Stirring control method and system for dry Rongalite production

Abstract

The present application provides a stirring control method and system for the production of dry Rongbai powder, which relates to the field of intelligent control. The method collects the heating temperature time series data set of the heating ring and the stirring speed time series data set of the spherical concentrate respectively, and uses the data analysis and processing technology based on deep learning to map the collected data to the Gram angle field, and then dynamically extracts the time series pattern, so as to intelligently recommend the stirring speed value at the next time point according to the dynamic memory attention interaction representation between the heating temperature time series dynamic pattern characteristics and the stirring speed time series dynamic pattern characteristics. In this way, the heating temperature and the stirring speed can be monitored in real time, and the complex dynamic relationship between the two can be captured to automatically adjust the stirring speed to adapt to the current optimal process conditions, ensuring the continuity and stability of production.

Description

Technical Field

[0001] The present application relates to the field of intelligent control, and more particularly, in an embodiment of the present application, to a stirring control method and system for the production of dry Rongalite. Background Art

[0002] Rongalite, chemically known as sodium bisulfite formaldehyde, is an important chemical raw material and reducing agent. It is widely used in industry, especially as a bleaching agent and reducing agent in the textile, printing and dyeing, papermaking and other industries. In order to further improve the quality of Rongalite, the invention with publication number CN117298618A proposes a production method and device for water-free and free formaldehyde-free vacuum stirring and drying Rongalite, which uses sodium pyrosulfite to replace sodium hydroxide to obtain a sodium bisulfite formaldehyde solution. Subsequently, the solution is heated and stirred in a sealed concentration mechanism and evaporated into a molten body. The molten body is then introduced into a crystallization mechanism in a vacuum state and stirred at high speed to crystallize into Rongalite. Finally, the completed Rongalite is discharged and packaged after pressure relief.

[0003] Concentration is an important step in the production process of Rongalite. Its main purpose is to increase the concentration of the target product, sodium bisulfite formaldehyde solution, by removing part of the solvent (such as water) in the solution, so as to form a more uniform and high-quality crystal structure in the subsequent vacuum crystallization step. Specifically, the concentration process in the above patent is to pour the synthesized solution into the concentration mechanism, heat it through the heating ring in a sealed state, and stir it with the spherical concentration piece to evaporate the solvent and concentrate it into a molten body.

[0004] In the above patent, the stirring speed of the spherical concentrator is set and controlled according to predetermined parameters. However, the static setting of the stirring speed in the above patent cannot respond to the dynamic changes in the production process in real time, making it difficult to automatically adjust to maintain the optimal process conditions. In addition, simple preset parameters cannot capture the complex dynamic relationship between heating temperature and stirring speed, and ignore the key impact of the interaction between the two on product quality. For example, temperature changes may cause the optimal stirring speed to deviate from the preset value, and static control cannot perceive and correct this deviation, thereby affecting the purity and stability of the final product.

[0005] Therefore, an optimized stirring control scheme for dry Rongalite production is desired. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a stirring control method and system for the production of dry Rongalite.

[0007] According to one aspect of the present application, a stirring control method for dry Rongalite production is provided, comprising:

[0008] Dissolving sodium metabisulfite in ice water and stirring to obtain a stirred solution; adding zinc powder to the stirred solution and heating it, then adding a reaction reagent to the heated solution and performing a constant temperature reaction to obtain a sodium bisulphite formaldehyde solution; concentrating the sodium bisulphite formaldehyde solution to obtain a concentrated sodium bisulphite formaldehyde solution; vacuum crystallizing the concentrated sodium bisulphite formaldehyde solution to obtain Rongalite; depressurizing and packaging the Rongalite to obtain the dry Rongalite; concentrating the sodium bisulphite formaldehyde solution to obtain a concentrated sodium bisulphite formaldehyde solution, comprising:

[0009] The heating temperature time series data set of the heating ring and the stirring speed time series data set of the spherical concentrating member are collected by the temperature sensor and the speed sensor respectively;

[0010] Performing Gram angular field transformation on the heating temperature time series data set and the stirring speed time series data set respectively to obtain a heating temperature mapping Gram angular field image and a stirring speed mapping Gram angular field image;

[0011] Extracting the temporal pattern dynamic features of the heating temperature mapping Gram's angle field image and the stirring speed mapping Gram's angle field image to obtain the heating temperature temporal pattern dynamic features and the stirring speed temporal pattern dynamic features;

[0012] Dynamic memory attention interaction is performed on the heating temperature timing dynamic pattern characteristics and the stirring speed timing dynamic pattern characteristics to obtain fine-grained interaction characteristics between thermal effect and stirring speed timing characteristics, and based on the fine-grained interaction characteristics between thermal effect and stirring speed timing characteristics, a recommended decoding value of the stirring speed at the next time point is obtained.

[0013] Furthermore, the heating temperature mapping Grammar angular field image and the stirring speed mapping Grammar angular field image are subjected to time series pattern dynamic feature extraction to obtain heating temperature time series dynamic pattern features and stirring speed time series dynamic pattern features, including: passing the heating temperature mapping Grammar angular field image and the stirring speed mapping Grammar angular field image through a time series pattern feature extractor based on a multi-scale convolutional LSTM to obtain a heating temperature time series dynamic pattern feature vector as the heating temperature time series dynamic pattern feature and a stirring speed time series dynamic pattern feature vector as the stirring speed time series dynamic pattern feature.

[0014] Furthermore, dynamic memory attention interaction is performed on the heating temperature time series dynamic mode feature and the stirring speed time series dynamic mode feature to obtain fine-grained interaction features between thermal effect and stirring speed time series features, and based on the fine-grained interaction features between thermal effect and stirring speed time series features, a recommended decoding value of the stirring speed at the next time point is obtained, including:

[0015] Performing dynamic memory attention interaction on the heating temperature time series dynamic mode feature vector and the stirring speed time series dynamic mode feature vector to obtain a fine-grained interaction feature vector between thermal effect and stirring speed time series features as the fine-grained interaction feature between thermal effect and stirring speed time series features;

[0016] Based on the fine-grained interactive feature vector between the thermal effect-stirring speed time series features, a recommended decoding value of the stirring speed at the next time point is obtained.

[0017] Furthermore, dynamic memory attention interaction is performed on the heating temperature time series dynamic mode feature vector and the stirring speed time series dynamic mode feature vector to obtain a fine-grained interaction feature vector between thermal effect and stirring speed time series features, including:

[0018] Inputting the heating temperature time series dynamic mode feature vector and the stirring speed time series dynamic mode feature vector into a collaborative feature extraction network to obtain an implicit collaborative encoding vector between the heating temperature and stirring speed time series features;

[0019] Writing the implicit collaborative coding vector between the heating temperature-stirring speed time series features into a dynamic memory unit to obtain a heating temperature-stirring speed time series dynamic key vector;

[0020] Based on the heating temperature-stirring speed timing dynamic key vector, feature attention modulation is performed on the heating temperature timing dynamic mode feature vector and the stirring speed timing dynamic mode feature vector respectively to obtain an optimized heating temperature timing dynamic mode feature vector and an optimized stirring speed timing dynamic mode feature vector;

[0021] The optimized heating temperature timing dynamic mode feature vector and the optimized stirring speed timing dynamic mode feature vector are input into a characteristic linear interaction network to obtain a fine-grained interaction feature vector between the thermal effect-stirring speed timing features.

[0022] Further, based on the heating temperature-stirring speed timing dynamic key vector, feature attention modulation is performed on the heating temperature timing dynamic mode feature vector and the stirring speed timing dynamic mode feature vector respectively to obtain an optimized heating temperature timing dynamic mode feature vector and an optimized stirring speed timing dynamic mode feature vector, including:

[0023] Extract the heating temperature-stirring speed timing dynamic key vector from the dynamic memory unit, and input the heating temperature timing dynamic mode feature vector and the heating temperature-stirring speed timing dynamic key vector into a feature attention modulation module based on the first converter structure to obtain the optimized heating temperature timing dynamic mode feature vector;

[0024] The heating temperature-stirring speed timing dynamic key vector is extracted from the dynamic memory unit, and the stirring speed timing dynamic mode feature vector and the heating temperature-stirring speed timing dynamic key vector are input into a feature attention modulation module based on a second converter structure to obtain the optimized stirring speed timing dynamic mode feature vector.

[0025] Furthermore, the optimized heating temperature timing dynamic pattern feature vector and the optimized stirring speed timing dynamic pattern feature vector are input into a characteristic linear interaction network to obtain a fine-grained interaction feature vector between the thermal effect-stirring speed timing features, including: calculating the position-weighted sum of the optimized heating temperature timing dynamic pattern feature vector and the optimized stirring speed timing dynamic pattern feature vector to obtain the fine-grained interaction feature vector between the thermal effect-stirring speed timing features; wherein the sum of the weighted coefficients of the optimized heating temperature timing dynamic pattern feature vector and the optimized stirring speed timing dynamic pattern feature vector is one.

[0026] Furthermore, based on the fine-grained interaction feature vector between the thermal effect-stirring speed timing characteristics, a recommended decoding value of the stirring speed at the next time point is obtained, including: passing the fine-grained interaction feature vector between the thermal effect-stirring speed timing characteristics through a decoder-based stirring controller to obtain the recommended decoding value of the stirring speed at the next time point.

[0027] Furthermore, the fine-grained interaction feature vector between the thermal effect-stirring speed timing features is passed through a decoder-based stirring controller to obtain a recommended decoding value of the stirring speed at the next time point, including: multiplying the decoding weight matrix of the decoder with the fine-grained interaction feature vector between the thermal effect-stirring speed timing features to obtain the decoded fine-grained interaction feature vector between the thermal effect-stirring speed timing features, and accumulating and summing all eigenvalues ​​of the decoded fine-grained interaction feature vector between the thermal effect-stirring speed timing features to obtain the recommended decoding value of the stirring speed at the next time point.

[0028] According to another aspect of the present application, a stirring control system for dry Rongalite production is provided, which is used to execute the stirring control method for dry Rongalite production, and comprises:

[0029] The dry Rongalite production data acquisition module is used to respectively acquire the heating temperature time series data set of the heating ring and the stirring speed time series data set of the spherical concentrate through the temperature sensor and the speed sensor;

[0030] A Gram angular field transformation module, used for performing Gram angular field transformation on the heating temperature time series data set and the stirring speed time series data set respectively to obtain a heating temperature mapping Gram angular field image and a stirring speed mapping Gram angular field image;

[0031] A time series pattern dynamic feature extraction module is used to extract time series pattern dynamic features from the heating temperature mapping Gram angle field image and the stirring speed mapping Gram angle field image to obtain heating temperature time series dynamic pattern features and stirring speed time series dynamic pattern features;

[0032] A dynamic memory attention interaction module is used to perform dynamic memory attention interaction on the heating temperature timing dynamic pattern characteristics and the stirring speed timing dynamic pattern characteristics to obtain fine-grained interaction characteristics between the thermal effect-stirring speed timing characteristics, and based on the fine-grained interaction characteristics between the thermal effect-stirring speed timing characteristics, obtain a recommended decoding value of the stirring speed at the next time point.

[0033] Furthermore, the timing pattern dynamic feature extraction module is used to: pass the heating temperature mapping Grammar angle field image and the stirring speed mapping Grammar angle field image through a timing pattern feature extractor based on multi-scale convolution LSTM to obtain a heating temperature timing dynamic pattern feature vector as the heating temperature timing dynamic pattern feature and a stirring speed timing dynamic pattern feature vector as the stirring speed timing dynamic pattern feature.

[0034] Compared with the prior art, the present application provides a stirring control method and system for the production of dry Rongbai powder, which collects the heating temperature time series data set of the heating ring and the stirring speed time series data set of the spherical concentrate respectively, and uses data analysis and processing technology based on deep learning to map the collected data to the Gram angle field, and then dynamically extracts the time series pattern, so as to intelligently recommend the stirring speed value at the next time point based on the dynamic memory attention interaction representation between the heating temperature time series dynamic pattern characteristics and the stirring speed time series dynamic pattern characteristics. In this way, the heating temperature and stirring speed can be monitored in real time, and the complex dynamic relationship between the two can be captured to automatically adjust the stirring speed to adapt to the current optimal process conditions, ensuring the continuity and stability of production. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0036] Figure 1The figure is a flow chart of a stirring control method for producing dry Rongalite according to an embodiment of the present application.

[0037] Figure 2 The present invention is a flow chart of concentrating the sodium bisulphite formaldehyde solution to obtain a concentrated sodium bisulphite formaldehyde solution in a stirring control method for producing dry Rongalite according to an embodiment of the present application.

[0038] Figure 3 The present invention is a data flow diagram of concentrating the sodium bisulphite formaldehyde solution to obtain a concentrated sodium bisulphite formaldehyde solution in a stirring control method for producing dry Rongalite according to an embodiment of the present application.

[0039] Figure 4 A flowchart is provided for performing dynamic memory attention interaction on the heating temperature timing dynamic pattern characteristics and the stirring speed timing dynamic pattern characteristics in the stirring control method for dry Rongalite production according to an embodiment of the present application to obtain fine-grained interaction characteristics between thermal effect and stirring speed timing characteristics, and obtaining a recommended decoding value of the stirring speed at the next time point based on the fine-grained interaction characteristics between the thermal effect and stirring speed timing characteristics.

[0040] Figure 5 The present invention is a flow chart of performing dynamic memory attention interaction on the heating temperature timing dynamic pattern feature vector and the stirring speed timing dynamic pattern feature vector in the stirring control method for dry Rongalite production according to an embodiment of the present application to obtain a fine-grained interaction feature vector between the thermal effect-stirring speed timing features.

[0041] Figure 6 System block diagram of a stirring control system for dry Rongalite production according to an embodiment of the present application. DETAILED DESCRIPTION

[0042] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0043] Rongalite (sodium bisulfite formaldehyde) is an important chemical raw material and reducing agent, widely used in textile, printing and dyeing, papermaking and other industries. In terms of improving the quality of Rongalite, the invention with publication number CN117298618A proposes an improved production method, using sodium metabisulfite instead of sodium hydroxide to react and generate sodium bisulfite formaldehyde solution. However, the stirring speed in the existing patents is statically set, fails to respond to dynamic changes in production in real time, and cannot be automatically adjusted to maintain optimal process conditions. In particular, the complex interaction between temperature and stirring speed has not been fully considered, resulting in the stirring speed deviating from the optimal value when the temperature changes, thereby affecting the purity and stability of the final product.

[0044] In view of the above technical problems, the present application proposes a stirring control method for the production of dry Rongalite. Figure 1 Flow chart of the stirring control method for dry Rongalite production according to an embodiment of the present application. Figure 1 As shown, the stirring control method for producing dry Rongalite according to the embodiment of the present application includes: S110, dissolving sodium pyrosulfite in ice water and stirring to obtain a stirred solution; S120, adding zinc powder to the stirred solution and heating it, then adding a reaction reagent to the heated solution and performing a constant temperature reaction to obtain a sodium bisulphite formaldehyde solution; S130, concentrating the sodium bisulphite formaldehyde solution to obtain a concentrated sodium bisulphite formaldehyde solution; S140, vacuum crystallizing the concentrated sodium bisulphite formaldehyde solution to obtain Rongalite; S150, depressurizing and packaging the Rongalite to obtain the dry Rongalite.

[0045] In the above-mentioned stirring control method for the production of dry Rongbai, the step S110 is to dissolve sodium pyrosulfite in ice water and stir to obtain a stirred solution. It should be understood that sodium pyrosulfite is a strong reducing agent and releases a certain amount of heat when dissolved. If it is directly dissolved in water at room temperature, the solution temperature may be too high, thereby affecting its stability and even causing some compounds to decompose. Dissolving sodium pyrosulfite in ice water and stirring it can control the temperature change during the dissolution process, thereby avoiding affecting the properties of the solution or producing unnecessary side reactions due to excessive heat release during dissolution, and ensuring that the dissolution process is more stable. In a specific embodiment of the present application, first, prepare enough ice water, the temperature of the ice water should be lower than 0°C, preferably a mixture of ice and water, to ensure that the solution is in a lower temperature environment. Then, sodium pyrosulfite is gradually added to the ice water and stirred slowly. At this time, due to the effect of low temperature, the dissolution rate of sodium pyrosulfite is relatively slow, but it can prevent the heat released during dissolution from accumulating too quickly. In addition, stirring can help sodium pyrosulfite molecules to better contact with water molecules and promote their dissolution. During the entire dissolution process, it is necessary to keep the ice water at a constant temperature, and the stirring should be kept uniform and moderate to avoid excessive stirring that causes the local temperature of the solution to rise, which affects the dissolution efficiency. Finally, as the sodium metabisulfite is completely dissolved, the resulting solution will be transparent and have a certain degree of stability.

[0046] Specifically, first, before starting, the required materials and equipment must be prepared. The required materials include a certain amount of sodium metabisulfite (Na2S2O5) and a sufficient amount of ice water. Ice water is maintained at a low temperature by adding ice cubes to water, which is very important for controlling the heat release that may be generated during the dissolution process. In addition, a container suitable for the dissolution process is required, which not only needs to have good sealing to prevent gas escape, but also should be equipped with a temperature monitoring device to monitor and adjust the temperature in real time. At the same time, a mechanical device that can provide effective stirring, such as an electric stirrer, is also essential. Then, slowly add an appropriate amount of ice water to the prepared container, followed by sodium metabisulfite. Among them, the feeding speed should not be too fast, so as to avoid excessive local concentration causing violent reactions or agglomeration. Once all solid materials are put into the container, start the stirring device immediately to start stirring. At this time, the stirring speed should be moderate, which can ensure that the sodium metabisulfite is fully dispersed and dissolved, but will not cause unnecessary foam or splashing due to excessive intensity. Continuous stirring helps to keep the temperature of the entire system uniform and promote the complete dissolution of sodium metabisulfite. Since sodium metabisulfite releases a certain amount of heat when it dissolves, the solution temperature needs to be closely monitored, and ice can be added if necessary to maintain a low temperature environment. Low temperature conditions are conducive to reducing the occurrence of side reactions and improving the selectivity of the target product, thus laying a good foundation for the next step. When all the added sodium metabisulfite is completely dissolved to form a clear or slightly colored solution, the desired stirred solution is obtained. At this time, the stirring can be stopped and ready to proceed to the next step.

[0047] It is worth noting that in order to ensure safety and product quality, the entire dissolution process should be carried out in a well-ventilated environment to avoid inhaling possible gases. In addition, the tools and containers used must be clean and pollution-free to avoid introducing impurities that affect the purity of the final product.

[0048] In the above-mentioned stirring control method for the production of dry Rongalite, in step S120, after adding zinc powder to the stirred solution and heating it, a reaction reagent is added to the heated solution and a constant temperature reaction is carried out to obtain a sodium bisulphite formaldehyde solution. Specifically, the reaction reagent is a catalyst, an activator, oxalic acid, and a formaldehyde solution. It should be understood that in the technical solution of the present application, zinc powder, as a reducing agent, can effectively reduce certain chemical substances in the solution and promote the reaction. By heating, the collision frequency between the reactant molecules can be increased and the reaction rate can be increased. When the solution is heated to an appropriate temperature, the required reaction reagents can be added to the solution, and these reagents are generally chemical substances that react specifically with sodium bisulphite and formaldehyde. After adding the reaction reagent, maintaining a constant temperature helps to ensure the stable progress of the reaction and enables the target product, sodium bisulphite formaldehyde solution, to be efficiently synthesized. At this temperature, the reaction of sodium bisulphite and formaldehyde will continue to form the expected compound, i.e., sodium bisulphite formaldehyde solution.

[0049] In the above-mentioned stirring control method for the production of dry Rongalite, the step S130 is to concentrate the sodium bisulphite formaldehyde solution to obtain a concentrated sodium bisulphite formaldehyde solution. It should be understood that the sodium bisulphite formaldehyde solution usually contains a certain proportion of solvent (such as water), and the concentration process can effectively remove the water in the solvent, increase the concentration of the effective ingredients, and make the solution more stable and efficient. In addition, the concentrated solution helps to improve the reaction activity, especially in the reaction that requires the participation of high concentration chemical substances, the concentrated solution can increase the reaction rate and yield.

[0050] Accordingly, in the process of concentrating the sodium bisulfite formaldehyde solution to obtain a concentrated sodium bisulfite formaldehyde solution, the technical concept of the present application is to collect the heating temperature time series data set of the heating ring and the stirring speed time series data set of the spherical concentrator through a temperature sensor and a speed sensor, respectively, and use data analysis and processing technology based on deep learning to map the heating temperature time series data and the stirring speed time series data set to the Gram angle field, and then dynamically extract the time series pattern of the mapped heating temperature mapping Gram angle field image and the stirring speed mapping Gram angle field image, so as to intelligently recommend the stirring speed value at the next time point based on the dynamic memory attention interaction representation between the heating temperature time series dynamic pattern characteristics and the stirring speed time series dynamic pattern characteristics. In this way, the heating temperature and the stirring speed can be monitored in real time, and the complex dynamic relationship between the two can be captured to automatically adjust the stirring speed to adapt to the current optimal process conditions, ensuring the continuity and stability of production.

[0051] Figure 2The present invention is a flow chart of concentrating the sodium bisulphite formaldehyde solution to obtain a concentrated sodium bisulphite formaldehyde solution in a stirring control method for producing dry Rongalite according to an embodiment of the present application. Figure 3 This is a data flow diagram of the concentrated sodium bisulphite formaldehyde solution in the stirring control method for producing dry Rongalite according to an embodiment of the present application. Figure 2 and Figure 3 As shown, in the step S130, the sodium bisulfite formaldehyde solution is concentrated to obtain a concentrated sodium bisulfite formaldehyde solution, including: S131, collecting the heating temperature time series data set of the heating ring and the stirring speed time series data set of the spherical concentrator through a temperature sensor and a speed sensor respectively; S132, performing Gram angular field transformation on the heating temperature time series data set and the stirring speed time series data set to obtain a heating temperature mapping Gram angular field image and a stirring speed mapping Gram angular field image; S133, performing time series pattern dynamic feature extraction on the heating temperature mapping Gram angular field image and the stirring speed mapping Gram angular field image to obtain a heating temperature time series dynamic pattern feature and a stirring speed time series dynamic pattern feature; S134, performing dynamic memory attention interaction on the heating temperature time series dynamic pattern feature and the stirring speed time series dynamic pattern feature to obtain a fine-grained interaction feature between the thermal effect and the stirring speed time series feature, and based on the fine-grained interaction feature between the thermal effect and the stirring speed time series feature, obtaining a recommended decoding value of the stirring speed at the next time point.

[0052] Specifically, in step S131, the heating temperature time series data set of the heating ring and the stirring speed time series data set of the spherical concentrator are collected by the temperature sensor and the speed sensor respectively. It should be understood that, considering that during the heating process, the temperature must be strictly controlled to avoid the decomposition or side reactions of chemical components caused by excessively high temperatures in the solution, especially formaldehyde and sodium bisulfite may decompose at excessively high temperatures. Therefore, the temperature data collected in real time by the temperature sensor can help to adjust the heating power in time to avoid the problem of excessively high or low temperatures. Among them, the continuous recording and analysis of the temperature time series data can ensure that the temperature is controlled within a reasonable range, making the concentration process more efficient and safe. And the stirring speed time series data set of the spherical concentrator is monitored by the speed sensor to ensure the uniformity of the solution during the stirring process. The role of stirring is to prevent the deposition and local overheating of high-concentration components in the solution, and to help the solvent (water) evaporate evenly. More specifically, considering that the stirring speed is too slow, the solution may be uneven, thereby affecting the concentration effect; and the speed is too fast, it may increase the bubble generation of the solution or cause excessive stirring. Therefore, the speed sensor collects stirring speed data in real time, which can help the control system dynamically adjust the stirring speed to ensure the uniformity of the solution and the concentration effect. In this way, the combination of temperature sensor and speed sensor can achieve more accurate concentration control by continuously monitoring and adjusting the heating temperature and stirring speed. This not only improves the efficiency and safety of the concentration process, but also avoids solution quality problems caused by overheating or uneven stirring.

[0053] Specifically, in the step S132, the heating temperature time series data set and the stirring speed time series data set are respectively subjected to Gram angular field transformation to obtain the heating temperature mapping Gram angular field image and the stirring speed mapping Gram angular field image. It should be understood that, considering that the heating temperature time series data set and the stirring speed time series data set of the spherical concentrate usually have complex time series dynamic characteristics, such as nonlinearity, non-stationarity and long-term and short-term dependencies, if the original heating temperature and stirring speed are directly processed, it may be difficult to capture these complex characteristics. Therefore, in the technical solution of the present application, the heating temperature time series data set of the heating ring and the stirring speed time series data set of the spherical concentrate are respectively subjected to Gram angular field transformation to obtain the heating temperature mapping Gram angular field image and the stirring speed mapping Gram angular field image. It can be understood that the Gram angular field transformation can convert one-dimensional time series data into a two-dimensional image representation. This conversion not only retains all the information of the original heating temperature and stirring speed time series data sets, but also can intuitively display the dynamic characteristics of the time series in the form of images, such as the changing trend of the heating temperature, the periodicity and fluctuation of the stirring speed, etc. Moreover, in two-dimensional space, similar time series patterns often correspond to similar image patterns. Through GAF transformation, the timing patterns in the time series can be more easily identified, which is very useful for monitoring the production process.

[0054] Specifically, in an embodiment of the present application, in the step S133, the heating temperature mapping Gram angular field image and the stirring speed mapping Gram angular field image are subjected to temporal pattern dynamic feature extraction to obtain the heating temperature temporal dynamic pattern feature and the stirring speed temporal dynamic pattern feature, including: the heating temperature mapping Gram angular field image and the stirring speed mapping Gram angular field image are subjected to a temporal pattern feature extractor based on a multi-scale convolutional LSTM to obtain a heating temperature temporal dynamic pattern feature vector as the heating temperature temporal dynamic pattern feature and a stirring speed temporal dynamic pattern feature vector as the stirring speed temporal dynamic pattern feature. It should be understood that, considering that there are significant temporal correlation characteristics and subtle differences between different time scales of the heating temperature mapping Gram angular field image and the stirring speed mapping Gram angular field image, although the traditional convolutional LSTM can process spatiotemporal data, capture local spatial structures and temporal dynamic changes, it is weak in perceiving multi-scale temporal characteristics. Therefore, in order to extract features at different levels from the Grammar angular field image, so as to understand the temporal characteristics of the heating temperature and the stirring speed more comprehensively and meticulously, in the technical solution of the present application, the heating temperature is mapped to the Grammar angular field image and the stirring speed is mapped to the Grammar angular field image. Through a temporal pattern feature extractor based on multi-scale convolutional LSTM, the characteristics of the multi-scale convolutional layer captured at different time scales are utilized to mine out subtle temporal features, and the heating temperature temporal dynamic pattern feature vector and the stirring speed temporal dynamic pattern feature vector are obtained. It is worth mentioning that the multi-scale convolutional LSTM structure is a deep learning architecture that combines the advantages of convolutional neural networks (CNNs) and long short-term memory networks (LSTMs), and is particularly suitable for processing data with complex spatiotemporal characteristics. This structure can effectively capture the multi-scale features and long-term dependencies in the data by applying convolution operations at different scales and inputting the results into the LSTM layer for time series modeling. In particular, multi-scale convolutional layers are able to process images at multiple resolution levels, ensuring that the model can both identify rapidly changing details and understand slowly evolving overall patterns, which greatly enhances the understanding and prediction capabilities of complex industrial processes.

[0055] Specifically, in step S134, the heating temperature timing dynamic pattern characteristics and the stirring speed timing dynamic pattern characteristics are dynamically memorized and interacted to obtain fine-grained interaction characteristics between thermal effect and stirring speed timing characteristics, and based on the fine-grained interaction characteristics between the thermal effect and stirring speed timing characteristics, the recommended decoding value of the stirring speed at the next time point is obtained. It should be understood that, considering that there is a complex nonlinear interaction between the heating temperature and the stirring speed in the industrial production process, that is, the heating temperature timing dynamic pattern characteristics and the stirring speed timing dynamic pattern characteristics present an interrelated dynamic pattern in the time dimension, and imply collaborative interaction feature information. Therefore, in order to effectively capture this nonlinear timing relationship and its subtle changes, the present application proposes a dynamic memory attention interaction mechanism to process the heating temperature timing dynamic pattern characteristics and the stirring speed timing dynamic pattern characteristics to obtain fine-grained interaction characteristics between the thermal effect and stirring speed timing characteristics. In other words, the dynamic memory module can remember long-term trends and short-term changes, ensuring that the model understands the past state and makes optimization decisions accordingly. At the same time, the attention mechanism allows the model to focus on the most relevant temporal features and avoid interference from irrelevant information, thereby achieving a fine-grained understanding of changes in heating temperature and stirring speed.

[0056] Figure 4 A flowchart of performing dynamic memory attention interaction on the heating temperature timing dynamic mode characteristics and the stirring speed timing dynamic mode characteristics in the stirring control method for dry Rongalite production according to an embodiment of the present application to obtain fine-grained interaction characteristics between thermal effect and stirring speed timing characteristics, and obtaining a recommended decoding value of the stirring speed at the next time point based on the fine-grained interaction characteristics between the thermal effect and stirring speed timing characteristics. Figure 4 As shown, in an embodiment of the present application, the step S134, performs dynamic memory attention interaction on the heating temperature timing dynamic pattern characteristics and the stirring speed timing dynamic pattern characteristics to obtain fine-grained interaction characteristics between thermal effect-stirring speed timing characteristics, and based on the fine-grained interaction characteristics between the thermal effect-stirring speed timing characteristics, obtains the recommended decoding value of the stirring speed at the next time point, including: S1341, performs dynamic memory attention interaction on the heating temperature timing dynamic pattern feature vector and the stirring speed timing dynamic pattern feature vector to obtain a fine-grained interaction feature vector between thermal effect-stirring speed timing characteristics as the fine-grained interaction feature between thermal effect-stirring speed timing characteristics; S1342, obtains the recommended decoding value of the stirring speed at the next time point based on the fine-grained interaction feature vector between the thermal effect-stirring speed timing characteristics.

[0057] Figure 5The present invention is a flow chart of performing dynamic memory attention interaction on the heating temperature time series dynamic mode feature vector and the stirring speed time series dynamic mode feature vector in the stirring control method for dry Rongalite production according to an embodiment of the present application to obtain a fine-grained interaction feature vector between the thermal effect and stirring speed time series features. Figure 5 As shown, in an embodiment of the present application, the step S1341 performs dynamic memory attention interaction on the heating temperature timing dynamic pattern feature vector and the stirring speed timing dynamic pattern feature vector to obtain a fine-grained interaction feature vector between thermal effect and stirring speed timing features, including: S1341-1, inputting the heating temperature timing dynamic pattern feature vector and the stirring speed timing dynamic pattern feature vector into a collaborative feature extraction network to obtain an implicit collaborative coding vector between heating temperature and stirring speed timing features; S1341-2, writing the implicit collaborative coding vector between the heating temperature and stirring speed timing features into a dynamic memory unit to obtain an implicit collaborative coding vector between the heating temperature and stirring speed timing features. Obtain the heating temperature-stirring speed timing dynamic key vector; S1341-3, based on the heating temperature-stirring speed timing dynamic key vector, respectively perform feature attention modulation on the heating temperature timing dynamic mode feature vector and the stirring speed timing dynamic mode feature vector to obtain the optimized heating temperature timing dynamic mode feature vector and the optimized stirring speed timing dynamic mode feature vector; S1341-4, input the optimized heating temperature timing dynamic mode feature vector and the optimized stirring speed timing dynamic mode feature vector into the feature linear interaction network to obtain the fine-grained interaction feature vector between the thermal effect-stirring speed timing features.

[0058] Specifically, in the step S1341-1, the heating temperature time series dynamic pattern feature vector and the stirring speed time series dynamic pattern feature vector are input into the collaborative feature extraction network to obtain the implicit collaborative coding vector between the heating temperature-stirring speed time series features. It should be understood that the collaborative feature extraction network is composed of a point convolution layer and an activation function, which aims to extract high-level abstract features from the original feature vector. Through a carefully designed network architecture, the collaborative feature extraction network is able to capture the deep-level association between the two feature vectors, the so-called "implicit collaboration". This implicit collaboration is not limited to simple linear combinations, but involves complex nonlinear interaction patterns. For example, through convolution operations, the network can identify local correlation patterns, and the use of activation functions such as ReLU or Leaky ReLU.

[0059] More specifically, the process of step S1341-1 can be expressed as follows:

[0060] ;

[0061] in, and represent the heating temperature time series dynamic mode characteristic vector and the stirring speed time series dynamic mode characteristic vector respectively, , and They are point-by-point addition, point-by-point subtraction, and point-by-point multiplication. For cascade processing, is the point convolutional encoding operation, is the Leaky ReLU activation function, is the implicit collaborative encoding vector between the heating temperature and stirring speed time series features. This ensures that the network can learn nonlinear feature representations, which is crucial for capturing the complex relationship between features.

[0062] Specifically, in step S1341-2, the implicit collaborative coding vector between the heating temperature-stirring speed timing features is written into the dynamic memory unit to obtain the heating temperature-stirring speed timing dynamic key vector. It should be understood that the dynamic memory unit is a special type of memory structure, and its working principle is similar to the memory unit in the recursive neural network, but with greater flexibility and adaptability. The dynamic memory unit can dynamically adjust its own memory content according to the current input state.

[0063] More specifically, the process of step S1341-2 can be expressed as follows:

[0064] ;

[0065] in, is a 1×1 convolution operation, is the heating temperature-stirring speed time series dynamic key vector. Therefore, when a new feature vector is input, the dynamic memory unit can adjust the memory according to the newly input data to better adapt to the context requirements of the current scene.

[0066] In an embodiment of the present application, the step S1341-3, based on the heating temperature-stirring speed timing dynamic key vector, respectively performs feature attention modulation on the heating temperature timing dynamic mode feature vector and the stirring speed timing dynamic mode feature vector to obtain the optimized heating temperature timing dynamic mode feature vector and the optimized stirring speed timing dynamic mode feature vector, including: S1341-31, extracting the heating temperature-stirring speed timing dynamic key vector from the dynamic memory unit, and inputting the heating temperature timing dynamic mode feature vector and the heating temperature-stirring speed timing dynamic key vector into a feature attention modulation module based on the first converter structure to obtain the optimized heating temperature timing dynamic mode feature vector; S1341-32, extracting the heating temperature-stirring speed timing dynamic key vector from the dynamic memory unit, and inputting the stirring speed timing dynamic mode feature vector and the heating temperature-stirring speed timing dynamic key vector into a feature attention modulation module based on the second converter structure to obtain the optimized stirring speed timing dynamic mode feature vector. It should be understood that in step S1341-3, the core of the feature attention modulation module based on the first converter structure and the feature attention modulation module based on the second converter structure is the self-attention mechanism, which allows the model to consider the influence of all other elements when processing each input element, not just adjacent elements. This global perspective helps the model capture long-distance dependencies, which is particularly important for understanding complex data.

[0067] Specifically, in step S1341-31, the heating temperature-stirring speed timing dynamic key vector is extracted from the dynamic memory unit, and the heating temperature timing dynamic mode feature vector and the heating temperature-stirring speed timing dynamic key vector are input into a feature attention modulation module based on the first converter structure to obtain the optimized heating temperature timing dynamic mode feature vector.

[0068] More specifically, the process of step S1341-31 can be expressed as follows:

[0069] ;

[0070] ;

[0071] ;

[0072] in, is the heating temperature query embedding matrix, is the first heating temperature bias vector, is matrix multiplication, is the heating temperature query vector, is the heating temperature value embedding matrix, is the second heating temperature bias vector, is the heating temperature value vector, yes The transposed vector of yes Length, yes function, is the characteristic vector of the dynamic mode of heating temperature timing after optimization.

[0073] Specifically, in the step S1341-32, the heating temperature-stirring speed timing dynamic key vector is extracted from the dynamic memory unit, and the stirring speed timing dynamic mode feature vector and the heating temperature-stirring speed timing dynamic key vector are input into the feature attention modulation module based on the second converter structure to obtain the optimized stirring speed timing dynamic mode feature vector. More specifically, the process of the above step S1341-32 can be expressed by the formula:

[0074] ;

[0075] ;

[0076] ;

[0077] in, is the stirring speed query embedding matrix, is the first stirring speed bias vector, is the stirring speed query vector, is the stirring speed value embedding matrix, is the second stirring speed bias vector, is the stirring speed value vector, is the characteristic vector of the dynamic mode of the stirring speed timing after optimization.

[0078] In this way, the feature attention modulation module based on the first converter structure and the feature attention modulation module based on the second converter structure can perform query attention-based feature modulation on the heating temperature timing dynamic mode feature vector and the stirring speed timing dynamic mode feature vector based on the heating temperature-stirring speed timing dynamic key vector, which is equivalent to using the heating temperature-stirring speed timing dynamic key vector as a common attention space, which not only enhances the expression of key information in the feature vector and reduces the influence of noise and irrelevant information, but also enables the optimized heating temperature timing dynamic mode feature vector and the optimized stirring speed timing dynamic mode feature vector to align the semantic space, which is conducive to the subsequent feature interaction and fusion between the two.

[0079] In an embodiment of the present application, the step S1341-4, inputting the optimized heating temperature timing dynamic pattern feature vector and the optimized stirring speed timing dynamic pattern feature vector into the characteristic linear interaction network to obtain the fine-grained interaction feature vector between the thermal effect-stirring speed timing features, includes: calculating the position-weighted sum of the optimized heating temperature timing dynamic pattern feature vector and the optimized stirring speed timing dynamic pattern feature vector to obtain the fine-grained interaction feature vector between the thermal effect-stirring speed timing features; wherein the sum of the weighted coefficients of the optimized heating temperature timing dynamic pattern feature vector and the optimized stirring speed timing dynamic pattern feature vector is one.

[0080] More specifically, the process of step S1341-4 can be expressed as follows:

[0081] ;

[0082] in, is the weighting coefficient, It is the fine-grained interaction feature vector between the thermal effect and stirring speed timing features. It should be understood that the main task of the feature linear interaction network is to further integrate the optimized heating temperature timing dynamic mode feature vector and the optimized stirring speed timing dynamic mode feature vector through linear transformation to generate the final thermal effect-stirring speed timing feature fine-grained interaction feature vector.

[0083] In an embodiment of the present application, the step S1342, based on the fine-grained interaction feature vector between the thermal effect and stirring speed timing features, obtains the recommended decoding value of the stirring speed at the next time point, including: passing the fine-grained interaction feature vector between the thermal effect and stirring speed timing features through a stirring controller based on a decoder to obtain the recommended decoding value of the stirring speed at the next time point. That is, the fine-grained interaction feature between the thermal effect and stirring speed timing features obtained by the dynamic memory attention interaction between the heating temperature timing dynamic mode feature vector and the stirring speed timing dynamic mode feature vector is decoded and processed to intelligently recommend the stirring speed value at the next time point. In this way, the heating temperature and stirring speed can be monitored in real time, and the complex dynamic relationship between the two can be captured to automatically adjust the stirring speed to adapt to the current optimal process conditions, thereby ensuring the continuity and stability of production.

[0084] Among them, the fine-grained interaction feature vector between the thermal effect-stirring speed timing features is passed through a decoder-based stirring controller to obtain the recommended decoding value of the stirring speed at the next time point, including: multiplying the decoding weight matrix of the decoder with the fine-grained interaction feature vector between the thermal effect-stirring speed timing features to obtain the decoded fine-grained interaction feature vector between the thermal effect-stirring speed timing features, and accumulating and summing all eigenvalues ​​of the decoded fine-grained interaction feature vector between the thermal effect-stirring speed timing features to obtain the recommended decoding value of the stirring speed at the next time point.

[0085] Here, when the heating temperature time series dynamic pattern feature vector and the stirring speed time series dynamic pattern feature vector respectively represent the multi-scale time series image semantic features of the Gram's angle field image corresponding to the heating temperature and the stirring speed, after performing dynamic memory attention interaction on them, the fine-grained interaction feature vector between the thermal effect-stirring speed time series features will also have significant attention interaction distribution spatial structure differences due to the differences in the semantic dynamic memory of the source data image, affecting the convergence consistency of the decoder, thereby affecting the accuracy of the recommended decoding value of the stirring speed at the next time point obtained by the stirring controller.

[0086] Therefore, in one example, the fine-grained interaction feature vector between the thermal effect-stirring speed time series features is optimized, and the optimization process includes:

[0087] First, the absolute value sum and the square root of the sum of squares of all eigenvalues ​​of the fine-grained interaction feature vectors between the thermal effect and stirring speed time series features are calculated to obtain the first fine-grained interaction spatial structure value between the thermal effect and stirring speed time series features and the second fine-grained interaction spatial structure value between the thermal effect and stirring speed time series features, that is:

[0088] ;

[0089] ;

[0090] in, represents the i-th eigenvalue of the fine-grained interaction feature vector between the thermal effect-stirring speed time series features, represents the absolute value function, ∑ represents the summation function, Represents the fine-grained interaction spatial structure value between the first thermal effect and stirring speed time series features, Represents the fine-grained interaction spatial structure value between the second thermal effect and stirring speed time series features.

[0091] Then, the number of eigenvalues ​​of all eigenvalues ​​of the fine-grained interaction feature vector between the thermal effect and stirring speed time series features is determined, that is, the length of the fine-grained interaction feature vector between the thermal effect and stirring speed time series features is determined. ;

[0092] Next, for each eigenvalue of the fine-grained interaction eigenvector between the thermal effect-stirring speed time series features, the fine-grained interaction long-range dependency value between the first thermal effect-stirring speed time series features is obtained by subtracting the product of the eigenvalue and the number of eigenvalues ​​from the first thermal effect-stirring speed time series features fine-grained interaction spatial structure value, that is:

[0093] ;

[0094] in, is the length of the fine-grained interaction feature vector between the thermal effect-stirring speed time series features, Represents the fine-grained interactive long-range dependency value between the first thermal effect and the stirring speed time series characteristics.

[0095] Next, the square root of the number of eigenvalues ​​multiplied by the product of the eigenvalues ​​minus the second thermal effect-stirring speed time series feature fine-grained interaction spatial structure value is calculated to obtain the second thermal effect-stirring speed time series feature fine-grained interaction long-range dependency value, that is:

[0096] ;

[0097] in, Represents the fine-grained interaction long-range dependency value between the second thermal effect and the stirring speed time series characteristics.

[0098] Furthermore, the index value calculated by taking the first thermal effect-stirring speed time series fine-grained interaction long-range dependency value as the exponent of the natural constant and the inverse of the second thermal effect-stirring speed time series fine-grained interaction long-range dependency value are weighted summed to obtain the optimized eigenvalue corresponding to each eigenvalue, that is:

[0099] ;

[0100] in, represents the fine-grained weighted value between the first features, represents a natural constant, represents the fine-grained weighted value between features, It represents the optimized eigenvalues ​​of each position in the fine-grained interaction eigenvector between the thermal effect and stirring speed timing features that constitute the optimized eigenvalues.

[0101] Finally, the optimized feature values ​​are combined into a fine-grained interaction feature vector between the optimized thermal effect-stirring speed timing features.

[0102] That is, in view of the possible spatial structure loss in the high-dimensional space of the feature set of the fine-grained interaction feature vector between the thermal effect-stirring speed time series features, which leads to the inconsistent convergence of the feature-based implicit inference spatial structure information of the decoder's weights, a long-distance feature dependency relationship is established based on the overall feature scale of the fine-grained interaction feature vector between the thermal effect-stirring speed time series features relative to the spatial structure representation of the fine-grained interaction feature vector between the thermal effect-stirring speed time series features, so as to establish the feature local connectivity of the fine-grained interaction feature vector between the thermal effect-stirring speed time series features, and to capture the spatial ambiguous information of the object feature value through the unstructured feature value point prediction of the fine-grained interaction feature vector between the thermal effect-stirring speed time series features, thereby improving the spatial inductive deviation perception ability of the feature set of the fine-grained interaction feature vector between the thermal effect-stirring speed time series features, improving the convergence consistency of the decoder, and improving the accuracy of the recommended decoding value of the stirring speed at the next time point obtained by the fine-grained interaction feature vector between the thermal effect-stirring speed time series features through the decoder-based stirring controller.

[0103] In the above-mentioned stirring control method for the production of dry Rongalite, the step S140 is to perform vacuum crystallization treatment on the concentrated sodium bisulphite formaldehyde solution to obtain Rongalite. It should be understood that vacuum crystallization treatment can perform crystallization at a lower temperature, reducing the decomposition or side reactions of heat-sensitive substances. For sodium bisulphite formaldehyde solution, vacuum crystallization can effectively remove excess water in the solution and promote the crystallization of the target substance to obtain a crystalline product with higher purity, that is, to obtain the solid product of Rongalite. Among them, the basic principle of vacuum crystallization treatment is to reduce the boiling point of the solution by reducing the gas pressure of the solution (using a vacuum device), thereby achieving the evaporation of the solvent (water) at a lower temperature. Since the solubility in the solution changes with the change of temperature, the evaporation of water in a vacuum environment increases the concentration of the solution. When it reaches a saturated or supersaturated state, the solute begins to precipitate and form crystals. In this process, due to the effect of low temperature, the decomposition of sodium bisulphite or formaldehyde at high temperature can be avoided, ensuring the stability of the crystallization process and the purity of the product. In the specific operation, the concentrated sodium bisulfite formaldehyde solution is first transferred to the crystallization tank or crystallizer, and then a vacuum environment is formed in the crystallizer. The vacuum pump reduces the air pressure in the crystallizer, evaporates the water in the solution, and the solute concentration in the solution gradually increases. As the concentration of the solution increases, when the solution reaches a supersaturated state, Rongalite begins to precipitate. In order to promote the crystallization process, the solution can be properly stirred during the crystallization process to ensure a uniform crystallization process and avoid incomplete crystallization or excessive crystallization.

[0104] It is worth mentioning that in the technical solution of the present application, after the sodium bisulphite formaldehyde solution is synthesized and concentrated in the early stage, a high-concentration melt is formed. In order to further process into high-quality solid Rongalite, it is necessary to remove the residual solvent through vacuum crystallization treatment and promote the target compound to precipitate in the form of crystals. Specifically, first of all, the vacuum crystallization system needs to be fully inspected and cleaned. This includes confirming that the functions of key components such as vacuum pumps, temperature sensors, stirring devices, etc. are intact, and preparing containers for storing finished products to ensure that they are dry and pollution-free. Then, the sodium bisulphite formaldehyde melt formed by the concentration treatment is introduced into a specially designed vacuum crystallizer. This transfer process needs to be as fast as possible and well sealed to prevent outside air from entering the system and affecting the crystallization effect. The melt is accurately transferred to the crystallizer through a pipeline or a special conveying device, and any operation that may cause changes in the properties of the solution should be avoided. Subsequently, the vacuum pump is immediately started to establish and maintain a stable vacuum environment inside the crystallizer. Vacuum conditions can effectively reduce the boiling point of the solvent, so that even at relatively low temperatures, rapid evaporation of the solvent can be achieved. This is particularly important for protecting heat-sensitive substances, as it can accelerate the crystallization process without destroying the structure of the target compound. With the vacuum environment established, the melt is then appropriately heated. Although the solvent evaporates more easily under vacuum conditions, the appropriate amount of heat input still helps to promote the escape of the solvent from the solution and accelerate the crystallization rate. The heating method can use an external heating coil or a built-in heating element to ensure that the heat is evenly distributed throughout the solution system. At the same time, the stirring device is started to keep the melt continuously and evenly mixed during the crystallization process. This stirring not only prevents the adverse consequences caused by local overheating, but also helps to form smaller and more uniform crystals, thereby improving the overall quality of the product. Modern crystallization equipment is usually equipped with an automated control system that can adjust various parameters such as temperature, pressure, stirring speed, etc. in real time to cope with possible changes. For example, if a slight increase in temperature is detected, the temperature can be quickly reduced by increasing the flow rate of the cooling medium; vice versa. In addition, it is also necessary to pay attention to the changes in the physical properties of the solution, such as color and viscosity, as an important basis for judging the degree of crystallization progress. Over time, the solute in the solution gradually precipitates to form tiny crystal particles. These particles will continue to grow under the action of continuous stirring until they reach the desired size range. When it is confirmed that the crystals have grown sufficiently, the heating and stirring can be stopped and the crystals can be separated. At this time, the crystals can be further precipitated by changing the vacuum conditions or adding a small amount of non-solvent. Then, the crystals are separated from the mother liquor using a centrifuge or other solid-liquid separation equipment to obtain pure Rongalite products.

[0105] In the above-mentioned stirring control method for the production of dry Rongalite, the step S150 is to release pressure and package the Rongalite to obtain the dry Rongalite. It should be understood that Rongalite, as a crystalline material, usually contains a certain amount of moisture or may be under high pressure, so it is necessary to release excess air pressure by pressure release to prevent the powder from absorbing moisture or deteriorating. The packaging process is to protect the Rongalite from external environmental influences, such as moisture, oxygen and pollution, to ensure its quality stability during storage and transportation. Specifically, the pressure release process refers to gradually restoring the Rongalite to normal atmospheric pressure after the crystallized Rongalite is taken out of the vacuum crystallizer. This process is usually carried out in a closed environment, and by slowly venting or gradually releasing the pressure, it is ensured that the Rongalite will not have quality problems due to sudden changes in air pressure during the pressure release process. During the pressure release process, it is necessary to pay close attention to the ambient humidity and temperature to ensure that the Rongalite will not absorb moisture or agglomerate. After the pressure release is completed, the Rongalite will be transferred to a dedicated drying device or packaging container, such as a vacuum packaging bag or a moisture-proof bag for packaging to avoid the intrusion of external moisture and pollutants. In order to ensure the dryness of Rongalite after packaging, the packaging material usually has good moisture-proof and isolation properties, which can prevent moisture from re-entering the powder and keep it dry. When packaging, nitrogen filling technology or vacuum packaging is used to remove air or oxygen from the packaging bag to extend the storage period of Rongalite and avoid oxidation or deterioration. At the same time, during the packaging process, it is necessary to ensure that Rongalite is not affected by the pressure or temperature of the external environment, which is crucial to preserve its physical and chemical properties and purity. After packaging, Rongalite is usually labeled with information such as production date, storage conditions and expiration date, and packaged according to needs. The final dry Rongalite can maintain stability for a long time under appropriate storage conditions and is easy to access and apply when needed.

[0106] In summary, the stirring control method for the production of dry Rongbai powder based on the embodiment of the present application is explained, which collects the heating temperature time series data set of the heating ring and the stirring speed time series data set of the spherical concentrate respectively, and uses the data analysis and processing technology based on deep learning to map the collected data to the Gram angle field, and then performs dynamic extraction of the time series pattern, so as to intelligently recommend the stirring speed value at the next time point according to the dynamic memory attention interaction representation between the heating temperature time series dynamic pattern characteristics and the stirring speed time series dynamic pattern characteristics. In this way, the heating temperature and the stirring speed can be monitored in real time, and the complex dynamic relationship between the two can be captured to automatically adjust the stirring speed to adapt to the current optimal process conditions, ensuring the continuity and stability of production.

[0107] Figure 6 System block diagram of a stirring control system for dry Rongalite production according to an embodiment of the present application. Figure 6As shown, according to an embodiment of the present application, a stirring control system 100 for the production of dry Rongalite includes: a dry Rongalite production data acquisition module 110, which is used to respectively acquire a heating temperature time series data set of a heating ring and a stirring speed time series data set of a spherical concentrate through a temperature sensor and a speed sensor; a Gram angular field transformation module 120, which is used to respectively perform a Gram angular field transformation on the heating temperature time series data set and the stirring speed time series data set to obtain a heating temperature mapping Gram angular field image and a stirring speed mapping Gram angular field image; a time series pattern dynamic feature extraction module 130, which is used to extract dynamic features of the ... The heating temperature mapping Grammar angle field image and the stirring speed mapping Grammar angle field image are subjected to time-series pattern dynamic feature extraction to obtain the heating temperature time-series dynamic pattern features and the stirring speed time-series dynamic pattern features; the dynamic memory attention interaction module 140 is used to perform dynamic memory attention interaction on the heating temperature time-series dynamic pattern features and the stirring speed time-series dynamic pattern features to obtain fine-grained interaction features between thermal effect and stirring speed time-series features, and based on the fine-grained interaction features between the thermal effect and stirring speed time-series features, obtain a recommended decoding value of the stirring speed at the next time point.

[0108] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned stirring control system for the production of dry Rongalite have been described in detail above. Figures 1 to 5 It has been described in detail in the description of the stirring control method for dry Rongalite production, and therefore, its repeated description will be omitted.

[0109] As described above, the stirring control system 100 for the production of dry Rongalite according to the embodiment of the present application can be implemented in various terminal devices. In one example, the stirring control system 100 for the production of dry Rongalite can be integrated into the terminal device as a software module and / or a hardware module. For example, the stirring control system 100 for the production of dry Rongalite can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the stirring control system 100 for the production of dry Rongalite can also be one of the many hardware modules of the terminal device.

[0110] Alternatively, in another example, the stirring control system 100 for the production of dry Rongalite and the terminal device may be separate devices, and the stirring control system 100 for the production of dry Rongalite 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.

Claims

1. A stirring control method for the production of dry Rongalite, comprising: Dissolving sodium metabisulfite in ice water and stirring to obtain a stirred solution; After adding zinc powder to the stirred solution and heating it, adding a reaction reagent to the heated solution and performing a constant temperature reaction to obtain a sodium bisulphite formaldehyde solution; concentrating the sodium bisulphite formaldehyde solution to obtain a concentrated sodium bisulphite formaldehyde solution; vacuum crystallizing the concentrated sodium bisulphite formaldehyde solution to obtain Rongalite; depressurizing and packaging the Rongalite to obtain the dry Rongalite; characterized in that the sodium bisulphite formaldehyde solution is concentrated to obtain a concentrated sodium bisulphite formaldehyde solution, comprising: The heating temperature time series data set of the heating ring and the stirring speed time series data set of the spherical concentrating member are collected by the temperature sensor and the speed sensor respectively; Performing Gram angular field transformation on the heating temperature time series data set and the stirring speed time series data set respectively to obtain a heating temperature mapping Gram angular field image and a stirring speed mapping Gram angular field image; Extracting the temporal pattern dynamic features of the heating temperature mapping Gram's angle field image and the stirring speed mapping Gram's angle field image to obtain the heating temperature temporal pattern dynamic features and the stirring speed temporal pattern dynamic features; Dynamic memory attention interaction is performed on the heating temperature timing dynamic pattern characteristics and the stirring speed timing dynamic pattern characteristics to obtain fine-grained interaction characteristics between thermal effect and stirring speed timing characteristics, and based on the fine-grained interaction characteristics between thermal effect and stirring speed timing characteristics, a recommended decoding value of the stirring speed at the next time point is obtained.

2. The stirring control method for dry Rongalite production according to claim 1, characterized in that: The heating temperature mapping Grammar angular field image and the stirring speed mapping Grammar angular field image are subjected to time series pattern dynamic feature extraction to obtain heating temperature time series dynamic pattern features and stirring speed time series dynamic pattern features, including: passing the heating temperature mapping Grammar angular field image and the stirring speed mapping Grammar angular field image through a time series pattern feature extractor based on a multi-scale convolutional LSTM to obtain a heating temperature time series dynamic pattern feature vector as the heating temperature time series dynamic pattern feature and a stirring speed time series dynamic pattern feature vector as the stirring speed time series dynamic pattern feature.

3. The stirring control method for dry Rongalite production according to claim 2, characterized in that: Dynamic memory attention interaction is performed on the heating temperature time series dynamic mode feature and the stirring speed time series dynamic mode feature to obtain a fine-grained interaction feature between the thermal effect and stirring speed time series features, and based on the fine-grained interaction feature between the thermal effect and stirring speed time series features, a recommended decoding value of the stirring speed at the next time point is obtained, including: Performing dynamic memory attention interaction on the heating temperature time series dynamic mode feature vector and the stirring speed time series dynamic mode feature vector to obtain a fine-grained interaction feature vector between thermal effect and stirring speed time series features as the fine-grained interaction feature between thermal effect and stirring speed time series features; Based on the fine-grained interactive feature vector between the thermal effect-stirring speed time series features, a recommended decoding value of the stirring speed at the next time point is obtained.

4. The stirring control method for dry Rongalite production according to claim 3, characterized in that: Dynamic memory attention interaction is performed on the heating temperature time series dynamic mode feature vector and the stirring speed time series dynamic mode feature vector to obtain a fine-grained interaction feature vector between thermal effect and stirring speed time series features, including: Inputting the heating temperature time series dynamic mode feature vector and the stirring speed time series dynamic mode feature vector into a collaborative feature extraction network to obtain an implicit collaborative encoding vector between the heating temperature and stirring speed time series features; Writing the implicit collaborative coding vector between the heating temperature-stirring speed time series features into a dynamic memory unit to obtain a heating temperature-stirring speed time series dynamic key vector; Based on the heating temperature-stirring speed timing dynamic key vector, feature attention modulation is performed on the heating temperature timing dynamic mode feature vector and the stirring speed timing dynamic mode feature vector respectively to obtain an optimized heating temperature timing dynamic mode feature vector and an optimized stirring speed timing dynamic mode feature vector; The optimized heating temperature timing dynamic mode feature vector and the optimized stirring speed timing dynamic mode feature vector are input into a characteristic linear interaction network to obtain a fine-grained interaction feature vector between the thermal effect-stirring speed timing features.

5. The stirring control method for dry Rongalite production according to claim 4, characterized in that: Based on the heating temperature-stirring speed timing dynamic key vector, feature attention modulation is performed on the heating temperature timing dynamic mode feature vector and the stirring speed timing dynamic mode feature vector respectively to obtain an optimized heating temperature timing dynamic mode feature vector and an optimized stirring speed timing dynamic mode feature vector, including: Extract the heating temperature-stirring speed timing dynamic key vector from the dynamic memory unit, and input the heating temperature timing dynamic mode feature vector and the heating temperature-stirring speed timing dynamic key vector into a feature attention modulation module based on the first converter structure to obtain the optimized heating temperature timing dynamic mode feature vector; The heating temperature-stirring speed timing dynamic key vector is extracted from the dynamic memory unit, and the stirring speed timing dynamic mode feature vector and the heating temperature-stirring speed timing dynamic key vector are input into a feature attention modulation module based on a second converter structure to obtain the optimized stirring speed timing dynamic mode feature vector.

6. The stirring control method for dry Rongalite production according to claim 5, characterized in that: The optimized heating temperature timing dynamic pattern feature vector and the optimized stirring speed timing dynamic pattern feature vector are input into a characteristic linear interaction network to obtain a fine-grained interaction feature vector between the thermal effect and stirring speed timing features, including: calculating the position-weighted sum of the optimized heating temperature timing dynamic pattern feature vector and the optimized stirring speed timing dynamic pattern feature vector to obtain the fine-grained interaction feature vector between the thermal effect and stirring speed timing features; wherein the sum of the weighted coefficient of the optimized heating temperature timing dynamic pattern feature vector and the weighted coefficient of the optimized stirring speed timing dynamic pattern feature vector is one.

7. The stirring control method for dry Rongalite production according to claim 6, characterized in that: Based on the fine-grained interaction feature vector between the thermal effect-stirring speed timing features, a recommended decoding value of the stirring speed at the next time point is obtained, including: passing the fine-grained interaction feature vector between the thermal effect-stirring speed timing features through a decoder-based stirring controller to obtain the recommended decoding value of the stirring speed at the next time point.

8. The stirring control method for dry Rongalite production according to claim 7, characterized in that: The fine-grained interaction feature vector between the thermal effect-stirring speed timing features is passed through a decoder-based stirring controller to obtain a recommended decoding value of the stirring speed at the next time point, including: multiplying the decoding weight matrix of the decoder with the fine-grained interaction feature vector between the thermal effect-stirring speed timing features to obtain the decoded fine-grained interaction feature vector between the thermal effect-stirring speed timing features, and accumulating and summing all eigenvalues ​​of the decoded fine-grained interaction feature vector between the thermal effect-stirring speed timing features to obtain the recommended decoding value of the stirring speed at the next time point.

9. A stirring control system for the production of dry Rongalite, used to execute the stirring control method for the production of dry Rongalite as claimed in claim 1, characterized in that: The stirring control system comprises: The dry Rongalite production data acquisition module is used to respectively acquire the heating temperature time series data set of the heating ring and the stirring speed time series data set of the spherical concentrate through the temperature sensor and the speed sensor; A Gram angular field transformation module, used for performing Gram angular field transformation on the heating temperature time series data set and the stirring speed time series data set respectively to obtain a heating temperature mapping Gram angular field image and a stirring speed mapping Gram angular field image; A time series pattern dynamic feature extraction module is used to extract time series pattern dynamic features from the heating temperature mapping Gram angle field image and the stirring speed mapping Gram angle field image to obtain heating temperature time series dynamic pattern features and stirring speed time series dynamic pattern features; A dynamic memory attention interaction module is used to perform dynamic memory attention interaction on the heating temperature timing dynamic pattern characteristics and the stirring speed timing dynamic pattern characteristics to obtain fine-grained interaction characteristics between the thermal effect-stirring speed timing characteristics, and based on the fine-grained interaction characteristics between the thermal effect-stirring speed timing characteristics, obtain a recommended decoding value of the stirring speed at the next time point.

10. The stirring control system for dry Rongalite production according to claim 9, characterized in that: The timing pattern dynamic feature extraction module is used to: pass the heating temperature mapping Gram angular field image and the stirring speed mapping Gram angular field image through a timing pattern feature extractor based on a multi-scale convolution LSTM to obtain a heating temperature timing dynamic pattern feature vector as the heating temperature timing dynamic pattern feature and a stirring speed timing dynamic pattern feature vector as the stirring speed timing dynamic pattern feature.

Images