A constant temperature control and early warning system for glycoside-containing fertilizer production adding link

Abstract

The application provides a glycoside-containing fertilizer production adding link constant temperature control and early warning system, which comprises multi-modal production monitoring and time stamp labeling on a glycoside-containing fertilizer production adding link, obtaining a production parameter sequence, and material viscosity analysis and foam thickness analysis on the production parameter sequence, obtaining a standard parameter sequence; a reduced thermal network model is constructed, the reduced thermal network model is dynamically corrected, and a standard temperature distribution model is obtained; based on the standard temperature distribution model, active thermal damage analysis is performed on the glycoside-containing fertilizer, an active attenuation index sequence is obtained, based on the standard parameter sequence, adhesion risk analysis is performed on a three-dimensional model of an adding bin, an adhesion risk index sequence is obtained; a feeding risk prediction model is constructed, future risk prediction and strategy updating are performed based on the feeding risk prediction model, a constant temperature adding strategy is obtained, and production adding is performed according to the constant temperature adding strategy.

Description

Technical Field

[0001] This invention relates to the fields of fertilizer production and reinforcement learning technology, and in particular to a constant temperature control and early warning system for the application of glycoside-containing fertilizers. Background Technology

[0002] With the development of functional fertilizers, adding bioactive components such as glycosides to traditional fertilizers has become an important way to improve fertilizer efficiency. The production efficiency and product quality of such glycoside-containing fertilizers are highly dependent on the process control during the addition and mixing stage, especially the uniformity and stability of the temperature field. However, existing temperature control methods in the addition process have significant shortcomings.

[0003] Traditional methods often rely on proportional-integral (PID) control at single or a few temperature measurement points, failing to detect the complex temperature distribution and material state changes within the mixing chamber. This makes it difficult to address the unique heat sensitivity, high viscosity, and foaming characteristics of glycoside-containing materials. Localized overheating can lead to glycoside deactivation, while uneven temperature or excessively high viscosity can cause material adhesion and scaling, affecting mixing uniformity and equipment stability. Existing control strategies lack online soft-sensing capabilities for key process states (such as real-time viscosity and foam thickness) and cannot quantitatively assess or predict the risks of thermal damage to active materials and equipment adhesion. This results in severe delays in production process control, allowing only passive adjustments after anomalies occur. It fails to provide early warning and proactive suppression of quality risks, hindering the improvement of activity retention and production stability in high-end glycoside-containing fertilizer products. Summary of the Invention

[0004] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a constant temperature control and early warning system for the application of glycoside-containing fertilizers. It has the advantages of real-time perception of multi-modal process states, high-precision prediction of three-dimensional temperature fields, quantitative assessment of activity and adhesion risks, and feedforward optimization control. It solves the problems of lack of key state perception, low accuracy of temperature field calculation, lag in control response, and difficulty in coordinating and optimizing multiple objectives in traditional methods.

[0005] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: This invention provides a constant temperature control and early warning system for the application of glycoside-containing fertilizers, comprising a data acquisition module, a thermal modeling module, an index calculation module, a risk early warning module, and a constant temperature control module, wherein: The data acquisition module performs multimodal production monitoring and timestamp annotation on the production and application of glycoside fertilizers to obtain a production parameter sequence. It then performs material viscosity analysis and foam thickness analysis on the production parameter sequence to obtain a standard parameter sequence. The production parameters include temperature distribution, humidity distribution, mixer speed, mixer torque, dosing pump frequency, and liquid level data with timestamp annotation. The thermal modeling module constructs a reduced-order thermal network model based on fluid mechanics and the discrete element method, and dynamically corrects the reduced-order thermal network model based on the standard parameter sequence to obtain a standard temperature distribution model. The index calculation module calculates a three-dimensional temperature field sequence based on the standard temperature distribution model, performs an activity thermal damage analysis on the glycoside-containing fertilizer based on the three-dimensional temperature field sequence to obtain an activity decay index sequence, and performs an adhesion risk analysis on the standard parameter sequence based on the three-dimensional temperature field sequence to obtain an adhesion risk index sequence. The risk warning module constructs a feeding risk prediction model based on the production parameter sequence, activity decay index sequence, and adhesion risk index sequence, and predicts future risks based on the feeding risk prediction model to obtain warning information. The constant temperature control module generates a warning addition strategy based on the warning information, updates the warning addition strategy into a constant temperature addition strategy based on the feeding risk prediction model, and performs production addition according to the constant temperature addition strategy.

[0006] According to a preferred embodiment of the present invention, when the data acquisition module performs material viscosity analysis and foam thickness analysis on the production parameter sequence to obtain a standard parameter sequence, it includes: The material viscosity sequence is obtained by analyzing the mixer speed and mixer torque of each production parameter in the production parameter sequence. Then, the material viscosity sequence is temperature compensated based on the temperature distribution of each production parameter to obtain the compensated viscosity sequence. The initial foam thickness sequence is calculated using the liquid level data of each production parameter in the production parameter sequence, and the stability of the initial foam thickness sequence is corrected based on the temperature and humidity distribution of each production parameter to obtain the corrected foam thickness sequence. Sputtering noise reduction is performed on the modified foam thickness sequence based on Kalman filtering to obtain a noise-reduced foam thickness sequence. Then, coupling error correction is performed on the noise-reduced foam thickness sequence based on the compensated viscosity sequence to obtain a standard foam thickness sequence. The compensated viscosity sequence and the standard foam thickness sequence are added to the production parameter sequence to obtain the standard parameter sequence.

[0007] According to another preferred embodiment of the present invention, when the thermal modeling module constructs a reduced-order thermal network model based on fluid dynamics and the discrete element method, it includes: A three-dimensional model of the mixing chamber containing the glycoside fertilizer is obtained. The movement of the fertilizer particles containing the glycoside fertilizer in the three-dimensional model of the chamber is simulated based on the discrete element method. The flow field of the glycoside liquid containing the glycoside fertilizer in the three-dimensional model of the chamber is simulated based on fluid dynamics to obtain the chamber dosing model. The heat transfer of frictional heat, adsorption heat and reaction heat was simulated for the loading model of the silo body to obtain a three-dimensional exothermic temperature field. The three-dimensional model of the chamber is spatially discretized to obtain an equivalent set of thermal nodes, and the set of nodal energy equations corresponding to the equivalent set of thermal nodes is calculated based on the three-dimensional exothermic temperature field. The set of node energy equations is subjected to equivalent clustering and merging, as well as intrinsic orthogonal decomposition, to obtain a reduced-order thermal network model.

[0008] According to another preferred embodiment of the present invention, when the thermal modeling module performs dynamic correction on the reduced-order thermal network model based on the standard parameter sequence to obtain a standard temperature distribution model, it includes: The compensated viscosity sequence and the standard foam thickness sequence are extracted from the standard parameter sequence. The compensated viscosity sequence is converted into a convective heat transfer coefficient sequence, and the standard foam thickness sequence is converted into an equivalent thermal resistance increment sequence. The convective heat transfer coefficient sequence and the equivalent thermal resistance increment sequence are input into the reduced-order thermal network model for time-domain recursive prediction to obtain the node predicted temperature sequence. Extract the node measured temperature sequence corresponding to the node predicted temperature sequence from each temperature distribution of the standard parameter sequence, and calculate the temperature residual between the node measured temperature sequence and the node predicted temperature sequence to obtain the node temperature residual sequence. The parameters of the reduced-order thermal network model are dynamically corrected based on the node temperature residual sequence to obtain a standard temperature distribution model.

[0009] According to another preferred embodiment of the present invention, when the index calculation module performs an activity thermal damage analysis on the glycoside-containing fertilizer based on the three-dimensional temperature field sequence to obtain an activity decay index sequence, it includes: Obtain the set of fertilizer movement trajectories of the glycoside-containing fertilizer, and combine them with the three-dimensional temperature field sequence to generate the spatiotemporal distribution sequence of fertilizer temperature corresponding to the fertilizer movement trajectory; Thermal damage analysis of the glycoside-containing fertilizer was performed based on the spatiotemporal distribution sequence of the fertilizer temperature to obtain the instantaneous thermal damage rate distribution sequence. The instantaneous thermal damage rate distribution sequence is integrated over time to obtain the cumulative thermal damage factor distribution sequence, and the activity retention rate is calculated based on the cumulative thermal damage factor distribution sequence to obtain the activity retention rate distribution sequence. Based on the volume distribution of the glycoside-containing fertilizer, the activity retention rate distribution sequence is weighted and quantified to obtain the activity decay index sequence.

[0010] According to another preferred embodiment of the present invention, when the index calculation module performs adhesion risk analysis on the standard parameter sequence based on the three-dimensional temperature field sequence to obtain the adhesion risk index sequence, it includes: Based on the three-dimensional temperature field sequence, the compensated viscosity sequence in the standard parameter sequence is reversed to obtain the local instantaneous viscosity field sequence. Based on the humidity distribution sequence in the standard parameter sequence, the adhesion risk is calculated for the local instantaneous viscosity field sequence to obtain a local adhesion risk factor distribution map sequence; The local adhesion risk factor distribution map sequence is globally aggregated to obtain a global adhesion risk index sequence, and the global adhesion risk index sequence is then filtered by moving average to obtain an adhesion risk index sequence.

[0011] According to another preferred embodiment of the present invention, when the risk warning module constructs a feeding risk prediction model based on the production parameter sequence, the activity decay index sequence, and the adhesion risk index sequence, it includes: The production parameter sequence is subjected to feature extraction and feature fusion to obtain a production parameter feature sequence. The activity decay index sequence and adhesion risk index sequence are vectorized and feature-concatenated to obtain a feeding risk feature sequence. The mutual information of the production parameter feature sequence and the material feeding risk feature sequence is calculated to obtain the mutual information coefficient. Based on the mutual information coefficient, the production parameter feature sequence is filtered to obtain the dimensionality-reduced production feature sequence. Using the reduced-dimensional production feature sequence as input and the time-shifted feeding risk feature sequence as label, the preset temporal neural network model is trained into a feeding risk prediction model.

[0012] According to another preferred embodiment of the present invention, when the risk warning module executes the feeding risk prediction model to predict future risks and obtains warning information, it includes: By taking the current reduced-dimensional production characteristics as input, the output is the material feeding risk characteristics at several future times, thus obtaining a predicted risk characteristic sequence. Decoding the predicted risk feature sequence yields the predicted activity decay index sequence and the predicted adhesion risk index sequence. When the predicted activity decay index is greater than the preset activity decay threshold or the predicted adhesion risk index is greater than the preset adhesion risk threshold, the corresponding time is taken as the warning time, the predicted activity decay index and the predicted adhesion risk index corresponding to the warning time are taken as the warning risk index, and the production parameters at the current time are taken as the warning production parameters. Early warning information is generated based on the warning time, warning risk index, and warning production parameters.

[0013] According to another preferred embodiment of the present invention, when the constant temperature control module updates the early warning dosing strategy to a constant temperature dosing strategy based on the feeding risk prediction model, it includes: Calculate the action score corresponding to the early warning deployment strategy, generate strategy production parameters based on the early warning information and the early warning deployment strategy, and extract strategy dimensionality reduction production features from the strategy production parameters; The strategy risk feature sequence corresponding to the strategy dimensionality reduction production characteristics is analyzed using the aforementioned material input risk prediction model. Calculate the strategy incentive value corresponding to the strategy risk feature sequence, and generate a strategy score based on the action score and the strategy incentive value; The early warning dosing strategy is updated using the strategy score to obtain the constant temperature dosing strategy.

[0014] This invention provides a method for constant temperature control and early warning in the application stage of glycoside-containing fertilizer production, comprising: Multimodal production monitoring and timestamp annotation were performed on the production and application of glycoside-containing fertilizers to obtain a production parameter sequence. Material viscosity analysis and foam thickness analysis were then performed on the production parameter sequence to obtain a standard parameter sequence. The production parameters included temperature distribution, humidity distribution, mixer speed, mixer torque, dosing pump frequency, and liquid level data with timestamp annotation. A reduced-order thermal network model was constructed based on fluid mechanics and the discrete element method, and the reduced-order thermal network model was dynamically corrected based on the standard parameter sequence to obtain a standard temperature distribution model. A three-dimensional temperature field sequence was calculated based on the standard temperature distribution model, and an activity thermal damage analysis was performed on the glycoside-containing fertilizer based on the three-dimensional temperature field sequence to obtain an activity decay index sequence. An adhesion risk analysis was performed on the standard parameter sequence based on the three-dimensional temperature field sequence to obtain an adhesion risk index sequence. A feeding risk prediction model is constructed based on the production parameter sequence, activity decay index sequence, and adhesion risk index sequence. Future risks are predicted based on the feeding risk prediction model to obtain early warning information. A warning dosing strategy is generated based on the warning information. The warning dosing strategy is then updated to a constant temperature dosing strategy based on the feeding risk prediction model. Production dosing is then carried out according to the constant temperature dosing strategy.

[0015] (III) Beneficial Effects Compared with the prior art, the present invention provides a constant temperature control and early warning system for the application of glycoside-containing fertilizers, which has the following beneficial effects: This isothermal control and early warning system for the dosing stage of glycoside-containing fertilizer production achieves dynamic estimation of the true rheological state of glycoside-containing materials by constructing viscosity inversion analysis based on stirring dynamics and exponential compensation analysis considering temperature effects. It calculates the initial foam thickness using liquid level difference and performs stability correction and state filtering by combining temperature gradient, humidity distribution, and viscosity coupling relationships, effectively eliminating interference from stirring splash and measurement noise on foam thickness identification. By introducing a Kalman state estimation coupled with viscosity change rate compensation mechanism, a physically consistent and temporally continuous standard foam thickness sequence is constructed, which not only improves the accuracy of boundary conditions in three-dimensional temperature field modeling but also provides highly reliable input data for subsequent prediction of active thermal damage and adhesion risk assessment. A set of nodal energy equations coupled with multiphysics fields is used to form a reduced-order thermal network model through spatial clustering and intrinsic orthogonal decomposition. This significantly reduces computational complexity while ensuring physical consistency, enabling the thermal simulation model to meet real-time control requirements. By introducing a dynamic correction mechanism based on standard parameter sequences, fluid correlation is used to convert compensated viscosity into convective heat transfer coefficient, and foam thickness is converted into equivalent thermal resistance correction boundary conditions. Combined with Kalman filtering for state estimation and adaptive parameter updates, the model can converge to the true temperature field in real time as production conditions fluctuate. This ensures the physical interpretability of the model and enhances its anti-disturbance capability and long-term stability, providing a high-precision, low-computational-load prediction basis for subsequent isothermal control and improving the computational accuracy of isothermal control.

[0016] This isothermal control and early warning system for the production and application of glycoside-containing fertilizers achieves dynamic characterization of the internal heat distribution of the mixing chamber by constructing a three-dimensional temperature field sequence. It couples temperature information with particle motion trajectory to establish a thermal damage kinetic model, which can accurately assess the degree of activity decay of glycoside-containing fertilizers during processing. By jointly modeling the temperature-corrected viscosity field and humidity distribution, an adhesion risk function is constructed to achieve real-time quantification of local and overall adhesion risks in the chamber. By unifying the thermal field, physical property changes, and structural risks under the same computational framework, it maintains the integrity of the physical causal chain while avoiding complex multi-field coupling solutions, reducing computational complexity, enhancing computational real-time performance, facilitating online monitoring and early warning control, and improving the stability and feasibility of isothermal control.

[0017] This isothermal control and early warning system for the glycoside fertilizer production and application process effectively reduces the dimensionality of production parameters and redundant information through feature extraction and mutual information screening mechanisms, thereby improving model training efficiency and real-time response capabilities. By constructing supervisory labels through time offsetting, the model gains forward-looking predictive capabilities, shifting from post-event evaluation to pre-event early warning. Employing LSTM for multivariate time-series modeling captures the long-term dependence of production parameters on activity decay and adhesion risk, improving prediction accuracy and stability. Through risk vector decoding and threshold determination mechanisms, a dual-risk joint early warning system is achieved, enhancing system safety redundancy. By unifying and integrating the results of temperature field analysis, activity decay analysis, and adhesion risk analysis, a complete risk closed-loop control system is constructed, improving production safety and product stability. Attached Figure Description

[0018] Figure 1 The diagram shown is a structural diagram of a constant temperature control and early warning system for the production and application of glycoside-containing fertilizers according to the present invention.

[0019] Figure 2 The flowchart shown is a method for constant temperature control and early warning in the production and application of glycoside-containing fertilizers according to the present invention. Detailed Implementation

[0020] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious modifications will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0021] It is understood that the term "a" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0022] Example 1: Please refer to Figure 1 This invention discloses a constant temperature control and early warning system for the application stage of glycoside-containing fertilizers. The system mainly includes a data acquisition module, a thermal modeling module, an index calculation module, a risk early warning module, and a constant temperature control module, wherein: The data acquisition module performs multimodal production monitoring and timestamp annotation on the production and application of glycoside-containing fertilizers to obtain a production parameter sequence. It then performs material viscosity analysis and foam thickness analysis on the production parameter sequence to obtain a standard parameter sequence.

[0023] The production parameters include time-stamped temperature distribution, humidity distribution, mixer speed, mixer torque, dosing pump frequency, and liquid level data. The production parameter sequence is a data sequence after data noise reduction, data standardization, and time-sequential arrangement of the production parameters at each time point. Temperature distribution can be collected using an array temperature sensor in multimodal production monitoring, reflecting the temperature data at various locations within the mixing chamber during the dosing process. Humidity distribution is collected using an array humidity sensor, reflecting the humidity data at various locations within the mixing chamber. Mixer speed refers to the real-time operating speed of the mixing motor within the mixing chamber, which can be collected through the inverter's digital communication interface. Mixer torque reflects the shear stress of the fertilizer material and can be obtained through a torque sensor of the mixing motor or calculated based on the mixer speed and operating current. Dosing pump frequency can be collected by reading the frequency converter of the dosing pump or through an electromagnetic flowmeter. The liquid level data, including foam level and material level, can be collected using an ultrasonic level gauge or radar level gauge, reflecting the foam thickness generated by the viscosity of glycosides during the dosing and mixing process.

[0024] Specifically, when the data acquisition module performs material viscosity analysis and foam thickness analysis on the production parameter sequence to obtain a standard parameter sequence, it includes: The material viscosity sequence is obtained by analyzing the mixer speed and mixer torque of each production parameter in the production parameter sequence. Then, the material viscosity sequence is temperature compensated based on the temperature distribution of each production parameter to obtain the compensated viscosity sequence. The initial foam thickness sequence is calculated using the liquid level data of each production parameter in the production parameter sequence, and the stability of the initial foam thickness sequence is corrected based on the temperature and humidity distribution of each production parameter to obtain the corrected foam thickness sequence. Sputtering noise reduction is performed on the modified foam thickness sequence based on Kalman filtering to obtain a noise-reduced foam thickness sequence. Then, coupling error correction is performed on the noise-reduced foam thickness sequence based on the compensated viscosity sequence to obtain a standard foam thickness sequence. The compensated viscosity sequence and the standard foam thickness sequence are added to the production parameter sequence to obtain the standard parameter sequence.

[0025] The calculation formula for material viscosity analysis is as follows: in, for The viscosity of the material at any given time. for The torque of the mixer at any given moment. For structural constants, for The speed of the mixer at any given time. The stirring flow regime correction index is determined experimentally. When the viscosity is Newtonian, it reflects the fluid state of the material. Since temperature affects the viscosity of the material, temperature compensation is required for the viscosity series. The compensation formula is as follows: in, for Compensation viscosity at any time, The activation energy of glycosides in glycoside-containing fertilizers was determined experimentally. This is the ideal gas constant, with a default value of 8.314. It is the mean of the temperature distribution. Temperature at corresponding location in the temperature distribution at any given time. The absolute zero constant is used to calculate the initial foam thickness. This involves extracting the foam level and material level from the liquid level data, and subtracting the material level from the foam level to obtain the initial foam thickness. Since the viscosity of glycosides produces foam, and foam can hinder temperature conduction, it is necessary to calculate the foam thickness to improve the accuracy of subsequent three-dimensional temperature field calculations.

[0026] Specifically, temperature affects the surface tension of the liquid, humidity affects the bubble evaporation rate, and temperature gradient affects the foam bursting rate. Therefore, stability correction is required, and the mathematical formula for stability correction is as follows: in, yes The foam thickness needs to be adjusted regularly. yes The initial foam thickness at time [time]. , , For experimental calibration coefficients, for The gradient of temperature distribution change over time. for Humidity distribution at any given time; due to the large amount of spurious fluctuations and random noise in the liquid level measurement during stirring, Kalman filtering is needed to remove the foam thickness noise caused by stirring and splashing. Since there is a correlation between material viscosity and foam thickness, high-viscosity materials will encapsulate a large number of tiny bubbles. Therefore, it is necessary to decouple the noise-reducing foam thickness sequence based on the compensated viscosity sequence to correct the coupling error. The corresponding mathematical formula is as follows: in, yes Standard foam thickness at any given time The preset viscous damping correction coefficient, The differential symbol, Yes The differential, For time coefficient The differential.

[0027] By constructing a viscosity inversion analysis based on stirring dynamics and an exponential compensation analysis considering the influence of temperature, dynamic estimation of the true rheological state of glycoside-containing materials was achieved. The initial foam thickness was calculated by liquid level difference, and stability correction and state filtering were performed by combining temperature gradient, humidity distribution and viscosity coupling relationship, which effectively eliminated the interference of stirring splash and measurement noise on foam thickness identification. By introducing a Kalman state estimation and viscosity change rate coupling compensation mechanism, a physically consistent and temporally continuous standard foam thickness sequence was constructed, which not only improved the accuracy of boundary conditions for three-dimensional temperature field modeling, but also provided highly reliable input data for subsequent active thermal damage prediction and adhesion risk assessment.

[0028] The thermal modeling module constructs a reduced-order thermal network model based on fluid mechanics and the discrete element method, and dynamically corrects the reduced-order thermal network model based on the standard parameter sequence to obtain a standard temperature distribution model.

[0029] Specifically, when the thermal modeling module constructs a reduced-order thermal network model based on fluid mechanics and the discrete element method, it includes: A three-dimensional model of the mixing chamber containing the glycoside fertilizer is obtained. The movement of the fertilizer particles containing the glycoside fertilizer in the three-dimensional model of the chamber is simulated based on the discrete element method. The flow field of the glycoside liquid containing the glycoside fertilizer in the three-dimensional model of the chamber is simulated based on fluid dynamics to obtain the chamber dosing model. The heat transfer of frictional heat, adsorption heat and reaction heat was simulated for the loading model of the silo body to obtain a three-dimensional exothermic temperature field. The three-dimensional model of the chamber is spatially discretized to obtain an equivalent set of thermal nodes, and the set of nodal energy equations corresponding to the equivalent set of thermal nodes is calculated based on the three-dimensional exothermic temperature field. The set of node energy equations is subjected to equivalent clustering and merging, as well as intrinsic orthogonal decomposition, to obtain a reduced-order thermal network model.

[0030] The 3D model of the storage chamber refers to the 3D model of the mixing chamber. Since glycoside-containing fertilizers are fed into the mixing chamber during production, and localized frictional heat generation occurs during mixing due to fertilizer particle friction, adsorption heat from the combination of fertilizer particles and glycoside liquid, and localized chemical reaction heat generation, motion simulation and flow field simulation are required to analyze the localized heat generation of the fertilizer inside the mixing chamber. Glycoside-containing fertilizers typically involve the mixing of liquid spray and solid particles; therefore, a two-way coupling algorithm is used to realize the momentum and energy exchange between particle motion and the flow field. Specifically, the amount of fertilizer particles fed is determined based on the pump frequency, and motion simulation is performed. The equations for the motion simulation are as follows: in, It refers to the first The quality of each fertilizer granule It refers to the spatial displacement vector of fertilizer particles. It is the contact force between fertilizer particles. It is the fluid resistance generated by glycoside liquids on fertilizer particles. It is the force of gravity acting on the fertilizer particles; The equations for the flow field simulation are as follows: in, It is the density of glycoside liquids. It is the sign of the partial derivative. It is the fluid velocity vector. It is a differential operator. For flow field pressure, To compensate for viscosity, The reaction force of fertilizer particles on glycoside liquid is a volume force. The three-dimensional exothermic temperature field is a heat transfer simulation of frictional heat, adsorption heat, dissolution heat, and local chemical reaction heat in the chamber addition model. This heat transfer can be simulated using the energy conservation equation. in, The specific heat capacity of glycoside-containing fertilizers, For local instantaneous temperature, Thermal conductivity of glycoside-containing fertilizers for At time, coordinates are The sum of heat generated at a given location, including frictional heat, adsorption heat, heat released from dissolution, and heat released from local chemical reactions; The spatial discretization refers to discretizing the three-dimensional model of the silo according to equal distances or equal grids, and taking each three-dimensional grid node as an equivalent set of thermal nodes. The calculation of the set of node energy equations refers to defining the heat capacity and thermal resistance of each equivalent thermal node, and calculating the node energy equations based on the heat capacity and thermal resistance. The corresponding formulas are as follows: in, It refers to the first The heat capacity of each equivalent thermal node. It refers to the first The density of glycosidic liquids at equivalent thermal nodes, It refers to the first Specific heat capacity of each equivalent thermal node It refers to the first The nodal volume of an equivalent hot node. It refers to the first The equivalent thermal node and the first Thermal resistance between equivalent thermal nodes It is the characteristic heat transfer distance between the centers of each equivalent thermal node. This represents the equivalent heat transfer area between equivalent thermal nodes. It is the first The temperature of the equivalent thermal node It refers to the first The temperature of the equivalent thermal node It is the first The equivalent thermal node's heat source power; equivalent clustering and merging refers to performing temperature time series correlation analysis on the equivalent thermal node set, constructing a correlation matrix between nodes, constructing a distance matrix between nodes based on the correlation matrix, and using spectral clustering or K-means clustering methods to spatially cluster and merge the equivalent thermal nodes according to the distance matrix between nodes to obtain several clustered node sets. For each clustered node set, parameter equivalence merging is performed, and the clustered thermal network model is converted into a state matrix in state space form. The state matrix is ​​then subjected to intrinsic orthogonal decomposition to construct a projection matrix, resulting in a reduced-order thermal network model. The construction of the reduced-order thermal network model can be performed before isothermal control, thereby improving the computational efficiency during isothermal control.

[0031] In detail, when the thermal modeling module performs dynamic correction on the reduced-order thermal network model based on the standard parameter sequence to obtain a standard temperature distribution model, it includes: The compensated viscosity sequence and the standard foam thickness sequence are extracted from the standard parameter sequence. The compensated viscosity sequence is converted into a convective heat transfer coefficient sequence, and the standard foam thickness sequence is converted into an equivalent thermal resistance increment sequence. The convective heat transfer coefficient sequence and the equivalent thermal resistance increment sequence are input into the reduced-order thermal network model for time-domain recursive prediction to obtain the node predicted temperature sequence. Extract the node measured temperature sequence corresponding to the node predicted temperature sequence from each temperature distribution of the standard parameter sequence, and calculate the temperature residual between the node measured temperature sequence and the node predicted temperature sequence to obtain the node temperature residual sequence. The parameters of the reduced-order thermal network model are dynamically corrected based on the node temperature residual sequence to obtain a standard temperature distribution model.

[0032] Specifically, fluid dynamics correlations can be used to transform the compensated viscosity sequence into a convective heat transfer coefficient sequence and the standard foam thickness sequence into an equivalent thermal resistance increment sequence. For example, the Reynolds number can be calculated using the compensated viscosity, and the Nusselt number corresponding to the Reynolds number can be calculated based on the Nusselt number correlation, and the convective heat transfer coefficient can be calculated using the Nusselt number. Spatial interpolation methods can be used to extract the node measured temperature sequence corresponding to the node predicted temperature sequence. The dynamic correction refers to completing the posterior state estimation through Kalman filtering, and adaptively updating the model boundary heat transfer coefficient or correction gain coefficient in the updated reduced-order heating network model based on the temperature residuals of each node, finally obtaining the standard temperature distribution model.

[0033] By constructing a set of nodal energy equations coupled with multiphysics fields, and then using spatial clustering and intrinsic orthogonal decomposition to form a reduced-order thermal network model, the computational complexity can be significantly reduced while ensuring physical consistency, enabling the thermal simulation model to meet real-time control requirements. By introducing a dynamic correction mechanism based on standard parameter sequences, the compensated viscosity is converted into the convective heat transfer coefficient using fluid correlation, and the foam thickness is converted into the equivalent thermal resistance correction boundary condition. Combined with Kalman filtering for state estimation and adaptive parameter updates, the model can converge to the real temperature field in real time as production conditions fluctuate. This ensures the physical interpretability of the model, enhances its anti-disturbance capability and long-term stability, provides a high-precision, low-computation prediction basis for subsequent isothermal control, and improves the computational accuracy of isothermal control.

[0034] The index calculation module calculates a three-dimensional temperature field sequence based on the standard temperature distribution model, performs an activity thermal damage analysis on the glycoside-containing fertilizer based on the three-dimensional temperature field sequence to obtain an activity decay index sequence, and performs an adhesion risk analysis on the standard parameter sequence based on the three-dimensional temperature field sequence to obtain an adhesion risk index sequence.

[0035] The calculation of the three-dimensional temperature field sequence based on the standard temperature distribution model refers to recalculating the temperature data at each location in the three-dimensional model of the mixing chamber based on each production parameter in the production parameter sequence, constructing the three-dimensional temperature field at the corresponding time, and assembling the three-dimensional temperature fields at each time into a three-dimensional temperature field sequence.

[0036] Specifically, when the index calculation module performs an activity thermal damage analysis on the glycoside-containing fertilizer based on the three-dimensional temperature field sequence to obtain the activity decay index sequence, it includes: Obtain the set of fertilizer movement trajectories of the glycoside-containing fertilizer, and combine them with the three-dimensional temperature field sequence to generate the spatiotemporal distribution sequence of fertilizer temperature corresponding to the fertilizer movement trajectory; Thermal damage analysis of the glycoside-containing fertilizer was performed based on the spatiotemporal distribution sequence of the fertilizer temperature to obtain the instantaneous thermal damage rate distribution sequence. The instantaneous thermal damage rate distribution sequence is integrated over time to obtain the cumulative thermal damage factor distribution sequence, and the activity retention rate is calculated based on the cumulative thermal damage factor distribution sequence to obtain the activity retention rate distribution sequence. Based on the volume distribution of the glycoside-containing fertilizer, the activity retention rate distribution sequence is weighted and quantified to obtain the activity decay index sequence.

[0037] The fertilizer trajectory set of glycoside-containing fertilizers can be obtained by simulating the motion of fertilizer particles and glycoside liquids and the flow field through the discrete element method in the thermal modeling module. The generation of the fertilizer temperature spatiotemporal distribution sequence refers to using the spatial location of the fertilizer as an index and sampling the temperature of the corresponding spatial location from the three-dimensional temperature field at the corresponding time in the sequence to obtain the fertilizer temperature spatiotemporal distribution sequence. Thermal damage analysis can be performed using the Arrhenius equation. The time integration refers to performing numerical integration of the instantaneous thermal damage rate at each location in the time dimension using the trapezoidal integral method. The cumulative thermal damage factor at each location in the cumulative thermal damage factor distribution sequence can be mapped by an exponential function to obtain the activity retention rate distribution sequence. The index can be quantified by taking the negative logarithm or the difference between 1 and the average value of the activity retention rate can be used as the activity decay index.

[0038] Specifically, when the index calculation module performs adhesion risk analysis on the standard parameter sequence based on the three-dimensional temperature field sequence to obtain the adhesion risk index sequence, it includes: Based on the three-dimensional temperature field sequence, the compensated viscosity sequence in the standard parameter sequence is reversed to obtain the local instantaneous viscosity field sequence. Based on the humidity distribution sequence in the standard parameter sequence, the adhesion risk is calculated for the local instantaneous viscosity field sequence to obtain a local adhesion risk factor distribution map sequence; The local adhesion risk factor distribution map sequence is globally aggregated to obtain a global adhesion risk index sequence, and the global adhesion risk index sequence is then filtered by moving average to obtain an adhesion risk index sequence.

[0039] The reverse correction method is consistent with the temperature compensation method of the data acquisition module, and will not be elaborated here. The adhesion risk calculation refers to calculating the corresponding adhesion risk using a preset empirical risk model. For example, nodes in the 3D model of the silo are selected as target nodes. The node humidity of the target node at the target time is selected from the humidity distribution sequence, and the node viscosity of the target node at the target time is selected from the local instantaneous viscosity field sequence. The corresponding local adhesion risk factor is calculated using the following adhesion risk function, and a local adhesion risk factor distribution map sequence is generated based on each local adhesion risk factor: in, It means Time and coordinates are Local adhesion risk factors at the location , For pre-trained adhesion weight factors, It means Time and coordinates are Nodal viscosity at the location It is a preset viscosity threshold. It means Time and coordinates are Node humidity at location It is a preset humidity threshold; global aggregation can be performed using spatial weighted integration.

[0040] By constructing a three-dimensional temperature field sequence, a dynamic characterization of the internal heat distribution of the mixing chamber was achieved. Temperature information was coupled with particle motion trajectories to establish a thermal damage kinetic model, enabling accurate assessment of the activity decay of glycoside-containing fertilizers during processing. Through joint modeling of the temperature-corrected viscosity field and humidity distribution, an adhesion risk function was constructed, enabling real-time quantification of local and overall adhesion risks within the chamber. By unifying the thermal field, physical property changes, and structural risks within the same computational framework, the integrity of the physical causal chain is maintained while avoiding complex multi-field coupling solutions. This reduces computational complexity, enhances real-time computation, facilitates online monitoring and early warning control, and improves the stability and feasibility of constant temperature control.

[0041] The risk warning module constructs a feeding risk prediction model based on the production parameter sequence, activity decay index sequence, and adhesion risk index sequence, and predicts future risks based on the feeding risk prediction model to obtain warning information.

[0042] Specifically, when the risk warning module constructs a feeding risk prediction model based on the production parameter sequence, activity decay index sequence, and adhesion risk index sequence, it includes: The production parameter sequence is subjected to feature extraction and feature fusion to obtain a production parameter feature sequence. The activity decay index sequence and adhesion risk index sequence are vectorized and feature-concatenated to obtain a feeding risk feature sequence. The mutual information of the production parameter feature sequence and the material feeding risk feature sequence is calculated to obtain the mutual information coefficient. Based on the mutual information coefficient, the production parameter feature sequence is filtered to obtain the dimensionality-reduced production feature sequence. Using the reduced-dimensional production feature sequence as input and the time-shifted feeding risk feature sequence as label, the preset temporal neural network model is trained into a feeding risk prediction model.

[0043] Feature extraction of the production parameter sequence refers to extracting statistical parameters such as the mean, variance, and maximum value of the production parameters at each time step in the production parameter sequence. Feature fusion refers to concatenating the vectorized production parameters and the vectorized statistical parameters to obtain the production parameter feature sequence. Mutual information calculation refers to calculating the mutual information between the feature sequences corresponding to each parameter in the production feature sequence and the material feeding risk feature sequence. Kernel density estimation can be used for mutual information calculation. Feature filtering refers to removing features of parameters with mutual information coefficients less than a preset mutual information threshold from the production parameter feature sequence, thereby reducing feature dimensionality and subsequent model computation. Since the material feeding risk prediction model is a time-series prediction model, the material feeding risk features need to be shifted by time steps. That is, the dimensionality-reduced production features at the current time step are used as the input for this batch, and the material feeding risk features at several time steps after the current time step are used as the label for this batch. The time-series neural network model can be a Long Short-Term Memory (LSTM) network. The training process of Memory (LSTM) refers to using a temporal neural network model to extract the temporal features corresponding to the dimensionality-reduced production features, and then outputting the predicted risk features corresponding to the temporal features through the output layer. The mean squared error loss between the predicted risk features and the label is calculated, and the model parameters of the temporal neural network model are iteratively optimized using Adam based on the mean squared error loss to obtain the material feeding risk prediction model.

[0044] In detail, when the risk warning module executes the material feeding risk prediction model to predict future risks and obtains warning information, it includes: By taking the current reduced-dimensional production characteristics as input, the output is the material feeding risk characteristics at several future times, thus obtaining a predicted risk characteristic sequence. Decoding the predicted risk feature sequence yields the predicted activity decay index sequence and the predicted adhesion risk index sequence. When the predicted activity decay index is greater than the preset activity decay threshold or the predicted adhesion risk index is greater than the preset adhesion risk threshold, the corresponding time is taken as the warning time, the predicted activity decay index and the predicted adhesion risk index corresponding to the warning time are taken as the warning risk index, and the production parameters at the current time are taken as the warning production parameters. Early warning information is generated based on the warning time, warning risk index, and warning production parameters.

[0045] The decoding method is the inverse step of the corresponding vectorization method.

[0046] By employing feature extraction and mutual information filtering mechanisms, the dimensionality of production parameters is effectively reduced, redundant information is minimized, and model training efficiency and real-time response capabilities are improved. Supervision labels are constructed using time offsets, enabling the model to possess forward-looking predictive capabilities and shifting from post-event evaluation to pre-event early warning. LSTM is used for multivariate time-series modeling to capture the long-term dependence of production parameters on activity decay and adhesion risks, improving prediction accuracy and stability. Risk vector decoding and threshold determination mechanisms enable joint early warning of dual risks, enhancing system safety redundancy. By unifying and integrating the results of temperature field analysis, activity decay analysis, and adhesion risk analysis, a complete risk closed-loop control system is constructed, improving production safety and product stability.

[0047] The constant temperature control module generates a warning addition strategy based on the warning information, updates the warning addition strategy into a constant temperature addition strategy based on the feeding risk prediction model, and performs production addition according to the constant temperature addition strategy.

[0048] Specifically, generating a warning-based dosing strategy based on the warning information involves normalizing the warning information and concatenating it into an environmental state vector; constructing an action space based on the parameter range of the production parameter sequence; and generating a warning-based dosing strategy based on the action space. The action space refers to the parameter range of production parameters that can be changed during the dosing process, such as the adjustment range of the dosing pump frequency and the adjustment range of the stirring speed. The action space is a discrete combination of all adjustment ranges. Generating the warning-based dosing strategy involves using a reinforcement learning (Actor) network to generate action probability distribution parameters based on the environmental state vector, and sampling in the action space based on this probability distribution to generate the corresponding warning-based dosing strategy. The warning-based dosing strategy is an adjustment scheme for production parameters, including temperature setting adjustment, humidity adjustment, stirring speed adjustment, and dosing pump frequency adjustment.

[0049] Specifically, when the constant temperature control module updates the early warning dosing strategy to a constant temperature dosing strategy based on the feeding risk prediction model, it includes: Calculate the action score corresponding to the early warning deployment strategy, generate strategy production parameters based on the early warning information and the early warning deployment strategy, and extract strategy dimensionality reduction production features from the strategy production parameters; The strategy risk feature sequence corresponding to the strategy dimensionality reduction production characteristics is analyzed using the aforementioned material input risk prediction model. Calculate the strategy incentive value corresponding to the strategy risk feature sequence, and generate a strategy score based on the action score and the strategy incentive value; The early warning dosing strategy is updated using the strategy score to obtain the constant temperature dosing strategy.

[0050] The calculation of action score refers to the rate of change of the corresponding parameters between the strategy production parameters and the early warning production parameters in the early warning information. The average rate of change of all parameters is used as the standard rate of change, and the negative value of the standard rate of change is used as the action score. The strategy production parameter refers to the production parameter at the current moment corresponding to the early warning deployment strategy. That is, the early warning production parameter is obtained by adding the adjustment amount in the early warning deployment strategy to the early warning production parameter. The method of extracting the strategy dimensionality reduction production feature is the same as the method of extracting the dimensionality reduction production feature sequence from the production parameter sequence, and will not be repeated here. The calculation of strategy incentive value refers to extracting the strategy activity decay index sequence and the strategy adhesion risk index sequence from the strategy risk feature sequence. The difference between the activity decay threshold and each strategy activity decay index in the strategy activity decay index sequence is calculated, and the difference between the adhesion risk threshold and each strategy adhesion risk index in the strategy adhesion risk index sequence is calculated. The sum of the differences is used as the primary incentive value. The sum of the squared deviations of the strategy activity decay index sequence and the sum of the squared deviations of the strategy adhesion risk index sequence is used as the standard sum of squared deviations. The strategy incentive value is obtained by subtracting the weighted standard sum of squared deviations from the primary incentive value.

[0051] Specifically, the policy score is equal to the weighted sum of the policy incentive value and the action score. Feedforward update refers to adjusting the sampling rules of the Actor network using linear correction or gradient correction methods and updating the warning application policy until the number of iterations is reached or the optimal policy score is obtained, at which point the warning application policy corresponding to the optimal policy score is used as the isothermal application policy.

[0052] By generating early warning feeding strategies through reinforcement learning and using a feeding risk prediction model for feedforward risk assessment, the coupling optimization of strategy generation and risk prediction is achieved. By constructing action scoring constraints on parameter variation, the impact of drastic fluctuations in production parameters on isothermal stability is avoided. By considering both the degree of risk reduction and risk volatility in the strategy incentive value and introducing a deviation sum of squares penalty term, the stability and safety of the control process are improved. Through weighted fusion of strategy scores, a balance optimization between risk and return and process constraints is achieved, improving production stability and risk suppression capabilities, and enhancing the system's adaptive adjustment capabilities.

[0053] Example 2: Please refer to Figure 2 This invention discloses a method for constant temperature control and early warning in the production and application of glycoside-containing fertilizers, comprising the following steps: Multimodal production monitoring and timestamp annotation were performed on the production and application of glycoside-containing fertilizers to obtain a production parameter sequence. Material viscosity analysis and foam thickness analysis were then performed on the production parameter sequence to obtain a standard parameter sequence. The production parameters included temperature distribution, humidity distribution, mixer speed, mixer torque, dosing pump frequency, and liquid level data with timestamp annotation. A reduced-order thermal network model was constructed based on fluid mechanics and the discrete element method, and the reduced-order thermal network model was dynamically corrected based on the standard parameter sequence to obtain a standard temperature distribution model. A three-dimensional temperature field sequence was calculated based on the standard temperature distribution model, and an activity thermal damage analysis was performed on the glycoside-containing fertilizer based on the three-dimensional temperature field sequence to obtain an activity decay index sequence. An adhesion risk analysis was performed on the standard parameter sequence based on the three-dimensional temperature field sequence to obtain an adhesion risk index sequence. A feeding risk prediction model is constructed based on the production parameter sequence, activity decay index sequence, and adhesion risk index sequence. Future risks are predicted based on the feeding risk prediction model to obtain early warning information. A warning dosing strategy is generated based on the warning information. The warning dosing strategy is then updated to a constant temperature dosing strategy based on the feeding risk prediction model. Production dosing is then carried out according to the constant temperature dosing strategy.

[0054] The processes described above with reference to the flowcharts in the embodiments disclosed in this invention can be implemented as computer software programs. The embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), it performs the functions defined in the methods of this application. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wire segments, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless segments, wire segments, optical fibers, RF, etc., or any suitable combination thereof.

[0055] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0056] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been shown and explained in the embodiments. Without departing from the stated principles, the implementation of the present invention may have any variations or modifications.

Claims

1. A constant temperature control and early warning system for the application stage of glycoside-containing fertilizers, characterized in that, The system includes a data acquisition module, a thermal modeling module, an index calculation module, a risk warning module, and a constant temperature control module, wherein: The data acquisition module performs multimodal production monitoring and timestamp annotation on the production and application of glycoside fertilizers to obtain a production parameter sequence. It then performs material viscosity analysis and foam thickness analysis on the production parameter sequence to obtain a standard parameter sequence. The production parameters include temperature distribution, humidity distribution, mixer speed, mixer torque, dosing pump frequency, and liquid level data with timestamp annotation. The thermal modeling module constructs a reduced-order thermal network model based on fluid mechanics and the discrete element method, and dynamically corrects the reduced-order thermal network model based on the standard parameter sequence to obtain a standard temperature distribution model. The index calculation module calculates a three-dimensional temperature field sequence based on the standard temperature distribution model, performs an activity thermal damage analysis on the glycoside-containing fertilizer based on the three-dimensional temperature field sequence to obtain an activity decay index sequence, and performs an adhesion risk analysis on the standard parameter sequence based on the three-dimensional temperature field sequence to obtain an adhesion risk index sequence. The risk warning module constructs a feeding risk prediction model based on the production parameter sequence, activity decay index sequence, and adhesion risk index sequence, and predicts future risks based on the feeding risk prediction model to obtain warning information. The constant temperature control module generates a warning addition strategy based on the warning information, updates the warning addition strategy into a constant temperature addition strategy based on the feeding risk prediction model, and performs production addition according to the constant temperature addition strategy.

2. The constant temperature control and early warning system for the application stage of glycoside-containing fertilizer production according to claim 1, characterized in that, When the data acquisition module performs material viscosity analysis and foam thickness analysis on the production parameter sequence to obtain a standard parameter sequence, it includes: The material viscosity sequence is obtained by analyzing the mixer speed and mixer torque of each production parameter in the production parameter sequence. Then, the material viscosity sequence is temperature compensated based on the temperature distribution of each production parameter to obtain the compensated viscosity sequence. The initial foam thickness sequence is calculated using the liquid level data of each production parameter in the production parameter sequence, and the stability of the initial foam thickness sequence is corrected based on the temperature and humidity distribution of each production parameter to obtain the corrected foam thickness sequence. Sputtering noise reduction is performed on the modified foam thickness sequence based on Kalman filtering to obtain a noise-reduced foam thickness sequence. Then, coupling error correction is performed on the noise-reduced foam thickness sequence based on the compensated viscosity sequence to obtain a standard foam thickness sequence. The compensated viscosity sequence and the standard foam thickness sequence are added to the production parameter sequence to obtain the standard parameter sequence.

3. The constant temperature control and early warning system for the application stage of glycoside-containing fertilizer production according to claim 2, characterized in that, When the thermal modeling module executes the construction of a reduced-order thermal network model based on fluid mechanics and the discrete element method, it includes: A three-dimensional model of the mixing chamber containing the glycoside fertilizer is obtained. The movement of the fertilizer particles containing the glycoside fertilizer in the three-dimensional model of the chamber is simulated based on the discrete element method. The flow field of the glycoside liquid containing the glycoside fertilizer in the three-dimensional model of the chamber is simulated based on fluid dynamics to obtain the chamber dosing model. The heat transfer of frictional heat, adsorption heat and reaction heat was simulated for the loading model of the silo body to obtain a three-dimensional exothermic temperature field. The three-dimensional model of the chamber is spatially discretized to obtain an equivalent set of thermal nodes, and the set of nodal energy equations corresponding to the equivalent set of thermal nodes is calculated based on the three-dimensional exothermic temperature field. The set of node energy equations is subjected to equivalent clustering and merging, as well as intrinsic orthogonal decomposition, to obtain a reduced-order thermal network model.

4. The constant temperature control and early warning system for the application stage of glycoside-containing fertilizer production according to claim 3, characterized in that, When the thermal modeling module performs dynamic correction on the reduced-order thermal network model based on the standard parameter sequence to obtain a standard temperature distribution model, it includes: The compensated viscosity sequence and the standard foam thickness sequence are extracted from the standard parameter sequence. The compensated viscosity sequence is converted into a convective heat transfer coefficient sequence, and the standard foam thickness sequence is converted into an equivalent thermal resistance increment sequence. The convective heat transfer coefficient sequence and the equivalent thermal resistance increment sequence are input into the reduced-order thermal network model for time-domain recursive prediction to obtain the node predicted temperature sequence. Extract the node measured temperature sequence corresponding to the node predicted temperature sequence from each temperature distribution of the standard parameter sequence, and calculate the temperature residual between the node measured temperature sequence and the node predicted temperature sequence to obtain the node temperature residual sequence. The parameters of the reduced-order thermal network model are dynamically corrected based on the node temperature residual sequence to obtain a standard temperature distribution model.

5. The constant temperature control and early warning system for the application stage of glycoside-containing fertilizer production according to claim 1, characterized in that, When the index calculation module performs an activity thermal damage analysis on the glycoside-containing fertilizer based on the three-dimensional temperature field sequence to obtain the activity decay index sequence, it includes: Obtain the set of fertilizer movement trajectories of the glycoside-containing fertilizer, and combine them with the three-dimensional temperature field sequence to generate the spatiotemporal distribution sequence of fertilizer temperature corresponding to the fertilizer movement trajectory; Thermal damage analysis of the glycoside-containing fertilizer was performed based on the spatiotemporal distribution sequence of the fertilizer temperature to obtain the instantaneous thermal damage rate distribution sequence. The instantaneous thermal damage rate distribution sequence is integrated over time to obtain the cumulative thermal damage factor distribution sequence, and the activity retention rate is calculated based on the cumulative thermal damage factor distribution sequence to obtain the activity retention rate distribution sequence. Based on the volume distribution of the glycoside-containing fertilizer, the activity retention rate distribution sequence is weighted and quantified to obtain the activity decay index sequence.

6. The constant temperature control and early warning system for the application stage of glycoside-containing fertilizer production according to claim 5, characterized in that, When the index calculation module performs adhesion risk analysis on the standard parameter sequence based on the three-dimensional temperature field sequence to obtain the adhesion risk index sequence, it includes: Based on the three-dimensional temperature field sequence, the compensated viscosity sequence in the standard parameter sequence is reversed to obtain the local instantaneous viscosity field sequence. Based on the humidity distribution sequence in the standard parameter sequence, the adhesion risk is calculated for the local instantaneous viscosity field sequence to obtain a local adhesion risk factor distribution map sequence; The local adhesion risk factor distribution map sequence is globally aggregated to obtain a global adhesion risk index sequence, and the global adhesion risk index sequence is then filtered by moving average to obtain an adhesion risk index sequence.

7. The constant temperature control and early warning system for the application stage of glycoside-containing fertilizer production according to claim 1, characterized in that, When the risk warning module executes the process of constructing a feeding risk prediction model based on the production parameter sequence, activity decay index sequence, and adhesion risk index sequence, it includes: The production parameter sequence is subjected to feature extraction and feature fusion to obtain the production parameter feature sequence. The activity decay index sequence and adhesion risk index sequence are vectorized and feature spliced ​​to obtain the feeding risk feature sequence. The mutual information of the production parameter feature sequence and the material feeding risk feature sequence is calculated to obtain the mutual information coefficient. Based on the mutual information coefficient, the production parameter feature sequence is filtered to obtain the dimensionality-reduced production feature sequence. Using the reduced-dimensional production feature sequence as input and the time-shifted feeding risk feature sequence as label, the preset temporal neural network model is trained into a feeding risk prediction model.

8. The constant temperature control and early warning system for the application stage of glycoside-containing fertilizer production according to claim 6, characterized in that, When the risk warning module executes the material feeding risk prediction model to predict future risks and obtains warning information, it includes: By taking the current reduced-dimensional production characteristics as input, the output is the material feeding risk characteristics at several future times, thus obtaining a predicted risk characteristic sequence. Decoding the predicted risk feature sequence yields the predicted activity decay index sequence and the predicted adhesion risk index sequence. When the predicted activity decay index is greater than the preset activity decay threshold or the predicted adhesion risk index is greater than the preset adhesion risk threshold, the corresponding time is taken as the warning time, the predicted activity decay index and the predicted adhesion risk index corresponding to the warning time are taken as the warning risk index, and the production parameters at the current time are taken as the warning production parameters. Early warning information is generated based on the warning time, warning risk index, and warning production parameters.

9. A constant temperature control and early warning system for the application stage of glycoside-containing fertilizer production according to claim 8, characterized in that, When the constant temperature control module updates the early warning dosing strategy to a constant temperature dosing strategy based on the feeding risk prediction model, it includes: Calculate the action score corresponding to the early warning deployment strategy, generate strategy production parameters based on the early warning information and the early warning deployment strategy, and extract strategy dimensionality reduction production features from the strategy production parameters; The strategy risk feature sequence corresponding to the strategy dimensionality reduction production characteristics is analyzed using the aforementioned material input risk prediction model. Calculate the strategy incentive value corresponding to the strategy risk feature sequence, and generate a strategy score based on the action score and the strategy incentive value; The early warning dosing strategy is updated using the strategy score to obtain the constant temperature dosing strategy.

10. A method for constant temperature control and early warning in the application stage of glycoside-containing fertilizer production, characterized in that, The method includes: Multimodal production monitoring and timestamp annotation were performed on the production and application of glycoside-containing fertilizers to obtain a production parameter sequence. Material viscosity analysis and foam thickness analysis were then performed on the production parameter sequence to obtain a standard parameter sequence. The production parameters included temperature distribution, humidity distribution, mixer speed, mixer torque, dosing pump frequency, and liquid level data with timestamp annotation. A reduced-order thermal network model was constructed based on fluid mechanics and the discrete element method, and the reduced-order thermal network model was dynamically corrected based on the standard parameter sequence to obtain a standard temperature distribution model. A three-dimensional temperature field sequence was calculated based on the standard temperature distribution model, and an activity thermal damage analysis was performed on the glycoside-containing fertilizer based on the three-dimensional temperature field sequence to obtain an activity decay index sequence. An adhesion risk analysis was performed on the standard parameter sequence based on the three-dimensional temperature field sequence to obtain an adhesion risk index sequence. A feeding risk prediction model is constructed based on the production parameter sequence, activity decay index sequence, and adhesion risk index sequence. Future risks are predicted based on the feeding risk prediction model to obtain early warning information. A warning dosing strategy is generated based on the warning information. The warning dosing strategy is then updated to a constant temperature dosing strategy based on the feeding risk prediction model. Production dosing is then carried out according to the constant temperature dosing strategy.

Images