Compound fertilizer high-efficiency production proportioning and granulation linkage adaptive control system
Patent Information
- Application Number
- CN202610876873.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-17
- Publication Date
- 2026-09-15
Smart Images
Figure CN122755432A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of compound fertilizer production control technology, specifically a high-efficiency compound fertilizer production batching and granulation linkage adaptive control system. Background Technology
[0002] Compound fertilizers refer to fertilizers containing at least two of the three nutrients—nitrogen, phosphorus, and potassium—in specified amounts, produced by chemical methods or physical mixing processes. As a core fertilizer in agricultural production, the uniformity of the finished granules during the granulation process is crucial for ensuring stable fertilizer efficacy and improving product grade and market adaptability.
[0003] Current compound fertilizer production often employs a segmented, independent control model for ingredient batching and granulation. In the ingredient batching stage, ingredients are added according to a fixed formula ratio, while the granulation stage relies on manual experience to adjust process parameters. A quantitative correlation model between the properties of the raw materials and the granulation effect has not been established. Different batches of ingredients exhibit natural differences due to variations in origin, moisture content, particle size, viscosity, and other physical properties. Even when produced according to the same formula, quality fluctuations are common, making it difficult to maintain stability within the high-quality range. This directly impacts fertilizer efficacy stability and product quality consistency. Traditional production control systems isolate the ingredient batching, granulation, and quality inspection stages. Quality inspection is merely a post-production sampling process, unable to provide real-time feedback of granulation quality test results to the ingredient batching stage. Furthermore, it cannot use quality inspection data to infer the ingredient combination status and dynamically optimize the existing ingredient combination formula within the processing plant. It also cannot optimize granulation deviations caused by differences in raw material characteristics, resulting in weak anti-interference capabilities during production and difficulty in achieving stable and uniform quality across different batches of finished products. Summary of the Invention
[0004] This invention aims to solve at least one of the technical problems existing in the prior art. To this end, this invention proposes an adaptive control system for the linkage between batching and granulation in the high-efficiency production of compound fertilizers, which can effectively solve the problem of insufficient uniformity in compound fertilizer products.
[0005] To achieve the above objectives, the present invention adopts the following technical solution: An adaptive control system for high-efficiency compound fertilizer production, including a data acquisition module, a granulation detection and correlation analysis module, an adaptive optimization control module, and a control execution module; The data acquisition module collects basic information on compound fertilizer ingredients in the processing plant and establishes files. It also establishes files for compound fertilizer granulation equipment in the processing plant and establishes a collaborative control database through MySQL to input compound fertilizer granulation equipment files and compound fertilizer ingredient files. The granulation detection and correlation analysis module performs quality detection on the granules produced by different compound fertilizer granulation equipment to obtain the physical and chemical properties of the granules, establishes a correlation model between physical properties and ingredient attributes, and outputs a predicted value of the combined ingredient attributes based on the physical properties of the granules. The adaptive optimization control module, based on the compound fertilizer granulation equipment files and compound fertilizer ingredient files in the collaborative control database, outputs the estimated value of the current batch number combination ingredient attributes through the physical characteristics and ingredient attribute association model according to the quality detection results of granulation during processing, and associates the content of the main components. Combined with the compound fertilizer production standards, it optimizes and matches the ingredient combinations of different batch numbers in the collaborative control database to obtain the optimal combination ingredient ratio. The control execution module delivers the optimal combination of ingredients with the optimal ratio to each granulation equipment in the processing plant for compound fertilizer processing.
[0006] Furthermore, the data acquisition module includes the following steps: Collect basic information on compound fertilizer ingredients in the current processing plant and establish a file. The basic information on compound fertilizer ingredients includes the input serial number, model, batch, storage time, and remaining quantity of the ingredients. Collect the equipment ID and model of the compound fertilizer granulation equipment in the current processing plant. Establish a collaborative control database through MySQL, create files with the compound fertilizer granulation equipment ID as the index, and group the files of compound fertilizer granulation equipment of the same model into the same folder. Synchronously enter the compound fertilizer ingredient files in the processing plant into the collaborative control database.
[0007] Furthermore, the particle formation detection and correlation analysis module includes the following steps: The quality of compound fertilizer granules produced in the processing plant is tested to obtain the physical and chemical properties of the granules. The physical properties include the median particle size, compressive strength, roundness, bulk density, angle of repose, and original moisture content of the compound fertilizer granules. The chemical properties include the content of each major component of the compound fertilizer. The quality test results of each batch of compound fertilizer granules are recorded in the collaborative control database, and the quality test results are associated with the current compound fertilizer granulation material file and granulation equipment file in the collaborative control database. Quality testing is performed on the combined ingredients corresponding to each batch of compound fertilizer granulation to obtain the combined ingredient attributes, including the fineness, moisture content, particle size distribution, and fine powder content of the combined ingredients. A dataset is established by combining the physical characteristics of the corresponding batch of compound fertilizer granulation, and a correlation model between physical characteristics and ingredient attributes is constructed. By using a model that correlates physical properties with ingredient attributes, the estimated value of combined ingredient attributes is output based on the physical properties of granulation.
[0008] Furthermore, the quality inspection of the compound fertilizer granules produced in the processing plant is carried out to obtain the physical and chemical properties of the granules. The physical properties include the median particle size, compressive strength, roundness, bulk density, angle of repose, and initial moisture content of the compound fertilizer granules. The chemical properties include the content of each major component of the compound fertilizer. The quality inspection results of each batch of compound fertilizer granules are recorded in the collaborative control database, and the quality inspection results are associated with the current compound fertilizer granulation material file and granulation equipment file in the collaborative control database. Specifically, this includes the following steps: For compound fertilizer granules produced in the processing plant, physical property vectors and chemical property vectors are obtained by testing the compound fertilizer granules of each production batch i. The physical property vector includes the median particle size, compressive strength, roundness, bulk density, angle of repose and original moisture content of the granules. The chemical property vector includes the content of each major component of the current compound fertilizer. Establish archives in the collaborative control database to record the quality test results of each batch of compound fertilizer granulation. Based on the ingredients and granulation equipment used in the batch of compound fertilizer granulation, link the ingredient archives and granulation equipment archives in the collaborative control database to the quality test archives of each batch of compound fertilizer granulation.
[0009] Furthermore, the quality testing of the combined ingredients corresponding to each batch of compound fertilizer granulation is performed to obtain the combined ingredient attributes, including the fineness, moisture content, particle size distribution, and fine powder content of the combined ingredients. A dataset is then established based on the physical characteristics of the corresponding batch of compound fertilizer granulation, and a model relating physical characteristics to ingredient attributes is constructed. This specifically includes the following steps: By using the ingredient files and granulation equipment files associated with each batch of compound fertilizer granulation in the collaborative control database, the quality of the combined ingredients during granulation processing is tested to obtain the combined ingredient attributes and obtain the ingredient attribute vector. The ingredient attribute vector includes the fineness, moisture content, particle size distribution, and fine powder content of the combined ingredients. Z-score normalization is used to normalize the parameters in the physical property vector and the ingredient attribute vector to construct the training dataset. Where N represents the total number of effective training batches, These represent the standardized physical property vector of the i-th batch and the standardized ingredient attribute vector of the i-th batch, respectively; A multi-output Gaussian process regression model is used to construct a correlation model between physical properties and ingredient attributes. Multi-output Gaussian process regression is then used to establish a nonlinear mapping relationship between granulation physical properties and combined ingredient attributes, where the k-th dimension of the ingredient attribute output... The particle formation physical property vector p is taken as input, and follows a Gaussian process distribution. ,in Let be the mean function of the k-th output. Let be the covariance kernel function of the k-th output, where the kernel function adopts the combination of radial basis function and white noise. Define the set of model hyperparameters corresponding to the k-th dimension output. Optimize the hyperparameters with the goal of maximizing the log marginal likelihood function to complete the training of the Gaussian process regression model. The kernel function employs a combination structure of radial basis functions superimposed with white noise, and its specific expression is as follows: ; in For signal variance, It is a diagonal matrix with a length scale. The length scale hyperparameter represents the physical feature of the k-th dimension output corresponding to the j-th dimension input. Represents noise variance. Represents the Kronecker function; The set of model hyperparameters corresponding to the k-th dimension ingredient attribute output is as follows: ; The model training process aims to maximize the log-marginal likelihood function and iteratively solves for the optimal hyperparameters. The expression for the log-marginal likelihood function is: ,in Let k be the column vector formed by the k-th dimension of the standardized ingredient attributes of all batches in the training set. To standardize the input feature matrix, Represents the N×N order training sample covariance matrix. The inverse matrix representing the covariance matrix. Let N be the determinant of the covariance matrix, and N be the total number of valid training batch samples.
[0010] Furthermore, the step of using a physical property-ingredient attribute correlation model to output a combined ingredient attribute estimate based on the physical properties of granulation includes the following steps: For the trained physical property and ingredient attribute association model, by inputting the standardized granulation physical property vector of the new batch, the Gaussian prediction distribution of the standardized ingredient attributes is obtained by using the trained physical property and ingredient attribute association model. The output result is restored to the actual physical dimensions by inverse Z-score normalization to obtain the combined ingredient attribute prediction vector.
[0011] Furthermore, the adaptive optimization control module includes the following steps: Based on the quality test results of different batch combinations of ingredients in the current granulation equipment of the processing plant, the estimated value of the ingredients of the current batch combination is output based on the correlation model between physical characteristics and ingredients attributes, and the content of the main components is correlated. In the collaborative control database, archives are created using batch number as an index, and the estimated values of combined ingredient attributes and the content of main components are recorded. Based on the estimated values of combined ingredient attributes, the proportion of main component content, and the remaining amount of combined ingredients, the combined ingredients of different batches in the collaborative control database are optimized and matched in conjunction with the compound fertilizer production standards to generate the mixing ratio of combined ingredients of different batches.
[0012] Furthermore, the quality inspection results of granulation for different batch combinations of ingredients in the current processing plant's granulation equipment are used to output the estimated value of the current batch combination of ingredients based on a physical property and ingredient attribute correlation model, and to correlate the content of major components. Specifically, this includes the following steps: The statistical collaborative control database contains the pelleting quality inspection results of different batch combinations of ingredients in the current processing plant's pelleting equipment. A physical property-ingredient attribute correlation model is used to output the predicted value of the current batch combination of ingredients. And associate the content of the main components in the granulation of the current batch number combination ingredients. Where m represents the number of major components, This represents the content of the p-th main ingredient in the q-th batch of ingredients.
[0013] Furthermore, the step of establishing files in the collaborative control database using batch numbers as indexes and recording the estimated values of combined ingredient attributes and the content of main components, and optimizing the combination of ingredients from different batches in the collaborative control database based on the estimated values of combined ingredient attributes, the proportion of main component content, and the remaining amount of combined ingredients, combined with the compound fertilizer production standards, to generate the mixing ratio of different batches of combined ingredients, specifically includes the following steps: Based on the upper and lower limits of compound fertilizer production standards, target values for main components and combined ingredient attributes are set. The range of the main component target is the positive and negative offset of the midpoint value of the standard interval corresponding to each target. The ideal value of the combined ingredient attribute target is the midpoint of the upper and lower limit interval, and the range is the positive and negative offset of the midpoint value of the upper and lower limit interval. Set the decision variable to the amount of ingredients used in each batch combination. , where x q Let represent the mass of the q-th batch combination of ingredients used in this production, and n represent the total number of batches of the combination ingredients. The objective function of the optimization model is set to minimize the weighted sum of squared relative deviations between the physical properties after mixing and the ideal values, that is: ; Where, λ k The value represents the weight coefficient of the k-th combined ingredient attribute in the objective function, and T represents the total amount of combined ingredients in this plan. This represents the ideal value of the attribute of the k-th ingredient combination. This represents the estimated value of the q-th batch number and the k-th combination ingredient attribute. The constraints include: the total usage of the combination ingredients for each batch number is equal to the planned total amount of the combination ingredients; the usage of the combination ingredients for each batch number does not exceed its inventory balance; the average content of each main component after mixing is within the target range of the main components; and the average value of each combination ingredient attribute after mixing is within the target range of the combination ingredient attribute. The interior point method is used to solve for the mass ratio of each batch number combination ingredient in the mixture, and the mixing ratio of generating different batch number combination ingredients is obtained as the optimal combination ingredient ratio.
[0014] Furthermore, the control execution module, based on the generated optimal combination ingredient ratio, extracts the compound fertilizer ingredient files of different ingredients in each batch of the optimal combination ingredient ratio from the collaborative control database, extracts the ingredients through the automatic ingredient batching system and combines them into a combination ingredient, and then mixes the combination ingredients according to the optimal combination ingredient ratio to obtain the optimal combination ingredient.
[0015] Compared with the prior art, the beneficial effects of the present invention are: In this invention, by analyzing the quality inspection results of the granulation process of compound fertilizer granulation equipment in the processing plant, and by constructing the attribute association between the ingredients and the granulation, the invention enables the optimized combination control of different batches of ingredients based on the granulation effect during the compound fertilizer processing, thereby improving the uniformity of compound fertilizer products under different batches of ingredients in the actual production process. In this invention, by combining and controlling different batches of ingredients, the differences in physical property fluctuations of different batches of ingredients are compensated by optimizing the combination, ensuring that the granulation quality of different batches of finished products remains stable within the optimized range. This reduces the number of compound fertilizer products that are close to the upper and lower limits of the standard, stabilizes the quality of the finished product in the central optimized zone of the standard range, improves the anti-interference ability in compound fertilizer processing and production, and enhances its functionality. In this invention, by statistically analyzing the quality inspection results of granulation equipment, the state of the combined ingredients is inferred, thereby further optimizing the combined ingredients. The optimization control of compound fertilizer production is integrated into the quality inspection steps of the processing plant, realizing the linkage optimization of compound fertilizer ingredient granulation in the production process, which meets the actual production needs of the processing plant, ensures the stable output of high-efficiency compound fertilizer, and enhances practicality. Attached Figure Description
[0016] Figure 1 This is an overall block diagram of an adaptive control system for the production of high-efficiency compound fertilizer, involving the mixing of raw materials and granulation. Figure 2 This is a flowchart of the granulation detection and correlation analysis module of the adaptive control system for high-efficiency compound fertilizer production ingredient batching and granulation according to the present invention. Figure 3This is a flowchart of the adaptive optimization control module of an adaptive control system for the production of high-efficiency compound fertilizers, which links ingredient mixing and granulation. Detailed Implementation
[0017] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0018] Example 1: like Figure 1 As shown, a compound fertilizer high-efficiency production batching and granulation linkage adaptive control system includes a data acquisition module, a granulation detection and correlation analysis module, an adaptive optimization control module, and a control execution module. The data acquisition module collects basic information on compound fertilizer ingredients in the processing plant and establishes files. It also establishes files for the compound fertilizer granulation equipment in the plant. A collaborative control database is established using MySQL, and the files for the compound fertilizer granulation equipment and compound fertilizer ingredients are entered. This includes the following steps: Collect basic information on compound fertilizer ingredients in the current processing plant and establish a file. The basic information on compound fertilizer ingredients includes the input serial number, model, batch, storage time, and remaining quantity of the ingredients. Collect the equipment ID and model of the compound fertilizer granulation equipment in the current processing plant. Establish a collaborative control database through MySQL, create files with the compound fertilizer granulation equipment ID as the index, and group the files of compound fertilizer granulation equipment of the same model into the same folder. Synchronously enter the compound fertilizer ingredient files in the processing plant into the collaborative control database.
[0019] It should be noted that the entered serial number is a unique identifier for the current ingredient, which is used to identify the ingredient when selecting ingredients in subsequent collaborative control.
[0020] Example 2: like Figure 2 As shown, the granulation detection and correlation analysis module performs quality testing on compound fertilizer granules produced by different compound fertilizer granulation equipment to obtain the physical and chemical properties of the granules, establishes a correlation model between physical properties and ingredient attributes, and outputs a predicted value of the combined ingredient attributes based on the physical properties of the granules. This includes the following steps: The quality of compound fertilizer granules produced in the processing plant is tested to obtain their physical and chemical properties. Physical properties include median particle size, compressive strength, roundness, bulk density, angle of repose, and initial moisture content. Chemical properties include the content of each major component of the compound fertilizer. The quality test results for each batch of compound fertilizer granules are recorded in a collaborative control database, and these results are then linked to the current compound fertilizer granulation batching and granulation equipment files. The specific steps include: For compound fertilizer granules produced in the processing plant, physical property vectors and chemical property vectors are obtained by testing the compound fertilizer granules of each production batch i. The physical property vector includes the median particle size, compressive strength, roundness, bulk density, angle of repose and original moisture content of the granules. The chemical property vector includes the content of each major component of the current compound fertilizer. It should be noted that compound fertilizer granules produced within a specified time window from the same batch of ingredients, the same set of process parameters, and continuous and stable production are considered as one production batch. The time window is set to 2-4 hours of continuous production to balance production efficiency and data accuracy. The constraint of this window value is that it is not less than the quasi-steady-state establishment time of the granulation section, and not greater than the maximum continuous time during which the key granulation characteristics of the production line can still be considered approximately stable without structural slow drift deviation under the feeding and thermal modes. In the granulation quality inspection process, the physical characteristics are inspected as follows: After sieving on a vibrating screen for 5-10 minutes using a standard set of sieves, the mass of each sieve layer is weighed, and the mass percentage is calculated to obtain the median particle size. A single granule >2mm is randomly selected and placed on the test platform. The maximum pressure value at the moment of crushing is recorded by a pressure sensor. The average value of 30 granules is taken as the compressive strength. Samples of granules are taken and laid flat, and high-definition images are taken with an industrial camera. The roundness of each particle is calculated by an image processing algorithm, and the average value is taken as the roundness. The closer the current roundness value is to 1, the rounder it is. The granules are dropped freely from a fixed height into a container of known volume. After being leveled, they are weighed to calculate the bulk density. The granules are then slowly flowed down through a funnel and naturally piled into a cone on a horizontal surface. The height and base radius of the cone are measured to calculate the angle of repose. A certain mass of sample is dried in an oven at 105℃ to constant weight to calculate the moisture content. When testing chemical properties, samples need to be prepared according to national standards, and then the total nitrogen should be determined by the Kjeldahl method, the available phosphorus by the quinoline phosphomolybdate gravimetric method, and the potassium by the sodium tetraphenylborate gravimetric method, or near-infrared spectroscopy can be used for rapid analysis to determine the content of each component in the compound fertilizer. The average value of parallel samples is taken as the content of each component in the current batch of granules.
[0021] Establish archives in the collaborative control database to record the quality test results of each batch of compound fertilizer granulation. Based on the ingredients and granulation equipment used in the batch of compound fertilizer granulation, link the ingredient archives and granulation equipment archives in the collaborative control database to the quality test archives of each batch of compound fertilizer granulation.
[0022] It should be noted that by linking the quality inspection batches of compound fertilizer granulation with the ingredient files and granulation equipment files, a collaborative data record of batch-ingredient-equipment-quality is formed. This facilitates the subsequent construction of an original sample set to train the association model by combining the combined ingredient test results corresponding to each batch of compound fertilizer granulation, as well as subsequent optimization of compound fertilizer processing.
[0023] Quality testing is performed on the combined ingredients corresponding to each batch of compound fertilizer granulation to obtain the combined ingredient attributes, including the fineness, moisture content, particle size distribution, and fine powder content of the combined ingredients. A dataset is established based on the physical characteristics of the corresponding batch of compound fertilizer granulation, and a correlation model between physical characteristics and ingredient attributes is constructed. The specific steps include: By using the ingredient files and granulation equipment files associated with each batch of compound fertilizer granulation in the collaborative control database, the quality of the combined ingredients during granulation processing is tested to obtain the combined ingredient attributes and obtain the ingredient attribute vector. The ingredient attribute vector includes the fineness, moisture content, particle size distribution, and fine powder content of the combined ingredients. It should be noted that the combined ingredients refer to the pre-mixed ingredients of the current compound fertilizer before granulation. Before granulation, the combined ingredients are quality tested to facilitate the establishment of a sample set. The combined ingredients are taken and sieved using a standard sieve. The mass of the material passing through the sieve is weighed and the percentage of passing through is calculated as the fineness. The moisture content of the combined ingredients is determined using a halogen lamp or infrared moisture meter. A small amount of the combined ingredients is dispersed and its full particle size distribution is measured using a laser particle size analyzer. The non-uniformity coefficient and curvature coefficient are calculated to obtain the particle size distribution. The combined ingredients are sieved using a 0.1 mm or 0.075 mm sieve. The mass of the material passing through the sieve is weighed and its percentage of the total mass is calculated to obtain the fine powder content.
[0024] Z-score normalization is used to normalize the parameters in the physical property vector and the ingredient attribute vector to construct the training dataset. Where N represents the total number of effective training batches, These represent the standardized physical property vector of the i-th batch and the standardized ingredient attribute vector of the i-th batch, respectively; It should be noted that the valid batches are the data after abnormal batches have been identified and removed from the original data using the 3sigma criterion.
[0025] A multi-output Gaussian process regression model is used to construct a correlation model between physical properties and ingredient attributes. Multi-output Gaussian process regression is then used to establish a nonlinear mapping relationship between granulation physical properties and combined ingredient attributes, where the k-th dimension of the ingredient attribute output... The particle formation physical property vector p is taken as input, and follows a Gaussian process distribution. ,in Let be the mean function of the k-th output. Let be the covariance kernel function of the k-th output, where the kernel function adopts the combination of radial basis function and white noise. Define the set of model hyperparameters corresponding to the k-th dimension output. Optimize the hyperparameters with the goal of maximizing the log marginal likelihood function to complete the training of the Gaussian process regression model. It should be noted that, Representing the k-th ingredient attribute, it is a function that takes the granulation physical characteristic vector p as input. The mean function is usually set to 0 to simplify calculations. The covariance kernel function is used to characterize the correlation between different input samples, measuring the correlation between two input points p and p'. This represents a Gaussian process.
[0026] The kernel function employs a combination structure of radial basis functions superimposed with white noise, and its specific expression is as follows: ; in For signal variance, It is a diagonal matrix with a length scale. The length scale hyperparameter represents the physical feature of the k-th dimension output corresponding to the j-th dimension input. Represents noise variance. Represents the Kronecker function; It should be noted that the Kronecker function takes a value of 1 when the input vectors p and p' are equal, and a value of 0 in other cases. The signal variance is used to characterize the overall fluctuation range of the output features. It is a diagonal matrix with a length scale. The length-scale hyperparameter represents the physical feature of the k-th output corresponding to the j-th input. It measures the decay rate of the feature's influence on the output and is used to adjust the model fitting sensitivity of the single-dimensional feature. Represents the noise variance, used to adapt to industrial testing noise and random disturbances in production.
[0027] The set of model hyperparameters corresponding to the k-th dimension ingredient attribute output is as follows: ; The model training process aims to maximize the log-marginal likelihood function and iteratively solves for the optimal hyperparameters. The expression for the log-marginal likelihood function is: ,in Let k be the column vector formed by the k-th dimension of the standardized ingredient attributes of all batches in the training set. To standardize the input feature matrix, Represents the N×N order training sample covariance matrix. The inverse matrix representing the covariance matrix. Let N be the determinant of the covariance matrix, and N be the total number of valid training batch samples.
[0028] It should be noted that, among them The input feature matrix is standardized, with each row corresponding to a standardized granulation physical property vector for a single batch. , This represents the N×N covariance matrix of the training samples. The matrix elements are calculated using a kernel function, where the elements are... In the set of model hyperparameters corresponding to the output of the k-th dimension ingredient attribute, The length scale parameters represent the six input features in the physical property vector. Represents the standard deviation of noise. This represents the standard deviation of the signal.
[0029] By using a correlation model between physical properties and ingredient attributes, the estimated value of combined ingredient attributes is output based on the physical properties of granulation, including the following steps: For the trained physical property and ingredient attribute association model, by inputting the standardized granulation physical property vector of the new batch, the Gaussian prediction distribution of the standardized ingredient attributes is obtained by using the trained physical property and ingredient attribute association model. The output result is restored to the actual physical dimensions by inverse Z-score normalization to obtain the combined ingredient attribute prediction vector.
[0030] It should be noted that the standardized granulation physical property vector of the new production batch By inputting the trained physical property and ingredient attribute association model, we can obtain the Gaussian probability prediction distribution of the corresponding standardized ingredient attributes. ,in Let be the predicted mean of the k-th dimension standardized ingredient attribute. The predicted variance of the standardized ingredient attribute in the k-th dimension is as follows: ; ; in, This represents the covariance vector between the test samples and all training samples. Each element of the vector is calculated using a kernel function. The standardized prediction results are then subjected to inverse Z-score normalization to restore the original dimensions. ,in The final predicted value represents the k-th dimension of the ingredient attribute. , Let represent the mean and standard deviation of the k-th dimension of the original ingredient attribute in the training set, respectively, to obtain the ingredient attribute prediction vector. During model training, the training dataset is divided into a training set and a test set in a ratio of 8:2. A set of candidate models is obtained by training on the training set. The root mean square error and coefficient of determination for each model are calculated on the test set. The model with the highest average coefficient of determination and the smallest standard deviation, as well as the lowest average root mean square error and the smallest standard deviation on the test set, is selected as the final model. When the cumulative production batches reach 50 batches, the system automatically triggers the model retraining process. The incremental learning mode is adopted, and the latest batch data is appended to the original training set. The forgetting factor method is used to exponentially weight the historical data and give the new samples higher weights. Then, the log marginal likelihood function is re-solved to update the hyperparameter set, thereby realizing the dynamic calibration of the model. Before the system goes live, a 3-day trial production is conducted to collect and record real physical properties and ingredient data synchronously with a fixed formula to form an initial dataset. The initial dataset is then used to complete the first offline training of the Gaussian process regression model and establish the initial mapping relationship.
[0031] Example 3: like Figure 3 As shown, the adaptive optimization control module, based on the compound fertilizer granulation equipment files and compound fertilizer ingredient files in the collaborative control database, outputs the estimated value of the current batch combination ingredient attributes through a physical characteristic and ingredient attribute association model based on the quality inspection results of granulation during processing. It also associates the content of major components and optimizes the matching of different batch combination ingredients in the collaborative control database in conjunction with compound fertilizer production standards to obtain the optimal combination ingredient ratio. This includes the following steps: Based on the quality inspection results of granulation from different batch combinations of ingredients in the current granulation equipment of the processing plant, the system outputs the estimated value of the ingredients' attributes for the current batch combination based on a correlation model between physical properties and ingredient attributes, and correlates it with the content of major components. The specific steps include: The statistical collaborative control database contains the pelleting quality inspection results of different batch combinations of ingredients in the current processing plant's pelleting equipment. A physical property-ingredient attribute correlation model is used to output the predicted value of the current batch combination of ingredients. And associate the content of the main components in the granulation of the current batch number combination ingredients. Where m represents the number of major components, This represents the content of the p-th main ingredient in the q-th batch of ingredients.
[0032] It should be noted that the batch number of the combined ingredients is the unique identifier of the combined ingredients that are mixed from multiple basic ingredients in a set ratio under different specific compound fertilizer ingredient files in the processing plant. The batch number includes the compound fertilizer ingredient file number included in the current combined ingredients and the ratio information of each basic ingredient. It needs to be recorded synchronously with the production batch when the compound fertilizer is granulated by the granulation equipment in the processing plant.
[0033] In the collaborative control database, archives are created using batch numbers as indexes, recording the estimated values of combined ingredient attributes and the content of major components. Based on the estimated values of combined ingredient attributes, the proportion of major component content, and the remaining amount of combined ingredients, the combined ingredients from different batches in the collaborative control database are optimized and matched in conjunction with compound fertilizer production standards to generate the mixing ratio of combined ingredients from different batches. The specific steps include: Based on the upper and lower limits of compound fertilizer production standards, target values for main components and combined ingredient attributes are set. The range of the main component target is the positive and negative offset of the midpoint value of the standard interval corresponding to each target. The ideal value of the combined ingredient attribute target is the midpoint of the upper and lower limit interval, and the range is the positive and negative offset of the midpoint value of the upper and lower limit interval. It should be noted that by setting an offset to narrow the upper and lower limits of the production standard, the generated mixing ratio fluctuates near the midpoint optimization zone of the standard value, thereby optimizing the overall uniformity of the compound fertilizer product. The offset is set to half the distance from the upper and lower limits to the midpoint. The upper and lower limits of the compound fertilizer production standard need to be determined according to the actual production requirements of compound fertilizer, such as the relevant industry standards set by the national compound fertilizer GB / T 15063-2020.
[0034] Set the decision variable to the amount of ingredients used in each batch combination. , where x q Let represent the mass of the q-th batch combination of ingredients used in this production, and n represent the total number of batches of the combination ingredients. The objective function of the optimization model is set to minimize the weighted sum of squared relative deviations between the physical properties after mixing and the ideal values, that is: ; Where, λ k The value represents the weight coefficient of the k-th combined ingredient attribute in the objective function, and T represents the total amount of combined ingredients in this plan. This represents the ideal value of the attribute of the k-th ingredient combination. This represents the estimated value of the q-th batch number and the k-th combination ingredient attribute. The constraints include: the total usage of the combination ingredients for each batch number is equal to the planned total amount of the combination ingredients; the usage of the combination ingredients for each batch number does not exceed its inventory balance; the average content of each main component after mixing is within the target range of the main components; and the average value of each combination ingredient attribute after mixing is within the target range of the combination ingredient attribute. The interior point method is used to solve for the mass ratio of each batch number combination ingredient in the mixture, and the mixing ratio of generating different batch number combination ingredients is obtained as the optimal combination ingredient ratio.
[0035] It should be noted that λ k The weight coefficients for the k-th combination ingredient attribute in the objective function are denoted as _____. The weight coefficients for fineness, moisture content, particle size distribution, and fine powder content are set to 0.35, 0.30, 0.20, and 0.15, respectively. Fineness directly determines the surface area and reactivity of the particles; excessive coarseness or fineness can lead to subsequent disintegration or excessive dust. Because it is greatly affected by the physical properties of the raw materials, it is given the highest weight to prioritize achieving the target in optimization. Too low a moisture content results in insufficient granulation, while too high a moisture content leads to agglomeration and clumping, which has a decisive impact on storage and transportation safety. The weight is set as the second highest. Particle size distribution mainly affects the appearance of the product and the filling rate of the packaging. Fine powder content mainly affects the dust environmental protection index at the production site. Fineness control can limit the amount of fine powder within a certain range. Therefore, in the objective function, the weights of the two are decreasing as auxiliary constraints. When there is no feasible solution, the total amount of combined ingredients T is adjusted and the solution is solved again until a feasible solution is obtained. The total amount of combined ingredients T needs to be preset according to the demand of compound fertilizer production orders in the actual production process. The inventory balance is obtained through the ingredient files in the collaborative control database.
[0036] The control and execution module delivers the optimal combination of ingredients with the best ratio to the granulation equipment in the processing plant for compound fertilizer processing. Based on the generated optimal combination ingredient ratio, the compound fertilizer ingredient files of different ingredients in each batch of the optimal combination ingredient ratio are extracted from the collaborative control database. After the ingredients are extracted and combined into a combination ingredient by the automatic batching system, the combination ingredients are mixed according to the optimal combination ingredient ratio to obtain the optimal combination ingredient.
[0037] It should be noted that the automatic batching system includes a crane grab bucket, a batching scale, and a screw conveyor. According to the instructions, the system sequentially goes to the designated storage silo or warehouse location, extracts the corresponding batching materials according to the precise quality, premixes them to obtain the combined batching materials, and after all batches of combined batching materials have been extracted and combined in the correct quantities, they are mixed by a mixer to obtain the optimal combined batching materials, and then transported to each granulation equipment for compound fertilizer production.
[0038] This invention discloses an adaptive control system for the linkage between batching and granulation in the production of high-efficiency compound fertilizer. During operation, by analyzing the granulation quality inspection results from the compound fertilizer granulation equipment in the processing plant, and establishing an attribute correlation between batching and granulation, the system optimizes the combination control of different batches of raw materials based on granulation effects during compound fertilizer processing. This improves the uniformity of compound fertilizer products from different batches during actual production. By combining and controlling different batches of raw materials, the system compensates for differences in physical characteristics across batches through optimized combinations, ensuring that the granulation quality of different batches of finished products remains stable within the optimized range. This reduces the number of compound fertilizer products approaching the upper and lower limits of the standard, stabilizing the finished product quality within the central optimized zone of the standard range. This enhances the anti-interference capability in compound fertilizer processing and production, and strengthens its functionality.
[0039] In the embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the method in this embodiment according to actual needs.
[0040] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A high-efficiency compound fertilizer production process with linked and adaptive control system for batching and granulation, characterized in that: It includes a data acquisition module, a granulation detection and correlation analysis module, an adaptive optimization control module, and a control execution module; The data acquisition module collects basic information on compound fertilizer ingredients in the processing plant and establishes files. It also establishes files for compound fertilizer granulation equipment in the processing plant and establishes a collaborative control database through MySQL to input compound fertilizer granulation equipment files and compound fertilizer ingredient files. The granulation detection and correlation analysis module performs quality detection on the granules produced by different compound fertilizer granulation equipment to obtain the physical and chemical properties of the granules, establishes a correlation model between physical properties and ingredient attributes, and outputs a predicted value of the combined ingredient attributes based on the physical properties of the granules. The adaptive optimization control module, based on the compound fertilizer granulation equipment files and compound fertilizer ingredient files in the collaborative control database, outputs the estimated value of the current batch number combination ingredient attributes through the physical characteristics and ingredient attribute association model according to the quality detection results of granulation during processing, and associates the content of the main components. Combined with the compound fertilizer production standards, it optimizes and matches the ingredient combinations of different batch numbers in the collaborative control database to obtain the optimal combination ingredient ratio. The control execution module delivers the optimal combination of ingredients with the optimal ratio to each granulation equipment in the processing plant for compound fertilizer processing.
2. The adaptive control system for the production, batching, and granulation of high-efficiency compound fertilizer as described in claim 1, characterized in that: The data acquisition module includes the following steps: Collect basic information on compound fertilizer ingredients in the current processing plant and establish a file. The basic information on compound fertilizer ingredients includes the input serial number, model, batch, storage time, and remaining quantity of the ingredients. Collect the equipment ID and model of the compound fertilizer granulation equipment in the current processing plant. Establish a collaborative control database through MySQL, create files with the compound fertilizer granulation equipment ID as the index, and group the files of compound fertilizer granulation equipment of the same model into the same folder. Synchronously enter the compound fertilizer ingredient files in the processing plant into the collaborative control database.
3. The adaptive control system for high-efficiency compound fertilizer production, ingredient mixing, and granulation as described in claim 2, is characterized in that: The granulation detection and correlation analysis module includes the following steps: The quality of compound fertilizer granules produced in the processing plant is tested to obtain the physical and chemical properties of the granules. The physical properties include the median particle size, compressive strength, roundness, bulk density, angle of repose, and original moisture content of the compound fertilizer granules. The chemical properties include the content of each major component of the compound fertilizer. The quality test results of each batch of compound fertilizer granules are recorded in the collaborative control database, and the quality test results are associated with the current compound fertilizer granulation material file and granulation equipment file in the collaborative control database. Quality testing is performed on the combined ingredients corresponding to each batch of compound fertilizer granulation to obtain the combined ingredient attributes, including the fineness, moisture content, particle size distribution, and fine powder content of the combined ingredients. A dataset is established by combining the physical characteristics of the corresponding batch of compound fertilizer granulation, and a correlation model between physical characteristics and ingredient attributes is constructed. By using a model that correlates physical properties with ingredient attributes, the estimated value of combined ingredient attributes is output based on the physical properties of granulation.
4. The adaptive control system for high-efficiency compound fertilizer production, including ingredient mixing and granulation, as described in claim 3, is characterized in that: The process involves quality testing of the compound fertilizer granules produced in the processing plant to obtain their physical and chemical properties. Physical properties include median particle size, compressive strength, roundness, bulk density, angle of repose, and initial moisture content. Chemical properties include the content of each major component of the compound fertilizer. The quality test results for each batch of compound fertilizer granules are recorded in a collaborative control database, and these results are then linked to the current compound fertilizer granulation material and granulation equipment files within the database. This process specifically includes the following steps: For compound fertilizer granules produced in the processing plant, physical property vectors and chemical property vectors are obtained by testing the compound fertilizer granules of each production batch i. The physical property vector includes the median particle size, compressive strength, roundness, bulk density, angle of repose and original moisture content of the granules. The chemical property vector includes the content of each major component of the current compound fertilizer. Establish archives in the collaborative control database to record the quality test results of each batch of compound fertilizer granulation. Based on the ingredients and granulation equipment used in the batch of compound fertilizer granulation, link the ingredient archives and granulation equipment archives in the collaborative control database to the quality test archives of each batch of compound fertilizer granulation.
5. The adaptive control system for the production, batching, and granulation of high-efficiency compound fertilizer as described in claim 4, characterized in that: The process involves quality testing of the constituent ingredients for each batch of compound fertilizer granulation to obtain the composition attributes, including the fineness, moisture content, particle size distribution, and fine powder content of the constituent ingredients. A dataset is then established based on the physical characteristics of the corresponding batch of compound fertilizer granulation, and a model relating physical characteristics to ingredient attributes is constructed. This process specifically includes the following steps: By using the ingredient files and granulation equipment files associated with each batch of compound fertilizer granulation in the collaborative control database, the quality of the combined ingredients during granulation processing is tested to obtain the combined ingredient attributes and obtain the ingredient attribute vector. The ingredient attribute vector includes the fineness, moisture content, particle size distribution, and fine powder content of the combined ingredients. Z-score normalization is used to normalize the parameters in the physical property vector and the ingredient attribute vector to construct the training dataset. Where N represents the total number of effective training batches, and represent the standardized physical property vector of the i-th batch and the standardized ingredient attribute vector of the i-th batch, respectively; A model relating physical properties to ingredient attributes is constructed using a multi-output Gaussian process regression model. The nonlinear mapping relationship between granulation physical properties and combined ingredient attributes is established using multi-output Gaussian process regression. The kernel function is a combination of radial basis functions and white noise. The set of model hyperparameters corresponding to the k-th dimension output is defined. The hyperparameters are optimized with the goal of maximizing the logarithmic marginal likelihood function, thus completing the training of the Gaussian process regression model.
6. The adaptive control system for high-efficiency compound fertilizer production, including ingredient mixing and granulation, as described in claim 5, is characterized in that: The method of using a correlation model between physical properties and ingredient attributes to output a predicted value of combined ingredient attributes based on the physical properties of granulation includes the following steps: For the trained physical property and ingredient attribute association model, by inputting the standardized granulation physical property vector of the new batch, the Gaussian prediction distribution of the standardized ingredient attributes is obtained by using the trained physical property and ingredient attribute association model. The output result is restored to the actual physical dimensions by inverse Z-score normalization to obtain the combined ingredient attribute prediction vector.
7. The adaptive control system for the production, batching, and granulation of high-efficiency compound fertilizer as described in claim 6, characterized in that: The adaptive optimization control module includes the following steps: Based on the quality test results of different batch combinations of ingredients in the current granulation equipment of the processing plant, the estimated value of the ingredients of the current batch combination is output based on the correlation model between physical characteristics and ingredients attributes, and the content of the main components is correlated. In the collaborative control database, archives are created using batch number as an index, and the estimated values of combined ingredient attributes and the content of main components are recorded. Based on the estimated values of combined ingredient attributes, the proportion of main component content, and the remaining amount of combined ingredients, the combined ingredients of different batches in the collaborative control database are optimized and matched in conjunction with the compound fertilizer production standards to generate the mixing ratio of combined ingredients of different batches.
8. The adaptive control system for high-efficiency compound fertilizer production, ingredient mixing, and granulation as described in claim 7, characterized in that: The quality inspection results of granulation for different batch combinations of ingredients in the current processing plant's granulation equipment are used to output the estimated value of the current batch combination of ingredients based on a physical property and ingredient attribute correlation model, and to correlate the content of major components. Specifically, this includes the following steps: The statistical collaborative control database contains the granulation quality test results of different batch combinations of ingredients in the current processing plant's granulation equipment. The data is then used to output the estimated value of the current batch combination of ingredients through a physical property and ingredient attribute association model, and to associate the content of the main components in the granulated product of the current batch combination of ingredients.
9. The adaptive control system for the production of high-efficiency compound fertilizers, involving both batching and granulation, as described in claim 8, is characterized in that: The process of establishing files in the collaborative control database using batch numbers as indexes and recording the estimated values of combined ingredient attributes and the content of major components, and optimizing the combination of ingredients from different batches in the collaborative control database based on the estimated values of combined ingredient attributes, the proportion of major component content, and the remaining amount of combined ingredients, combined with compound fertilizer production standards, to generate the mixing ratio of different batches of combined ingredients, specifically includes the following steps: Based on the upper and lower limits of compound fertilizer production standards, target values for main components and combined ingredient attributes are set. The range of the main component target is the positive and negative offset of the midpoint value of the standard interval corresponding to each target. The ideal value of the combined ingredient attribute target is the midpoint of the upper and lower limit interval, and the range is the positive and negative offset of the midpoint value of the upper and lower limit interval. The decision variable is set as the usage amount of each batch of combined ingredients, and the objective function of the optimization model is set as minimizing the weighted relative squared deviation of the physical properties after mixing from the ideal value. The constraints include that the total usage of each batch of combined ingredients is equal to the planned total amount of combined ingredients, the usage of each batch of combined ingredients does not exceed its inventory balance, the average content of each main component after mixing is within the target range of the main components, and the average value of each combined ingredient attribute after mixing is within the target range of the combined ingredient attribute. The interior point method is used to solve for the mass ratio of each batch of combined ingredients in the mixing, and the mixing ratio of generating different batches of combined ingredients is obtained as the optimal combined ingredient ratio.
10. The adaptive control system for the production, batching, and granulation of high-efficiency compound fertilizer as described in claim 1, characterized in that: The control execution module, based on the generated optimal combination ratio, extracts the compound fertilizer ingredient files of different ingredients in each batch of the optimal combination ratio from the collaborative control database. After extracting the ingredients through the automatic batching system and combining them into a combination, the combination is mixed according to the optimal combination ratio to obtain the optimal combination.