WO2024045247A1 - 用于氟化铵生产的生产管理控制系统及其控制方法 - Google Patents

用于氟化铵生产的生产管理控制系统及其控制方法 Download PDF

Info

Publication number
WO2024045247A1
WO2024045247A1 PCT/CN2022/121231 CN2022121231W WO2024045247A1 WO 2024045247 A1 WO2024045247 A1 WO 2024045247A1 CN 2022121231 W CN2022121231 W CN 2022121231W WO 2024045247 A1 WO2024045247 A1 WO 2024045247A1
Authority
WO
WIPO (PCT)
Prior art keywords
flow rate
vector
feature
matrix
reaction
Prior art date
Application number
PCT/CN2022/121231
Other languages
English (en)
French (fr)
Inventor
华鹏
郑琦
丘贵龙
邹珍妮
石凌斌
Original Assignee
福建天甫电子材料有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 福建天甫电子材料有限公司 filed Critical 福建天甫电子材料有限公司
Publication of WO2024045247A1 publication Critical patent/WO2024045247A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B19/00—Programme-control systems
    • G05B19/02—Programme-control systems electric
    • G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
    • G05B19/41885—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by modeling, simulation of the manufacturing system
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B2219/00—Program-control systems
    • G05B2219/30—Nc systems
    • G05B2219/32—Operator till task planning
    • G05B2219/32339—Object oriented modeling, design, analysis, implementation, simulation language
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Definitions

  • the present invention relates to the field of intelligent production management, and more specifically, to a production management control system for ammonium fluoride production and a control method thereof.
  • Ammonium fluoride the molecular formula is NH4F, the relative molecular mass is 37.04, the relative density is 1.015 (25°C), colorless leaf-like or needle-like crystals, after sublimation, it becomes hexagonal columnar crystals; easy to deliquesce and agglomerate, soluble in cold water, slightly Soluble in alcohol, insoluble in acetone and liquid ammonia. When heated or exposed to hot water, it decomposes and loses ammonia and converts it into more stable ammonium fluoride.
  • Ammonium fluoride is widely used, such as as a glass etching agent, a chemical polishing agent for metal surfaces, a wood and wine preservative, a disinfectant, a mordant for fibers, and a solvent for extracting rare elements. It can also be used as a component for ion detection in chemical analysis.
  • the traditional production method of ammonium fluoride is the liquid phase method: put a certain amount of hydrofluoric acid into a lead or plastic container. Cool with water outside the container, and slowly introduce ammonia gas while stirring until the pH value of the reaction solution reaches about 4. The reaction solution is cooled and crystallized, centrifuged, and air-dried to obtain ammonium fluoride product.
  • Ammonium fluoride produced by the traditional liquid phase method has shortcomings such as high water content, easy caking, and inability to be stored for a long time. Therefore, an optimized preparation scheme for ammonium fluoride is expected.
  • the embodiment of the present application provides a production management control system and a control method for ammonium fluoride production, which adopts artificial intelligence control technology to add flow rate values of liquid ammonia at multiple predetermined time points within a predetermined time period.
  • the addition flow rate value of anhydrous hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the pH value of the reaction liquid are used as input data, and a deep neural network model is used as a feature extractor to add liquid ammonia and anhydrous hydrogen fluoride into the reaction tank.
  • the real-time dynamic characteristics of the flow rate and the change characteristics of the reaction temperature are coordinated to carry out dynamic intelligent control of the flow rate of the cooling water.
  • the change characteristic information of the pH value of the reaction liquid is also added as the basis for the final result to improve the reaction efficiency and product quality.
  • a production management control system for ammonium fluoride production which includes:
  • the production parameter acquisition module is used to obtain the addition flow rate value of liquid ammonia, the addition flow rate value of anhydrous hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the pH value of the reaction liquid at multiple predetermined time points within a predetermined time period; adding The speed structured correlation module is used to arrange the adding flow rate values of liquid ammonia and the adding flow rate values of anhydrous hydrogen fluoride at multiple predetermined time points within the predetermined time period into the first adding flow rate vector and the second adding flow rate value according to the time dimension. After adding the flow velocity vector, calculate the flow velocity control correlation matrix between the first added flow velocity vector and the second added flow velocity vector; add a velocity feature extraction module for converting the flow velocity control correlation matrix to each other through adjacent layers. Transpose the first convolutional neural network of the convolution kernel to obtain the flow rate control feature matrix;
  • the reaction data encoding module is used to arrange the reaction temperature values and cooling water flow rate values at multiple predetermined time points within the predetermined time period into input vectors according to the time dimension, and then pass the first temporal encoding including a one-dimensional convolution layer
  • the processor is used to obtain the temperature feature vector and the cooling feature vector;
  • the difference module is used to calculate the difference feature vector between the temperature feature vector and the cooling feature vector;
  • the state data encoding module is used to encode multiple data within the predetermined time period.
  • the pH value of the reaction solution at a predetermined time point is passed through a second temporal encoder including a one-dimensional convolution layer to obtain a PH temporal feature vector;
  • a fusion module is used to compare the differential feature vector with the flow rate control feature matrix. Multiply to obtain a reaction feature vector;
  • a responsiveness estimation module used to calculate the responsiveness estimate of the PH timing feature vector relative to the reaction feature vector to obtain a classification feature vector; and
  • the production management control result generation module is used to pass the classification feature vector through a classifier to obtain a classification result.
  • the classification result is used to indicate that the flow rate value of the cooling water at the current point in time should be increased or decreased.
  • the adding speed structured correlation module includes: a vector correlation unit for calculating the first adding flow rate vector and the second adding flow rate vector according to the following formula: The flow rate control correlation matrix between flow rate vectors; wherein, the formula is:
  • V 1 represents the first added flow velocity vector
  • V 2 represents the second added flow velocity vector
  • M represents the flow rate control correlation matrix
  • the adding speed feature extraction module includes: a matrix correction unit for controlling the flow rate based on the position information of each position in the flow rate control correlation matrix.
  • the correlation matrix is modified to obtain the corrected flow velocity control correlation matrix;
  • the convolution coding unit is used to pass the corrected flow velocity control correlation matrix through the first convolution neural network of the adjacent layers using mutually transposed convolution kernels. network to obtain the flow rate control characteristic matrix.
  • the matrix correction unit is further used to: based on the position information of each position in the flow rate control correlation matrix, perform the following formula on the flow rate control correlation matrix: Correct to obtain the corrected flow rate control correlation matrix; wherein, the formula is:
  • M represents the flow rate control correlation matrix
  • M' represents the corrected flow rate control correlation matrix
  • Cov 1 () and Cov 2 () are both single convolution layers, Used to map two-dimensional position coordinates to one-dimensional values
  • P M represents the (x, y) coordinate matrix of matrix M
  • ⁇ represents the position-wise dot multiplication.
  • the convolution coding unit includes: a shallow feature map extraction subunit for extracting shallow layers from the Mth layer of the first convolutional neural network Feature matrix, M is an even number; the deep feature map extraction subunit is used to extract the deep feature matrix from the Nth layer of the first convolutional neural network, where N is an even number, and N is greater than 2 times of M; and, A feature map fusion subunit is used to fuse the shallow feature map and the deep feature map to generate the flow rate control feature matrix.
  • the reaction data encoding module includes: an input vector construction unit for converting the reaction temperature values and cooling water values at multiple predetermined time points within the predetermined time period.
  • the flow rate values are arranged into temperature input vectors and flow rate input vectors according to the time dimension;
  • a fully connected encoding unit is used to use the fully connected layer of the first temporal encoder to respectively encode the temperature input vector and the flow rate using the following formula
  • the input vector is fully connected to encode to respectively extract the high-dimensional hidden features of the eigenvalues of each position in the temperature input vector and the flow velocity input vector, where the formula is: where X is the input vector, Y is the output vector, W is the weight matrix, and B is the bias vector, represents a matrix multiplication;
  • a one-dimensional convolution coding unit used to perform one-dimensional convolution coding on the temperature input vector and the flow rate input vector using the following formula using the one-dimensional convolution layer of the first temporal encoder to respectively High-dimensional implicit correlation features between
  • a is the width of the convolution kernel in the x direction
  • F is the convolution kernel parameter vector
  • G is the local vector matrix that operates with the convolution kernel function
  • w is the size of the convolution kernel
  • X represents the input vector.
  • the differential module is further used to: calculate the differential eigenvector between the temperature eigenvector and the cooling eigenvector according to the following formula; wherein, The formula is:
  • V t represents the temperature eigenvector
  • V c represents the cooling eigenvector
  • V d represents the differential eigenvector
  • the status data encoding module is further used to: arrange the pH values of the reaction liquid at multiple predetermined time points within the predetermined time period into one according to the time dimension. dimensional input vector; use the fully connected layer of the second temporal encoder to perform fully connected encoding on the input vector with the following formula to extract the high-dimensional hidden features of the eigenvalues of each position in the input vector, where, The formula is: where X is the input vector, Y is the output vector, W is the weight matrix, and B is the bias vector, represents matrix multiplication; use the one-dimensional convolution layer of the second temporal encoder to perform one-dimensional convolution encoding on the input vector with the following formula to extract the high-dimensional implicit correlation between the feature values of each position in the input vector Characteristics, where the formula is:
  • a is the width of the convolution kernel in the x direction
  • F is the convolution kernel parameter vector
  • G is the local vector matrix that operates with the convolution kernel function
  • w is the size of the convolution kernel
  • X represents the input vector.
  • the responsiveness estimation module is further used to: calculate the responsiveness estimate of the pH timing feature vector relative to the reaction feature vector using the following formula to obtain the The classification feature vector; where, the formula is
  • s 1 represents the PH time series feature vector
  • s 2 represents the reaction feature vector
  • s 3 represents the classification feature vector
  • ⁇ represents the vector dot product
  • ⁇ -1 represents the reciprocal of the value of each position of the vector.
  • the production management control result generation module is further configured to: use the classifier to process the classification feature vector with the following formula to obtain the classification result , where the formula is: softmax ⁇ (W n ,B n ):...:(W 1 ,B 1 )
  • a control method for a production management control system for ammonium fluoride production includes: obtaining the addition flow rate values of liquid ammonia at multiple predetermined time points within a predetermined time period, the flow rate of anhydrous hydrogen fluoride Add the flow rate value, the reaction temperature value, the flow rate value of the cooling water and the PH value of the reaction liquid; add the flow rate values of liquid ammonia and the addition flow rate of anhydrous hydrogen fluoride at multiple predetermined time points within the predetermined time period according to the time respectively
  • the flow control correlation matrix between the first added flow velocity vector and the second added flow velocity vector is calculated; the flow control correlation matrix is passed through the adjacent
  • the first convolutional neural network layer uses mutually transposed convolution kernels to obtain the flow rate control feature matrix; the reaction temperature values and the flow rate values of the cooling water at multiple predetermined time points within the predetermined time period are calculated according to the time dimension.
  • the temperature feature vector and the cooling feature vector are obtained through the first temporal encoder including a one-dimensional convolution layer; calculating the differential feature vector between the temperature feature vector and the cooling feature vector;
  • the PH values of the reaction solution at multiple predetermined time points within a predetermined time period are passed through a second temporal encoder including a one-dimensional convolution layer to obtain a PH temporal feature vector;
  • the differential feature vector is compared with the flow rate control feature matrix. Multiply to obtain a reaction feature vector; calculate the responsiveness estimate of the PH timing feature vector relative to the reaction feature vector to obtain a classification feature vector; and
  • the classification feature vector is passed through a classifier to obtain a classification result, which is used to indicate that the flow rate value of the cooling water at the current point in time should be increased or decreased.
  • calculating the flow rate control correlation matrix between the first addition flow rate vector and the second addition flow rate vector includes: calculating the flow rate with the following formula The flow rate control correlation matrix between the first added flow rate vector and the second added flow rate vector; wherein, the formula is:
  • V 1 represents the first added flow velocity vector
  • V 2 represents the second added flow velocity vector
  • M represents the flow rate control correlation matrix
  • the flow rate control correlation matrix is passed through the first convolutional neural network using mutually transposed convolution kernels in adjacent layers to obtain the flow rate control characteristics.
  • the matrix includes: based on the position information of each position in the flow rate control correlation matrix, modifying the flow speed control correlation matrix to obtain a corrected flow speed control correlation matrix; passing the corrected flow speed control correlation matrix through the adjacent
  • the first convolutional neural network layer uses mutually transposed convolution kernels to obtain the flow rate control feature matrix.
  • the flow rate control correlation matrix is modified to obtain a corrected flow rate control correlation matrix
  • the method includes: based on the position information of each position in the flow rate control correlation matrix, modifying the flow speed control correlation matrix with the following formula to obtain the corrected flow speed control correlation matrix; wherein the formula is:
  • M represents the flow rate control correlation matrix
  • M' represents the corrected flow rate control correlation matrix
  • Cov 1 () and Cov 2 () are both single convolution layers, Used to map two-dimensional position coordinates to one-dimensional values
  • P M represents the (x, y) coordinate matrix of matrix M
  • ⁇ represents the position-wise dot multiplication.
  • the corrected flow rate control correlation matrix is passed through the adjacent layer using the first convolutional neural network of mutually transposed convolution kernels to Obtaining the flow rate control feature matrix includes: extracting a shallow feature matrix from the Mth layer of the first convolutional neural network, where M is an even number; extracting a deep feature matrix from the Nth layer of the first convolutional neural network , where N is an even number, and N is greater than 2 times of M; and, the shallow feature map and the deep feature map are fused to generate the flow rate control feature matrix.
  • the reaction temperature values and cooling water flow rate values at multiple predetermined time points within the predetermined time period are arranged into input vectors according to the time dimension and then passed
  • the first temporal encoder including a one-dimensional convolution layer to obtain the temperature feature vector and the cooling feature vector includes: converting the reaction temperature values and cooling water flow rate values at multiple predetermined time points within the predetermined time period according to the time dimension respectively.
  • a is the width of the convolution kernel in the x direction
  • F is the convolution kernel parameter vector
  • G is the local vector matrix that operates with the convolution kernel function
  • w is the size of the convolution kernel
  • X represents the input vector.
  • calculating the differential feature vector between the temperature feature vector and the cooling feature vector includes: calculating the temperature feature vector and the cooling feature vector with the following formula The differential eigenvector between the cooling eigenvectors; wherein, the formula is:
  • V t represents the temperature eigenvector
  • V c represents the cooling eigenvector
  • V d represents the differential eigenvector
  • a is the width of the convolution kernel in the x direction
  • F is the convolution kernel parameter vector
  • G is the local vector matrix that operates with the convolution kernel function
  • w is the size of the convolution kernel
  • X represents the input vector.
  • calculating the responsiveness estimate of the pH time series feature vector relative to the reaction feature vector to obtain the classification feature vector includes: calculating the following formula The responsiveness of the PH time series feature vector relative to the reaction feature vector is estimated to obtain the classification feature vector; where, the formula is
  • s 1 represents the PH time series feature vector
  • s 2 represents the reaction feature vector
  • s 3 represents the classification feature vector
  • ⁇ represents the vector dot product
  • ⁇ -1 represents the reciprocal of the value of each position of the vector.
  • passing the classification feature vector through a classifier to obtain a classification result includes: using the classifier to process the classification feature vector with the following formula To obtain the classification result, the formula is: softmax ⁇ (W n ,B n ):...:(W 1 ,B 1 )
  • the production management control system and its control method for ammonium fluoride production adopt artificial intelligence control technology and add liquid ammonia at multiple predetermined time points within a predetermined time period.
  • the flow rate value, the addition flow rate value of anhydrous hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the pH value of the reaction liquid are used as input data, and a deep neural network model is used as a feature extractor to add the reaction between liquid ammonia and anhydrous hydrogen fluoride.
  • the real-time dynamic characteristics of the flow rate in the tank and the change characteristics of the reaction temperature are coordinated to carry out dynamic intelligent control of the flow rate of the cooling water.
  • the change characteristic information of the pH value of the reaction liquid is also added as the basis for the final result to improve the reaction. efficiency and product quality.
  • Figure 1A is a flow chart of a preparation process of a production management control system for ammonium fluoride production according to an embodiment of the present application.
  • Figure 1B is an application scenario diagram of a production management control system for ammonium fluoride production according to an embodiment of the present application.
  • Figure 2 is a block diagram of a production management control system for ammonium fluoride production according to an embodiment of the present application.
  • Figure 3 is a block diagram of a speed feature extraction module added to a production management control system for ammonium fluoride production according to an embodiment of the present application.
  • Figure 4 is a flow chart of a control method of a production management control system for ammonium fluoride production according to an embodiment of the present application.
  • Figure 5 is an architectural schematic diagram of a control method of a production management control system for ammonium fluoride production according to an embodiment of the present application.
  • ammonium fluoride has a molecular formula of NH4F, a relative molecular mass of 37.04, and a relative density of 1.015 (25°C). It is colorless leaf-like or needle-like crystals and turns into hexagonal columnar crystals after sublimation. It is easy to deliquesce and agglomerate, and can Soluble in cold water, slightly soluble in alcohol, insoluble in acetone and liquid ammonia. When heated or exposed to hot water, it decomposes and loses ammonia and converts it into more stable ammonium fluoride.
  • Ammonium fluoride is widely used, such as as a glass etching agent, a chemical polishing agent for metal surfaces, a wood and wine preservative, a disinfectant, a mordant for fibers, and a solvent for extracting rare elements. It can also be used as a component for ion detection in chemical analysis.
  • the traditional production method of ammonium fluoride is the liquid phase method: put a certain amount of hydrofluoric acid into a lead or plastic container. Cool with water outside the container, and slowly introduce ammonia gas while stirring until the pH value of the reaction solution reaches about 4. The reaction solution is cooled and crystallized, centrifuged, and air-dried to obtain ammonium fluoride product.
  • Ammonium fluoride produced by the traditional liquid phase method has shortcomings such as high water content, easy caking, and inability to be stored for a long time.
  • the preparation process is:
  • Step 1 Add the mother liquor to the reaction tank, then add liquid ammonia and anhydrous hydrogen fluoride under stirring to react;
  • Step 2 Prepare the reaction solution by cooling, crystallizing, centrifuging and drying to prepare ammonium fluoride.
  • the mother liquid is the liquid obtained by centrifugal separation of the reaction liquid in step 2.
  • the ammonium fluoride produced has the advantages of low product moisture content, not easy to agglomerate, durable in storage, and of high quality.
  • the mother liquid is a liquid obtained after centrifugal separation of the reaction liquid, and its main components are ammonium fluoride and ammonia water.
  • a certain amount of mother liquor can be prepared in advance to start the preparation process.
  • the liquid obtained by centrifugation of the reaction liquid can be recycled as the mother liquid.
  • the liquid obtained by centrifugation of the reaction liquid can be retained. Make the mother liquor needed for the next preparation, no need to prepare another mother liquor.
  • liquid ammonia and anhydrous hydrogen fluoride in the following order: first add 50 to 60 kg of liquid ammonia, then add 100 to 110 kg of anhydrous hydrogen fluoride, and finally add the remaining liquid ammonia and anhydrous hydrogen fluoride at the same time.
  • anhydrous hydrogen fluoride has a high density. If added first, it will sink to the bottom, resulting in uneven reactions.
  • Adding a certain amount of liquid ammonia first and then adding a certain amount of anhydrous hydrogen fluoride can effectively avoid uneven reactions while ensuring When the reaction is uniform, adding the remaining liquid ammonia and anhydrous hydrogen fluoride at the same time will help improve production efficiency and avoid excessively long production cycles and increased production costs.
  • liquid ammonia and anhydrous hydrogen fluoride should be added slowly to control the reaction temperature between 90 and 110°C.
  • a cooling water pipe can also be set up on the reaction tank, supplemented by cooling water for cooling. If the reaction temperature If the reaction rises too fast, it can be adjusted by reducing the feed amount or increasing the cooling water.
  • the reaction pressure should be controlled at normal pressure so that the reaction proceeds continuously, uniformly, slowly and stably.
  • the pH value of the reaction end point is preferably controlled at 5 to 6.
  • the specific control method can be carried out in the following way: when there is still 5% difference from the end point of the feed (calculated as liquid ammonia), use pH test paper or other pH detection device to detect the pH value of the reaction solution. pH value, and then adjust the remaining amount of liquid ammonia and anhydrous hydrogen fluoride accordingly according to the test results, so that the pH value at the end of the reaction is controlled at 5 to 6.
  • the inventor of the present application found that in the above preparation scheme, the synergy between the flow rate control of liquid ammonia and anhydrous hydrogen fluoride added to the reaction tank and the reaction temperature is of great significance for improving reaction efficiency and product quality. Therefore, in the technical solution of this application, it is expected to comprehensively carry out real-time dynamic control of the reaction through the addition flow rate of liquid ammonia, the addition flow rate of anhydrous hydrogen fluoride, the flow rate of cooling water, and the reaction temperature value, and through the pH detection device Detect the pH value of the reaction solution to determine the final end time, thereby improving production efficiency and product quality.
  • the addition flow rate values of liquid ammonia and the addition of anhydrous hydrogen fluoride at multiple predetermined time points within a predetermined time period are obtained through various sensors, such as flow rate sensors, temperature sensors and pH value sensors. Flow rate value, reaction temperature value, cooling water flow rate value and pH value of the reaction solution. Then, for the addition flow rate of liquid ammonia and the addition flow rate of anhydrous hydrogen fluoride at the plurality of predetermined time points, there is some hidden value between the addition flow rate of liquid ammonia and the addition flow rate of anhydrous hydrogen fluoride.
  • the adding flow rate values of liquid ammonia and the adding flow rate value of anhydrous hydrogen fluoride at multiple predetermined time points within the predetermined time period are arranged as the first adding flow rate according to the time dimension.
  • vector and the second addition flow rate vector to integrate the addition flow rate value of liquid ammonia and the addition flow rate value information of anhydrous hydrogen fluoride at each time point, and further calculate the first addition flow rate vector and the second addition flow rate vector. For example, the product between the transpose vector of the first added flow velocity vector and the second added flow velocity vector is calculated to obtain the flow rate control correlation matrix.
  • the flow rate control correlation matrix having the correlation information of the addition flow rate value of liquid ammonia and the addition flow rate value of anhydrous hydrogen fluoride at each time point can be passed through adjacent layers using mutually transposed convolution kernels.
  • Feature extraction is performed in the first convolutional neural network to obtain the flow rate control feature matrix.
  • processing using a convolutional neural network model in which adjacent convolutional layers are convolution kernels that are transposed to each other can simultaneously update network parameters and search for network parameter structures suitable for specific data structures during training.
  • correlation features of the addition flow rate of anhydrous hydrogen fluoride that are more suitable for expressing the addition flow rate value of liquid ammonia and the addition flow rate of anhydrous hydrogen fluoride can be extracted, thereby improving the accuracy of subsequent classification.
  • the first convolutional neural network using mutually transposed convolution kernels in adjacent layers can search for a network parameter structure suitable for a specific local data structure in the flow rate control correlation matrix, this It also makes the parameter configuration of the convolution kernel of the first convolutional neural network more focused on the local flow rate correlation value in the flow rate control correlation matrix, so that the flow rate control characteristic matrix may be added to the liquid ammonia flow rate value.
  • the global correlation characteristics with the added flow rate value of the anhydrous hydrogen fluoride are insufficiently expressed.
  • the correlation characteristic expression characteristic is related to the predetermined numerical value of the addition flow rate value of the liquid ammonia and the addition flow rate value of the anhydrous hydrogen fluoride, and the flow rate control characteristic matrix To optimize, that is:
  • Cov 1 () and Cov 2 () are both single convolutional layers, Used to map two-dimensional position coordinates to one-dimensional values, P M represents the (x, y) coordinate matrix of matrix M.
  • the coordinate transformation function is used
  • the position information obtained by ⁇ is used as a proposal to reason about the global scene semantics of the flow control feature matrix through the local perceptual field of the convolutional layer, and by adding the bias of the structural transposition, the structural distribution of the associated feature information is realized. Local-global migration.
  • the optimized flow rate control feature matrix can obtain better global expression of the characteristic correlation characteristics of the added flow rate value of liquid ammonia and the added flow rate value of anhydrous hydrogen fluoride.
  • a first temporal encoder including a one-dimensional convolution layer is used to respectively analyze the reaction temperature values and cooling values of multiple predetermined time points within the predetermined time period.
  • the water flow rate value is encoded to obtain a temperature feature vector and a cooling feature vector.
  • the temporal encoder consists of alternately arranged fully connected layers and one-dimensional convolutional layers, which respectively extract the reaction temperature value and the value of the cooling water through one-dimensional convolutional coding.
  • the correlation of the flow rate values in the time series dimension and the high-dimensional hidden features of the reaction temperature value and the flow rate value of the cooling water are respectively extracted through fully connected coding.
  • the multiple predetermined time points within the predetermined time period are The pH value of the reaction solution at the time point is passed through the second temporal encoder including a one-dimensional convolution layer to obtain the pH temporal feature vector.
  • the differential feature vector and the flow rate control feature matrix can be multiplied to fuse the two feature information to obtain a reaction feature vector.
  • the characteristic scales of the dynamic change characteristics of the pH value of the reaction liquid and the implicit dynamic characteristics associated with the parameters are different, and the dynamic characteristics of the pH value of the reaction liquid are in a high-dimensional feature space. It can be regarded as a responsive feature to the change of the parameter correlation. Therefore, in order to better integrate the PH timing feature vector and the reaction feature vector, the relationship between the PH timing feature vector and the reaction feature vector is further calculated. Responsiveness is estimated to obtain categorical feature vectors. In this way, the classification feature vector can be passed through a classifier to obtain a classification result indicating that the flow rate value of the cooling water at the current time point should be increased or decreased.
  • this application proposes a production management control system for the production of ammonium fluoride, which includes: a production parameter acquisition module, used to obtain the addition flow rate value of liquid ammonia at multiple predetermined time points within a predetermined time period, without The addition flow rate value of water hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the pH value of the reaction liquid; the addition speed structured correlation module is used to add the flow rate of liquid ammonia at multiple predetermined time points within the predetermined time period After the value and the addition flow rate value of anhydrous hydrogen fluoride are arranged into the first addition flow rate vector and the second addition flow rate vector according to the time dimension, the flow rate control correlation between the first addition flow rate vector and the second addition flow rate vector is calculated.
  • a production parameter acquisition module used to obtain the addition flow rate value of liquid ammonia at multiple predetermined time points within a predetermined time period, without The addition flow rate value of water hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the pH value of the reaction liquid
  • Matrix add a velocity feature extraction module for passing the flow rate control correlation matrix through the first convolutional neural network using mutually transposed convolution kernels in adjacent layers to obtain the flow rate control feature matrix; a reaction data encoding module, with After arranging the reaction temperature values and cooling water flow rate values at multiple predetermined time points within the predetermined time period into input vectors according to the time dimension, the temperature feature vector is obtained through the first temporal encoder including a one-dimensional convolution layer.
  • a difference module used to calculate the difference feature vector between the temperature feature vector and the cooling feature vector
  • a state data encoding module used to encode the responses at multiple predetermined time points within the predetermined time period
  • the pH value of the liquid passes through a second temporal encoder including a one-dimensional convolution layer to obtain a PH temporal feature vector
  • a fusion module is used to multiply the differential feature vector and the flow rate control feature matrix to obtain a reaction feature vector
  • Responsiveness estimation module used to calculate the responsiveness estimate of the PH timing feature vector relative to the reaction feature vector to obtain a classification feature vector
  • a production management control result generation module used to pass the classification feature vector through
  • a classifier is used to obtain a classification result, which is used to indicate that the flow rate value of the cooling water at the current point in time should be increased or decreased.
  • FIG. 1B illustrates an application scenario diagram of a production management control system for ammonium fluoride production according to an embodiment of the present application.
  • the reaction tank for example, R as shown in Figure 1B
  • PH value sensor T3 the addition flow rate value of liquid ammonia, the addition flow rate value of anhydrous hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the PH value of the reaction liquid at multiple predetermined time points within a predetermined time period.
  • the obtained addition flow rate values of liquid ammonia, the addition flow rate value of anhydrous hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the pH value of the reaction liquid at multiple predetermined time points within the predetermined time period are input to the deployment
  • a server with a production management control algorithm for ammonium fluoride production for example, the cloud server S as shown in Figure 1B
  • the server can control the said production management control algorithm with the production management control algorithm for ammonium fluoride production.
  • the addition flow rate values of liquid ammonia, the addition flow rate value of anhydrous hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the pH value of the reaction liquid at multiple predetermined time points within a predetermined time period are processed to generate a representation of the current time.
  • the classification result of the cooling water flow rate value of the point should be increased or decreased.
  • Figure 2 illustrates a block diagram of a production management control system for ammonium fluoride production according to an embodiment of the present application.
  • the production management control system 200 for ammonium fluoride production according to the embodiment of the present application includes: a production parameter collection module 210, used to obtain the addition of liquid ammonia at multiple predetermined time points within a predetermined time period.
  • the adding speed structured correlation module 220 is used to combine the values of multiple predetermined time points within the predetermined time period.
  • the first addition flow rate vector and the second addition flow rate vector are calculated.
  • the reaction data encoding module 240 is used to arrange the reaction temperature values and cooling water flow rate values at multiple predetermined time points within the predetermined time period into input vectors according to the time dimension, and then pass them through the first time series including a one-dimensional convolution layer.
  • the difference module 250 is used to calculate the difference feature vector between the temperature feature vector and the cooling feature vector;
  • the state data encoding module 260 is used to convert the predetermined time
  • the PH values of the reaction solution at multiple predetermined time points within the segment are passed through a second temporal encoder including a one-dimensional convolution layer to obtain a PH temporal feature vector;
  • the fusion module 270 is used to combine the differential feature vector with the flow rate control
  • the feature matrices are multiplied to obtain a reaction feature vector;
  • the responsiveness estimation module 280 is used to calculate the responsiveness estimate of the PH timing feature vector relative to the reaction feature vector to obtain a classification feature vector;
  • production management control result generation Module 290 is used to pass the classification feature vector through a classifier to obtain a classification result.
  • the classification result is used to indicate that the flow rate value of the cooling water at the current point in time should be increased or decreased.
  • the production parameter collection module 210 and the adding speed structured correlation module 220 are used to obtain the adding flow rate values of liquid ammonia and anhydrous at multiple predetermined time points within a predetermined time period.
  • the addition flow rate value of hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the pH value of the reaction liquid, and the addition flow rate value of liquid ammonia and the addition flow rate value of anhydrous hydrogen fluoride at multiple predetermined time points within the predetermined time period After arranging the first added flow velocity vector and the second added flow velocity vector according to the time dimension, the flow rate control correlation matrix between the first added flow velocity vector and the second added flow velocity vector is calculated.
  • the addition flow rate values of liquid ammonia and anhydrous value at multiple predetermined time points within a predetermined time period are obtained through various sensors, such as a flow rate sensor, a temperature sensor and a pH value sensor.
  • the addition flow rate of hydrogen fluoride, the reaction temperature, the flow rate of cooling water and the pH value of the reaction solution are obtained through various sensors, such as a flow rate sensor, a temperature sensor and a pH value sensor.
  • the addition flow rate of liquid ammonia and the addition flow rate of anhydrous hydrogen fluoride at the plurality of predetermined time points there is some hidden value between the addition flow rate of liquid ammonia and the addition flow rate of anhydrous hydrogen fluoride.
  • the adding flow rate values of liquid ammonia and the adding flow rate value of anhydrous hydrogen fluoride at multiple predetermined time points within the predetermined time period are arranged as the first adding flow rate according to the time dimension.
  • vector and the second addition flow rate vector to integrate the addition flow rate value of liquid ammonia and the addition flow rate value information of anhydrous hydrogen fluoride at each time point, and further calculate the sum of the first addition flow rate vector and the second addition flow rate vector. For example, the product of the transpose vector of the first added flow velocity vector and the second added flow velocity vector is calculated to obtain the flow rate control correlation matrix.
  • the adding speed structured correlation module includes: a vector correlation unit, used to calculate the relationship between the first adding flow rate vector and the second adding flow rate vector using the following formula The flow rate control correlation matrix of ; wherein, the formula is:
  • V 1 represents the first added flow velocity vector
  • V 2 represents the second added flow velocity vector
  • M represents the flow rate control correlation matrix
  • the added velocity feature extraction module 230 is used to pass the flow velocity control correlation matrix through the first convolutional neural network using mutually transposed convolution kernels in adjacent layers to obtain Flow rate control characteristic matrix. That is, in the technical solution of the present application, the flow rate control correlation matrix having the correlation information of the addition flow rate value of liquid ammonia and the addition flow rate value of anhydrous hydrogen fluoride at each time point can be passed through the adjacent layer Feature extraction is performed in the first convolutional neural network using mutually transposed convolution kernels to obtain a flow rate control feature matrix.
  • processing using a convolutional neural network model in which adjacent convolutional layers are convolution kernels that are transposed to each other can simultaneously update network parameters and search for network parameter structures suitable for specific data structures during training.
  • correlation features of the addition flow rate of anhydrous hydrogen fluoride that are more suitable for expressing the addition flow rate value of liquid ammonia and the addition flow rate of anhydrous hydrogen fluoride can be extracted, thereby improving the accuracy of subsequent classification.
  • the adding speed feature extraction module includes: first, based on the position information of each position in the flow speed control correlation matrix, correcting the flow speed control correlation matrix to obtain the corrected Flow rate control correlation matrix.
  • the first convolutional neural network can search for a network parameter structure suitable for a specific local data structure in the flow rate control correlation matrix. , which also makes the parameter configuration of the convolution kernel of the first convolutional neural network more focused on the local flow rate correlation value in the flow rate control correlation matrix, so that the flow rate control characteristic matrix may be used when the liquid ammonia is added.
  • the flow speed control correlation matrix is modified with the following formula to obtain the corrected flow speed control correlation matrix; wherein, The formula is:
  • M represents the flow rate control correlation matrix
  • M' represents the corrected flow rate control correlation matrix
  • Cov 1 () and Cov 2 () are both single convolution layers, Used to map two-dimensional position coordinates to one-dimensional values
  • P M represents the (x, y) coordinate matrix of matrix M
  • ⁇ represents the position-wise dot multiplication.
  • the coordinate transformation function is used
  • the position information obtained by ⁇ is used as a proposal to reason about the global scene semantics of the flow control feature matrix through the local perceptual field of the convolutional layer, and by adding the bias of the structural transposition, the structural distribution of the associated feature information is realized. Local-global migration.
  • the optimized flow rate control feature matrix can obtain better global expression of the characteristic correlation characteristics of the added flow rate value of liquid ammonia and the added flow rate value of anhydrous hydrogen fluoride.
  • the convolutional coding unit includes: a shallow feature map extraction subunit, used to extract a shallow feature matrix from the Mth layer of the first convolutional neural network, where M is an even number; a deep feature map extraction subunit, used to extract a deep feature matrix from the Nth layer of the first convolutional neural network, where N is an even number, and N is greater than 2 times M; and the feature map fusion subunit , used to fuse the shallow feature map and the deep feature map to generate the flow rate control feature matrix.
  • Figure 3 illustrates a block diagram of the addition speed feature extraction module in the automatic batching system for the preparation of lithium hexafluorophosphate according to an embodiment of the present application.
  • the added speed feature extraction module 230 includes: a matrix correction unit 231, used to correct the flow rate control correlation matrix based on the position information of each position in the flow speed control correlation matrix to obtain a correction.
  • the post-flow velocity control correlation matrix; the convolution encoding unit 232 is used to pass the corrected flow velocity control correlation matrix through the adjacent layer using the first convolutional neural network of mutually transposed convolution kernels to obtain the flow velocity. Control feature matrix.
  • the reaction data encoding module 240 is used to arrange the reaction temperature values and cooling water flow rate values at multiple predetermined time points within the predetermined time period into input vectors according to the time dimension. Then, the temperature feature vector and the cooling feature vector are obtained through the first temporal encoder containing a one-dimensional convolution layer.
  • a first temporal encoder including a one-dimensional convolution layer is used to respectively encode the predetermined time period.
  • reaction temperature values and cooling water flow rate values at multiple predetermined time points are encoded to obtain a temperature feature vector and a cooling feature vector.
  • the temporal encoder consists of alternately arranged fully connected layers and one-dimensional convolutional layers, which respectively extract the reaction temperature value and the value of the cooling water through one-dimensional convolutional coding. Correlation of flow rate values in the time series dimension and extraction of high-dimensional hidden features of the reaction temperature value and the flow rate value of the cooling water through fully connected coding
  • the reaction data encoding module includes: an input vector construction unit for converting the reaction temperature values and cooling water flow rate values at multiple predetermined time points within the predetermined time period, respectively.
  • the temperature input vector and the flow rate input vector are arranged according to the time dimension; a fully connected encoding unit is used to use the fully connected layer of the first temporal encoder to fully perform the following formula on the temperature input vector and the flow rate input vector.
  • the codes are connected to respectively extract the high-dimensional hidden features of the eigenvalues of each position in the temperature input vector and the flow velocity input vector, where the formula is: where X is the input vector, Y is the output vector, W is the weight matrix, and B is the bias vector, represents a matrix multiplication; a one-dimensional convolution coding unit, used to perform one-dimensional convolution coding on the temperature input vector and the flow rate input vector using the following formula using the one-dimensional convolution layer of the first temporal encoder to respectively High-dimensional implicit correlation features between the eigenvalues of each position in the temperature input vector and the flow velocity input vector are respectively extracted, where the formula is:
  • a is the width of the convolution kernel in the x direction
  • F is the convolution kernel parameter vector
  • G is the local vector matrix that operates with the convolution kernel function
  • w is the size of the convolution kernel
  • X represents the input vector.
  • the difference module 250 is used to calculate a difference feature vector between the temperature feature vector and the cooling feature vector. It should be understood that since this application uses cooling water for cooling, if the reaction temperature rises too fast, it can be adjusted by increasing the flow rate of the cooling water. Therefore, there is a gap between the reaction temperature and the flow rate of the cooling water. The opposite characteristic relationship, therefore, in the technical solution of the present application, a differential characteristic vector between the temperature characteristic vector and the cooling characteristic vector is further calculated.
  • the differential module is further configured to: calculate the differential eigenvector between the temperature eigenvector and the cooling eigenvector using the following formula; wherein, the formula is :
  • V t represents the temperature eigenvector
  • V c represents the cooling eigenvector
  • V d represents the differential eigenvector
  • the state data encoding module 260 is used to encode the pH values of the reaction liquid at multiple predetermined time points within the predetermined time period through a second temporal encoding including a one-dimensional convolution layer. device to obtain the PH timing feature vector.
  • a second temporal encoding including a one-dimensional convolution layer. device to obtain the PH timing feature vector.
  • the pH values of the reaction liquid at multiple predetermined time points within the predetermined time period also have implicit dynamic correlation characteristics in the time series dimension. Therefore, similarly, similarly, the predetermined time period is The pH values of the reaction solution at multiple predetermined time points are passed through a second temporal encoder including a one-dimensional convolution layer to obtain a pH temporal feature vector.
  • the state data encoding module is further configured to: arrange the pH values of the reaction liquid at multiple predetermined time points within the predetermined time period into a one-dimensional input vector according to the time dimension. ; Use the fully connected layer of the second temporal encoder to perform fully connected encoding on the input vector with the following formula to extract high-dimensional implicit features of the eigenvalues of each position in the input vector, where the formula is: where X is the input vector, Y is the output vector, W is the weight matrix, and B is the bias vector, represents matrix multiplication; use the one-dimensional convolution layer of the second temporal encoder to perform one-dimensional convolution encoding on the input vector with the following formula to extract the high-dimensional implicit correlation between the feature values of each position in the input vector Characteristics, where the formula is:
  • a is the width of the convolution kernel in the x direction
  • F is the convolution kernel parameter vector
  • G is the local vector matrix that operates with the convolution kernel function
  • w is the size of the convolution kernel
  • X represents the input vector.
  • the fusion module 270, the responsiveness estimation module 280 and the production management control result generation module 290 are used to perform a fusion between the differential feature vector and the flow rate control feature matrix. Multiply to obtain a reaction feature vector, and calculate the responsiveness estimate of the PH time series feature vector relative to the reaction feature vector to obtain a classification feature vector, and then pass the classification feature vector through a classifier to obtain a classification result.
  • the classification result is used to indicate that the flow rate value of the cooling water at the current point in time should be increased or decreased. That is, in the technical solution of the present application, further, the differential feature vector and the flow rate control feature matrix can be multiplied to fuse the two feature information to obtain a reaction feature vector.
  • the characteristic scales of the dynamic change characteristics of the pH value of the reaction liquid and the implicit dynamic characteristics associated with the parameters are different, and the dynamic characteristics of the pH value of the reaction liquid are in a high-dimensional feature space. It can be regarded as a responsive feature to the change of the parameter correlation. Therefore, in order to better integrate the PH timing feature vector and the reaction feature vector, the relationship between the PH timing feature vector and the reaction feature vector is further calculated. Responsiveness is estimated to obtain categorical feature vectors. In this way, the classification feature vector can be passed through a classifier to obtain a classification result indicating that the flow rate value of the cooling water at the current time point should be increased or decreased.
  • the classifier is used to process the classification feature vector with the following formula to obtain the classification result, wherein the formula is: softmax ⁇ (W n ,B n ):... :(W 1 ,B 1 )
  • the responsiveness estimation module is further configured to: calculate the responsiveness estimate of the PH timing feature vector relative to the reaction feature vector using the following formula to obtain the classification feature vector ;wherein, the formula is
  • s 1 represents the PH time series feature vector
  • s 2 represents the reaction feature vector
  • s 3 represents the classification feature vector
  • ⁇ represents the vector dot product
  • ⁇ -1 represents the reciprocal of the value of each position of the vector.
  • the production management control system 200 for ammonium fluoride production based on the embodiment of the present application has been clarified, which uses artificial intelligence control technology to add flow rates of liquid ammonia at multiple predetermined time points within a predetermined time period.
  • value, the addition flow rate value of anhydrous hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the PH value of the reaction liquid are used as input data, and a deep neural network model is used as a feature extractor to add liquid ammonia and anhydrous hydrogen fluoride to the reaction tank.
  • the real-time dynamic characteristics of the flow rate and the change characteristics of the reaction temperature are coordinated to carry out dynamic intelligent control of the flow rate of the cooling water.
  • the change characteristic information of the pH value of the reaction liquid is also added as the basis for the final result to improve the reaction efficiency. and product quality.
  • the production management control system 200 for ammonium fluoride production according to the embodiment of the present application can be implemented in various terminal devices, such as a server for the production management control algorithm of ammonium fluoride production, etc.
  • the production management control system 200 for ammonium fluoride production according to the embodiment of the present application can be integrated into the terminal device as a software module and/or a hardware module.
  • the production management control system 200 for ammonium fluoride production can be a software module in the operating system of the terminal equipment, or can be an application program developed for the terminal equipment; of course, the production management control system 200 for ammonium fluoride production can be a software module in the operating system of the terminal equipment.
  • the production management control system 200 for ammonium chemical production can also be one of the many hardware modules of the terminal equipment.
  • the production management control system 200 for ammonium fluoride production and the terminal equipment may also be separate devices, and the production management control system 200 for ammonium fluoride production may be connected via a wired And/or a wireless network is connected to the terminal device, and the interactive information is transmitted according to the agreed data format.
  • Figure 4 illustrates a flow chart of a control method of a production management control system for ammonium fluoride production.
  • the control method of the production management control system for ammonium fluoride production includes the steps: S110, obtaining the adding flow rate value of liquid ammonia at multiple predetermined time points within a predetermined time period, The addition flow rate value of anhydrous hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the pH value of the reaction liquid; S120, add the flow rate value of liquid ammonia and the addition flow rate value of anhydrous hydrogen fluoride at multiple predetermined time points within the predetermined time period.
  • the flow rate control correlation matrix is passed through the first convolutional neural network using mutually transposed convolution kernels in adjacent layers to obtain the flow rate control feature matrix; S140, combine the reaction temperature values of multiple predetermined time points within the predetermined time period.
  • cooling water flow rate values are arranged as input vectors according to the time dimension and then passed through the first temporal encoder including a one-dimensional convolution layer to obtain the temperature feature vector and the cooling feature vector; S150, calculate the temperature feature vector and the cooling Difference eigenvectors between eigenvectors; S160, pass the PH values of the reaction liquid at multiple predetermined time points within the predetermined time period through the second temporal encoder including a one-dimensional convolution layer to obtain the PH temporal eigenvector; S170 , multiply the differential feature vector and the flow rate control feature matrix to obtain a reaction feature vector; S180, calculate the responsiveness estimate of the PH timing feature vector relative to the reaction feature vector to obtain a classification feature vector; and , S190. Pass the classification feature vector through a classifier to obtain a classification result. The classification result is used to indicate that the flow rate value of the cooling water at the current point in time should be increased or decreased.
  • Figure 5 illustrates an architectural schematic diagram of a control method of a production management control system for ammonium fluoride production according to an embodiment of the present application.
  • the network architecture of the control method of the production management control system for ammonium fluoride production first, the addition flow rate of liquid ammonia at multiple predetermined time points within the predetermined time period is obtained.
  • the value (for example, P1 as shown in Figure 5) and the addition flow rate value of anhydrous hydrogen fluoride (for example, P2 as shown in Figure 5) are respectively arranged according to the time dimension as the first addition flow rate vector (for example, as shown in Figure 5
  • V1 as shown in Figure 5 After V1 as shown in Figure 5) and the second added flow rate vector (for example, V2 as shown in Figure 5), calculate the flow rate control correlation matrix between the first added flow rate vector and the second added flow rate vector ( For example, M1 as shown in Figure 5); then, the flow rate control correlation matrix is passed through the first convolutional neural network using mutually transposed convolution kernels in adjacent layers (for example, as shown in Figure 5 CNN1) to obtain the flow rate control feature matrix (for example, MF1 as shown in Figure 5); then, the reaction temperature values obtained at multiple predetermined time points within the predetermined time period (for example, as shown in Figure 5
  • the schematic P3) and the flow rate value of the cooling water (for example, the schematic P4 in Figure 5) are respectively
  • step S110 and step S120 the addition flow rate value of liquid ammonia, the addition flow rate value of anhydrous hydrogen fluoride, the reaction temperature value, the flow rate value of cooling water and the reaction liquid at multiple predetermined time points within the predetermined time period are obtained PH value, and arrange the adding flow rate values of liquid ammonia and anhydrous hydrogen fluoride at multiple predetermined time points within the predetermined time period according to the time dimension into the first adding flow rate vector and the second adding flow rate vector. , calculate the flow rate control correlation matrix between the first added flow rate vector and the second added flow rate vector.
  • the addition flow rate values of liquid ammonia and anhydrous value at multiple predetermined time points within a predetermined time period are obtained through various sensors, such as a flow rate sensor, a temperature sensor and a pH value sensor.
  • the addition flow rate of hydrogen fluoride, the reaction temperature, the flow rate of cooling water and the pH value of the reaction solution are obtained through various sensors, such as a flow rate sensor, a temperature sensor and a pH value sensor.
  • the addition flow rate of liquid ammonia and the addition flow rate of anhydrous hydrogen fluoride at the plurality of predetermined time points there is some hidden value between the addition flow rate of liquid ammonia and the addition flow rate of anhydrous hydrogen fluoride.
  • the adding flow rate values of liquid ammonia and the adding flow rate value of anhydrous hydrogen fluoride at multiple predetermined time points within the predetermined time period are arranged as the first adding flow rate according to the time dimension.
  • vector and the second addition flow rate vector to integrate the addition flow rate value of liquid ammonia and the addition flow rate value information of anhydrous hydrogen fluoride at each time point, and further calculate the first addition flow rate vector and the second addition flow rate vector. For example, the product between the transpose vector of the first added flow velocity vector and the second added flow velocity vector is calculated to obtain the flow rate control correlation matrix.
  • the flow rate control correlation matrix is passed through the first convolutional neural network using mutually transposed convolution kernels in adjacent layers to obtain the flow rate control feature matrix. That is, in the technical solution of the present application, the flow rate control correlation matrix having the correlation information of the addition flow rate value of liquid ammonia and the addition flow rate value of anhydrous hydrogen fluoride at each time point can be passed through the adjacent layer Feature extraction is performed in the first convolutional neural network using mutually transposed convolution kernels to obtain a flow rate control feature matrix. It should be understood that processing using a convolutional neural network model in which adjacent convolutional layers are convolution kernels that are transposed to each other can simultaneously update network parameters and search for network parameter structures suitable for specific data structures during training. In this way, correlation features of the addition flow rate of anhydrous hydrogen fluoride that are more suitable for expressing the addition flow rate value of liquid ammonia and the addition flow rate of anhydrous hydrogen fluoride can be extracted, thereby improving the accuracy of subsequent classification.
  • step S140 the reaction temperature values and cooling water flow rate values at multiple predetermined time points within the predetermined time period are respectively arranged into input vectors according to the time dimension and then passed through the first step including a one-dimensional convolution layer.
  • Timing encoder to obtain temperature feature vector and cooling feature vector.
  • reaction temperature values and cooling water flow rate values at multiple predetermined time points are encoded to obtain a temperature feature vector and a cooling feature vector.
  • the temporal encoder consists of alternately arranged fully connected layers and one-dimensional convolutional layers, which respectively extract the reaction temperature value and the value of the cooling water through one-dimensional convolutional coding. Correlation of flow rate values in the time series dimension and extraction of high-dimensional hidden features of the reaction temperature value and the flow rate value of the cooling water through fully connected coding
  • step S150 a differential feature vector between the temperature feature vector and the cooling feature vector is calculated. It should be understood that since this application uses cooling water for cooling, if the reaction temperature rises too fast, it can be adjusted by increasing the flow rate of the cooling water. Therefore, there is a gap between the reaction temperature and the flow rate of the cooling water. The opposite characteristic relationship, therefore, in the technical solution of the present application, a differential characteristic vector between the temperature characteristic vector and the cooling characteristic vector is further calculated.
  • step S160 the PH values of the reaction liquid at multiple predetermined time points within the predetermined time period are passed through a second temporal encoder including a one-dimensional convolution layer to obtain a PH temporal feature vector.
  • a second temporal encoder including a one-dimensional convolution layer to obtain a PH temporal feature vector.
  • the differential feature vector is multiplied by the flow rate control feature matrix to obtain a reaction feature vector, and the PH timing feature vector relative to the reaction is calculated.
  • the responsiveness of the feature vector is estimated to obtain the classification feature vector, and then the classification feature vector is passed through the classifier to obtain the classification result.
  • the classification result is used to indicate that the flow rate value of the cooling water at the current time point should be increased or decreased. . That is, in the technical solution of the present application, further, the differential feature vector and the flow rate control feature matrix can be multiplied to fuse the two feature information to obtain a reaction feature vector.
  • the characteristic scales of the dynamic change characteristics of the pH value of the reaction liquid and the implicit dynamic characteristics associated with the parameters are different, and the dynamic characteristics of the pH value of the reaction liquid are in a high-dimensional feature space. It can be regarded as a responsive feature to the change of the parameter correlation. Therefore, in order to better integrate the PH timing feature vector and the reaction feature vector, the relationship between the PH timing feature vector and the reaction feature vector is further calculated. Responsiveness is estimated to obtain categorical feature vectors. In this way, the classification feature vector can be passed through a classifier to obtain a classification result indicating that the flow rate value of the cooling water at the current time point should be increased or decreased.
  • control method of the production management control system for ammonium fluoride production based on the embodiments of the present application is clarified, which uses artificial intelligence control technology to control liquid ammonia at multiple predetermined time points within a predetermined time period.
  • the real-time dynamic characteristics of the flow rate in the reaction tank and the change characteristics of the reaction temperature are coordinated to carry out dynamic intelligent control of the flow rate of the cooling water.
  • the change characteristics of the pH value of the reaction liquid are also added as the basis for the final results to improve reaction efficiency and product quality.
  • each component or each step can be decomposed and/or recombined. These decompositions and/or recombinations shall be considered equivalent versions of this application.

Landscapes

  • Engineering & Computer Science (AREA)
  • Manufacturing & Machinery (AREA)
  • General Engineering & Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Feedback Control In General (AREA)
  • Treating Waste Gases (AREA)

Abstract

一种用于氟化铵生产的生产管理控制系统及其控制方法，采用人工智能控制技术，通过对于预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的pH值作为输入数据，使用深度神经网络模型作为特征提取器，根据液氨与无水氟化氢加入反应槽中的流速实时动态特征与反应温度的变化特征的协同来进行冷却水的流速动态智能控制，并在此过程中还加入了反应液的pH值变化特征信息作为最终的结果依据，以提高反应效率和产品质量。

Description

用于氟化铵生产的生产管理控制系统及其控制方法 技术领域
本发明涉及智能生产管理的领域，且更为具体地，涉及一种用于氟化铵生产的生产管理控制系统及其控制方法。
背景技术
氟化铵，分子式为NH4F，相对分子质量37.04，相对密度为1.015(25℃)，无色叶状或针状结晶，升华后为六角柱状晶体；易潮解易结块，可溶于冷水，微溶于醇，不溶于丙酮和液氨。受热或遇热水即分解失去氨转化成更稳定的氟化铵。氟化铵用途广泛，如作为玻璃刻蚀剂、金属表面的化学抛光剂、木材及酿酒防腐剂、消毒剂、纤维的媒染剂及提取稀有元素的溶剂等，还可作为化学分析中离子检测的掩蔽剂、酿酒的消毒剂、防腐剂、纤维的媒染剂等。
传统的氟化铵生产方法为液相法：在铅制或塑料容器中，投入定量氢氟酸。在容器外用水冷却，在搅拌下缓慢通入氨气，直至反应液PH值达4左右为止。反应液经冷却结晶、离心分离、气流干燥，制得氟化铵产品。传统液相法生产的氟化铵存在产品含水量高、易结块、不能长期储存等缺点。因此，期待一种优化的用于氟化铵的制备方案。
发明内容
为了解决上述技术问题，提出了本申请。本申请的实施例提供了一种用于氟化铵生产的生产管理控制系统及其控制方法，其采用人工智能控制技术，通过对于预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值作为输入数据，使用深度神经网络模型作为特征提取器，以根据液氨与无水氟化氢加入反应槽中的流速实时动态特征与反应温度的变化特征的协同来进行冷却水的流速动态智能控制，并在此过程中还加入了反应液的PH值变化特征信息作为最终的结果依据，以提高反应效率和产品质量。
根据本申请的一个方面，提供了一种用于氟化铵生产的生产管理控制系统，其包括：
生产参数采集模块，用于获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值；加入速度结构化关联模块，用于将所述预定时间段内多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值分别按照时间维度排列为第一加入流速向量和第二加入流速向量后，计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵；加入速度特征提取模块，用于将所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络以得到流速控制特征矩阵；
反应数据编码模块，用于将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为输入向量后通过包含一维卷积层的第一时序编码器以得到温度特征向量和冷却特征向量；差分模块，用于计算所述温度特征向量和所述冷却特征向量之间的差分特征向量；状态数据编码模块，用于将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二时序编码器以得到PH时序特征向量；融合模块，用于将所述差分特征向量与所述流速控制特征矩阵进行相乘以得到反应特征向量；响应性估计模块，用于计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量；以及
生产管理控制结果生成模块，用于将所述分类特征向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却水的流速值应增大或应减小。
在上述用于氟化铵生产的生产管理控制系统中，所述加入速度结构化关联模块，包括：向量关联单元，用于以如下公式来计算所述第一加入流速向量和所述第二加入流速向量之间的所述流速控制关联矩阵；其中，所述公式为：
Figure PCTCN2022121231-appb-000001
其中V 1表示所述第一加入流速向量，
Figure PCTCN2022121231-appb-000002
表示所述第一加入流速向量的转置向量，V 2表示所述第二加入流速向量，M表示所述流速控制关联矩阵，
Figure PCTCN2022121231-appb-000003
表示向量相乘。
在上述用于氟化铵生产的生产管理控制系统中，所述加入速度特征提取模块，包括：矩阵校正单元，用于基于所述流速控制关联矩阵中各个位置的位置信息，对所述流速控制关联矩阵进行修正以得到校正后流速控制关联矩阵；卷积编码单元，用于将所述校正后流速控制关联矩阵通过所述相邻层使用互为转置的卷积核的第一卷积神经网络以得到所述流速控制特征矩阵。
在上述用于氟化铵生产的生产管理控制系统中，所述矩阵校正单元，进一步用于：基于所述流速控制关联矩阵中各个位置的位置信息，以如下公式对所述流速控制关联矩阵进行修正以得到所述校正后流速控制关联矩阵；其中，所述公式为：
Figure PCTCN2022121231-appb-000004
其中M表示所述流速控制关联矩阵,M'表示所述校正后流速控制关联矩阵，Cov 1()和Cov 2()均为单个卷积层，
Figure PCTCN2022121231-appb-000005
用于将二维位置坐标映射为一维数值，P M表示矩阵M的(x,y)坐标 矩阵，
Figure PCTCN2022121231-appb-000006
表示特征矩阵的按位置加法，⊙表示按位置点乘。
在上述用于氟化铵生产的生产管理控制系统中，所述卷积编码单元，包括：浅层特征图提取子单元，用于从所述第一卷积神经网络的第M层提取浅层特征矩阵，M是偶数；深层特征图提取子单元，用于从所述第一卷积神经网络的第N层提取深层特征矩阵，其中，N为偶数，且N大于M的2倍；以及，特征图融合子单元，用于融合所述浅层特征图和所述深层特征图以生成所述流速控制特征矩阵。
在上述用于氟化铵生产的生产管理控制系统中，所述反应数据编码模块，包括：输入向量构造单元，用于将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为温度输入向量和流速输入向量；全连接编码单元，用于使用所述第一时序编码器的全连接层以如下公式分别对所述温度输入向量和所述流速输入向量进行全连接编码以分别提取出所述温度输入向量和所述流速输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000007
其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
Figure PCTCN2022121231-appb-000008
表示矩阵乘；一维卷积编码单元，用于使用所述第一时序编码器的一维卷积层以如下公式分别对所述温度输入向量和所述流速输入向量进行一维卷积编码以分别提取出所述温度输入向量和所述流速输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000009
其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸，X表示所述输入向量。
在上述用于氟化铵生产的生产管理控制系统中，所述差分模块，进一步用于：以如下公式计算所述温度特征向量和所述冷却特征向量之间的所述差分特征向量；其中，所述公式为：
Figure PCTCN2022121231-appb-000010
其中V t表示所述温度特征向量，V c表示所述冷却特征向量，V d表示所述差分特征向量，
Figure PCTCN2022121231-appb-000011
表示特征向量的按位置减法。
在上述用于氟化铵生产的生产管理控制系统中，所述状态数据编码模块，进一步用于：将所述预定时间段内多个预定时间点的反应液的PH值按照时间维度排列为一维的输入向量；使用所述第二时序编码器的全连接层以如下公式对所述输入向量进行全连接编码以提取出所述输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000012
其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
Figure PCTCN2022121231-appb-000013
表示矩阵乘；使用所述第二时序编码器的一维卷积层以如下公式对所述输入向量进行一维卷积编码以提取出所述输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000014
其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸，X表示所述输入向量。
在上述用于氟化铵生产的生产管理控制系统中，所述响应性估计模块，进一步用于：以如下公式计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到所述分类特征向量；其中，所述公式为
s 3＝s 2⊙s 1 ⊙-1
其中s 1表示所述PH时序特征向量，s 2表示所述反应特征向量，s 3表示所述分类特征向量，⊙表示向量点乘，⊙-1表示对向量的每个位置的值取倒数。
在上述用于氟化铵生产的生产管理控制系统中，所述生产管理控制结果生成模块，进一步用于：使用所述分类器以如下公式对所述分类特征向量进行处理以获得所述分类结果，其中，所述公式为：softmax{(W n,B n):…:(W 1,B 1)|X}，其中，W 1到W n为权重矩阵，B 1到B n为偏置向量，X为所述分类特征向量。
根据本申请的另一方面，一种用于氟化铵生产的生产管理控制系统的控制方法，其包括：获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值；将所述预定时间段内多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值分别按照时间维度排列为第一加入流速向量和第二加入流速向量后，计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵；将所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络以得到流速控制特征矩阵；将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为输入向量后通过包含一维卷积层的第一时序编码器以得到温度特征向量和冷却特征向量；计算所述温度特征向量和所述冷却特征向量之间的差分特征向量；将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二 时序编码器以得到PH时序特征向量；将所述差分特征向量与所述流速控制特征矩阵进行相乘以得到反应特征向量；计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量；以及
将所述分类特征向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却水的流速值应增大或应减小。
在上述用于氟化铵生产的生产管理控制系统的控制方法中，计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵，包括：以如下公式来计算所述第一加入流速向量和所述第二加入流速向量之间的所述流速控制关联矩阵；其中，所述公式为：
Figure PCTCN2022121231-appb-000015
其中V 1表示所述第一加入流速向量，
Figure PCTCN2022121231-appb-000016
表示所述第一加入流速向量的转置向量，V 2表示所述第二加入流速向量，M表示所述流速控制关联矩阵，
Figure PCTCN2022121231-appb-000017
表示向量相乘。
在上述用于氟化铵生产的生产管理控制系统的控制方法中，将所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络以得到流速控制特征矩阵，包括：基于所述流速控制关联矩阵中各个位置的位置信息，对所述流速控制关联矩阵进行修正以得到校正后流速控制关联矩阵；将所述校正后流速控制关联矩阵通过所述相邻层使用互为转置的卷积核的第一卷积神经网络以得到所述流速控制特征矩阵。
在上述用于氟化铵生产的生产管理控制系统的控制方法中，基于所述流速控制关联矩阵中各个位置的位置信息，对所述流速控制关联矩阵进行修正以得到校正后流速控制关联矩阵，包括：基于所述流速控制关联矩阵中各个位置的位置信息，以如下公式对所述流速控制关联矩阵进行修正以得到所述校正后流速控制关联矩阵；其中，所述公式为：
Figure PCTCN2022121231-appb-000018
其中M表示所述流速控制关联矩阵,M'表示所述校正后流速控制关联矩阵，Cov 1()和Cov 2()均为单个卷积层，
Figure PCTCN2022121231-appb-000019
用于将二维位置坐标映射为一维数值，P M表示矩阵M的(x,y)坐标矩阵，
Figure PCTCN2022121231-appb-000020
表示特征矩阵的按位置加法，⊙表示按位置点乘。
在上述用于氟化铵生产的生产管理控制系统的控制方法中，将所述校正后流速控制关联矩阵通过所述相邻层使用互为转置的卷积核的第一卷积神经网络以得到所述流速控制特征矩阵，包括：从所述第一卷积神经网络的第M层提取浅层特征矩阵，M是偶数；从所述第一卷积神经网络的第N层提取深层特征矩阵，其中，N为偶数，且N大于M的2倍；以及，融合所述浅层特征图和所述深层特征图以生成所述流速控制特征矩阵。
在上述用于氟化铵生产的生产管理控制系统的控制方法中，将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为输入向量后通过包含一维卷积层的第一时序编码器以得到温度特征向量和冷却特征向量，包括：将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为温度输入向量和流速输入向量；使用所述第一时序编码器的全连接层以如下公式分别对所述温度输入向量和所述流速输入向量进行全连接编码以分别提取出所述温度输入向量和所述流速输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000021
其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
Figure PCTCN2022121231-appb-000022
表示矩阵乘；使用所述第一时序编码器的一维卷积层以如下公式分别对所述温度输入向量和所述流速输入向量进行一维卷积编码以分别提取出所述温度输入向量和所述流速输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000023
其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸，X表示所述输入向量。
在上述用于氟化铵生产的生产管理控制系统的控制方法中，计算所述温度特征向量和所述冷却特征向量之间的差分特征向量，包括：以如下公式计算所述温度特征向量和所述冷却特征向量之间的所述差分特征向量；其中，所述公式为：
Figure PCTCN2022121231-appb-000024
其中V t表示所述温度特征向量，V c表示所述冷却特征向量，V d表示所述差分特征向量，
Figure PCTCN2022121231-appb-000025
表示特征向量的按位置减法。
在上述用于氟化铵生产的生产管理控制系统的控制方法中，将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二时序编码器以得到PH时序特征向量，包括：将所述预定时间段内多个预定时间点的反应液的PH值按照时间维度排列为一维的输入向量；使用所述第二时序编码器的全连接层以如下公式对所述输入向量进行全连接编码以提取出所述输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000026
其中X是所述输入向量，Y是输出 向量，W是权重矩阵，B是偏置向量，
Figure PCTCN2022121231-appb-000027
表示矩阵乘；使用所述第二时序编码器的一维卷积层以如下公式对所述输入向量进行一维卷积编码以提取出所述输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000028
其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸，X表示所述输入向量。
在上述用于氟化铵生产的生产管理控制系统的控制方法中，计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量，包括：以如下公式计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到所述分类特征向量；其中，所述公式为
s 3＝s 2⊙s 1 ⊙-1
其中s 1表示所述PH时序特征向量，s 2表示所述反应特征向量，s 3表示所述分类特征向量，⊙表示向量点乘，⊙-1表示对向量的每个位置的值取倒数。
在上述用于氟化铵生产的生产管理控制系统的控制方法中，将所述分类特征向量通过分类器以得到分类结果，包括：使用所述分类器以如下公式对所述分类特征向量进行处理以获得所述分类结果，其中，所述公式为：softmax{(W n,B n):…:(W 1,B 1)|X}，其中，W 1到W n为权重矩阵，B 1到B n为偏置向量，X为所述分类特征向量。
与现有技术相比，本申请提供的用于氟化铵生产的生产管理控制系统及其控制方法，其采用人工智能控制技术，通过对于预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值作为输入数据，使用深度神经网络模型作为特征提取器，以根据液氨与无水氟化氢加入反应槽中的流速实时动态特征与反应温度的变化特征的协同来进行冷却水的流速动态智能控制，并在此过程中还加入了反应液的PH值变化特征信息作为最终的结果依据，以提高反应效率和产品质量。
附图说明
通过结合附图对本申请实施例进行更详细的描述，本申请的上述以及其他目的、特征和优势将变得更加明显。附图用来提供对本申请实施例的进一步理解，并且构成说明书的一部分，与本申请实施例一起用于解释本申请，并不构成对本申请的限制。在附图中，相同的参考标号通常代表相同部件或步骤。
图1A为根据本申请实施例的用于氟化铵生产的生产管理控制系统的制备过程的流程图。
图1B为根据本申请实施例的用于氟化铵生产的生产管理控制系统的应用场景图。
图2为根据本申请实施例的用于氟化铵生产的生产管理控制系统的框图。
图3为根据本申请实施例的用于氟化铵生产的生产管理控制系统中加入速度特征提取模块的框图。
图4为根据本申请实施例的用于氟化铵生产的生产管理控制系统的控制方法的流程图。
图5为根据本申请实施例的用于氟化铵生产的生产管理控制系统的控制方法的架构示意图。
具体实施方式
下面，将参考附图详细地描述根据本申请的示例实施例。显然，所描述的实施例仅仅是本申请的一部分实施例，而不是本申请的全部实施例，应理解，本申请不受这里描述的示例实施例的限制。
场景概述
如前所述，氟化铵，分子式为NH4F，相对分子质量37.04，相对密度为1.015(25℃)，无色叶状或针状结晶，升华后为六角柱状晶体；易潮解易结块，可溶于冷水，微溶于醇，不溶于丙酮和液氨。受热或遇热水即分解失去氨转化成更稳定的氟化铵。氟化铵用途广泛，如作为玻璃刻蚀剂、金属表面的化学抛光剂、木材及酿酒防腐剂、消毒剂、纤维的媒染剂及提取稀有元素的溶剂等，还可作为化学分析中离子检测的掩蔽剂、酿酒的消毒剂、防腐剂、纤维的媒染剂等。
传统的氟化铵生产方法为液相法：在铅制或塑料容器中，投入定量氢氟酸。在容器外用水冷却，在搅拌下缓慢通入氨气，直至反应液PH值达4左右为止。反应液经冷却结晶、离心分离、气流干燥，制得氟化铵产品。传统液相法生产的氟化铵存在产品含水量高、易结块、不能长期储存等缺点。
因此，期待一种优化的用于氟化铵的制备方案。
如图1A所示，在一种制备方案中，其制备过程为：
步骤1：在反应槽中加入母液，然后在搅拌状态下加入液氨和无水氟化氢进行反应；步骤2：将反应液经过冷却结晶、离心分离和干燥后制得氟化铵。
所述母液为步骤2中的反应液经离心分离后得到的液体。制得的氟化铵具有产品含水量低、不易 结块、耐储存、品质高等优点。所述母液是反应液在离心分离后得到的液体，其主要成分是氟化铵和氨水。在制备初期，可预先配制好一定量的母液以启动制备流程，然后在制备过程中即可循环利用反应液离心分离后得到的液体作为母液，制备结束后，反应液离心分离得到的液体可留作下次制备所需的母液，无需再另行配制母液。
这样，通过在反应槽中加入母液，避免设备损坏和杜绝污染，因为空槽时直接加入无水氟化氢，会产生污染，损坏设备。且由于所述母液只需在制备初期配制一次，制备过程中及后续制备均可循环利用，无需另行配制，因此大大降低了生产成本，简化生产工艺。
加入母液后，在搅拌状态下向反应槽中加入液氨与无水氟化氢进行反应，由于反应生成的氟化铵容易产生分层现象，导致酸度不均匀，因此在反应过程中不断对反应槽内的反应液进行搅拌，防止不合格产品的产生及取样分析不准确。搅拌可由反应槽中自带的电动搅拌装置完成，也可另行增加搅拌装置以加强搅拌效果。
特别地，液氨与无水氟化氢按照以下加入顺序效果最佳：先加入50～60kg的液氨，再加入100～110kg的无水氟化氢，最后同时加入剩余的液氨和无水氟化氢。这是因为无水氟化氢密度大，先加入会沉入底部，导致反应不均匀，而先加入一定量的液氨再加入一定量的无水氟化氢则可以有效地避免反应不均匀的现象，在确保反应均匀的情况下，最后将剩余的液氨和无水氟化氢同时加入则有利于提高生产效率，避免生产周期过长而提高生产成本。在反应过程中，液氨与无水氟化氢的加入要缓慢，以控制反应温度在90～110℃之间为最佳，反应槽上还可设置冷却水管，辅以冷却水进行降温，如果反应温升过快，则可以采用减少进料量或开大冷却水来调节，反应压力控制在常压为宜，使反应在连续、均匀、缓慢、稳定中进行。反应终点的PH值控制在5～6为宜，具体控制方法可按照以下方式进行：在离投料终点还差5％时(以液氨计)，用PH试纸或其他PH检测装置检测反应液的PH值，然后根据检测结果对液氨和无水氟化氢的剩余加入量进行相应的调整，使反应终点的PH值控制在5～6。
基于此，本申请发明人发现在上述制备方案中，液氨与无水氟化氢加入反应槽中的流速控制与反应温度的协同对于提高反应效率和提高产品质量具有重要意义。因此，在本申请的技术方案中，期望通过液氨的加入流速值、无水氟化氢的加入流速值和冷却水的流速值以及反应温度值来综合进行反应的实时动态控制，并通过PH检测装置检测反应液的PH值来确定最终的结束时间，进而来提高生产的效率和产品的质量。
具体地，在本申请的技术方案中，首先，通过各个传感器，例如流速传感器、温度传感器和PH值传感器获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值。然后，对于所述多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值，由于所述液氨的加入流速和所述无水氟化氢的加入流速之间存在着某种隐藏的关联，因此为了深层挖掘出这种关联关系，将所述预定时间段内多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值分别按照时间维度排列为第一加入流速向量和第二加入流速向量以整合所述各个时间点的液氨的加入流速值和无水氟化氢的加入流速值信息，进一步再计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵，例如计算所述第一加入流速向量的转置向量和所述第二加入流速向量之间的乘积以得到所述流速控制关联矩阵。
这样，就可以将具有所述各个时间点的液氨的加入流速值和无水氟化氢的加入流速值的关联信息的所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络中进行特征提取，以得到流速控制特征矩阵。应可以理解，使用邻卷积层为互为转置的卷积核的卷积神经网络模型进行处理，能够在训练时能够同时更新网络参数的更新和适合特定数据结构的网络参数结构的搜索，以更提取出更适于表达所述液氨的加入流速值所述无水氟化氢的加入流速的关联特征，进而提高后续分类的准确性。
但是，由于相邻层使用互为转置的卷积核的所述第一卷积神经网络能够对所述流速控制关联矩阵中的特定局部数据结构进行适于其的网络参数结构的搜索，这也使得所述第一卷积神经网络的卷积核的参数配置更聚焦于所述流速控制关联矩阵中的局部流速关联值，使得所述流速控制特征矩阵可能在所述液氨的加入流速值和所述无水氟化氢的加入流速值的全局关联特性上表达不足。
因此，针对所述流速控制特征矩阵的每个位置的特征值对所述液氨的加入流速值和所述无水氟化氢的加入流速值的预定数值关联的关联特征表达特性，对流速控制特征矩阵进行优化，即：
Figure PCTCN2022121231-appb-000029
Figure PCTCN2022121231-appb-000030
Cov 1()和Cov 2()均为单个卷积层，
Figure PCTCN2022121231-appb-000031
用于将二维位置坐标映射为一维数值，P M表示矩阵M的(x,y)坐标矩阵。
也就是，为了全面融合所述流速控制特征矩阵内由所述第一卷积神经网络的互为转置的卷积核所 捕获的局部语义，并进一步从其衍生全局语义，使用由坐标转换函数Φ得到的位置信息作为提议，通过卷积层的局部感知场来对所述流速控制特征矩阵的全局场景语义进行推理，并且通过添加结构转置的偏置，来实现关联特征信息的结构分布的局部-全局迁移。这样，优化后的所述流速控制特征矩阵可以获得更好的所述液氨的加入流速值和所述无水氟化氢的加入流速值的特征关联特性的全局表达能力。
对于所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值，考虑到所述反应温度值和所述冷却水的流速值在时间维度上具有着隐含的特征信息，因此为了更为充分地提取出这种动态变化的隐含关联特征，使用包含一维卷积层的第一时序编码器分别对所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值进行编码，以得到温度特征向量和冷却特征向量。相应地，在一个具体示例中，所述时序编码器由交替设置的全连接层和一维卷积层组成，其通过一维卷积编码分别提取出所述反应温度值和所述冷却水的流速值在时序维度上的关联和通过全连接编码分别提取所述反应温度值和所述冷却水的流速值的高维隐含特征。
然后，考虑到由于本申请辅以冷却水进行降温，如果反应温度上升过快，可以采用增大冷却水的流速来调节，因此所述反应温度和所述冷却水的流速之间存在着相反的特征关系，因此，进一步计算所述温度特征向量和所述冷却特征向量之间的差分特征向量。
对于所述预定时间段内多个预定时间点的反应液的PH值，由于其在时序维度上也存在着隐含的动态关联特征，因此，同样地，将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二时序编码器以得到PH时序特征向量。
进一步地，就可以将所述差分特征向量与所述流速控制特征矩阵进行相乘以融合这两者的特征信息得到反应特征向量。应可以理解，由于所述反应液的PH值的动态变化特征与所述参数关联的隐含动态特征之间的特征尺度不同，并且所述反应液的PH值的动态特征在高维特征空间中可以看作是对于所述参数关联变化的响应性特征，因此为了更好地融合所述PH时序特征向量和所述反应特征向量，进一步计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量。这样，就可以将所述分类特征向量通过分类器以得到用于表示当前时间点的冷却水的流速值应增大或应减小的分类结果。
基于此，本申请提出了一种用于氟化铵生产的生产管理控制系统，其包括：生产参数采集模块，用于获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值；加入速度结构化关联模块，用于将所述预定时间段内多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值分别按照时间维度排列为第一加入流速向量和第二加入流速向量后，计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵；加入速度特征提取模块，用于将所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络以得到流速控制特征矩阵；反应数据编码模块，用于将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为输入向量后通过包含一维卷积层的第一时序编码器以得到温度特征向量和冷却特征向量；差分模块，用于计算所述温度特征向量和所述冷却特征向量之间的差分特征向量；状态数据编码模块，用于将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二时序编码器以得到PH时序特征向量；融合模块，用于将所述差分特征向量与所述流速控制特征矩阵进行相乘以得到反应特征向量；响应性估计模块，用于计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量；以及，生产管理控制结果生成模块，用于将所述分类特征向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却水的流速值应增大或应减小。
图1B图示了根据本申请实施例的用于氟化铵生产的生产管理控制系统的应用场景图。如图1B所示，在该应用场景中，首先，通过部署于反应槽(例如，如图1B中所示意的R)的各个传感器(例如，如图1B中所示意的流速传感器T1、温度传感器T2和PH值传感器T3)获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值。然后，将获取的所述预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值输入至部署有用于氟化铵生产的生产管理控制算法的服务器中(例如，如图1B中所示意的云服务器S)，其中，所述服务器能够以用于氟化铵生产的生产管理控制算法对所述预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值进行处理，以生成用于表示当前时间点的冷却水的流速值应增大或应减小的分类结果。
在介绍了本申请的基本原理之后，下面将参考附图来具体介绍本申请的各种非限制性实施例。
示例性系统
图2图示了根据本申请实施例的用于氟化铵生产的生产管理控制系统的框图。如图2所示，根据 本申请实施例的用于氟化铵生产的生产管理控制系统200，包括：生产参数采集模块210，用于获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值；加入速度结构化关联模块220，用于将所述预定时间段内多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值分别按照时间维度排列为第一加入流速向量和第二加入流速向量后，计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵；加入速度特征提取模块230，用于将所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络以得到流速控制特征矩阵；反应数据编码模块240，用于将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为输入向量后通过包含一维卷积层的第一时序编码器以得到温度特征向量和冷却特征向量；差分模块250，用于计算所述温度特征向量和所述冷却特征向量之间的差分特征向量；状态数据编码模块260，用于将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二时序编码器以得到PH时序特征向量；融合模块270，用于将所述差分特征向量与所述流速控制特征矩阵进行相乘以得到反应特征向量；响应性估计模块280，用于计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量；以及，生产管理控制结果生成模块290，用于将所述分类特征向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却水的流速值应增大或应减小。
具体地，在本申请实施例中，所述生产参数采集模块210和所述加入速度结构化关联模块220，用于获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值，并将所述预定时间段内多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值分别按照时间维度排列为第一加入流速向量和第二加入流速向量后，计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵。如前所述，由于液氨与无水氟化氢加入反应槽中的流速控制与反应温度的协同对于提高反应效率和提高产品质量具有重要意义。因此，在本申请的技术方案中，期望通过液氨的加入流速值、无水氟化氢的加入流速值和冷却水的流速值以及反应温度值来综合进行反应的实时动态控制，并通过PH检测装置检测反应液的PH值来确定最终的结束时间，进而来提高生产的效率和产品的质量。
也就是，具体地，在本申请的技术方案中，首先，通过各个传感器，例如流速传感器、温度传感器和PH值传感器获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值。然后，对于所述多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值，由于所述液氨的加入流速和所述无水氟化氢的加入流速之间存在着某种隐藏的关联，因此为了深层挖掘出这种关联关系，将所述预定时间段内多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值分别按照时间维度排列为第一加入流速向量和第二加入流速向量以整合所述各个时间点的液氨的加入流速值和无水氟化氢的加入流速值信息，进一步再计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵，例如计算所述第一加入流速向量的转置向量和所述第二加入流速向量之间的乘积以得到所述流速控制关联矩阵。
更具体地，在本申请实施例中，所述加入速度结构化关联模块，包括：向量关联单元，用于以如下公式来计算所述第一加入流速向量和所述第二加入流速向量之间的所述流速控制关联矩阵；其中，所述公式为：
Figure PCTCN2022121231-appb-000032
其中V 1表示所述第一加入流速向量，
Figure PCTCN2022121231-appb-000033
表示所述第一加入流速向量的转置向量，V 2表示所述第二加入流速向量，M表示所述流速控制关联矩阵，
Figure PCTCN2022121231-appb-000034
表示向量相乘。
具体地，在本申请实施例中，所述加入速度特征提取模块230，用于将所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络以得到流速控制特征矩阵。也就是，在本申请的技术方案中，进一步就可以将具有所述各个时间点的液氨的加入流速值和无水氟化氢的加入流速值的关联信息的所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络中进行特征提取，以得到流速控制特征矩阵。应可以理解，使用邻卷积层为互为转置的卷积核的卷积神经网络模型进行处理，能够在训练时能够同时更新网络参数的更新和适合特定数据结构的网络参数结构的搜索，以更提取出更适于表达所述液氨的加入流速值所述无水氟化氢的加入流速的关联特征，进而提高后续分类的准确性。
更具体地，在本申请实施例中，所述加入速度特征提取模块，包括：首先，基于所述流速控制关联矩阵中各个位置的位置信息，对所述流速控制关联矩阵进行修正以得到校正后流速控制关联矩阵。应可以理解，由于相邻层使用互为转置的卷积核的所述第一卷积神经网络能够对所述流速控制关联矩阵中的特定局部数据结构进行适于其的网络参数结构的搜索，这也使得所述第一卷积神经网络的卷积核的参数配置更聚焦于所述流速控制关联矩阵中的局部流速关联值，使得所述流速控制特征矩阵可能在所述液氨的加入流速值和所述无水氟化氢的加入流速值的全局关联特性上表达不足。因此，在本申请的技术方案中，针对所述流速控制特征矩阵的每个位置的特征值对所述液氨的加入流速值和所述无 水氟化氢的加入流速值的预定数值关联的关联特征表达特性，对所述流速控制特征矩阵进行优化。
相应地，在一个具体示例中，基于所述流速控制关联矩阵中各个位置的位置信息，以如下公式对所述流速控制关联矩阵进行修正以得到所述校正后流速控制关联矩阵；其中，所述公式为：
Figure PCTCN2022121231-appb-000035
Figure PCTCN2022121231-appb-000036
其中M表示所述流速控制关联矩阵,M'表示所述校正后流速控制关联矩阵，Cov 1()和Cov 2()均为单个卷积层，
Figure PCTCN2022121231-appb-000037
用于将二维位置坐标映射为一维数值，P M表示矩阵M的(x,y)坐标矩阵，
Figure PCTCN2022121231-appb-000038
表示特征矩阵的按位置加法，⊙表示按位置点乘。也就是，为了全面融合所述流速控制特征矩阵内由所述第一卷积神经网络的互为转置的卷积核所捕获的局部语义，并进一步从其衍生全局语义，使用由坐标转换函数Φ得到的位置信息作为提议，通过卷积层的局部感知场来对所述流速控制特征矩阵的全局场景语义进行推理，并且通过添加结构转置的偏置，来实现关联特征信息的结构分布的局部-全局迁移。这样，优化后的所述流速控制特征矩阵可以获得更好的所述液氨的加入流速值和所述无水氟化氢的加入流速值的特征关联特性的全局表达能力。
然后，将所述校正后流速控制关联矩阵通过所述相邻层使用互为转置的卷积核的第一卷积神经网络以得到所述流速控制特征矩阵。具体地，在本申请实施例中，所述卷积编码单元，包括：浅层特征图提取子单元，用于从所述第一卷积神经网络的第M层提取浅层特征矩阵，M是偶数；深层特征图提取子单元，用于从所述第一卷积神经网络的第N层提取深层特征矩阵，其中，N为偶数，且N大于M的2倍；以及，特征图融合子单元，用于融合所述浅层特征图和所述深层特征图以生成所述流速控制特征矩阵。
图3图示了根据本申请实施例的用于六氟磷酸锂制备的自动配料系统中加入速度特征提取模块的框图。如图3所示，所述加入速度特征提取模块230，包括：矩阵校正单元231，用于基于所述流速控制关联矩阵中各个位置的位置信息，对所述流速控制关联矩阵进行修正以得到校正后流速控制关联矩阵；卷积编码单元232，用于将所述校正后流速控制关联矩阵通过所述相邻层使用互为转置的卷积核的第一卷积神经网络以得到所述流速控制特征矩阵。
具体地，在本申请实施例中，所述反应数据编码模块240，用于将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为输入向量后通过包含一维卷积层的第一时序编码器以得到温度特征向量和冷却特征向量。应可以理解，对于所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值，考虑到所述反应温度值和所述冷却水的流速值在时间维度上具有着隐含的特征信息，因此为了更为充分地提取出这种动态变化的隐含关联特征，在本申请的技术方案中，使用包含一维卷积层的第一时序编码器分别对所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值进行编码，以得到温度特征向量和冷却特征向量。相应地，在一个具体示例中，所述时序编码器由交替设置的全连接层和一维卷积层组成，其通过一维卷积编码分别提取出所述反应温度值和所述冷却水的流速值在时序维度上的关联和通过全连接编码分别提取所述反应温度值和所述冷却水的流速值的高维隐含特征
更具体地，在本申请实施例中，所述反应数据编码模块，包括：输入向量构造单元，用于将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为温度输入向量和流速输入向量；全连接编码单元，用于使用所述第一时序编码器的全连接层以如下公式分别对所述温度输入向量和所述流速输入向量进行全连接编码以分别提取出所述温度输入向量和所述流速输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000039
其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
Figure PCTCN2022121231-appb-000040
表示矩阵乘；一维卷积编码单元，用于使用所述第一时序编码器的一维卷积层以如下公式分别对所述温度输入向量和所述流速输入向量进行一维卷积编码以分别提取出所述温度输入向量和所述流速输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000041
其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸，X表示所述输入向量。
具体地，在本申请实施例中，所述差分模块250，用于计算所述温度特征向量和所述冷却特征向量之间的差分特征向量。应可以理解，考虑到由于本申请辅以冷却水进行降温，如果反应温度上升过快，可以采用增大冷却水的流速来调节，因此所述反应温度和所述冷却水的流速之间存在着相反的特征关系，因此，在本申请的技术方案中，进一步计算所述温度特征向量和所述冷却特征向量之间的差分特征向量。
更具体地，在本申请实施例中，所述差分模块，进一步用于：以如下公式计算所述温度特征向量和所述冷却特征向量之间的所述差分特征向量；其中，所述公式为：
Figure PCTCN2022121231-appb-000042
其中V t表示所述温度特征向量，V c表示所述冷却特征向量，V d表示所述差分特征向量，
Figure PCTCN2022121231-appb-000043
表示特征向量的按位置减法。
具体地，在本申请实施例中，所述状态数据编码模块260，用于将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二时序编码器以得到PH时序特征向量。应可以理解，对于所述预定时间段内多个预定时间点的反应液的PH值，由于其在时序维度上也存在着隐含的动态关联特征，因此，同样地，将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二时序编码器以得到PH时序特征向量。
更具体地，在本申请实施例中，所述状态数据编码模块，进一步用于：将所述预定时间段内多个预定时间点的反应液的PH值按照时间维度排列为一维的输入向量；使用所述第二时序编码器的全连接层以如下公式对所述输入向量进行全连接编码以提取出所述输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000044
其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
Figure PCTCN2022121231-appb-000045
表示矩阵乘；使用所述第二时序编码器的一维卷积层以如下公式对所述输入向量进行一维卷积编码以提取出所述输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
Figure PCTCN2022121231-appb-000046
其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸，X表示所述输入向量。
具体地，在本申请实施例中，所述融合模块270、所述响应性估计模块280和所述生产管理控制结果生成模块290，用于将所述差分特征向量与所述流速控制特征矩阵进行相乘以得到反应特征向量，并计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量，再将所述分类特征向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却水的流速值应增大或应减小。也就是，在本申请的技术方案中，进一步地，就可以将所述差分特征向量与所述流速控制特征矩阵进行相乘以融合这两者的特征信息得到反应特征向量。应可以理解，由于所述反应液的PH值的动态变化特征与所述参数关联的隐含动态特征之间的特征尺度不同，并且所述反应液的PH值的动态特征在高维特征空间中可以看作是对于所述参数关联变化的响应性特征，因此为了更好地融合所述PH时序特征向量和所述反应特征向量，进一步计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量。这样，就可以将所述分类特征向量通过分类器以得到用于表示当前时间点的冷却水的流速值应增大或应减小的分类结果。
相应地，在一个具体示例中，使用所述分类器以如下公式对所述分类特征向量进行处理以获得所述分类结果，其中，所述公式为：softmax{(W n,B n):…:(W 1,B 1)|X}，其中，W 1到W n为权重矩阵，B 1到B n为偏置向量，X为所述分类特征向量。
更具体地，在本申请实施例中，所述响应性估计模块，进一步用于：以如下公式计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到所述分类特征向量；其中，所述公式为
s 3＝s 2⊙s 1 ⊙-1
其中s 1表示所述PH时序特征向量，s 2表示所述反应特征向量，s 3表示所述分类特征向量，⊙表示向量点乘，⊙-1表示对向量的每个位置的值取倒数。
综上，基于本申请实施例的所述用于氟化铵生产的生产管理控制系统200被阐明，其采用人工智能控制技术，通过对于预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值作为输入数据，使用深度神经网络模型作为特征提取器，以根据液氨与无水氟化氢加入反应槽中的流速实时动态特征与反应温度的变化特征的协同来进行冷却水的流速动态智能控制，并在此过程中还加入了反应液的PH值变化特征信息作为最终的结果依据，以提高反应效率和产品质量。
如上所述，根据本申请实施例的用于氟化铵生产的生产管理控制系统200可以实现在各种终端设备中，例如用于氟化铵生产的生产管理控制算法的服务器等。在一个示例中，根据本申请实施例的用于氟化铵生产的生产管理控制系统200可以作为一个软件模块和/或硬件模块而集成到终端设备中。例如，该用于氟化铵生产的生产管理控制系统200可以是该终端设备的操作系统中的一个软件模块，或者可以是针对于该终端设备所开发的一个应用程序；当然，该用于氟化铵生产的生产管理控制系统200同样可以是该终端设备的众多硬件模块之一。
替换地，在另一示例中，该用于氟化铵生产的生产管理控制系统200与该终端设备也可以是分 立的设备，并且该用于氟化铵生产的生产管理控制系统200可以通过有线和/或无线网络连接到该终端设备，并且按照约定的数据格式来传输交互信息。
示例性方法
图4图示了用于氟化铵生产的生产管理控制系统的控制方法的流程图。如图4所示，根据本申请实施例的用于氟化铵生产的生产管理控制系统的控制方法，包括步骤：S110，获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值；S120，将所述预定时间段内多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值分别按照时间维度排列为第一加入流速向量和第二加入流速向量后，计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵；S130，将所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络以得到流速控制特征矩阵；S140，将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为输入向量后通过包含一维卷积层的第一时序编码器以得到温度特征向量和冷却特征向量；S150，计算所述温度特征向量和所述冷却特征向量之间的差分特征向量；S160，将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二时序编码器以得到PH时序特征向量；S170，将所述差分特征向量与所述流速控制特征矩阵进行相乘以得到反应特征向量；S180，计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量；以及，S190，将所述分类特征向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却水的流速值应增大或应减小。
图5图示了根据本申请实施例的用于氟化铵生产的生产管理控制系统的控制方法的架构示意图。如图5所示，在所述用于氟化铵生产的生产管理控制系统的控制方法的网络架构中，首先，将获得的所述预定时间段内多个预定时间点的液氨的加入流速值(例如，如图5中所示意的P1)和无水氟化氢的加入流速值(例如，如图5中所示意的P2)分别按照时间维度排列为第一加入流速向量(例如，如图5中所示意的V1)和第二加入流速向量(例如，如图5中所示意的V2)后，计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵(例如，如图5中所示意的M1)；接着，将所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络(例如，如图5中所示意的CNN1)以得到流速控制特征矩阵(例如，如图5中所示意的MF1)；然后，将获得的所述预定时间段内多个预定时间点的反应温度值(例如，如图5中所示意的P3)和冷却水的流速值(例如，如图5中所示意的P4)分别按照时间维度排列为输入向量后通过包含一维卷积层的第一时序编码器(例如，如图5中所示意的E1)以得到温度特征向量(例如，如图5中所示意的VF1)和冷却特征向量(例如，如图5中所示意的VF2)；接着，计算所述温度特征向量和所述冷却特征向量之间的差分特征向量(例如，如图5中所示意的VF3)；然后，将所述预定时间段内多个预定时间点的反应液的PH值(例如，如图5中所示意的P5)通过包含一维卷积层的第二时序编码器(例如，如图5中所示意的E2)以得到PH时序特征向量(例如，如图5中所示意的VF4)；接着，将所述差分特征向量与所述流速控制特征矩阵进行相乘以得到反应特征向量(例如，如图5中所示意的VF5)；然后，计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量(例如，如图5中所示意的VF)；以及，最后，将所述分类特征向量通过分类器(例如，如图5中所示意的圈S)以得到分类结果，所述分类结果用于表示当前时间点的冷却水的流速值应增大或应减小。
更具体地，在步骤S110和步骤S120中，获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值，并将所述预定时间段内多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值分别按照时间维度排列为第一加入流速向量和第二加入流速向量后，计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵。应可以理解，由于液氨与无水氟化氢加入反应槽中的流速控制与反应温度的协同对于提高反应效率和提高产品质量具有重要意义。因此，在本申请的技术方案中，期望通过液氨的加入流速值、无水氟化氢的加入流速值和冷却水的流速值以及反应温度值来综合进行反应的实时动态控制，并通过PH检测装置检测反应液的PH值来确定最终的结束时间，进而来提高生产的效率和产品的质量。
也就是，具体地，在本申请的技术方案中，首先，通过各个传感器，例如流速传感器、温度传感器和PH值传感器获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值。然后，对于所述多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值，由于所述液氨的加入流速和所述无水氟化氢的加入流速之间存在着某种隐藏的关联，因此为了深层挖掘出这种关联关系，将所述预定时间段内多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值分别按照时间维度排列为第一加入流速向量和第二加入流速向量以整合所述各个时间点的液氨的加入流速值和无水氟化氢的加入流速值信息，进一步再计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵，例如计算所述第一加入流 速向量的转置向量和所述第二加入流速向量之间的乘积以得到所述流速控制关联矩阵。
更具体地，在步骤S130中，将所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络以得到流速控制特征矩阵。也就是，在本申请的技术方案中，进一步就可以将具有所述各个时间点的液氨的加入流速值和无水氟化氢的加入流速值的关联信息的所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络中进行特征提取，以得到流速控制特征矩阵。应可以理解，使用邻卷积层为互为转置的卷积核的卷积神经网络模型进行处理，能够在训练时能够同时更新网络参数的更新和适合特定数据结构的网络参数结构的搜索，以更提取出更适于表达所述液氨的加入流速值所述无水氟化氢的加入流速的关联特征，进而提高后续分类的准确性。
更具体地，在步骤S140中，将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为输入向量后通过包含一维卷积层的第一时序编码器以得到温度特征向量和冷却特征向量。应可以理解，对于所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值，考虑到所述反应温度值和所述冷却水的流速值在时间维度上具有着隐含的特征信息，因此为了更为充分地提取出这种动态变化的隐含关联特征，在本申请的技术方案中，使用包含一维卷积层的第一时序编码器分别对所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值进行编码，以得到温度特征向量和冷却特征向量。相应地，在一个具体示例中，所述时序编码器由交替设置的全连接层和一维卷积层组成，其通过一维卷积编码分别提取出所述反应温度值和所述冷却水的流速值在时序维度上的关联和通过全连接编码分别提取所述反应温度值和所述冷却水的流速值的高维隐含特征
更具体地，在步骤S150中，计算所述温度特征向量和所述冷却特征向量之间的差分特征向量。应可以理解，考虑到由于本申请辅以冷却水进行降温，如果反应温度上升过快，可以采用增大冷却水的流速来调节，因此所述反应温度和所述冷却水的流速之间存在着相反的特征关系，因此，在本申请的技术方案中，进一步计算所述温度特征向量和所述冷却特征向量之间的差分特征向量。
更具体地，在步骤S160中，将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二时序编码器以得到PH时序特征向量。应可以理解，对于所述预定时间段内多个预定时间点的反应液的PH值，由于其在时序维度上也存在着隐含的动态关联特征，因此，同样地，将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二时序编码器以得到PH时序特征向量。
更具体地，在步骤S170、步骤S180和步骤S190中，将所述差分特征向量与所述流速控制特征矩阵进行相乘以得到反应特征向量，并计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量，再将所述分类特征向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却水的流速值应增大或应减小。也就是，在本申请的技术方案中，进一步地，就可以将所述差分特征向量与所述流速控制特征矩阵进行相乘以融合这两者的特征信息得到反应特征向量。应可以理解，由于所述反应液的PH值的动态变化特征与所述参数关联的隐含动态特征之间的特征尺度不同，并且所述反应液的PH值的动态特征在高维特征空间中可以看作是对于所述参数关联变化的响应性特征，因此为了更好地融合所述PH时序特征向量和所述反应特征向量，进一步计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量。这样，就可以将所述分类特征向量通过分类器以得到用于表示当前时间点的冷却水的流速值应增大或应减小的分类结果。
综上，基于本申请实施例的所述用于氟化铵生产的生产管理控制系统的控制方法被阐明，其采用人工智能控制技术，通过对于预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值作为输入数据，使用深度神经网络模型作为特征提取器，以根据液氨与无水氟化氢加入反应槽中的流速实时动态特征与反应温度的变化特征的协同来进行冷却水的流速动态智能控制，并在此过程中还加入了反应液的PH值变化特征信息作为最终的结果依据，以提高反应效率和产品质量。
以上结合具体实施例描述了本申请的基本原理，但是，需要指出的是，在本申请中提及的优点、优势、效果等仅是示例而非限制，不能认为这些优点、优势、效果等是本申请的各个实施例必须具备的。另外，上述公开的具体细节仅是为了示例的作用和便于理解的作用，而非限制，上述细节并不限制本申请为必须采用上述具体的细节来实现。
本申请中涉及的器件、装置、设备、系统的方框图仅作为例示性的例子并且不意图要求或暗示必须按照方框图示出的方式进行连接、布置、配置。如本领域技术人员将认识到的，可以按任意方式连接、布置、配置这些器件、装置、设备、系统。诸如“包括”、“包含”、“具有”等等的词语是开放性词汇，指“包括但不限于”，且可与其互换使用。这里所使用的词汇“或”和“和”指词汇“和/或”，且可与其互换使用，除非上下文明确指示不是如此。这里所使用的词汇“诸如”指词组“诸如但不限于”，且可与其互换使用。
还需要指出的是，在本申请的装置、设备和方法中，各部件或各步骤是可以分解和/或重新组合 的。这些分解和/或重新组合应视为本申请的等效方案。
提供所公开的方面的以上描述以使本领域的任何技术人员能够做出或者使用本申请。对这些方面的各种修改对于本领域技术人员而言是非常显而易见的，并且在此定义的一般原理可以应用于其他方面而不脱离本申请的范围。因此，本申请不意图被限制到在此示出的方面，而是按照与在此公开的原理和新颖的特征一致的最宽范围。
为了例示和描述的目的已经给出了以上描述。此外，此描述不意图将本申请的实施例限制到在此公开的形式。尽管以上已经讨论了多个示例方面和实施例，但是本领域技术人员将认识到其某些变型、修改、改变、添加和子组合。

Claims (10)

  1. 一种用于氟化铵生产的生产管理控制系统，其特征在于，包括：生产参数采集模块，用于获取预定时间段内多个预定时间点的液氨的加入流速值、无水氟化氢的加入流速值、反应温度值、冷却水的流速值以及反应液的PH值；加入速度结构化关联模块，用于将所述预定时间段内多个预定时间点的液氨的加入流速值和无水氟化氢的加入流速值分别按照时间维度排列为第一加入流速向量和第二加入流速向量后，计算所述第一加入流速向量和所述第二加入流速向量之间的流速控制关联矩阵；加入速度特征提取模块，用于将所述流速控制关联矩阵通过相邻层使用互为转置的卷积核的第一卷积神经网络以得到流速控制特征矩阵；反应数据编码模块，用于将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为输入向量后通过包含一维卷积层的第一时序编码器以得到温度特征向量和冷却特征向量；差分模块，用于计算所述温度特征向量和所述冷却特征向量之间的差分特征向量；状态数据编码模块，用于将所述预定时间段内多个预定时间点的反应液的PH值通过包含一维卷积层的第二时序编码器以得到PH时序特征向量；融合模块，用于将所述差分特征向量与所述流速控制特征矩阵进行相乘以得到反应特征向量；响应性估计模块，用于计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到分类特征向量；以及生产管理控制结果生成模块，用于将所述分类特征向量通过分类器以得到分类结果，所述分类结果用于表示当前时间点的冷却水的流速值应增大或应减小。
  2. 根据权利要求1所述的用于氟化铵生产的生产管理控制系统，其特征在于，所述加入速度结构化关联模块，包括：向量关联单元，用于以如下公式来计算所述第一加入流速向量和所述第二加入流速向量之间的所述流速控制关联矩阵；其中，所述公式为：
    Figure PCTCN2022121231-appb-100001
    其中V 1表示所述第一加入流速向量，
    Figure PCTCN2022121231-appb-100002
    表示所述第一加入流速向量的转置向量，V 2表示所述第二加入流速向量，M表示所述流速控制关联矩阵，
    Figure PCTCN2022121231-appb-100003
    表示向量相乘。
  3. 根据权利要求2所述的用于氟化铵生产的生产管理控制系统，其特征在于，所述加入速度特征提取模块，包括：矩阵校正单元，用于基于所述流速控制关联矩阵中各个位置的位置信息，对所述流速控制关联矩阵进行修正以得到校正后流速控制关联矩阵；卷积编码单元，用于将所述校正后流速控制关联矩阵通过所述相邻层使用互为转置的卷积核的第一卷积神经网络以得到所述流速控制特征矩阵。
  4. 根据权利要求3所述的用于氟化铵生产的生产管理控制系统，其特征在于，所述矩阵校正单元，进一步用于：基于所述流速控制关联矩阵中各个位置的位置信息，以如下公式对所述流速控制关联矩阵进行修正以得到所述校正后流速控制关联矩阵；其中，所述公式为：
    Figure PCTCN2022121231-appb-100004
    Figure PCTCN2022121231-appb-100005
    其中M表示所述流速控制关联矩阵,M'表示所述校正后流速控制关联矩阵，Cov 1()和Cov 2()均为单个卷积层，
    Figure PCTCN2022121231-appb-100006
    用于将二维位置坐标映射为一维数值，P M表示矩阵M的(x,y)坐标矩阵，
    Figure PCTCN2022121231-appb-100007
    表示特征矩阵的按位置加法，⊙表示按位置点乘。
  5. 根据权利要求4所述的用于氟化铵生产的生产管理控制系统，其特征在于，所述卷积编码单元，包括：浅层特征图提取子单元，用于从所述第一卷积神经网络的第M层提取浅层特征矩阵，M是偶数；深层特征图提取子单元，用于从所述第一卷积神经网络的第N层提取深层特征矩阵，其中，N为偶数，且N大于M的2倍；以及特征图融合子单元，用于融合所述浅层特征图和所述深层特征图以生成所述流速控制特征矩阵。
  6. 根据权利要求5所述的用于氟化铵生产的生产管理控制系统，其特征在于，所述反应数据编码模块，包括：输入向量构造单元，用于将所述预定时间段内多个预定时间点的反应温度值和冷却水的流速值分别按照时间维度排列为温度输入向量和流速输入向量；全连接编码单元，用于使用所述第一时序编码器的全连接层以如下公式分别对所述温度输入向量和所述流速输入向量进行全连接编码以分别提取出所述温度输入向量和所述流速输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
    Figure PCTCN2022121231-appb-100008
    其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
    Figure PCTCN2022121231-appb-100009
    表示矩阵乘；一维卷积编码单元，用于使用所述第一时序编码器的一维卷积层以如下公式分别对所述温度输入向量和所述流速输入向量进行一维卷积编码以分别提取出所述温度输入向量和所述流速输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
    Figure PCTCN2022121231-appb-100010
    Figure PCTCN2022121231-appb-100011
    其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸，X表示所述输入向量。
  7. 根据权利要求6所述的用于氟化铵生产的生产管理控制系统，其特征在于，所述差分模块，进一步用于：以如下公式计算所述温度特征向量和所述冷却特征向量之间的所述差分特征向量；其中，所述公式为：
    Figure PCTCN2022121231-appb-100012
    其中V t表示所述温度特征向量，V c表示所述冷却特征向量，V d表示所述差分特征向量，
    Figure PCTCN2022121231-appb-100013
    表示特征向量的按位置减法。
  8. 根据权利要求7所述的用于氟化铵生产的生产管理控制系统，其特征在于，所述状态数据编 码模块，进一步用于：将所述预定时间段内多个预定时间点的反应液的PH值按照时间维度排列为一维的输入向量；使用所述第二时序编码器的全连接层以如下公式对所述输入向量进行全连接编码以提取出所述输入向量中各个位置的特征值的高维隐含特征，其中，所述公式为：
    Figure PCTCN2022121231-appb-100014
    其中X是所述输入向量，Y是输出向量，W是权重矩阵，B是偏置向量，
    Figure PCTCN2022121231-appb-100015
    表示矩阵乘；使用所述第二时序编码器的一维卷积层以如下公式对所述输入向量进行一维卷积编码以提取出所述输入向量中各个位置的特征值间的高维隐含关联特征，其中，所述公式为：
    Figure PCTCN2022121231-appb-100016
    其中，a为卷积核在x方向上的宽度、F为卷积核参数向量、G为与卷积核函数运算的局部向量矩阵，w为卷积核的尺寸，X表示所述输入向量。
  9. 根据权利要求8所述的用于氟化铵生产的生产管理控制系统，其特征在于，所述响应性估计模块，进一步用于：以如下公式计算所述PH时序特征向量相对于所述反应特征向量的响应性估计以得到所述分类特征向量；其中，所述公式为s 3＝s 2⊙s 1 ⊙-1其中s 1表示所述PH时序特征向量，s 2表示所述反应特征向量，s 3表示所述分类特征向量，⊙表示向量点乘，⊙-1表示对向量的每个位置的值取倒数。
  10. 根据权利要求9所述的用于氟化铵生产的生产管理控制系统，其特征在于，所述生产管理控制结果生成模块，进一步用于：使用所述分类器以如下公式对所述分类特征向量进行处理以获得所述分类结果，其中，所述公式为：softmax{(W n,B n):…:(W 1,B 1)|X}，其中，W 1到W n为权重矩阵，B 1到B n为偏置向量，X为所述分类特征向量。
PCT/CN2022/121231 2022-08-31 2022-09-26 用于氟化铵生产的生产管理控制系统及其控制方法 WO2024045247A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211054120.0 2022-08-31
CN202211054120.0A CN115453990B (zh) 2022-08-31 2022-08-31 用于氟化铵生产的生产管理控制系统及其控制方法

Publications (1)

Publication Number Publication Date
WO2024045247A1 true WO2024045247A1 (zh) 2024-03-07

Family

ID=84301684

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/121231 WO2024045247A1 (zh) 2022-08-31 2022-09-26 用于氟化铵生产的生产管理控制系统及其控制方法

Country Status (2)

Country Link
CN (1) CN115453990B (zh)
WO (1) WO2024045247A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117884032A (zh) * 2024-03-11 2024-04-16 山东森杰清洁科技有限公司 一种污水净化用消毒剂及其制备方法
CN118226762A (zh) * 2024-05-27 2024-06-21 浙江高腾机电制造股份有限公司 一种无缝针织内衣机的自适应控制系统及其方法

Families Citing this family (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115936682B (zh) * 2022-12-21 2023-11-07 江西有源工业废物回收处理有限公司 印制线路板的废料回收系统及其方法
LU503370B1 (de) * 2023-01-17 2024-07-22 Tomeisei Projektgesellschaft mbH Verfahren zur Ausführung bei der Herstellung von Endprodukten
CN116110507B (zh) * 2023-02-15 2024-02-06 浙江宏丰炉料有限公司 镁碳砖智能化生产方法及系统
CN116825215B (zh) * 2023-02-28 2024-04-16 福建省龙德新能源有限公司 用于六氟磷酸锂制备的流体循环反应控制系统及其方法
CN115845428B (zh) * 2023-02-28 2023-04-25 福建省龙德新能源有限公司 用于诱导六氟磷酸结晶的超声波装置及其控制方法
CN117046140B (zh) * 2023-02-28 2024-03-08 福建天甫电子材料有限公司 用于电子级氢氟酸精馏的装置
CN116059681B (zh) * 2023-03-07 2023-06-09 福建省龙德新能源有限公司 用于控制结晶成核的超声控制方法及其系统
CN116825217B (zh) * 2023-03-15 2024-05-14 福建省德旭新材料有限公司 制备高纯五氟化磷的方法
CN116101958B (zh) * 2023-03-22 2024-07-23 德玛克(嘉兴)智能装备有限公司 大桶饮用水智能化灌装方法及其系统
CN116859830B (zh) * 2023-03-27 2024-01-26 福建天甫电子材料有限公司 用于电子级氟化铵生产的生产管理控制系统
CN116048031B (zh) * 2023-03-31 2023-08-04 克拉玛依市蓝润环保科技有限责任公司 用于石油助剂生产的控制系统及其方法
CN116047987B (zh) * 2023-03-31 2023-06-16 福建天甫电子材料有限公司 用于电子级缓冲氧化物蚀刻液生产的智能控制系统
CN116605617A (zh) * 2023-05-17 2023-08-18 长兴南方水泥有限公司 一种智能化放料控制方法及其系统
CN116854111B (zh) * 2023-05-23 2024-05-14 福建省龙德新能源有限公司 电子级六氟磷酸锂的纯化方法及其系统
CN116603390B (zh) * 2023-07-05 2023-10-10 浙江浙能迈领环境科技有限公司 船舶尾气洗涤系统及其方法
CN116859848B (zh) * 2023-07-12 2024-01-26 普洛赛斯工程技术(宁波)有限公司 甲基异丁基酮生产装置的控制系统及其方法
CN117311298B (zh) * 2023-11-29 2024-02-09 江苏一家园健康科技有限公司 结合pH值控制的产品优化生产方法与系统
CN117668484B (zh) * 2023-12-05 2024-07-30 东莞市成铭胶粘剂有限公司 胶黏剂生产过程的实时监测系统及方法

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20010051128A1 (en) * 1994-01-07 2001-12-13 Joe G. Hoffman On-site generation of ultra-high-purity buffered-hf and ammonium fluoride
CN102557076A (zh) * 2010-12-08 2012-07-11 上海华谊微电子材料有限公司 一种生产电子级氟化铵水溶液的方法
CN106560446A (zh) * 2016-06-29 2017-04-12 宁夏海诚电化信息科技有限公司 一种氟化铵生产工艺
CN108595602A (zh) * 2018-04-20 2018-09-28 昆明理工大学 基于浅层模型与深度模型结合的问句文本分类方法
US20210333172A1 (en) * 2020-04-28 2021-10-28 Buckman Laboratories International, Inc. Contextual data modeling and dynamic process intervention for industrial plants
CN114191953A (zh) * 2021-12-07 2022-03-18 国网河北能源技术服务有限公司 一种基于卷积神经网络和XGBoost的烟气脱硫脱硝控制方法
CN114870416A (zh) * 2022-04-28 2022-08-09 福建德尔科技股份有限公司 用于电子级一氟甲烷制备的精馏控制系统及精馏控制方法

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103303941B (zh) * 2013-05-30 2014-12-17 福建省邵武市永飞化工有限公司 一种氟化铵制备方法
JP6873600B2 (ja) * 2016-03-04 2021-05-19 キヤノン株式会社 画像認識装置、画像認識方法及びプログラム
CN113479908B (zh) * 2021-07-15 2022-10-21 江西省东沿药业有限公司 一种氟化铵或氟化氢铵的制备方法
CN114779819A (zh) * 2022-05-23 2022-07-22 泰昌科技(杭州)有限公司 基于无人机移动式机场的高精度升降控制系统及控制方法
CN114897889B (zh) * 2022-06-27 2023-01-31 浙江旭派动力科技有限公司 电池组点焊自动化全检方法及其系统

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20010051128A1 (en) * 1994-01-07 2001-12-13 Joe G. Hoffman On-site generation of ultra-high-purity buffered-hf and ammonium fluoride
CN102557076A (zh) * 2010-12-08 2012-07-11 上海华谊微电子材料有限公司 一种生产电子级氟化铵水溶液的方法
CN106560446A (zh) * 2016-06-29 2017-04-12 宁夏海诚电化信息科技有限公司 一种氟化铵生产工艺
CN108595602A (zh) * 2018-04-20 2018-09-28 昆明理工大学 基于浅层模型与深度模型结合的问句文本分类方法
US20210333172A1 (en) * 2020-04-28 2021-10-28 Buckman Laboratories International, Inc. Contextual data modeling and dynamic process intervention for industrial plants
CN114191953A (zh) * 2021-12-07 2022-03-18 国网河北能源技术服务有限公司 一种基于卷积神经网络和XGBoost的烟气脱硫脱硝控制方法
CN114870416A (zh) * 2022-04-28 2022-08-09 福建德尔科技股份有限公司 用于电子级一氟甲烷制备的精馏控制系统及精馏控制方法

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117884032A (zh) * 2024-03-11 2024-04-16 山东森杰清洁科技有限公司 一种污水净化用消毒剂及其制备方法
CN117884032B (zh) * 2024-03-11 2024-05-28 山东森杰清洁科技有限公司 一种污水净化用消毒剂及其制备方法
CN118226762A (zh) * 2024-05-27 2024-06-21 浙江高腾机电制造股份有限公司 一种无缝针织内衣机的自适应控制系统及其方法

Also Published As

Publication number Publication date
CN115453990B (zh) 2023-02-17
CN115453990A (zh) 2022-12-09

Similar Documents

Publication Publication Date Title
WO2024045247A1 (zh) 用于氟化铵生产的生产管理控制系统及其控制方法
WO2024026993A1 (zh) 氟化铵制备用的自动配料控制系统及其控制方法
WO2024000800A1 (zh) 用于六氟磷酸锂制备的能源管理控制系统及其控制方法
CN115586755B (zh) 用于六氟磷酸锂制备的生产管理控制系统及其方法
WO2024045244A1 (zh) 用于无水氟化氢生产的能源管理控制系统及其控制方法
WO2024021258A1 (zh) 电子级氢氧化钾的智慧产线的控制系统及其控制方法
WO2024021259A1 (zh) 用于缓冲氧化物蚀刻液生产的自动配料系统及其配料方法
WO2024000798A1 (zh) 用于电子级氢氟酸制备的生产管理控制系统及其控制方法
WO2022170840A1 (zh) 基于二部图的后期融合多视图聚类机器学习方法及系统
US20190228517A1 (en) Method for separating out a defect image from a thermogram sequence based on feature extraction and multi-objective optimization
CN114870416B (zh) 用于电子级一氟甲烷制备的精馏控制系统及精馏控制方法
WO2024021254A1 (zh) 电子级三氟化氯的智能分离纯化系统
WO2023226236A1 (zh) 用于电子级氢氟酸制备的能源管理控制系统及其控制方法
CN114475350A (zh) 智能充电系统及其工作方法
WO2023226227A1 (zh) 用于电子级氢氟酸制备的自动配料系统及其配料方法
CN108563690A (zh) 一种基于面向对象聚类的协同过滤推荐方法
CN109754008B (zh) 基于矩阵分解的高维对称稀疏网络缺失信息的估计方法
CN116955675B (zh) 基于细粒度相似关系对比学习的哈希图像检索方法及网络
CN115983126A (zh) 异麦芽糖酐铁1000的智能化制备方法
CN113538535A (zh) 一种基于双分支特征交互的3d点云配准网络模型及配准方法
CN110175642A (zh) 一种基于pca降维和特征二元化的菊花相似性计算方法
CN104346456B (zh) 基于空间依赖性度量的数字图像多语义标注方法
CN117455451A (zh) 一种草甘膦水解溶剂的回收方法
CN114168782B (zh) 一种基于三元组网络的深度哈希图像检索方法
CN115860056A (zh) 一种用于混合气体浓度预测的传感器阵列神经网络方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22957066

Country of ref document: EP

Kind code of ref document: A1