Temperature intelligent control method and device for wall-scraping crystallization kettle with multiple sensors

Abstract

The application discloses a temperature intelligent control method and device for a wall-scraping crystallization kettle with fusion multi-sensing, relates to the field of industrial control, and comprises the following steps: after nitrate material enters the wall-scraping crystallization kettle, a scraper is started to mix, and a real-time concentration value and a real-time temperature value are monitored; an actual nucleation induction temperature and an actual main growth temperature are recorded and acquired; a predicted particle size distribution and a predicted scarring risk index are obtained; a temperature setting value adjustment amount and a scraper rotating speed adjustment amount are obtained through optimization calculation; and the temperature setting value and the scraper rotating speed of the wall-scraping crystallization kettle after the main growth temperature node are adjusted. The technical problems that the existing temperature control method for the wall-scraping crystallization kettle is difficult to adapt to the dynamic change of process requirements in the crystallization process, leading to uneven crystal particle size distribution, poor product batch stability, and serious scarring of the inner wall of the crystallization kettle are solved.

Description

Technical Field

[0001] This application relates to the field of industrial control technology, specifically to a method and device for intelligent temperature control of a scraped-wall crystallizer that integrates multiple sensors. Background Technology

[0002] In modern industrial production, the crystallization process of fine chemical products such as nitrates places increasingly higher demands on product quality and production efficiency. As the core equipment in the crystallization process, the precision and intelligence of the scraped-wall crystallizer's temperature control directly affect the quality of the final product and the continuity of production.

[0003] However, traditional temperature control methods for scraped-wall crystallizers often rely on empirically set fixed temperature curves or simple feedback control based on a single temperature sensor. These methods are easily affected by various factors such as the initial state of the material, stirring intensity, and environmental disturbances. A single sensor cannot fully reflect the true state of the material inside the vessel, and a fixed temperature curve cannot adapt to the dynamically changing process requirements during crystallization. This often leads to problems such as uneven crystal particle size distribution, poor batch stability of products, and severe scaling on the inner wall of the crystallizer. This not only reduces heat transfer efficiency and increases energy consumption but may also affect the normal operation of the scraper, even requiring frequent shutdowns for cleaning, severely restricting the improvement of production efficiency and product quality. Summary of the Invention

[0004] This application provides a method and device for intelligent temperature control of a scraped wall crystallizer that integrates multiple sensors. This solves the technical problems of existing scraped wall crystallizer temperature control methods being unable to adapt to the dynamically changing process requirements during crystallization, resulting in uneven crystal particle size distribution, poor batch stability of products, and severe scaling on the inner wall of the crystallizer.

[0005] The technical solution to the above-mentioned technical problems in this application is as follows: In a first aspect, this application provides a method for intelligent temperature control of a wall-scraping crystallizer integrating multiple sensors, the method comprising: After the nitrate material enters the scraped wall crystallizer, the scraper is activated to mix it, and the real-time concentration and temperature of the nitrate material are monitored in real time by multiple sensors. The nitrate crystallization process is continuously monitored, and based on the real-time concentration and temperature values, the nucleation induction temperature node and the main growth temperature node are identified, and the corresponding actual nucleation induction temperature and actual main growth temperature are recorded and obtained. The actual nucleation induction temperature and the actual main growth temperature are input into the digital twin model, and the predicted particle size distribution and predicted scarring risk index are obtained synchronously through simulation. Based on the premise that the predicted scarring risk index meets the preset safety constraints, the temperature setpoint adjustment amount and scraper speed adjustment amount required to make the predicted particle size distribution approach the target particle size distribution are obtained through optimization calculation. Based on the temperature setpoint adjustment amount and the scraper rotation speed adjustment amount, the temperature setpoint and scraper rotation speed of the scraper crystallizer after the main growth temperature node are adjusted.

[0006] Secondly, this application provides an intelligent temperature control device for a wall-scraping crystallizer integrating multiple sensors, including: The material mixing module is used to start the scraper to mix the nitrate material after it enters the scraped crystallizer, and to monitor the real-time concentration and temperature of the nitrate material in real time through multiple sensors. The temperature monitoring module is used to continuously monitor the nitrate crystallization process, and based on the real-time concentration value and real-time temperature value, identify the nucleation induction temperature node and the main growth temperature node, record and obtain the corresponding actual nucleation induction temperature and actual main growth temperature. The model training module is used to input the actual nucleation induction temperature and the actual main growth temperature into the digital twin model, and synchronously output the predicted particle size distribution and the predicted scarring risk index through simulation. The adjustment calculation module is used to calculate, through optimization, the temperature setpoint adjustment amount and scraper speed adjustment amount required to make the predicted particle size distribution approach the target particle size distribution, based on the condition that the predicted scar risk index meets the preset safety constraints. The adjustment execution module is used to adjust the temperature setpoint and scraper speed of the scraper crystallizer after the main growth temperature node according to the temperature setpoint adjustment amount and scraper speed adjustment amount.

[0007] This application provides one or more technical solutions, which have at least the following technical effects or advantages: This application provides a method and apparatus for intelligent temperature control of a wall-scraped crystallizer that integrates multiple sensors. First, it uses multiple sensors to monitor the real-time concentration and temperature of nitrate materials, capturing dynamic changes in the material within the vessel. Second, it identifies nucleation induction temperature nodes and main growth temperature nodes based on the real-time concentration and temperature values, making the stage division of the crystallization process more scientific and reasonable. Then, it inputs the actual nucleation induction temperature and actual main growth temperature into a digital twin model, simultaneously outputting predicted particle size distribution and predicted scaling risk indicators, achieving accurate simulation and multi-objective prediction of the crystallization process. Next, based on the condition that the predicted scaling risk indicator meets preset safety constraints, it calculates the adjustment amount of the temperature setpoint and the scraper rotation speed through optimization calculations. Under the premise of ensuring production safety, it makes the predicted particle size distribution approach the target particle size distribution, effectively solving the problems of uneven crystal particle size distribution and severe scaling. Finally, it adjusts the temperature setpoint and scraper rotation speed of the wall-scraped crystallizer according to the adjustment amounts, realizing dynamic and intelligent control of the crystallization process, improving product quality and production efficiency, and enhancing the continuity and stability of production.

[0008] Through the above technical solution, this application can achieve precise and intelligent control of the temperature of the scraped crystallizer, effectively overcoming the shortcomings of traditional control methods that rely on experience and have poor adaptability, improving the product quality stability and production continuity of fine chemical products such as nitrates in the crystallization process, and reducing the energy consumption increase and downtime cleaning costs caused by scaling. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic flowchart of the intelligent temperature control method for a wall-scraping crystallizer integrating multiple sensors provided in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of the intelligent temperature control device for the wall-scraping crystallizer that integrates multiple sensors, provided in the embodiments of this application.

[0011] The components represented by each number in the attached diagram are explained below: Material mixing module 11, temperature monitoring module 12, model training module 13, adjustment amount calculation module 14, and adjustment execution module 15. Detailed Implementation

[0012] This application provides a method and apparatus for intelligent temperature control of a scraped-wall crystallizer that integrates multiple sensors. This method addresses the technical problems of existing scraped-wall crystallizer temperature control methods being unable to adapt to the dynamic process requirements during crystallization, resulting in uneven crystal particle size distribution, poor batch stability of products, and severe scaling on the inner wall of the crystallizer.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0015] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid unnecessarily obscuring the description of this application. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0016] Example 1, as Figure 1 As shown in the embodiments of this application, a method for intelligent temperature control of a wall-scraping crystallizer integrating multiple sensors is provided, including: S10: After the nitrate material enters the scraped crystallizer, the scraper is started to mix it, and the real-time concentration and temperature of the nitrate material are monitored in real time by multiple sensors. In this embodiment, a material flow sensor is installed at the feed inlet of the scraped crystallizer to monitor the feed rate of nitrate material in real time, so as to dynamically adjust the initial mixing intensity of the scraper according to the feed rate.

[0017] Specifically, the multi-sensor system includes at least three temperature sensors evenly distributed circumferentially along the inner wall of the crystallizer, and a concentration sensor located in the middle of the vessel. The temperature sensors cover the expected temperature range of the nitrate crystallization process with an accuracy of ±0.1℃. The concentration sensor uses the laser scattering principle to output the mass percentage concentration of the material in real time, with a response time of no more than 2 seconds. When the scraper is started, the initial rotation speed is set to 50-80 rpm to ensure that the material is initially uniformly mixed within 10-15 minutes. At this time, the multi-sensor system begins to synchronously collect data and transmits the data to the central control system via industrial Ethernet. The data sampling frequency is set to once per second to ensure timely capture of changes in the material's state.

[0018] Specifically, step S10 in the method includes: After the nitrate material is fed into the scraped-wall crystallizer, the scraper is activated to mix it. The real-time concentration of nitrate material in the scraped-wall crystallizer is measured using an online concentration meter. The real-time temperature values ​​of the nitrate material are simultaneously collected by an array of temperature sensors placed at different depths within the scraped-wall crystallizer.

[0019] In this embodiment, firstly, after the nitrate material enters the internal cavity of the scraped-wall crystallizer through the feed pipe, the central control system sends a start signal to the scraper drive motor. The drive motor drives the scraper to rotate at a preset initial speed, stirring and mixing the nitrate material entering the reactor. The initial speed setting needs to comprehensively consider factors such as the initial viscosity of the material, the feed rate, and the reactor volume, and is generally set in the range of 50 to 80 revolutions per minute. This speed range can ensure that the material forms a preliminary uniform mixture in a short time, usually 10 to 15 minutes, avoiding excessive local concentration or temperature differences that may adversely affect the subsequent crystallization process.

[0020] Simultaneously, the online concentration meter installed on the circulating pipeline of the crystallizer begins operation. This online concentration meter uses advanced laser scattering or ultrasonic attenuation principles to penetrate the pipeline and perform non-contact real-time measurement of the nitrate material flowing inside, thereby obtaining the real-time concentration value of the material. Its measurement accuracy can reach ±0.5%, and the response time is no more than 2 seconds.

[0021] Furthermore, an array of temperature sensors is deployed at different depths along the inner wall of the scraped-wall crystallizer. This array typically includes at least three high-precision platinum resistance temperature sensors. The sensors are installed along the axial direction of the vessel, near the bottom, the middle, and near the liquid surface, respectively. At the same depth, the sensors may also be evenly distributed circumferentially to eliminate temperature measurement deviations that might be caused by dead zones in the stirring. The temperature sensors cover the entire temperature range expected during the nitrate crystallization process, from initial mixing to final crystallization, with a measurement accuracy of ±0.1℃, ensuring accurate capture of temperature changes.

[0022] S20: Continuously monitor the nitrate crystallization process, and based on the real-time concentration value and real-time temperature value, identify the nucleation induction temperature node and the main growth temperature node, record and obtain the corresponding actual nucleation induction temperature and actual main growth temperature; In this embodiment of the application, the nitrate crystallization process is continuously monitored, and based on the real-time concentration value and real-time temperature value, the nucleation induction temperature node and the main growth temperature node are identified, and the corresponding actual nucleation induction temperature and actual main growth temperature are recorded and obtained.

[0023] Specifically, during continuous monitoring, the concentration and temperature values ​​collected in real time are dynamically analyzed. The identification of the nucleation induction temperature node primarily involves observing the rate of change of the real-time concentration value. When the nitrate material is cooled, its concentration gradually approaches its solubility curve. If a significant, sudden, small jump in the real-time concentration value occurs at this point, the temperature at that moment is identified as the nucleation induction temperature node, and the actual nucleation induction temperature at that instant is recorded.

[0024] Furthermore, the identification of the main growth temperature node considers the trend of temperature change and the stability of concentration change. After nucleation induction, the crystal enters the growth stage, at which point the solution concentration will continue to decrease slowly, but the rate of decrease gradually stabilizes. When it is monitored that during the continuous decrease of real-time temperature, the rate of concentration decrease remains within a relatively stable range for 5 consecutive minutes, for example, the fluctuation range of the concentration decrease rate is less than ±0.05% / minute, and the slope of the temperature curve is also basically stable, then the temperature at that moment is determined as the main growth temperature node, and the corresponding actual main growth temperature is recorded. The main growth temperature node marks the transition of the crystal from the nucleation stage to the main crystal growth stage.

[0025] Specifically, step S20 in the method includes: The supersaturation of nitrate materials is calculated using a thermodynamic model based on real-time concentration and temperature values. When the supersaturation first reaches the nucleation induction threshold preset for the nitrate crystallization process, the current moment is determined to be the nucleation induction temperature node, and the temperature of the nitrate material at this time is recorded as the actual nucleation induction temperature. After the nucleation induction temperature node, the nitrate crystallization process was continuously monitored. The change rate of crystal number was monitored and recorded in real time by a focused beam reflectometer, and the growth status of average crystal size was monitored and recorded in real time by a particle video microscope. When the rate of change of the number of crystals is lower than the preset rate of change threshold and the average crystal size growth enters a stable stage, the current moment is determined to be the main growth temperature node, and the temperature of the nitrate material at this time is recorded as the actual main growth temperature.

[0026] In this embodiment, firstly, based on real-time concentration and temperature values, and combined with nitrate solubility data, the supersaturation of the nitrate material is accurately calculated using a constructed thermodynamic model. This thermodynamic model considers the effects of temperature, concentration, and trace impurities that may exist in the solution on solubility. The saturation concentration under current conditions is obtained by looking up a table. Then, the difference between the real-time concentration and the saturation concentration is divided by the saturation concentration to obtain the supersaturation value.

[0027] Secondly, when the calculated supersaturation first reaches the preset nucleation induction threshold for the nitrate crystallization process, the current moment is determined as the nucleation induction temperature node. The nucleation induction threshold is determined by fitting previous experimental data based on the crystallization kinetics of a specific nitrate; its value is usually slightly higher than the theoretical nucleation point to ensure stable triggering of the nucleation process. At this time, the real-time temperature of the nitrate material is recorded as the actual nucleation induction temperature.

[0028] Secondly, after the nucleation induction temperature node, the crystallization process enters the crystal formation and growth stage. The nitrate crystallization process is continuously monitored. During this stage, two online monitoring devices are introduced: a focused beam reflectance meter and a particle video microscope. The focused beam reflectance meter emits a laser beam into the material and uses the reflection signal from the particles to monitor and record the rate of change in the number of crystals in real time, capturing the generation and aggregation of tiny crystals. The particle video microscope directly captures images of the crystals inside the reactor using a built-in camera, and through image analysis algorithms, monitors and records the growth status of the average crystal size in real time, intuitively reflecting the crystal growth trend.

[0029] Specifically, when the rate of change in the number of crystals monitored by the focused beam reflectometer is lower than the preset rate of change threshold for 3 consecutive minutes (e.g., the crystal number growth rate is less than 0.5% / minute), and the particle video microscope monitors that the average crystal size growth has entered a stable stage (i.e., the fluctuation range of the average crystal size growth rate is less than ±0.2μm / minute for 5 consecutive minutes), the current moment is determined to be the main growth temperature node. At this time, the real-time temperature of the nitrate material is recorded as the actual main growth temperature.

[0030] Furthermore, the determination of the main growth temperature node signifies that the crystallization process has smoothly transitioned from a stage dominated by nucleation to a stage dominated by crystal growth.

[0031] S30: Input the actual nucleation induction temperature and the actual main growth temperature into the digital twin model, and obtain the predicted particle size distribution and predicted scarring risk index through simulation. In this embodiment, the digital twin model is a multi-scale coupled model constructed based on the physical parameters of the scraped-wall crystallizer, the crystallization kinetics mechanism, and historical production data. The model first uses 3D modeling software to reconstruct the geometry of the crystallizer, including the diameter, height, jacket structure, scraper shape, and installation position. The dimensional accuracy of the geometric model is controlled within ±1mm to ensure the accuracy of the flow field simulation.

[0032] In terms of physical field modeling, heat transfer, mass transfer, fluid dynamics, and crystal nucleation and growth kinetics models are integrated. Simulations are performed by inputting the actual nucleation induction temperature and the actual main growth temperature into the digital twin model. During the nucleation induction stage, the model calculates the initial number and rate of nuclei based on the actual nucleation induction temperature and the corresponding supersaturation, predicting the formation and initial growth of crystal nuclei. After entering the main growth stage, the model simulates the crystal's trajectory, collisions, aggregation, and fragmentation behavior based on the actual main growth temperature and the flow field distribution. The evolution of crystal grain size distribution at different times is obtained by solving the PBE equation.

[0033] Meanwhile, the digital twin model predicts scar risk indicators by coupling a scar formation mechanism model. The scar formation mechanism model considers the effects of factors such as temperature gradient, supersaturation distribution, wall shear stress, and residence time on scar formation.

[0034] Specifically, step S30 in the method includes: The actual nucleation induction temperature and the actual main growth temperature are input into the digital twin model; In the digital twin model, the complete process from the nucleation induction temperature node to the end of crystallization is simulated based on the nitrate crystallization kinetic model, and the predicted particle size distribution is output. In the digital twin model, the scaling process on the inner wall of the scraped crystallizer is simulated based on a parallel scaling dynamics model, and the scaling risk index is output synchronously.

[0035] In this embodiment, the actual nucleation induction temperature and the actual main growth temperature are first used as key input parameters and imported into a pre-constructed and calibrated digital twin model. This digital twin model is constructed based on a 1:1 mapping of the physical entity of the scraped-wall crystallizer and includes precise physical parameters such as the vessel structure, scraper parameters, jacket heat transfer characteristics, as well as the thermodynamic and kinetic characteristics of the crystallization process.

[0036] In the digital twin model, the built-in nitrate crystallization kinetic model is first invoked. This kinetic model integrates the nucleation rate equation, crystal growth rate equation, and aggregation and fragmentation model for a specific nitrate. Starting from the actual nucleation induction temperature, and combining real-time monitoring of concentration and temperature changes, the model simulates the complete dynamic process from the nucleation induction temperature node, through the main growth stage, to the end of the crystallization process.

[0037] Simultaneously, within the digital twin model, a scaling dynamics model of the wall-scraping crystallizer is run in parallel. This model models the scaling process on the inner wall of the crystallizer, considering multiple factors such as the local supersaturation of the supersaturated solution near the wall during crystallization, the temperature difference between the wall and the bulk solution, the shear intensity of the scraper on the wall, and the residence time of the material on the wall.

[0038] In the digital twin model, the complete process from the nucleation induction temperature node to the end of crystallization is simulated based on the nitrate crystallization kinetics model, and the predicted particle size distribution is output, including: Using the actual nucleation induction temperature and the actual main growth temperature as boundaries, the crystallization process is divided into a nucleation stage and a main growth stage that proceed sequentially. In the nucleation stage, the supersaturation is calculated based on the actual nucleation induction temperature and the real-time concentration value. The crystal nucleation rate is calculated based on the supersaturation, and an initial crystal population containing information on the number of crystals and their initial size is simulated and generated. During the main growth stage, the crystal growth rate is calculated based on the actual main growth temperature and real-time concentration value, and the growth process of the initial crystal population is simulated to generate crystal size distribution data. Based on the crystal size distribution data, a predicted particle size distribution is output, which includes the proportion of crystals in different particle size ranges.

[0039] In this embodiment, the entire nitrate crystallization process is first clearly divided into a nucleation stage and a main growth stage, which proceed sequentially, using the actual nucleation induction temperature and the actual main growth temperature as boundary conditions. The nucleation stage begins after the nitrate material has completed initial mixing and multiple sensors have started to collect data stably, and ends at the identified actual nucleation induction temperature. Immediately afterward, the main growth stage begins and continues until the crystallization process is completed. Its starting point is the actual nucleation induction temperature, while the actual main growth temperature marks the period within this stage when crystal growth enters a stable and efficient phase.

[0040] During the nucleation stage, the digital twin model calculates the supersaturation of the nitrate material based on the actual nucleation induction temperature and the real-time concentration value collected and transmitted to the system in real time, using a pre-constructed thermodynamic model. Supersaturation is the key thermodynamic driving force for crystal nucleation, and the calculation formula follows the solubility characteristics of specific nitrates.

[0041] Next, based on the calculated supersaturation, the model calls the built-in nucleation rate equation, considering temperature, supersaturation, and possible impurity influence coefficients, to calculate the number of crystal nuclei generated per unit time and unit volume, i.e., the crystal nucleation rate. Subsequently, based on the nucleation rate and the duration of the nucleation stage, i.e., the time interval from the start of cooling to reaching the actual nucleation induction temperature, the model simulates and generates an initial crystal population containing detailed information such as the initial number of crystals and the initial grain size distribution of each crystal.

[0042] Furthermore, upon entering the main growth stage, the model, based on real-time monitoring of the actual main growth temperature and the corresponding real-time concentration value, and combined with the current supersaturation, calculates the linear growth rate or mass growth rate of the crystal using the crystal growth rate equation. The crystal growth rate equation considers factors such as the influence of temperature on the interfacial reaction rate, the solute diffusion coefficient, and supersaturation. Based on this growth rate, the model simulates the growth process of the initial crystal population generated during the nucleation stage. During the simulation, the motion of each crystal particle in the flow field, collisions with other particles, possible aggregation or fragmentation behaviors, and the dynamic changes in the growth rate under different temperature and concentration conditions are simulated. Through statistical analysis of a large number of crystal particles, time-varying crystal size distribution data is generated, recording the quantity or mass percentage of crystals within different particle size ranges.

[0043] Finally, based on the crystal size distribution data generated by the above simulation, the digital twin model outputs a predicted particle size distribution report containing multiple continuous particle size intervals and their corresponding crystal quantity percentages or mass percentages.

[0044] Furthermore, in the digital twin model, the scaling process on the inner wall of the scraped crystallizer is simulated based on a parallel scaling kinetic model, and the scaling risk indicators are output synchronously, including: Based on the actual nucleation induction temperature and the actual main growth temperature, the temperature distribution on the inner wall of the scraped crystallizer was obtained by heat conduction simulation calculation using the scaling dynamics model of the scraped crystallizer. Based on the temperature distribution and real-time concentration values, the local supersaturation at the inner wall of the scraped crystallizer is calculated. Based on the local supersaturation and the corresponding wall temperature, the crystal growth rate of the crystal layer in each region of the inner wall of the scraped crystallizer is calculated using the crystallization kinetic equation including the Arrhenius correction term. The scraping rate of the crystal layer by the scraper is calculated based on the scraper rotation speed and mechanical working parameters of the scraper in the wall-scraped crystallizer. The mechanical working parameters include at least the scraper contact pressure parameter. The difference between the crystal growth rate and the scraping rate is calculated to obtain the net cumulative rate, which characterizes the instantaneous growth trend of the crystal layer. Based on the net cumulative rate, the crystal layer thickness at different times is calculated by time integration, forming a dynamic change sequence of crystal layer thickness. Based on the dynamic change sequence of the crystal layer thickness, the cumulative percentage of time during which the crystal layer thickness exceeds a preset safety threshold is calculated as an indicator for predicting the risk of scarring.

[0045] In this embodiment, firstly, based on the actual nucleation induction temperature and the actual main growth temperature, the scaling kinetic model of the scraped crystallizer calculates the temperature distribution on the inner wall of the scraped crystallizer using a heat conduction simulation module. This simulation considers the heat transfer efficiency of the jacket, the thermal conductivity of the vessel material, and the influence of the latent heat of crystallization released during crystallization on the wall temperature. Specifically, the actual nucleation induction temperature and the actual main growth temperature are used as characteristic temperature boundary conditions for the bulk solution. Combined with the temperature and flow rate of the cooling / heating medium in the jacket, the heat conduction equation is solved using the finite element method to obtain the temperature values ​​at different locations on the inner wall of the crystallizer at different times, forming a continuous temperature distribution cloud map.

[0046] Secondly, the scaling kinetic model of the scraped crystallizer further calculates the local supersaturation at the inner wall surface based on the calculated temperature distribution of the inner wall and the real-time monitored concentration of the bulk solution. Since there may be slight differences between the wall temperature and the bulk solution temperature, especially in areas of intense heat exchange or with localized stagnant zones, the solubility of the solution near the wall may differ from that of the bulk solution. The model obtains the saturated concentration corresponding to the temperature at each point on the inner wall by querying the solubility data of the nitrate at different temperatures. Then, the difference between the real-time concentration value and the wall saturated concentration is divided by the wall saturated concentration to obtain the local supersaturation in each region of the inner wall.

[0047] Furthermore, the scaling kinetic model of the wall-scraping crystallizer calculates the crystal growth rate in each region of the wall surface using a crystallization kinetic equation incorporating an Arrhenius correction term, based on the local supersaturation and corresponding wall temperature in each region. This equation not only considers local supersaturation as a driving force for crystallization but also introduces the influence of temperature on the crystallization rate constant through the Arrhenius correction term. The formula is: Crystal growth rate = k0 × exp(-Ea / (R × Tw)) × (S-1) n Where k0 is the pre-exponential factor, Ea is the crystallization activation energy, R is the gas constant, Tw is the wall temperature, S is the local supersaturation, and n is the reaction order. The parameters k0, Ea, and n were determined through preliminary experiments for specific nitrates and crystallization vessel materials.

[0048] Then, the scaling kinetic model of the wall-scraping crystallizer calculates the scraping rate of the crystal layer by the scraper based on the scraper rotation speed and mechanical working parameters of the scraper. The scraper rotation speed directly affects the number of times the scraper scrapes a specific area of ​​the wall per unit time. The scraper contact pressure parameter in the mechanical working parameters determines the tightness of the contact between the scraper and the wall and the scraping force. The scraping rate is proportional to the scraper rotation speed and positively correlated with the scraper contact pressure, and can be obtained by fitting empirical formulas or experimental data, for example, scraping rate = a × N × P + b, where a and b are fitting coefficients, N is the scraper rotation speed, and P is the scraper contact pressure.

[0049] Subsequently, the scaling kinetic model of the wall-scraping crystallizer was used to calculate the difference between the crystal growth rate and the scraping rate, obtaining the net accumulation rate, which characterizes the instantaneous growth trend of the crystal layer. When the net accumulation rate is positive, it indicates that the crystal layer thickness in this region is increasing; when the net accumulation rate is negative or zero, it indicates that the crystal layer has been effectively scraped away or is in a dynamic equilibrium state.

[0050] Based on this net cumulative rate, the crystal layer thickness at different times can be calculated by integrating over time, forming dynamic sequence data of crystal layer thickness changing with time. For example, at a certain monitoring time point t, the crystal layer thickness is equal to the initial thickness plus the integral of the net cumulative rate from the initial time to time t over time.

[0051] Finally, the scaling kinetic model of the scraped crystallizer calculates the cumulative percentage of time when the crystal layer thickness exceeds a preset safety threshold based on the dynamic change sequence of the crystal layer thickness, serving as a predictive indicator of scaling risk. The preset safety threshold is set according to the crystallizer's heat transfer efficiency requirements, the maximum scraping capacity of the scraper, and production experience. For example, when the crystal layer thickness exceeds 0.5 mm, it may affect heat transfer or cause the scraper to jam. The model statistically analyzes the total time during which the crystal layer thickness in each region of the inner wall exceeds this safety threshold throughout the entire crystallization cycle, divides this percentage by the total crystallization time, and obtains the scaling risk indicator for that region.

[0052] Furthermore, by weighting and averaging the risk indicators of each region or taking the maximum value, a comprehensive predicted scaling risk indicator for the entire inner wall of the crystallizer can be obtained. Thus, the scaling kinetic model of the wall-scraping crystallizer transforms the traditionally difficult-to-quantify and experience-dependent scaling problem into a calculable and predictable process variable: on the one hand, it calculates the crystal growth rate based on the local supersaturation derived from the real-time process conditions; on the other hand, it calculates the scraping rate based on the equipment operating parameters. By comparing and integrating these two values ​​through a dynamic equilibrium equation, it can simulate the evolution sequence of the crystal thickness and obtain an accurate and quantifiable comprehensive predicted scaling risk indicator.

[0053] S40: Taking the predicted scar risk index as a condition that the preset safety constraints are met, the temperature setpoint adjustment amount and scraper speed adjustment amount required to make the predicted particle size distribution approach the target particle size distribution are obtained by optimization calculation. In this embodiment of the application, under the premise that the predicted risk index of scab does not exceed the preset safety threshold, the key operating parameters are adjusted by the optimization algorithm so that the predicted particle size distribution is as close as possible to the target particle size distribution required by production. That is, the adjustment amount of temperature setpoint and scraper speed is obtained by optimization.

[0054] Specifically, step S40 in the method includes: Based on the actual nucleation induction temperature, the actual main growth temperature, and the preset crystallization process curve, a benchmark process trajectory is constructed that extends from the main growth temperature node to the end of crystallization. Based on the benchmark process trajectory, within the preset operational constraints, an initial candidate parameter set is generated, which includes a combination of multiple temperature setpoint sequences and scraper speed setpoint sequences. Each set of parameter sequences in the initial candidate parameter set is sequentially input into the digital twin model to simulate the crystallization process after the main growth temperature node, and obtain the corresponding simulated predicted particle size distribution and simulated predicted scarring risk index. Based on the simulated predicted scar risk index and the simulated predicted granularity distribution, the initial candidate parameter set is comprehensively scored and ranked. Guided by the comprehensive score and ranking results, the top-ranked part of the initial candidate parameter set is retained as a high-quality solution. Based on the high-quality solution, a new candidate parameter set is generated by parameter recombination and the introduction of controllable perturbations, forming a new generation of candidate parameter set. The new generation of candidate parameter set is input into the digital twin model, and the process of simulating, scoring, sorting, and generating the new generation of candidate parameter set is repeated until the preset optimization termination condition is met. When the optimization termination condition is reached, the temperature setpoint sequence and scraper speed setpoint sequence corresponding to the candidate parameter set with the highest comprehensive score are compared with the corresponding actual temperature setpoint and actual scraper speed, respectively. The difference between the calculated temperature setpoint sequence is used as the temperature setpoint adjustment amount, and the difference between the scraper speed setpoint sequence is used as the scraper speed adjustment amount.

[0055] In this embodiment, firstly, based on the actual nucleation induction temperature, the actual main growth temperature, and a preset crystallization process curve, a baseline process trajectory extending from the main growth temperature node to the end of crystallization is constructed. The preset crystallization process curve is a standard temperature-time relationship curve pre-defined based on historical production experience, target product quality requirements, and equipment characteristics, specifying the ideal temperature change path from the start to the end of crystallization. The actual nucleation induction temperature and the actual main growth temperature are key node temperatures actually monitored and identified during the crystallization process of the current batch. The two actual temperature nodes are matched and calibrated with the corresponding stages in the preset process curve, and based on this, a smooth transition and extension are formed to create a complete baseline process trajectory starting from the main growth temperature node and covering subsequent cooling, heat preservation, and crystallization completion.

[0056] Secondly, based on the baseline process trajectory, within the preset operational constraints, an initial candidate parameter set is generated, comprising multiple combinations of temperature setpoint sequences and scraper speed setpoint sequences. The operational constraints are determined by the equipment's safe operating limits, process requirements, and historical optimization experience. For example, the adjustment range of the temperature setpoint may be limited to ±5℃ of the temperature at the corresponding moment in the baseline process trajectory, and the adjustment range of the scraper speed may be limited to 5-30 rpm. The initial candidate parameter set is generated using methods such as Latin hypercube sampling or uniform sampling to ensure that the parameter combinations have good distribution within the constraint space, covering possible optimization directions. For example, for the temperature setpoint sequence, several discrete adjustment options can be set at each time step of the baseline trajectory; similarly, different speed level options can be set for the scraper speed setpoint sequence. Then, the temperature and speed sequences are combined to form numerous initial candidate parameter combinations.

[0057] Next, each set of parameter sequences from the initial candidate parameter set is sequentially input into the digital twin model to simulate the crystallization process after the main growth temperature node. For each set of parameters, the temperature control strategy of the crystallizer and the scraper operation status are dynamically adjusted based on the input temperature setpoint sequence and scraper rotation speed setpoint sequence. During the simulation, the model calculates key variables such as supersaturation, crystal growth rate, crystal layer growth rate, and scraping rate in real time, and simulates and generates crystal size distribution data and dynamic change sequence of crystal layer thickness under the given parameter conditions, thereby obtaining the corresponding simulated predicted particle size distribution and simulated predicted scaling risk index.

[0058] Subsequently, the initial candidate parameter set is comprehensively scored and ranked based on the simulated predicted scarring risk index and the simulated predicted granularity distribution. The scoring system employs a multi-objective optimization strategy, whereby the simulated predicted scarring risk index must be strictly controlled below a preset safety threshold; if it exceeds the threshold, the parameter combination is directly eliminated. For parameter combinations that meet the scarring risk standard, the similarity between the simulated predicted granularity distribution and the target granularity distribution is then scored. Specifically, the root mean square error (RMSE) can be used to quantify the difference between the two; the smaller the difference, the higher the score for that item. The comprehensive score can be obtained by combining the scarring risk compliance status with the granularity distribution similarity score. For example, based on the scarring risk compliance status, the granularity distribution similarity score can be used as the primary ranking criterion.

[0059] Furthermore, guided by the comprehensive scoring and ranking results, the top-ranked portion of the initial candidate parameter set, such as the top 20%, is retained as high-quality solutions. Based on these high-quality solutions, a new set of candidate parameters is generated through parameter recombination and the introduction of controllable perturbations, forming a new generation of candidate parameter sets. Parameter recombination can employ the crossover operation in genetic algorithms; for example, the temperature setpoint sequences of two high-quality solutions can be cross-combined at a certain time point, or a similar crossover can be performed on the scraper speed setpoint sequence. Controllable perturbations involve introducing small random changes near the parameter values ​​of high-quality solutions; for example, increasing or decreasing a small random amount at each time step in the temperature setpoint sequence to explore a better parameter space.

[0060] Specifically, the next-generation candidate parameter set is input into the digital twin model, and the crystallization process simulation, scoring and ranking based on the simulation results, and generation of a new generation of candidate parameter sets are repeatedly executed. The iterative process continues until a preset optimization termination condition is met. The optimization termination condition may be reaching a preset maximum number of iterations, such as 50 iterations, or the highest value of the comprehensive score no longer significantly improves over multiple generations, such as an improvement of less than 0.1% over 5 consecutive generations.

[0061] Furthermore, when the optimization termination condition is met, the temperature setpoint sequence and scraper speed setpoint sequence corresponding to the candidate parameter set with the highest comprehensive score are compared with the currently executing actual temperature setpoint and actual scraper speed, respectively. The difference between the setpoint at each time step in the calculated temperature setpoint sequence and the corresponding actual setpoint is used as the temperature setpoint adjustment amount; similarly, the difference between the setpoint at each time step in the scraper speed setpoint sequence and the corresponding actual scraper speed is used as the scraper speed adjustment amount. These adjustment amounts will guide subsequent crystallization process control, achieving dynamic optimization of temperature and scraper speed.

[0062] Specifically, the initial candidate parameter set is comprehensively scored and ranked based on whether the simulated predicted scar risk index meets preset safety constraints and the degree of fit between the simulated predicted granularity distribution and the target granularity distribution, including: Calculate the relative error between the simulated predicted particle size distribution and the target particle size distribution in terms of target particle size characteristics, and use it as the first relative error; The relative error between the simulated predicted granularity distribution and the target granularity distribution in terms of distribution width characteristics is calculated as the second relative error; The first relative error and the second relative error are weighted and summed to obtain the granularity distribution matching score; Determine whether the simulated predicted scar risk index exceeds a preset safety threshold. If it does not exceed the threshold, the scar risk penalty score is zero. If it exceeds the threshold, calculate the value by which the simulated predicted scar risk index exceeds the preset safety threshold, and multiply the value by a preset penalty coefficient to obtain the scar risk penalty score. The difference between the granularity distribution matching score and the scar risk penalty score is calculated and used as the comprehensive score of the candidate parameter set; All candidate parameter sets are sorted in descending order based on the comprehensive score.

[0063] In this embodiment, firstly, the relative error between the simulated predicted particle size distribution and the target particle size distribution in terms of the target particle size characteristic is calculated, which is taken as the first relative error. The target particle size characteristic is usually selected as an indicator that has a significant impact on product quality, such as the median particle size D50. Specifically, the corresponding target particle size characteristic values ​​are extracted from the simulated predicted particle size distribution and the target particle size distribution, respectively, and denoted as Ds and Dt. Then, the first relative error is calculated by the formula (|Ds-Dt| / Dt)×100%. The smaller this error value, the closer the simulated predicted target particle size is to the expected target.

[0064] Secondly, the relative error between the simulated predicted granularity distribution and the target granularity distribution in terms of distribution width characteristics is calculated as the second relative error. The distribution width characteristic is used to measure the dispersion of the granularity distribution, such as the distribution width Span. The span Ss of the simulated predicted granularity distribution and the span St of the target granularity distribution are calculated separately, and then the second relative error is calculated using the formula (|Ss-St| / St)×100%. The smaller this error value, the better the concentration of the simulated predicted granularity distribution matches the target distribution.

[0065] Next, the first relative error and the second relative error are weighted and summed to obtain the granularity distribution matching score. To reflect the weight of different characteristics on product quality, different weight coefficients w1 and w2 can be assigned to the first and second relative errors according to process requirements or historical data, and w1 + w2 = 1.

[0066] For example, if the accuracy of the average particle size is of greater concern in production, w1 can be set to 0.6 and w2 to 0.4. The formula for calculating the particle size distribution matching score is: matching score = 100 - (w1 × first relative error + w2 × second relative error). The higher the score, the better the overall matching degree between the simulated predicted particle size distribution and the target distribution.

[0067] Next, it is determined whether the simulated predicted scaling risk index exceeds the preset safety threshold. The preset safety threshold is a critical value for scaling risk set based on factors such as the heat transfer efficiency of the crystallizer and the safety of scraper operation. For example, it is set to 5%, meaning that the cumulative time percentage of the crystal layer thickness exceeding the safety threshold does not exceed 5%. If the simulated predicted scaling risk index does not exceed this threshold, the scaling risk penalty score is zero, indicating that the parameter combination meets the safety requirements in terms of scaling control. If it exceeds the threshold, the value by which the simulated predicted scaling risk index exceeds the preset safety threshold is calculated, i.e., excess amount = simulated value - threshold. This excess amount is then multiplied by the preset penalty coefficient k to obtain the scaling risk penalty score. The value of the penalty coefficient k is determined according to the degree of harm of scaling risk to production. The greater the harm, the higher the value of k. For example, it is set to 10. In this case, the penalty score = k × excess amount, and this score will be directly deducted from the particle size distribution matching score.

[0068] Finally, the difference between the granularity distribution matching score and the scarring risk penalty score is calculated as the comprehensive score for the candidate parameter set, i.e., Comprehensive Score = Granularity Distribution Matching Score - Scarring Risk Penalty Score. The comprehensive score considers both granularity distribution quality and scarring risk control, providing a comprehensive evaluation of the candidate parameter set's merits. All candidate parameter sets are then sorted in descending order based on the comprehensive score. The higher-ranked candidate parameter sets indicate that they achieve a granularity distribution close to the target while ensuring controllable scarring risk.

[0069] Furthermore, guided by the scoring and ranking results, the top-ranked candidate parameters in the initial candidate parameter set are retained as high-quality solutions. Based on these high-quality solutions, a new set of candidate parameters is generated through parameter recombination and the introduction of controllable perturbations, forming a new generation of candidate parameter sets, including: Calculate the standard deviation of the comprehensive score of all candidate parameter sets, and dynamically determine the retention ratio based on the standard deviation, wherein the retention ratio is inversely proportional to the magnitude of the standard deviation; Based on the ranking results of the comprehensive score, the candidate parameter set with the retention ratio mentioned above is retained as the high-quality solution; Two candidate parameter sets are randomly selected from the high-quality solutions as the first parent parameter set and the second parent parameter set; A random exchange time point is generated within the time range of the crystallization process. At the exchange time point, the subsequent segments of the temperature setpoint sequence and the scraper speed setpoint sequence of the first parent parameter set and the second parent parameter set are exchanged to generate two new child parameter sets. A controllable perturbation is applied to the two new child parameter sets and the high-quality solution, wherein the controllable perturbation is implemented by the perturbation amplitude, which decreases with the number of iterations and increases when the optimization stalls; The parameter set after applying controllable perturbation is combined with a new temperature setpoint sequence and a new scraper speed setpoint sequence randomly generated according to the retention ratio to form a new generation of candidate parameter set.

[0070] In this embodiment, firstly, the standard deviation of the comprehensive score of all current candidate parameter sets is calculated, and the retention ratio is dynamically determined based on this standard deviation. The retention ratio is inversely proportional to the magnitude of the standard deviation. Specifically, the standard deviation reflects the dispersion of the comprehensive score of the current candidate parameter sets. The larger the standard deviation, the more significant the performance differences between the parameter sets. In this case, the retention ratio can be appropriately reduced, for example, retaining the top 10%-15% of the parameter sets as high-quality solutions to centrally select the better-performing individuals. If the standard deviation is small, it indicates that the overall performance of the parameter sets is relatively similar. To avoid getting trapped in local optima, the retention ratio can be appropriately increased, such as retaining the top 25%-30% of the parameter sets to retain more diverse high-quality solutions.

[0071] Secondly, based on the descending sorting results of the comprehensive scores, the candidate parameter set with the retention ratio mentioned above is retained as the high-quality solution. For example, if the current candidate parameter set has a total of 100 parameters and the dynamically determined retention ratio is 20%, then the parameter set with the top 20 comprehensive scores is selected as the high-quality solution for this iteration. The high-quality solution represents the better solution in the current parameter space.

[0072] Next, two candidate parameter sets are randomly selected from the high-quality solutions to serve as the first and second parent parameter sets. This random selection increases the diversity of gene combinations and avoids the optimization direction becoming singular due to fixed pairings.

[0073] Then, a random exchange time point is generated within the time range of the crystallization process. For example, assuming the total crystallization process duration after the main growth temperature node is 100 minutes, a random exchange time point of 30 minutes is generated. At this exchange time point, the subsequent segments of the temperature setpoint sequence and scraper speed setpoint sequence of the first parent parameter set and the second parent parameter set are exchanged.

[0074] Specifically, the temperature and rotation speed sequences of the first parent parameter set remain unchanged before the exchange time point, while the sequences after the exchange time point are replaced with the sequences of the corresponding time segments of the second parent parameter set; and vice versa, thus generating two new child parameter sets.

[0075] Subsequently, a controlled perturbation is applied to the two new child parameter sets and the high-quality solution. The controlled perturbation is implemented through the perturbation amplitude, which decreases with the number of iterations and increases when the optimization stalls.

[0076] For example, in the initial optimization stage, the number of iterations is small, and the perturbation amplitude can be set relatively large, such as ±2℃ for the temperature setpoint sequence and ±5 rpm for the scraper speed setpoint sequence, to promote extensive exploration of the parameter space. As the number of iterations increases, the perturbation amplitude gradually decreases, such as reducing the temperature perturbation amplitude to ±0.5℃ and the speed perturbation amplitude to ±1 rpm, to finely adjust the parameters and converge towards the optimal solution. If the highest value of the comprehensive score does not significantly improve over multiple generations, indicating optimization stagnation, the perturbation amplitude is temporarily increased, such as restoring it to the initial larger perturbation level, to escape the local optimum and explore new parameter combinations.

[0077] Finally, the parameter set after controlled perturbation, including the offspring parameter set and the perturbated high-quality solutions, is combined with a new set of temperature setpoints and a new set of scraper speed setpoints randomly generated according to the retention ratio to form a new generation of candidate parameter sets. Introducing new parameter combinations is to maintain population diversity and prevent the optimization process from prematurely converging due to an over-concentration of high-quality solutions.

[0078] For example, if the retention ratio is 20% and the total number of the new generation of candidate parameter sets is the same as the initial candidate parameter sets, then the number of new parameter combinations can account for 20% of the total number of the new generation of candidate parameter sets.

[0079] S50: Adjust the temperature setpoint and scraper speed of the scraper crystallizer after the main growth temperature node according to the temperature setpoint adjustment amount and scraper speed adjustment amount.

[0080] In this embodiment, the aforementioned optimized temperature setpoint adjustment and scraper rotation speed adjustment are converted into instructions that act on the actual control actuator of the scraper crystallizer.

[0081] Specifically, after the digital twin model completes iterative optimization and outputs the temperature setpoint sequence and scraper speed setpoint sequence corresponding to the candidate parameter set with the highest comprehensive score, the optimized setpoint sequence is compared with the actual temperature setpoint and actual scraper speed setpoint used in the crystallizer during the current execution, i.e., the previous control cycle, step by step. For the temperature setpoint, the difference between the optimized setpoint and the current actual setpoint is calculated at each corresponding time step, and this difference is the temperature setpoint adjustment amount for that time step.

[0082] Similarly, for the scraper speed setpoint sequence, the difference between it and the current actual scraper speed at each corresponding time step is calculated and used as the scraper speed adjustment amount. This adjustment amount serves as the basis for adjusting the temperature control system and scraper drive system in the future, such as in the next control cycle.

[0083] In summary, compared to existing technologies, this application constructs a closed-loop mechanism from parameter optimization to dynamic control by deeply integrating multi-sensor data with a digital twin model. High-precision simulation of the crystallization process is achieved through the digital twin model, combined with a parameter optimization strategy using a genetic algorithm. This ensures that the target granularity distribution is matched while maintaining controllable scarring risk, overcoming the problems of low efficiency and blind parameter adjustment inherent in traditional trial-and-error methods. Furthermore, a comprehensive scoring mechanism is introduced, combining granularity distribution matching with scarring risk penalties, achieving synergistic optimization of quality objectives and production safety. The dynamically adjusted retention ratio and perturbation amplitude strategy enhances the adaptability and global search capability of the parameter optimization process, effectively avoiding local optima traps.

[0084] In summary, the embodiments of this application have at least the following technical effects: This application provides an intelligent temperature control method for a wall-scraping crystallizer that integrates multiple sensors. First, it uses multiple sensors to monitor the real-time concentration and temperature of nitrate materials, capturing dynamic changes in the material within the vessel. Second, it identifies nucleation induction temperature nodes and main growth temperature nodes based on the real-time concentration and temperature values, making the stage division of the crystallization process more scientific and reasonable. Then, it inputs the actual nucleation induction temperature and actual main growth temperature into a digital twin model, simultaneously outputting predicted particle size distribution and predicted scaling risk indicators, achieving accurate simulation and multi-objective prediction of the crystallization process. Next, based on the condition that the predicted scaling risk indicator meets preset safety constraints, it calculates the adjustment amount of the temperature setpoint and the scraper rotation speed through optimization calculations. Under the premise of ensuring production safety, it makes the predicted particle size distribution approach the target particle size distribution, effectively solving the problems of uneven crystal particle size distribution and severe scaling. Finally, it adjusts the temperature setpoint and scraper rotation speed of the wall-scraping crystallizer according to the adjustment amounts, realizing dynamic and intelligent control of the crystallization process, improving product quality and production efficiency, and enhancing the continuity and stability of production.

[0085] Through the above technical solution, this application can achieve precise and intelligent control of the temperature of the scraped crystallizer, effectively overcoming the shortcomings of traditional control methods that rely on experience and have poor adaptability, improving the product quality stability and production continuity of fine chemical products such as nitrates in the crystallization process, and reducing the energy consumption increase and downtime cleaning costs caused by scaling.

[0086] Example 2, as Figure 2 As shown, based on the same inventive concept as the intelligent temperature control method for a wall-scraping crystallizer integrating multiple sensors provided in Embodiment 1, this application also provides an intelligent temperature control device for a wall-scraping crystallizer integrating multiple sensors, including: The material mixing module 11 is used to start the scraper to mix the nitrate material after it enters the scraped crystallizer, and to monitor the real-time concentration and temperature of the nitrate material in real time through multiple sensors. Temperature monitoring module 12 is used to continuously monitor the nitrate crystallization process, and based on the real-time concentration value and real-time temperature value, identify the nucleation induction temperature node and the main growth temperature node, record and obtain the corresponding actual nucleation induction temperature and actual main growth temperature. Model training module 13 is used to input the actual nucleation induction temperature and the actual main growth temperature into the digital twin model, and synchronously output the predicted particle size distribution and the predicted scarring risk index through simulation. The adjustment calculation module 14 is used to calculate, through optimization, the temperature setpoint adjustment amount and the scraper speed adjustment amount required to make the predicted particle size distribution approach the target particle size distribution, based on the condition that the predicted scar risk index meets the preset safety constraints. The adjustment execution module 15 is used to adjust the temperature setpoint and scraper speed of the scraper crystallizer after the main growth temperature node according to the temperature setpoint adjustment amount and scraper speed adjustment amount.

[0087] In one embodiment, the material mixing module 11 is specifically used for: After the nitrate material is fed into the scraped-wall crystallizer, the scraper is activated to mix it. The real-time concentration of nitrate material in the scraped-wall crystallizer is measured using an online concentration meter. The real-time temperature values ​​of the nitrate material are simultaneously collected by an array of temperature sensors placed at different depths within the scraped-wall crystallizer.

[0088] In one embodiment, the temperature monitoring module 12 is specifically used for: The supersaturation of nitrate materials is calculated using a thermodynamic model based on real-time concentration and temperature values. When the supersaturation first reaches the nucleation induction threshold preset for the nitrate crystallization process, the current moment is determined to be the nucleation induction temperature node, and the temperature of the nitrate material at this time is recorded as the actual nucleation induction temperature. After the nucleation induction temperature node, the nitrate crystallization process was continuously monitored. The change rate of crystal number was monitored and recorded in real time by a focused beam reflectometer, and the growth status of average crystal size was monitored and recorded in real time by a particle video microscope. When the rate of change of the number of crystals is lower than the preset rate of change threshold and the average crystal size growth enters a stable stage, the current moment is determined to be the main growth temperature node, and the temperature of the nitrate material at this time is recorded as the actual main growth temperature.

[0089] In one embodiment, the model training module is specifically used for: The actual nucleation induction temperature and the actual main growth temperature are input into the digital twin model; In the digital twin model, the complete process from the nucleation induction temperature node to the end of crystallization is simulated based on the nitrate crystallization kinetic model, and the predicted particle size distribution is output. In the digital twin model, the scaling process on the inner wall of the scraped crystallizer is simulated based on a parallel scaling dynamics model, and the scaling risk index is output synchronously.

[0090] Furthermore, in one embodiment of the application, the digital twin model simulates the entire process from the nucleation induction temperature node to the end of crystallization based on a nitrate crystallization kinetic model, and outputs a predicted particle size distribution, including: Using the actual nucleation induction temperature and the actual main growth temperature as boundaries, the crystallization process is divided into a nucleation stage and a main growth stage that proceed sequentially. In the nucleation stage, the supersaturation is calculated based on the actual nucleation induction temperature and the real-time concentration value. The crystal nucleation rate is calculated based on the supersaturation, and an initial crystal population containing information on the number of crystals and their initial size is simulated and generated. During the main growth stage, the crystal growth rate is calculated based on the actual main growth temperature and real-time concentration value, and the growth process of the initial crystal population is simulated to generate crystal size distribution data. Based on the crystal size distribution data, a predicted particle size distribution is output, which includes the proportion of crystals in different particle size ranges.

[0091] Furthermore, in one embodiment of the application, in the digital twin model, the scaling process on the inner wall surface of the scraped crystallizer is simulated based on a parallel scaling kinetic model, and a predictive scaling risk index is output simultaneously, including: Based on the actual nucleation induction temperature and the actual main growth temperature, the temperature distribution on the inner wall of the scraped crystallizer was obtained by heat conduction simulation calculation using the scaling dynamics model of the scraped crystallizer. Based on the temperature distribution and real-time concentration values, the local supersaturation at the inner wall of the scraped crystallizer is calculated. Based on the local supersaturation and the corresponding wall temperature, the crystal growth rate of the crystal layer in each region of the inner wall of the scraped crystallizer is calculated using the crystallization kinetic equation including the Arrhenius correction term. The scraping rate of the crystal layer by the scraper is calculated based on the scraper rotation speed and mechanical working parameters of the scraper in the wall-scraped crystallizer. The mechanical working parameters include at least the scraper contact pressure parameter. The difference between the crystal growth rate and the scraping rate is calculated to obtain the net cumulative rate, which characterizes the instantaneous growth trend of the crystal layer. Based on the net cumulative rate, the crystal layer thickness at different times is calculated by time integration, forming a dynamic change sequence of crystal layer thickness. Based on the dynamic change sequence of the crystal layer thickness, the cumulative percentage of time during which the crystal layer thickness exceeds a preset safety threshold is calculated as an indicator for predicting the risk of scarring.

[0092] In one embodiment, the adjustment calculation module 14 is specifically used for: Based on the actual nucleation induction temperature, the actual main growth temperature, and the preset crystallization process curve, a benchmark process trajectory is constructed that extends from the main growth temperature node to the end of crystallization. Based on the benchmark process trajectory, within the preset operational constraints, an initial candidate parameter set is generated, which includes a combination of multiple temperature setpoint sequences and scraper speed setpoint sequences. Each set of parameter sequences in the initial candidate parameter set is sequentially input into the digital twin model to simulate the crystallization process after the main growth temperature node, and obtain the corresponding simulated predicted particle size distribution and simulated predicted scarring risk index. Based on the simulated predicted scar risk index and the simulated predicted granularity distribution, the initial candidate parameter set is comprehensively scored and ranked. Guided by the comprehensive score and ranking results, the top-ranked part of the initial candidate parameter set is retained as a high-quality solution. Based on the high-quality solution, a new candidate parameter set is generated by parameter recombination and the introduction of controllable perturbations, forming a new generation of candidate parameter set. The new generation of candidate parameter set is input into the digital twin model, and the process of simulating, scoring, sorting, and generating the new generation of candidate parameter set is repeated until the preset optimization termination condition is met. When the optimization termination condition is reached, the temperature setpoint sequence and scraper speed setpoint sequence corresponding to the candidate parameter set with the highest comprehensive score are compared with the corresponding actual temperature setpoint and actual scraper speed, respectively. The difference between the calculated temperature setpoint sequence is used as the temperature setpoint adjustment amount, and the difference between the scraper speed setpoint sequence is used as the scraper speed adjustment amount.

[0093] Furthermore, based on whether the simulated predicted scar risk index meets preset safety constraints and the degree of fit between the simulated predicted granularity distribution and the target granularity distribution, the initial candidate parameter set is comprehensively scored and ranked, including: Calculate the relative error between the simulated predicted particle size distribution and the target particle size distribution in terms of target particle size characteristics, and use it as the first relative error; The relative error between the simulated predicted granularity distribution and the target granularity distribution in terms of distribution width characteristics is calculated as the second relative error; The first relative error and the second relative error are weighted and summed to obtain the granularity distribution matching score; Determine whether the simulated predicted scar risk index exceeds a preset safety threshold. If it does not exceed the threshold, the scar risk penalty score is zero. If it exceeds the threshold, calculate the value by which the simulated predicted scar risk index exceeds the preset safety threshold, and multiply the value by a preset penalty coefficient to obtain the scar risk penalty score. The difference between the granularity distribution matching score and the scar risk penalty score is calculated and used as the comprehensive score of the candidate parameter set; All candidate parameter sets are sorted in descending order based on the comprehensive score.

[0094] Furthermore, guided by the scoring and ranking results, the top-ranked candidate parameters in the initial candidate parameter set are retained as high-quality solutions. Based on these high-quality solutions, a new set of candidate parameters is generated through parameter recombination and the introduction of controllable perturbations, forming a new generation of candidate parameter sets, including: Calculate the standard deviation of the comprehensive score of all candidate parameter sets, and dynamically determine the retention ratio based on the standard deviation, wherein the retention ratio is inversely proportional to the magnitude of the standard deviation; Based on the ranking results of the comprehensive score, the candidate parameter set with the retention ratio mentioned above is retained as the high-quality solution; Two candidate parameter sets are randomly selected from the high-quality solutions as the first parent parameter set and the second parent parameter set; A random exchange time point is generated within the time range of the crystallization process. At the exchange time point, the subsequent segments of the temperature setpoint sequence and the scraper speed setpoint sequence of the first parent parameter set and the second parent parameter set are exchanged to generate two new child parameter sets. A controllable perturbation is applied to the two new child parameter sets and the high-quality solution, wherein the controllable perturbation is implemented by the perturbation amplitude, which decreases with the number of iterations and increases when the optimization stalls; The parameter set after applying controllable perturbation is combined with a new temperature setpoint sequence and a new scraper speed setpoint sequence randomly generated according to the retention ratio to form a new generation of candidate parameter set.

[0095] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0096] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0097] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for intelligent temperature control of a wall-scraping crystallizer integrating multiple sensors, characterized in that, The method includes: After the nitrate material enters the scraped crystallizer, the scraper is activated to mix it, and the real-time concentration and temperature of the nitrate material are monitored in real time by multiple sensors. The nitrate crystallization process is continuously monitored, and based on the real-time concentration and temperature values, the nucleation induction temperature node and the main growth temperature node are identified, and the corresponding actual nucleation induction temperature and actual main growth temperature are recorded and obtained. The actual nucleation induction temperature and the actual main growth temperature are input into the digital twin model, and the predicted particle size distribution and predicted scarring risk index are obtained synchronously through simulation. Based on the premise that the predicted scarring risk index meets the preset safety constraints, the temperature setpoint adjustment amount and scraper speed adjustment amount required to make the predicted particle size distribution approach the target particle size distribution are obtained through optimization calculation. Based on the temperature setpoint adjustment amount and the scraper rotation speed adjustment amount, the temperature setpoint and scraper rotation speed of the scraper crystallizer after the main growth temperature node are adjusted.

2. The intelligent temperature control method for a wall-scraping crystallizer integrating multiple sensors according to claim 1, characterized in that, After the nitrate material enters the scraped-wall crystallizer, the scraper is activated for mixing, and the real-time concentration and temperature of the nitrate material are monitored in real time by multiple sensors, including: After the nitrate material is fed into the scraped-wall crystallizer, the scraper is activated to mix it. The real-time concentration of nitrate material in the scraped-wall crystallizer is measured using an online concentration meter. The real-time temperature values ​​of the nitrate material are simultaneously collected by an array of temperature sensors placed at different depths within the scraped-wall crystallizer.

3. The intelligent temperature control method for a wall-scraping crystallizer integrating multiple sensors according to claim 1, characterized in that, The nitrate crystallization process is continuously monitored, and based on the real-time concentration and temperature values, nucleation induction temperature nodes and main growth temperature nodes are identified. The corresponding actual nucleation induction temperature and actual main growth temperature are recorded and obtained, including: The supersaturation of nitrate materials is calculated using a thermodynamic model based on real-time concentration and temperature values. When the supersaturation first reaches the nucleation induction threshold preset for the nitrate crystallization process, the current moment is determined to be the nucleation induction temperature node, and the temperature of the nitrate material at this time is recorded as the actual nucleation induction temperature. After the nucleation induction temperature node, the nitrate crystallization process was continuously monitored. The change rate of crystal number was monitored and recorded in real time by a focused beam reflectometer, and the growth status of average crystal size was monitored and recorded in real time by a particle video microscope. When the rate of change of the number of crystals is lower than the preset rate of change threshold and the average crystal size growth enters a stable stage, the current moment is determined to be the main growth temperature node, and the temperature of the nitrate material at this time is recorded as the actual main growth temperature.

4. The intelligent temperature control method for a wall-scraping crystallizer integrating multiple sensors according to claim 1, characterized in that, The actual nucleation induction temperature and the actual main growth temperature are input into a digital twin model, and the predicted particle size distribution and predicted scarring risk indicators are output synchronously, including: The actual nucleation induction temperature and the actual main growth temperature are input into the digital twin model; In the digital twin model, the complete process from the nucleation induction temperature node to the end of crystallization is simulated based on the nitrate crystallization kinetic model, and the predicted particle size distribution is output. In the digital twin model, the scaling process on the inner wall of the scraped crystallizer is simulated based on a parallel scaling dynamics model, and the scaling risk index is output synchronously.

5. The intelligent temperature control method for a wall-scraping crystallizer integrating multiple sensors according to claim 4, characterized in that, In the digital twin model, the complete process from the nucleation induction temperature node to the end of crystallization is simulated based on the nitrate crystallization kinetics model, and the predicted particle size distribution is output, including: Using the actual nucleation induction temperature and the actual main growth temperature as boundaries, the crystallization process is divided into a nucleation stage and a main growth stage that proceed sequentially. In the nucleation stage, the supersaturation is calculated based on the actual nucleation induction temperature and the real-time concentration value. The crystal nucleation rate is calculated based on the supersaturation, and an initial crystal population containing information on the number of crystals and their initial size is simulated and generated. During the main growth stage, the crystal growth rate is calculated based on the actual main growth temperature and real-time concentration value, and the growth process of the initial crystal population is simulated to generate crystal size distribution data. Based on the crystal size distribution data, a predicted particle size distribution is output, which includes the proportion of crystals in different particle size ranges.

6. The intelligent temperature control method for a wall-scraping crystallizer integrating multiple sensors according to claim 4, characterized in that, In the digital twin model, the scaling process on the inner wall of the scraped crystallizer is simulated based on a parallel scaling kinetic model, and the scaling risk indicators are output synchronously, including: Based on the actual nucleation induction temperature and the actual main growth temperature, the temperature distribution on the inner wall of the scraped crystallizer was obtained by heat conduction simulation calculation using the scaling dynamics model of the scraped crystallizer. Based on the temperature distribution and real-time concentration values, the local supersaturation at the inner wall of the scraped crystallizer is calculated. Based on the local supersaturation and the corresponding wall temperature, the crystal growth rate of the crystal layer in each region of the inner wall of the scraped crystallizer is calculated using the crystallization kinetic equation including the Arrhenius correction term. The scraping rate of the crystal layer by the scraper is calculated based on the scraper rotation speed and mechanical working parameters of the scraper in the wall-scraped crystallizer. The mechanical working parameters include at least the scraper contact pressure parameter. The difference between the crystal growth rate and the scraping rate is calculated to obtain the net cumulative rate, which characterizes the instantaneous growth trend of the crystal layer. Based on the net cumulative rate, the crystal layer thickness at different times is calculated by time integration, forming a dynamic change sequence of crystal layer thickness. Based on the dynamic change sequence of the crystal layer thickness, the cumulative percentage of time during which the crystal layer thickness exceeds a preset safety threshold is calculated as an indicator for predicting the risk of scarring.

7. The intelligent temperature control method for a wall-scraping crystallizer integrating multiple sensors according to claim 1, characterized in that, Based on the premise that the predicted scarring risk index meets the preset safety constraints, the adjustment amounts of the temperature setpoint and scraper rotation speed required to make the predicted particle size distribution approach the target particle size distribution are obtained through optimization calculations, including: Based on the actual nucleation induction temperature, the actual main growth temperature, and the preset crystallization process curve, a benchmark process trajectory is constructed that extends from the main growth temperature node to the end of crystallization. Based on the benchmark process trajectory, within the preset operational constraints, an initial candidate parameter set is generated, which includes a combination of multiple temperature setpoint sequences and scraper speed setpoint sequences. Each set of parameter sequences in the initial candidate parameter set is sequentially input into the digital twin model to simulate the crystallization process after the main growth temperature node, and obtain the corresponding simulated predicted particle size distribution and simulated predicted scarring risk index. Based on the simulated predicted scar risk index and the simulated predicted granularity distribution, the initial candidate parameter set is comprehensively scored and ranked. Guided by the comprehensive score and ranking results, the top-ranked part of the initial candidate parameter set is retained as a high-quality solution. Based on the high-quality solution, a new candidate parameter set is generated by parameter recombination and the introduction of controllable perturbations, forming a new generation of candidate parameter set. The new generation of candidate parameter set is input into the digital twin model, and the process of simulating, scoring, sorting, and generating the new generation of candidate parameter set is repeated until the preset optimization termination condition is met. When the optimization termination condition is reached, the temperature setpoint sequence and scraper speed setpoint sequence corresponding to the candidate parameter set with the highest comprehensive score are compared with the corresponding actual temperature setpoint and actual scraper speed, respectively. The difference between the calculated temperature setpoint sequence is used as the temperature setpoint adjustment amount, and the difference between the scraper speed setpoint sequence is used as the scraper speed adjustment amount.

8. The intelligent temperature control method for a wall-scraping crystallizer integrating multiple sensors according to claim 7, characterized in that, Based on whether the simulated and predicted scarring risk indicators meet preset safety constraints and the degree of fit between the simulated and predicted granularity distribution and the target granularity distribution, the initial candidate parameter set is comprehensively scored and ranked, including: Calculate the relative error between the simulated predicted particle size distribution and the target particle size distribution in terms of target particle size characteristics, and use it as the first relative error; The relative error between the simulated predicted granularity distribution and the target granularity distribution in terms of distribution width characteristics is calculated as the second relative error; The first relative error and the second relative error are weighted and summed to obtain the granularity distribution matching score; Determine whether the simulated predicted scar risk index exceeds a preset safety threshold. If it does not exceed the threshold, the scar risk penalty score is zero. If it exceeds the threshold, calculate the value by which the simulated predicted scar risk index exceeds the preset safety threshold, and multiply the value by a preset penalty coefficient to obtain the scar risk penalty score. The difference between the granularity distribution matching score and the scar risk penalty score is calculated and used as the comprehensive score of the candidate parameter set; All candidate parameter sets are sorted in descending order based on the comprehensive score.

9. The intelligent temperature control method for a wall-scraping crystallizer integrating multiple sensors according to claim 7, characterized in that, Guided by the scoring and ranking results, the top-ranked candidate parameters in the initial candidate parameter set are retained as high-quality solutions. Based on these high-quality solutions, a new set of candidate parameters is generated through parameter recombination and the introduction of controllable perturbations, forming a new generation of candidate parameter sets, including: Calculate the standard deviation of the comprehensive score of all candidate parameter sets, and dynamically determine the retention ratio based on the standard deviation, wherein the retention ratio is inversely proportional to the magnitude of the standard deviation; Based on the ranking results of the comprehensive score, the candidate parameter set with the retention ratio mentioned above is retained as the high-quality solution; Two candidate parameter sets are randomly selected from the high-quality solutions as the first parent parameter set and the second parent parameter set; A random exchange time point is generated within the time range of the crystallization process. At the exchange time point, the subsequent segments of the temperature setpoint sequence and the scraper speed setpoint sequence of the first parent parameter set and the second parent parameter set are exchanged to generate two new child parameter sets. A controllable perturbation is applied to the two new child parameter sets and the high-quality solution, wherein the controllable perturbation is implemented by the perturbation amplitude, which decreases with the number of iterations and increases when the optimization stalls; The parameter set after applying controllable perturbation is combined with a new temperature setpoint sequence and a new scraper speed setpoint sequence randomly generated according to the retention ratio to form a new generation of candidate parameter set.

10. A multi-sensor integrated intelligent temperature control device for a wall-scraping crystallizer, characterized in that, The method for intelligent temperature control of a wall-scraping crystallizer integrating multiple sensors as described in any one of claims 1-9 includes: The material mixing module is used to start the scraper to mix the nitrate material after it enters the scraped crystallizer, and to monitor the real-time concentration and temperature of the nitrate material in real time through multiple sensors. The temperature monitoring module is used to continuously monitor the nitrate crystallization process, and based on the real-time concentration value and real-time temperature value, identify the nucleation induction temperature node and the main growth temperature node, record and obtain the corresponding actual nucleation induction temperature and actual main growth temperature. The model training module is used to input the actual nucleation induction temperature and the actual main growth temperature into the digital twin model, and synchronously output the predicted particle size distribution and the predicted scarring risk index through simulation. The adjustment calculation module is used to calculate, through optimization, the temperature setpoint adjustment amount and scraper speed adjustment amount required to make the predicted particle size distribution approach the target particle size distribution, based on the condition that the predicted scar risk index meets the preset safety constraints. The adjustment execution module is used to adjust the temperature setpoint and scraper speed of the scraper crystallizer after the main growth temperature node according to the temperature setpoint adjustment amount and scraper speed adjustment amount.

Images