is used to: calculate the natural exponential function value with the natural constant e as the base and each slurry state-scraper speed similarity score in the set of slurry state-scraper speed similarity scores as the exponential power to obtain a set of slurry state-scraper speed similarity modulation scores; calculate the set of slurry state-scraper speed similarity modulation scores point by point to obtain the total sum of slurry state-scraper speed similarity scores; divide the set of slurry state-scraper speed similarity modulation scores by the total sum of slurry state-scraper speed similarity scores by position to obtain the set of slurry state-scraper speed weight coefficients.

[0019] In the above-mentioned inorganic panel production line management system based on digital twin, the slurry state-scraper speed feature mask secondary sub-unit is used to: multiply the transposed vector of the slurry state-scraper speed local query prompt semantic encoding vector with the scraper speed time series associated feature vector to obtain a slurry state-scraper speed prompt association matrix; divide the slurry state-scraper speed prompt association matrix by the square root of the scale of the slurry state-scraper speed prompt association matrix by position to obtain a slurry state-scraper speed prompt association modulation matrix; input the slurry state-scraper speed prompt association modulation matrix into the softmax function to obtain the slurry state-scraper speed cross-modal local feature mask weight matrix based on the prompt information.

[0020] In the above-mentioned inorganic panel production line management system based on digital twin, the speed management result generation unit is used to: input the slurry state-scraper speed cross-modal joint feature map into the classifier-based speed management engine to obtain the speed management result.

[0021] According to another aspect of the present application, a method for managing an inorganic panel production line based on digital twin is provided, comprising:

[0022] Constructing a digital twin of the inorganic panel production line;

[0023] A time queue for receiving a mixed slurry state image acquired by an industrial camera and a speed value of a scraper acquired by a speed sensor;

[0024] Mapping the mixed slurry state image and the time queue of the rotation speed value of the scraper to the digital twin of the inorganic plate production line;

[0025] Processing the mixed slurry state image and the time queue of the speed value of the scraper to obtain a speed management result, and the speed management result is used to indicate increasing the speed gear, reducing the speed gear or maintaining the speed gear, including: encoding the mixed slurry state image and the time queue of the speed value of the scraper respectively to obtain the semantic coding feature of the mixed slurry state and the time series association feature of the scraper speed; performing cross-modal continuous prompt joint encoding on the semantic coding feature of the mixed slurry state and the time series association feature of the scraper speed to obtain the slurry state-scraper speed cross-modal joint feature; obtaining the speed management result based on the slurry state-scraper speed cross-modal joint feature;

[0026] The rotation speed value of the scraper is controlled based on the rotation speed management result.

[0027] Compared with the prior art, the inorganic panel production line management system and method based on digital twin provided by the present application first constructs the digital twin of the inorganic panel production line, then receives the mixed slurry state image collected by the industrial camera and the time queue of the scraper speed value collected by the speed sensor, and maps these collected data to the digital twin, processes these images and time queue data in the digital twin to obtain the speed management result for indicating the increase, decrease or maintenance of the speed gear, and finally controls the scraper speed value according to the result. In this way, the intelligent management and optimization control of the inorganic panel production line is realized based on the digital twin, which significantly improves the real-time and accuracy of the production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0029] Figure 1This is a system block diagram of an inorganic panel production line management system based on digital twin according to an embodiment of the present application.

[0030] Figure 2 It is a block diagram of the twin intelligent center module in the inorganic panel production line management system based on digital twin according to an embodiment of the present application.

[0031] Figure 3 This is a schematic diagram of data flow of the twin intelligent center module in the inorganic panel production line management system based on digital twin according to an embodiment of the present application.

[0032] Figure 4 It is a block diagram of the state scraper speed analysis unit in the inorganic panel production line management system based on digital twin according to an embodiment of the present application.

[0033] Figure 5 It is a block diagram of the slurry scraper cross-modal joint unit in the digital twin-based inorganic panel production line management system according to an embodiment of the present application.

[0034] Figure 6 It is a flowchart of the inorganic panel production line management method based on digital twin according to an embodiment of the present application. DETAILED DESCRIPTION

[0035] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0036] Inorganic board is a board with inorganic materials as the main component. It has the advantages of fire resistance, moisture resistance, sound insulation and environmental protection. It is widely used in the fields of construction, decoration and fire protection. Its types mainly include calcium silicate board, fiber reinforced cement board and gypsum board. These boards meet the safety and environmental protection standards of modern buildings and provide a variety of functional options. In the production process of inorganic board, the slurry after the raw materials are mixed needs to be stirred to ensure the uniform distribution of the raw materials and prevent the generation of bubbles and voids, thereby improving the density and strength of the inorganic board.

[0037] However, the traditional stirring control method relies on the operator's experience and preset process parameters. This method can meet basic production needs in simple scenarios, but it has obvious shortcomings in complex production environments and high-precision requirements. Specifically, the traditional method relies on the operator's experience and visual inspection to judge the state of the mixed slurry, and it is difficult to capture the rapid changes in the slurry state in time, resulting in unstable production. In addition, the state changes of the mixed slurry are complex, including factors such as consistency, uniformity, and bubbles, which are difficult to fully and accurately evaluate manually, making it difficult to determine the optimal scraper speed adjustment point, which will affect the uniformity and consistency of the mixed slurry, and thus affect the performance of the final product.

[0038] Digital Twin is a technology that makes full use of data such as physical models, sensor updates, and operation history, integrates multi-disciplinary, multi-physical quantity, multi-scale, and multi-probability simulation processes, completes mapping in virtual space, and thus reflects the full life cycle process of the corresponding physical equipment. Simply put, it creates a digital copy of a physical entity to achieve real-time monitoring, simulation, prediction, and optimization of the entity. In the production of inorganic panels, real-time data (such as temperature, pressure, humidity, etc.) on the production line is collected through sensors. Digital twins can reflect the production status of inorganic panels and help managers to detect abnormal situations in a timely manner. At the same time, using digital twin models, production managers can simulate different production parameters to find the best production conditions, thereby improving product quality and production efficiency. In general, the application of digital twin technology provides a smarter production management solution, which is conducive to promoting the digital transformation of inorganic panel production.

[0039] Based on this, this application proposes an inorganic panel production line management system based on digital twin. Figure 1 : is a system block diagram of an inorganic panel production line management system based on digital twin according to an embodiment of the present application. Figure 1 As shown, in the inorganic panel production line management system 100 based on digital twin, it includes: a digital twin construction module 110, used to construct a digital twin of the inorganic panel production line; a sensor data receiving module 120, used to receive the mixed slurry state image collected by the industrial camera and the time queue of the scraper speed value collected by the speed sensor; a sensor data mapping module 130, used to map the mixed slurry state image and the time queue of the scraper speed value to the digital twin of the inorganic panel production line; a twin intelligent center module 140, used to process the mixed slurry state image and the time queue of the scraper speed value to obtain a speed management result, and the speed management result is used to indicate increasing the speed gear, reducing the speed gear or maintaining the speed gear; an instruction remote control module 150, used to control the speed value of the scraper based on the speed management result.

[0040] In the embodiment of the present application, the digital twin construction module 110 is used to construct the digital twin of the inorganic panel production line. It should be understood that the digital twin can be used as a virtual mapping of the physical production line to reflect the various states and parameters in the production process in real time. In order to realize the intelligent production management of the inorganic panel production line, the digital twin of the inorganic panel production line can be constructed. By constructing the digital twin of the production line, production managers can better understand and optimize the inorganic panel production process, identify potential problems and improve them, so as to improve the intelligent management level of the production line. In particular, in a specific implementation method of the embodiment of the present application, the digital twin of the inorganic panel production line is constructed, including: a. Demand analysis and goal setting: clearly constructing the goal of the digital twin of the inorganic panel production line and analyzing and determining the key parameters of the production line that need to be monitored and controlled; b. Digital twin modeling: selecting a suitable modeling tool (such as MATLAB, Simulink, Unity, etc.) to construct a digital twin, and adding the attribute information, operation status information, production parameter information, etc. of the production equipment in the inorganic panel production line to the digital twin to form a complete digital twin.

[0041] In a specific implementation method of the present application, a digital twin of the inorganic panel production line is constructed, including: setting a goal to monitor the state of the mixed slurry in real time to ensure the uniformity and consistency of the slurry. Key parameters include: consistency, uniformity, bubbles of the mixed slurry, and the speed of the scraper. Use Unity tools to build a virtual model of the production line. Add attribute information of the production equipment to the model, such as the model, specification, and operating status of the mixer.

[0042] In the embodiment of the present application, the sensor data receiving module 120 is used to receive the mixed slurry state image collected by the industrial camera and the time queue of the speed value of the scraper collected by the speed sensor. It should be understood that the mixed slurry state image collected by the camera reflects the consistency information, uniformity information, bubble conditions, etc. of the slurry, and the time queue of the scraper speed collected by the speed sensor can reflect the time series change information, fluctuation information, etc. of the speed. By combining and analyzing the time series data of the slurry state image and the scraper speed, the relationship between the speed change and the slurry state can be established, and then the optimal speed setting can be found to ensure the stability of the production process and product quality.

[0043] In an embodiment of the present application, the sensor data mapping module 130 is used to map the time queue of the mixed slurry state image and the speed value of the scraper to the digital twin of the inorganic panel production line. It should be understood that the core of the digital twin is to achieve data mapping and real-time synchronization between the physical world and the digital world. In order to ensure that the constructed digital twin of the inorganic panel production line can reflect the state and behavior of the production equipment on the production line in real time, it is necessary to map the time queue of the mixed slurry state image and the speed value of the scraper to the digital twin of the inorganic panel production line in the technical solution of the present application. Through the twin that has completed the data mapping, the data in the actual production process can be deeply analyzed to optimize the production decisions in the production process of the inorganic panel production line. In particular, in an embodiment of the present application, a specific implementation method of mapping the mixed slurry state image and the time queue of the scraper speed value to the digital twin of the inorganic panel production line can be: a. Data preprocessing: performing data preprocessing on the mixed slurry state image and the time queue of the scraper speed value collected by the industrial camera on the actual production line to improve the quality of the original collected data; b. Image data mapping: establishing a virtual image area corresponding to the mixed slurry state image in the digital twin, which can display the mixed slurry state on the actual production line in real time, and the preprocessed image area can be mapped to the digital twin. The mixed slurry state image after processing is transmitted in real time to the virtual image area of ​​the digital twin to realize image mapping; c. Time queue data mapping: a data interface corresponding to the scraper speed value is established in the digital twin to receive and process the time queue of the scraper speed value collected by the speed sensor, and the time queue of the preprocessed scraper speed value is transmitted in real time to the data interface of the digital twin to realize data mapping; d. Real-time synchronization and update: ensure that the virtual image area and data interface in the digital twin can receive and update the mixed slurry state image and the time queue of the scraper speed value from the actual production line in real time.

[0044] In a specific implementation method of the embodiment of the present application, the pre-processed mixed slurry state image is transmitted in real time to the virtual image area of ​​the digital twin to achieve image mapping, including: establishing a virtual image area in the digital twin, located above the mixer. Use the WebSocket protocol to transmit the pre-processed image data to the virtual image area of ​​the digital twin in real time. Ensure that the image data can be accurately mapped to the corresponding position of the virtual image area. Set the image data to be updated once a second to ensure that the image in the digital twin can reflect the latest status of the actual production line in real time.

[0045] In an embodiment of the present application, the twin intelligent center module 140 is used to process the time queue of the mixed slurry state image and the speed value of the scraper to obtain a speed management result, and the speed management result is used to indicate increasing the speed gear, reducing the speed gear or maintaining the speed gear.

[0046] Correspondingly, in the twin intelligent center module 140, the technical concept of the present application is to use image analysis and data processing technology based on deep learning to semantically encode the mixed slurry state image, and at the same time, to time-series associate the speed value of the scraper, so as to automatically obtain the speed management result according to the cross-modal prompt joint representation between the semantic encoding features of the mixed slurry state and the time-series association features of the scraper speed. In this way, it is possible to monitor and analyze the slight changes in the slurry state in real time and in detail, and to adaptively adjust the speed based on the dynamic association between the real-time scraper speed value and the slurry state to ensure that the slurry state is consistent in each production process, thereby reducing human intervention and improving the intelligence level of the inorganic plate production line.

[0047] Figure 2 It is a block diagram of the twin intelligent center module in the inorganic panel production line management system based on digital twin according to an embodiment of the present application. Figure 3 Schematic diagram of data flow of the twin intelligent center module in the inorganic panel production line management system based on digital twin according to the embodiment of the present application. Figure 2 and Figure 3 As shown, the twin intelligent center module 140 includes: a state scraper speed analysis unit 141, which is used to encode the time queue of the mixed slurry state image and the scraper speed value respectively to obtain the mixed slurry state semantic coding features and the scraper speed timing association features; a slurry scraper cross-modal joint unit 142, which is used to perform cross-modal continuous prompt joint encoding on the mixed slurry state semantic coding features and the scraper speed timing association features to obtain the slurry state-scraper speed cross-modal joint features; a speed management result generation unit 143, which is used to obtain the speed management result based on the slurry state-scraper speed cross-modal joint features.

[0048] In the embodiment of the present application, the state scraper speed analysis unit 141 is used to encode the mixed slurry state image and the time queue of the scraper speed value to obtain the mixed slurry state semantic coding features and the scraper speed time series correlation features. Specifically, Figure 4 FIG. 1 is a block diagram of a state scraper speed analysis unit in an inorganic panel production line management system based on digital twin according to an embodiment of the present application. Figure 4As shown, the state scraper speed analysis unit 141 includes: a mixed slurry state semantic coding feature extraction unit 1411, which is used to input the mixed slurry state image into a slurry state feature extractor based on a deep separable convolutional neural network model to obtain a mixed slurry state semantic coding feature map as the mixed slurry state semantic coding feature; a scraper speed timing association feature extraction unit 1412, which is used to input the time queue of the scraper speed value into a scraper speed timing encoder based on a forward LSTM model to obtain a scraper speed timing association feature vector as the scraper speed timing association feature.

[0049] In an embodiment of the present application, the mixed slurry state semantic coding feature extraction unit 1411 is used to input the mixed slurry state image into a slurry state feature extractor based on a deep separable convolutional neural network model to obtain a mixed slurry state semantic coding feature map as the mixed slurry state semantic coding feature. It should be understood that the mixed slurry state image contains visual information such as the consistency and uniformity of the slurry, which is crucial for speed management. Based on this, in order to capture and mine the semantic feature information about the slurry implicit in the mixed slurry state image, in the technical solution of the present application, the mixed slurry state image is input into a slurry state feature extractor based on a deep separable convolutional neural network model to obtain a mixed slurry state semantic coding feature map. Those of ordinary skill in the art should know that the deep separable convolutional neural network model decomposes the standard convolution into two steps: deep convolution and point-by-point convolution. Deep convolution applies a separate convolution kernel to each input channel, while point-by-point convolution uses a 1x1 convolution kernel to combine the output of the deep convolution. This decomposition operation significantly reduces the amount of calculation and model parameters while maintaining high performance. That is, using a deep separable convolutional neural network model to process the mixed slurry state image can significantly improve computational efficiency and flexibility while maintaining high performance.

[0050] In an embodiment of the present application, the scraper speed time series correlation feature extraction unit 1412 is used to input the time queue of the speed value of the scraper into the scraper speed time series encoder based on the forward LSTM model to obtain the scraper speed time series correlation feature vector as the scraper speed time series correlation feature. It should be understood that there is a correlation between the speed contexts between different time intervals in the time queue of the speed value of the scraper, and the forward LSTM model can effectively capture the long-term and short-term dependencies in the time series of the scraper speed value by introducing memory units and gating mechanisms. Therefore, in the technical solution of the present application, the time queue of the speed value of the scraper is input into the scraper speed time series encoder based on the forward LSTM model to capture the time series correlation of the speed values ​​between different time spans by utilizing the characteristics of the forward LSTM model that can capture long-term data dependencies, and obtain the scraper speed time series correlation feature vector, so that the trend and pattern of the scraper speed changing over time can be effectively captured, so as to better understand the working status and historical behavior of the scraper.

[0051] In the embodiment of the present application, the slurry scraper cross-modal joint unit 142 is used to perform cross-modal continuous prompt joint encoding on the semantic coding feature of the mixed slurry state and the scraper speed time series correlation feature to obtain the slurry state-scraper speed cross-modal joint feature. Specifically, Figure 5 FIG. 1 is a block diagram of a slurry scraper cross-modal joint unit in a digital twin-based inorganic panel production line management system according to an embodiment of the present application. Figure 5 As shown, the slurry scraper cross-modal union unit 142 includes: a mixed slurry state semantic coding local feature decomposition subunit 1421, which is used to perform local feature decomposition along the channel dimension on the mixed slurry state semantic coding feature map to obtain a set of mixed slurry state local feature matrices; a slurry state-scraper speed prompt perception subunit 1422, which is used to perform prompt perception cross-modal union on the set of scraper speed time series correlation feature vectors and the mixed slurry state local feature matrix to obtain the slurry state-scraper speed cross-modal joint feature.

[0052] It should be understood that the mixed slurry state semantic coding feature map contains the physical and chemical state of the mixed slurry in the production process and deep semantic visual information, such as the uniformity of the slurry, whether there are bubbles or impurities, etc. The scraper speed time series association feature vector records the change of the scraper speed over time during the production process, reflecting the working state and operation mode of the scraper. It is possible to consider these two aspects of information at the same time to obtain a more comprehensive description of the production status, thereby improving the control accuracy of the knife speed in the production process of inorganic boards. In the technical solution of the present application, the mixed slurry state semantic coding feature map and the scraper speed time series association feature vector are cross-modal continuously prompted and jointly encoded to obtain a slurry state-scraper speed cross-modal joint feature map as a slurry state-scraper speed cross-modal joint feature.

[0053] Specifically, first, the semantic encoding feature map of the mixed slurry state is subjected to local feature decomposition along the channel dimension to obtain a set of local feature matrices of the mixed slurry state. That is, through local feature decomposition, detailed information of the slurry state can be captured from different scales. For example, some channels may capture the large-scale structure of the slurry (such as uniformity), while other channels may capture small-scale details (such as bubbles and particle distribution), providing data support for subsequent fine-grained interactions.

[0054] Next, the scraper speed time-series associated feature vector is used as the query vector and each mixed slurry state local feature matrix in the set of mixed slurry state local feature matrices is used as the key matrix, which is input into the prompt learning network based on the converter structure to utilize the self-attention mechanism to capture the trend of scraper speed changes over time and its impact on the slurry state, so as to enhance the understanding of the key dynamic interaction between the slurry state and the scraper speed, and obtain a set of slurry state-scraper speed local query prompt semantic encoding vectors.

[0055] Then, the set of query hint semantic encoding vectors and the set of query hint semantic encoding vectors are input into the hint-based cross-modal mask weaving network to obtain a set of hint-based slurry state-scraper speed cross-modal local feature mask weight matrices. In other words, the hint-based cross-modal mask weaving network can selectively emphasize the most relevant part of the modal interaction, suppress noise and irrelevant features, and highlight and display the relevant and representative information in the interaction between the two, thereby improving the expressiveness of the features and the performance of the model.

[0056] After that, each mask weight matrix is ​​multiplied by each mixed slurry state local feature matrix by position point to obtain a set of slurry state-scraper speed cross-modal local granularity significant interaction matrices. In this way, those features that have a greater impact on the production process at a specific position can be enhanced, while irrelevant or weakly related features can be suppressed. The generated significant interaction matrix can reflect the significant interaction relationship between the slurry state and the scraper speed in different local areas.

[0057] Finally, the set of the slurry state-scraper speed cross-modal local granularity significant interaction matrices is subjected to feature aggregation to fuse multiple local granularity significant interaction matrices into a comprehensive slurry state-scraper speed cross-modal joint feature map, thereby more comprehensively and accurately reflecting various changes and states in the production process. Specifically, the feature aggregation is to splice each slurry state-scraper speed cross-modal local granularity significant interaction matrix in the set of slurry state-scraper speed cross-modal local granularity significant interaction matrices along the channel dimension to obtain a slurry state-scraper speed cross-modal joint feature map.

[0058] Specifically, in an embodiment of the present application, the slurry state-scraper speed prompt perception subunit 1422 includes: a slurry state-scraper speed query prompt secondary subunit, which is used to use the scraper speed time series association feature vector as the query vector and each mixed slurry state local feature matrix in the set of mixed slurry state local feature matrices as the key matrix, and input the query vector and the key matrix into a prompt learning network based on a converter structure to obtain a set of slurry state-scraper speed local query prompt semantic encoding vectors; a slurry state-scraper speed feature mask secondary subunit, which is used to input the scraper speed time series association feature vector and each slurry state-scraper speed local query prompt semantic encoding vector in the set of slurry state-scraper speed local query prompt semantic encoding vector into a cross-modal mask weaving network based on prompt information to obtain a prompt-based A set of slurry state-scraper speed cross-modal local feature mask weight matrices based on prompt information; a slurry state-scraper speed feature interaction secondary sub-unit, used to calculate the set of slurry state-scraper speed cross-modal local feature mask weight matrices based on prompt information and the set of mixed slurry state local feature matrices, and multiply the position points between each corresponding slurry state-scraper speed cross-modal local feature mask weight matrix based on prompt information and the mixed slurry state local feature matrix to obtain a set of slurry state-scraper speed cross-modal local granularity significant interaction matrices; a slurry state-scraper speed cross-modal joint feature generation secondary sub-unit, used to perform feature aggregation on the set of slurry state-scraper speed cross-modal local granularity significant interaction matrices to obtain a slurry state-scraper speed cross-modal joint feature map as the slurry state-scraper speed cross-modal joint feature.

[0059] More specifically, in an embodiment of the present application, the slurry state-scraper speed query prompt secondary sub-unit includes: a slurry scraper similarity score calculation tertiary sub-unit, which is used to calculate the similarity score between the scraper speed time series correlation feature vector and each row vector in the mixed slurry state local feature matrix to obtain a set of slurry state-scraper speed similarity scores; a slurry scraper weight calculation tertiary sub-unit, which is used to standardize each slurry state-scraper speed similarity score in the set of slurry state-scraper speed similarity scores to obtain a set of slurry state-scraper speed weight coefficients; a slurry state-scraper speed local weighted sum encoding tertiary sub-unit, which is used to calculate the weighted sum of each row vector in the mixed slurry state local feature matrix based on the set of slurry state-scraper speed weight coefficients to obtain the slurry state-scraper speed local query prompt semantic encoding vector.

[0060] More specifically, in an embodiment of the present application, the slurry scraper similarity score calculation three-level sub-unit is used to: calculate a norm of each row vector in the local feature matrix of the mixed slurry state to obtain a set of local norm values ​​of the mixed slurry state; calculate a norm of the scraper speed timing-related feature vector to obtain a norm value of the scraper speed timing-related feature vector; multiply the scraper speed timing-related feature vector norm value and the set of local norm values ​​of the mixed slurry state by position to obtain a set of slurry scraper joint norm values; calculate the product between the scraper speed timing-related feature vector and the transposed vector of each row vector in the local feature matrix of the mixed slurry state to obtain a set of slurry scraper related values; divide the set of the slurry scraper related values ​​and the set of the slurry scraper joint norm values ​​by the slurry scraper joint norm value for each group of corresponding slurry scraper related values ​​to obtain the set of slurry state-scraper speed similarity scores.

[0061] More specifically, in an embodiment of the present application, the slurry scraper weight calculation tertiary sub-unit is used to: calculate the natural exponential function value with the natural constant e as the base and each slurry state-scraper speed similarity score in the set of slurry state-scraper speed similarity scores as the exponential power to obtain a set of slurry state-scraper speed similarity modulation scores; calculate the set of slurry state-scraper speed similarity modulation scores point by point to obtain the total sum of slurry state-scraper speed similarity scores; divide the set of slurry state-scraper speed similarity modulation scores by the total sum of slurry state-scraper speed similarity scores by position to obtain the set of slurry state-scraper speed weight coefficients.

[0062] More specifically, in an embodiment of the present application, the slurry state-scraper speed feature mask secondary sub-unit is used to: multiply the transposed vector of the slurry state-scraper speed local query prompt semantic encoding vector with the scraper speed time series associated feature vector to obtain a slurry state-scraper speed prompt association matrix; divide the slurry state-scraper speed prompt association matrix by the square root of the scale of the slurry state-scraper speed prompt association matrix by position to obtain a slurry state-scraper speed prompt association modulation matrix; input the slurry state-scraper speed prompt association modulation matrix into a softmax function to obtain the slurry state-scraper speed cross-modal local feature mask weight matrix based on the prompt information.

[0063] In the embodiment of the present application, specifically, the slurry scraper cross-modal joint unit 142 is used to perform cross-modal continuous prompt joint encoding processing on the mixed slurry state semantic coding feature map and the scraper speed time series associated feature vector according to the following formula; wherein the formula is: ; ; ; ; ;in, is the decomposition function, is the semantic encoding feature map of the mixed slurry state, , , and are the first, second, and third local characteristic matrices of the mixed slurry state. and The local characteristic matrix of the mixed slurry state, , , and They are The first, second, and and row vectors, for coding, is the scraper speed time series correlation feature vector, yes The transposed vector of is matrix multiplication, is the one-norm of the vector, for and The slurry state-scraper speed similarity score between represents the exponential function value with the natural constant e as the base, yes Middle row vectors, for and The slurry state-scraper speed similarity score between for The corresponding slurry state-scraper speed weight coefficient, yes The number of row vectors in is the semantic encoding vector of the local query hint of slurry state-scraper speed, for The transposed vector of is the scale of the slurry state-scraper speed prompt correlation matrix, that is, the length of the matrix multiplied by the width of the matrix, yes function, yes The corresponding slurry state-scraper speed cross-modal local feature mask weight matrix based on the prompt information, , , are the first, second and third in the set of the slurry state-scraper speed cross-modal local feature mask weight matrices based on the prompt information. A slurry state-scraper speed cross-modal local feature mask weight matrix based on prompt information, For point-by-position multiplication, is the feature concatenation along the channel dimension, It is the cross-modal joint characteristic diagram of slurry state and scraper speed.

[0064] In an embodiment of the present application, the speed management result generation unit 143 is used to obtain the speed management result based on the slurry state-scraper speed cross-modal joint feature. Specifically, in an embodiment of the present application, the speed management result generation unit 143 is used to: input the slurry state-scraper speed cross-modal joint feature map into the classifier-based speed management engine to obtain the speed management result. That is, the slurry state-scraper speed cross-modal joint feature obtained by cross-modal joint encoding using the mixed slurry state semantic coding feature map and the scraper speed time series association feature vector is classified and processed to automatically obtain the speed management result. In this way, it is possible to monitor and analyze the slight changes in the slurry state in real time and in detail, and to adaptively adjust the speed based on the dynamic correlation between the real-time scraper speed value and the slurry state to ensure that the slurry state is consistent in each production process, thereby reducing human intervention and improving the intelligence level of the inorganic plate production line. In particular, in a specific implementable embodiment of the present application, the slurry state-scraper speed cross-modal joint feature map is input into a speed management engine based on a classifier to obtain the speed management result, including: expanding each slurry state-scraper speed cross-modal joint feature matrix in the slurry state-scraper speed cross-modal joint feature map into a one-dimensional feature vector according to a row vector or a column vector, and then cascading to obtain a slurry state-scraper speed cross-modal joint classification feature vector; using the fully connected layer of the classifier to fully connect encode the slurry state-scraper speed cross-modal joint classification feature vector to obtain a slurry state-scraper speed cross-modal joint fully connected encoded classification feature vector; inputting the slurry state-scraper speed cross-modal joint fully connected encoded classification feature vector into the Softmax classification function of the classifier to obtain the probability value of the slurry state-scraper speed cross-modal joint feature map belonging to each classification label, and the classification labels include those for indicating an increase in speed gear, for indicating a decrease in speed gear, and for indicating maintaining speed gear; the classification label corresponding to the largest of the probability values ​​is determined as the speed management result.

[0065] In particular, when the mixed slurry state semantic encoding feature map and the scraper speed temporal association feature vector respectively represent the image semantic features of the mixed slurry state image and the short-range-long-range bidirectional temporal association features of the scraper speed value, when performing cross-modal continuous prompt joint encoding, the cross-modal joint encoding feature representation of the slurry state-scraper speed cross-modal joint feature map will also have a long-range class regression understanding representation imbalance relative to the distribution field selective state space of the heteromodal data feature prompts, thereby deviating from the source function modal semantic distribution representation class understanding and affecting the accuracy of the classification results.

[0066] Preferably, the slurry state-scraper speed cross-modal joint feature map is input into a classifier-based speed management engine to obtain the speed management result, including: determining the number of zero eigenvalues ​​in the slurry state-scraper speed cross-modal joint feature map, and respectively calculating the reciprocal of the logarithm of the number of zero eigenvalues ​​with base two and the exponential value of the reciprocal of the number of zero eigenvalues ​​with base natural constants to obtain a first slurry state-scraper speed cross-modal joint representation value and a second slurry state-scraper speed cross-modal joint representation value; expanding the slurry state-scraper speed cross-modal joint feature map into a slurry state-scraper speed cross-modal joint feature vector; calculating a power function of each eigenvalue of the slurry state-scraper speed cross-modal joint feature vector with the reciprocal of the number of zero eigenvalues ​​as the exponent to obtain a slurry state-scraper speed cross-modal joint structural state vector; and The slurry state-scraper speed cross-modal joint structural state vector is dot-multiplied with the first slurry state-scraper speed cross-modal joint representation value to obtain a slurry state-scraper speed cross-modal joint information representation vector; the autocorrelation matrix of the slurry state-scraper speed cross-modal joint structural state vector is dot-multiplied with the second slurry state-scraper speed cross-modal joint representation value and the weight hyperparameter respectively to obtain a slurry state-scraper speed cross-modal joint regression understanding matrix; the slurry state-scraper speed cross-modal joint information representation vector is matrix-multiplied with the slurry state-scraper speed cross-modal joint regression understanding matrix to obtain an optimized slurry state-scraper speed cross-modal joint feature vector; the optimized slurry state-scraper speed cross-modal joint feature vector is input into the classifier-based speed management engine to obtain the speed management result.

[0067] Here, the first slurry state-scraper speed cross-modal joint representation value and the second slurry state-scraper speed cross-modal joint representation value are expressed as: ; ;in, is the number of zero eigenvalues ​​in the slurry state-scraper speed cross-modal joint characteristic map, represents an exponential function with the natural constant e as the base, is the first slurry state-scraper speed cross-modal joint representation value, It is the second slurry state-scraper speed cross-modal joint representation value.

[0068] And the optimized slurry state-scraper speed cross-modal joint feature vector is expressed as: ; ; ;in, is the slurry state-scraper speed cross-modal joint eigenvector, and is a row vector, is the number of zero eigenvalues ​​in the slurry state-scraper speed cross-modal joint characteristic map, is the first slurry state-scraper speed cross-modal joint representation value, It means point multiplication by position. To calculate each eigenvalue of the slurry state-scraper speed cross-modal joint eigenvector, a power function with the inverse of the number of zero eigenvalues ​​as the exponent, is the slurry state-scraper speed cross-modal joint information representation vector, is the weight hyperparameter, is the second slurry state-scraper speed cross-modal joint representation value, Expressed as The transposed vector of is the slurry state-scraper speed cross-modal joint regression understanding matrix, represents matrix multiplication, It is the optimized slurry state-scraper speed cross-modal joint eigenvector.

[0069] That is, the present application realizes effective long-range modeling of the slurry state-scraper speed cross-modal joint feature map under the field linear complexity by modeling the vector field zero dimension of the slurry state-scraper speed cross-modal joint feature vector obtained after the expansion of the slurry state-scraper speed cross-modal joint feature map as the global field redundant dependency. In this way, in the feature regression process of the slurry state-scraper speed cross-modal joint feature map, the structured correlation self-distillation between information representation and class understanding is performed in the selective state space based on the feature distribution, so as to maintain the overall information complexity of the feature distribution of the slurry state-scraper speed cross-modal joint feature map while promoting the long-range dynamic class understanding balance of its feature convergence, so as to improve the class understanding consistency between the classification regression and feature extraction processes, and improve the accuracy of the speed management result obtained by inputting the slurry state-scraper speed cross-modal joint feature map into the classifier-based speed management engine. In this way, it is possible to monitor and analyze minute changes in the slurry state in real time and detail, and to adaptively adjust the speed based on the dynamic relationship between the real-time scraper speed value and the slurry state to ensure that the slurry state is consistent during each production process, thereby reducing human intervention and improving the intelligence level of the inorganic panel production line.

[0070] Therefore, the twin intelligent center module 140 is explained, which uses image analysis and data processing technology based on deep learning to semantically encode the mixed slurry state image, and at the same time, the speed value of the scraper is time-series associated, so as to automatically obtain the speed management result according to the cross-modal prompt joint representation between the semantic encoding features of the mixed slurry state and the time-series association features of the scraper speed. In this way, it is possible to monitor and analyze the slight changes in the slurry state in real time and in detail, and to adaptively adjust the speed based on the dynamic association between the real-time scraper speed value and the slurry state to ensure that the slurry state is consistent in each production process, thereby reducing human intervention and improving the intelligence level of the inorganic plate production line.

[0071] In an embodiment of the present application, the command remote control module 150 is used to control the speed value of the scraper based on the speed management result. It should be understood that by transmitting the speed management result from the digital twin to the actual inorganic board production line, the interaction between the virtual world and the real world is realized. This interaction enables production managers to monitor and optimize the production line in a virtual environment, and then apply the optimization results to actual production to improve the efficiency and stability of the production line. In actual production, by accurately controlling the speed of the scraper, the stirring quality of the slurry can be ensured, thereby improving the quality of the inorganic board finally produced. In general, the combination of digital twin technology and real-time data analysis reduces manual intervention, improves the level of automation, and realizes more intelligent inorganic board production line management.

[0072] In summary, the inorganic panel production line management system 100 based on digital twins according to the embodiment of the present application is explained, which first constructs a digital twin of the inorganic panel production line, then receives the mixed slurry state image collected by the industrial camera and the time queue of the scraper speed value collected by the speed sensor, and maps these collected data to the digital twin, processes these images and time queue data in the digital twin to obtain a speed management result for indicating an increase, decrease or maintenance of the speed gear, and finally controls the scraper speed value according to the result. In this way, intelligent management and optimized control of the inorganic panel production line are realized based on digital twins, which significantly improves the real-time and accuracy of the production process.

[0073] As described above, the inorganic panel production line management system 100 based on digital twin according to the embodiment of the present application can be implemented in various terminal devices, such as a server for inorganic panel production line management based on digital twin, etc. In one example, the inorganic panel production line management system 100 based on digital twin according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the inorganic panel production line management system 100 based on digital twin can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the inorganic panel production line management system 100 based on digital twin can also be one of the many hardware modules of the terminal device.

[0074] Alternatively, in another example, the digital twin-based inorganic panel production line management system 100 and the terminal device may also be separate devices, and the digital twin-based inorganic panel production line management system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0075] Figure 6 Flow chart of the inorganic panel production line management method based on digital twin according to the embodiment of the present application. Figure 6 As shown, in the inorganic panel production line management method based on digital twin, it includes: S110, constructing a digital twin of the inorganic panel production line; S120, receiving a mixed slurry state image collected by an industrial camera and a time queue of a scraper speed value collected by a speed sensor; S130, mapping the mixed slurry state image and the time queue of the scraper speed value to the digital twin of the inorganic panel production line; S140, processing the mixed slurry state image and the time queue of the scraper speed value to obtain a speed management result, and the speed management result is used to indicate an increase in the speed gear. , lower the speed gear or maintain the speed gear; S150, control the speed value of the scraper based on the speed management result, wherein step S140 includes: respectively encoding the time queue of the mixed slurry state image and the scraper speed value to obtain the mixed slurry state semantic coding feature and the scraper speed timing association feature; cross-modal continuous prompt joint encoding of the mixed slurry state semantic coding feature and the scraper speed timing association feature to obtain the slurry state-scraper speed cross-modal joint feature; based on the slurry state-scraper speed cross-modal joint feature, obtain the speed management result.

[0076] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned inorganic panel production line management method based on digital twins have been referred to above. Figures 1 to 5It has been introduced in detail in the description of the inorganic panel production line management system based on digital twin, and therefore, its repeated description will be omitted.

[0077] In summary, the inorganic panel production line management method based on digital twins according to the embodiment of the present application is explained, which first constructs a digital twin of the inorganic panel production line, then receives the mixed slurry state image collected by the industrial camera and the time queue of the scraper speed value collected by the speed sensor, and maps these collected data to the digital twin, processes these images and time queue data in the digital twin to obtain a speed management result for indicating an increase, decrease or maintenance of the speed gear, and finally controls the scraper speed value according to the result. In this way, intelligent management and optimized control of the inorganic panel production line are realized based on digital twins, which significantly improves the real-time and accuracy of the production process.

[0078] In the several embodiments provided in this application, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0079] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0080] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0081] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application.

Claims

1. An inorganic panel production line management system based on digital twin, characterized in that: include: A digital twin construction module, used to construct a digital twin of the inorganic panel production line; A sensor data receiving module is used to receive a mixed slurry state image collected by an industrial camera and a time queue of a speed value of a scraper collected by a speed sensor; A sensor data mapping module, used for mapping the mixed slurry state image and the time queue of the rotation speed value of the scraper to the digital twin of the inorganic plate production line; A twin intelligent center module, used for processing the time queue of the mixed slurry state image and the speed value of the scraper to obtain a speed management result, and the speed management result is used to indicate increasing the speed gear, reducing the speed gear or maintaining the speed gear, wherein the twin intelligent center module includes: a state scraper speed analysis unit, used for encoding the time queue of the mixed slurry state image and the speed value of the scraper respectively to obtain the semantic coding feature of the mixed slurry state and the time series correlation feature of the scraper speed; a slurry scraper cross-modal joint unit, used for cross-modal continuous prompt joint encoding of the semantic coding feature of the mixed slurry state and the time series correlation feature of the scraper speed to obtain the slurry state-scraper speed cross-modal joint feature; a speed management result generation unit, used to obtain the speed management result based on the slurry state-scraper speed cross-modal joint feature; An instruction remote control module is used to control the speed value of the scraper based on the speed management result; Wherein, the slurry scraper cross-modal joint unit comprises: A mixed slurry state semantic coding local feature decomposition subunit, used for performing local feature decomposition along the channel dimension on the mixed slurry state semantic coding feature map to obtain a set of mixed slurry state local feature matrices; A slurry state-scraper speed prompt perception subunit, used for performing prompt perception cross-modal combination of the scraper speed time series correlation feature vector and the set of the mixed slurry state local feature matrix to obtain the slurry state-scraper speed cross-modal combination feature; Wherein, the slurry state-scraper speed prompt sensing subunit includes: A slurry state-scraper speed query prompt secondary subunit is used to use the scraper speed time series associated feature vector as a query vector and each mixed slurry state local feature matrix in the set of mixed slurry state local feature matrices as a key matrix, and input the query vector and the key matrix into a prompt learning network based on a converter structure to obtain a set of slurry state-scraper speed local query prompt semantic encoding vectors; A slurry state-scraper speed feature mask secondary subunit is used to input each slurry state-scraper speed local query prompt semantic coding vector in the set of the scraper speed time series association feature vector and the slurry state-scraper speed local query prompt semantic coding vector into a cross-modal mask weaving network based on prompt information to obtain a set of slurry state-scraper speed cross-modal local feature mask weight matrices based on prompt information; The slurry state-scraper speed feature interaction secondary subunit is used to calculate the set of the slurry state-scraper speed cross-modal local feature mask weight matrices based on the prompt information and the set of the mixed slurry state local feature matrices, and to obtain the set of slurry state-scraper speed cross-modal local granularity significant interaction matrices by multiplying the position points between each group of corresponding slurry state-scraper speed cross-modal local feature mask weight matrices based on the prompt information and the mixed slurry state local feature matrices; A slurry state-scraper speed cross-modal joint feature generation secondary subunit is used to perform feature aggregation on the set of the slurry state-scraper speed cross-modal local granularity significant interaction matrices to obtain a slurry state-scraper speed cross-modal joint feature map as the slurry state-scraper speed cross-modal joint feature; The slurry scraper cross-modal joint unit is used to perform cross-modal continuous prompt joint encoding processing on the mixed slurry state semantic coding feature map and the scraper speed time series associated feature vector according to the following formula; wherein the formula is: ， in, is the decomposition function, is the semantic encoding feature map of the mixed slurry state, , , and are the first, second, and third local characteristic matrices of the mixed slurry state. and The local characteristic matrix of the mixed slurry state, , , and They are The first, second, and and row vectors, for coding, is the scraper speed time series correlation feature vector, yes The transposed vector of is matrix multiplication, is the one-norm of the vector, for and The slurry state-scraper speed similarity score between represents the exponential function value with the natural constant e as the base, yes Middle row vectors, for and The slurry state-scraper speed similarity score between for The corresponding slurry state-scraper speed weight coefficient, yes The number of row vectors in is the semantic encoding vector of the local query hint of slurry state-scraper speed, for The transposed vector of is the scale of the slurry state-scraper speed prompt correlation matrix, that is, the length of the matrix multiplied by the width of the matrix, yes function, yes The corresponding slurry state-scraper speed cross-modal local feature mask weight matrix based on the prompt information, , , are the first, second and third in the set of the slurry state-scraper speed cross-modal local feature mask weight matrices based on the prompt information. A slurry state-scraper speed cross-modal local feature mask weight matrix based on prompt information, For point-by-position multiplication, is the feature concatenation along the channel dimension, It is the cross-modal joint characteristic diagram of slurry state and scraper speed.

2. The inorganic panel production line management system based on digital twin according to claim 1 is characterized in that: The state scraper speed analysis unit comprises: A mixed slurry state semantic coding feature extraction unit, used for inputting the mixed slurry state image into a slurry state feature extractor based on a deep separable convolutional neural network model to obtain a mixed slurry state semantic coding feature map as the mixed slurry state semantic coding feature; The scraper speed timing correlation feature extraction unit is used to input the time queue of the scraper speed value into the scraper speed timing encoder based on the forward LSTM model to obtain the scraper speed timing correlation feature vector as the scraper speed timing correlation feature.

3. The inorganic panel production line management system based on digital twin according to claim 2 is characterized in that: The slurry state-scraper speed query prompt secondary subunit includes: The slurry scraper similarity score calculation tertiary subunit is used to calculate the similarity scores between the scraper speed time series correlation feature vector and each row vector in the mixed slurry state local feature matrix to obtain a set of slurry state-scraper speed similarity scores; The slurry scraper weight calculation tertiary subunit is used to normalize each slurry state-scraper speed similarity score in the set of slurry state-scraper speed similarity scores to obtain a set of slurry state-scraper speed weight coefficients; The slurry state-scraper speed local weighted sum coding third-level sub-unit is used to calculate the weighted sum of each row vector in the mixed slurry state local feature matrix based on the set of slurry state-scraper speed weight coefficients to obtain the slurry state-scraper speed local query prompt semantic coding vector.

4. The inorganic panel production line management system based on digital twin according to claim 3 is characterized in that: The slurry scraper similarity score calculation three-level sub-unit is used to: Calculating a norm of each row vector in the local characteristic matrix of the mixed slurry state to obtain a set of local norm values ​​of the mixed slurry state; Calculating a norm of the scraper speed time series associated characteristic vector to obtain a norm value of the scraper speed time series associated characteristic vector; Multiplying the scraper speed time series associated feature vector norm value and the set of mixed slurry state local norm values ​​by position to obtain a set of slurry scraper joint norm values; Respectively calculating the product between the scraper speed time series associated characteristic vector and the transposed vector of each row vector in the mixed slurry state local characteristic matrix to obtain a set of slurry scraper related values; Each corresponding group of slurry scraper related values ​​and slurry scraper joint norm values ​​in the set of the slurry scraper related values ​​and the set of the slurry scraper joint norm values ​​is divided to obtain the set of slurry state-scraper speed similarity scores.

5. The inorganic panel production line management system based on digital twin according to claim 4 is characterized in that: The slurry scraper weight calculation three-level sub-unit is used for: Calculate the natural exponential function value with the natural constant e as the base and each slurry state-scraper speed similarity score in the set of slurry state-scraper speed similarity scores as the exponential power to obtain a set of slurry state-scraper speed similarity modulation scores; Calculating the set of the slurry state-scraper speed similarity modulation scores point by point to obtain a total sum of the slurry state-scraper speed similarity scores; The set of the slurry state-scraper speed similarity modulation scores is divided by the total sum of the slurry state-scraper speed similarity scores by position to obtain the set of the slurry state-scraper speed weight coefficients.

6. The inorganic panel production line management system based on digital twin according to claim 5 is characterized in that: The slurry state-scraper speed characteristic mask secondary subunit is used for: Multiplying the transposed vector of the slurry state-scraper speed local query prompt semantic encoding vector by the scraper speed time series associated feature vector to obtain a slurry state-scraper speed prompt association matrix; Dividing the slurry state-scraper speed prompt association matrix by the square root of the scale of the slurry state-scraper speed prompt association matrix by position to obtain a slurry state-scraper speed prompt association modulation matrix; The slurry state-scraper speed prompt association modulation matrix is ​​input into the softmax function to obtain the slurry state-scraper speed cross-modal local feature mask weight matrix based on the prompt information.

7. The inorganic panel production line management system based on digital twin according to claim 6 is characterized in that: The speed management result generating unit is used to input the slurry state-scraper speed cross-modal joint feature map into a classifier-based speed management engine to obtain the speed management result.

8. A method for managing an inorganic panel production line based on digital twins, using the inorganic panel production line management system based on digital twins according to claim 1, characterized in that: include: Constructing a digital twin of the inorganic panel production line; A time queue for receiving a mixed slurry state image acquired by an industrial camera and a speed value of a scraper acquired by a speed sensor; Mapping the mixed slurry state image and the time queue of the rotation speed value of the scraper to the digital twin of the inorganic plate production line; Processing the mixed slurry state image and the time queue of the speed value of the scraper to obtain a speed management result, and the speed management result is used to indicate increasing the speed gear, reducing the speed gear or maintaining the speed gear, including: encoding the mixed slurry state image and the time queue of the speed value of the scraper respectively to obtain the semantic coding feature of the mixed slurry state and the time series association feature of the scraper speed; performing cross-modal continuous prompt joint encoding on the semantic coding feature of the mixed slurry state and the time series association feature of the scraper speed to obtain the slurry state-scraper speed cross-modal joint feature; obtaining the speed management result based on the slurry state-scraper speed cross-modal joint feature; The rotation speed value of the scraper is controlled based on the rotation speed management result.

Images