A method and system for controlling the particle size of an adc blowing agent

Abstract

This application provides a particle size control method and system for ADC foaming agents, belonging to the field of ADC foaming agent production technology. The control method includes: generating initial control parameters based on a pre-trained particle size-process parameter mapping model according to the target particle size; real-time acquisition of multi-modal parameters of the oxidation reaction, predicting the crystallization nucleation rate, growth rate, and particle size change trend through an ADC crystallization kinetic model; employing a feedforward-feedback dual closed-loop control to collaboratively adjust three control loops: chlorination rate, stirring speed, and cooling flow rate, to achieve a dynamic balance between nucleation and growth rates; and determining the reaction endpoint through multi-parameter cross-validation. The intelligent control system includes six modules: particle size matching parameter preset, multi-modal parameter acquisition, crystallization kinetic prediction, multi-parameter collaborative control, adaptive endpoint determination, and batch iterative optimization. This application can obtain the target particle size product without subsequent screening processes, reducing material loss and process costs.

Description

Technical Field

[0001] This application relates to the field of ADC foaming preparation, and in particular to a method and system for controlling the particle size of an ADC foaming agent. Background Technology

[0002] ADC (azodicarbonamide) is the most widely used organic foaming agent. The consumption structure of ADC is roughly as follows: PVC 40%, PE 35%, PP 12%, rubber 5%, and others 8%. Different materials and application fields have different particle size requirements for ADC. In the field of footwear materials based on EVA or PE, ADC particle size tends to be 6–12 μm; in PVC foam products, ADC particle size tends to be 10–15 μm; in PE foam products, ADC particle size tends to be 8–12 μm; in PP foam products, ADC particle size tends to be 6–8 μm; and in the interior parts of new energy vehicles, ADC particle size tends to be 3–8 μm. Other fields also have their own specific requirements. During the production process of ADC, various factors influence the particle size. Currently, controlling the particle size of ADCs is relatively simple. For example, Chinese invention patent application number 201510267264.8, entitled "A New Method for Easily Controlling the Particle Size of ADC Products," employs a method that adjusts the particle size of the blowing agent by adding aldehyde compounds. This method still yields an ADC blowing agent with a relatively wide particle size range. According to its experimental results, D... 50 The diameters range from 6.3 to 41.3 mm. To obtain a smaller, more suitable range of ADC blowing agents, the ADC product needs to be sieved, and then the product within the corresponding range needs to be taken. Products outside the applicable particle size range are either recrystallized or stored for other uses. The problem with this method is that there are many related processes, and there is material waste during the sieving process. Summary of the Invention

[0003] In view of this, this application proposes a particle size control method for ADC foaming agent, which reduces the particle size distribution by implementing intelligent control during the production process, thereby obtaining ADC products with the required particle size and reducing process and material waste.

[0004] This application also proposes a particle size control system for ADC foaming agents.

[0005] A method for controlling the particle size of an ADC foaming agent includes the following steps: Based on the target particle size, a pre-trained particle size-process parameter mapping model is invoked to generate initial control parameters for the oxidation reaction; Multimodal parameters of the oxidation reaction process are collected in real time, and based on the collected multimodal parameters, a pre-trained ADC crystallization kinetic model is called to predict the crystallization nucleation rate, growth rate and particle size change trend. Based on the deviation between the particle size prediction result and the target value, a feedforward-feedback dual closed-loop control is adopted to coordinate the adjustment of three control loops: chlorination rate, stirring speed, and cooling flow rate, so as to achieve a dynamic balance between the ADC nucleation rate and growth rate. Based on online particle size detection data and parameters of the ADC crystallization kinetic model, multi-parameter cross-validation is used. When the ADC particle size reaches the target range and meets the preset conditions, the reaction endpoint is determined and chlorination is stopped.

[0006] A particle size control system for an ADC foaming agent, comprising: Particle size matching parameter preset module: used to generate initial control parameters according to target particle size requirements; Multimodal parameter acquisition module: used to acquire real-time parameters of the reaction process, such as temperature, pH, chlorine flow rate, torque, and particle size distribution. Crystallization kinetics prediction module: used to predict crystallization nucleation rate, growth rate, and grain size change trend; Multi-parameter coordinated control module: used to coordinate the control loops of chlorination rate, stirring speed, and cooling flow rate to achieve dynamic balance between nucleation and growth rates; Adaptive endpoint determination module: used to determine the response endpoint through multi-parameter cross-validation; Batch Iterative Optimization Module: Used to periodically optimize model parameters to adapt to changes in operating conditions.

[0007] The technical advantages of this application are as follows: This application achieves automatic matching of target requirements to process parameters through a particle size-process parameter mapping model, enabling the formulation of control parameters based on particle size range requirements; it realizes the crystallization state through real-time acquisition of multimodal parameters, providing a data foundation for dynamic control; it predicts particle size trends through crystallization kinetics prediction, preventing abrupt particle size changes caused by control lag; and it achieves a dynamic balance between nucleation and growth through feedforward-feedback dual closed-loop collaborative control. The ADC foaming agent prepared using the control method of this application requires virtually no sorting, reducing material loss and process costs in the screening process. Detailed Implementation

[0008] The embodiments of the technical solution of this application will be described in detail below. The following embodiments are only used to illustrate the technical solution of this application more clearly, and are therefore only examples, and should not be used to limit the scope of protection of this application.

[0009] A method for controlling the particle size of an ADC foaming agent includes the following steps: Based on the target particle size, a pre-trained particle size-process parameter mapping model is invoked to generate initial control parameters for the oxidation reaction; In a preferred embodiment, the initial control parameters for the oxidation reaction include the reaction temperature range, the chlorine flow rate curve, the stirring speed range, and the pH control range.

[0010] In a preferred embodiment, the particle size-process parameter mapping model is generated through training on historical production data. Specifically, the particle size-process parameter mapping model adopts a multi-input multi-output nonlinear mapping architecture. The input parameters are the target particle size range, the biuret feedstock particle size range, and the number of mother liquor cycles; the output parameters are the combination of initial control parameters for the oxidation reaction, including: Temperature parameters: initial reaction temperature, heating rate, and temperature control range for each stage; Chlorine parameters: initial chlorine rate, stage chlorine adjustment curve, chlorine excess coefficient; Mixing parameters: initial mixing speed, stage adjustment rate, and final mixing speed; Acidity parameters: initial pH value, acid replenishment threshold.

[0011] Afterwards, a training dataset was established to facilitate model training. This application uses orthogonal experiments to generate a basic dataset covering all working conditions, selecting five influencing factors: reaction temperature, chlorination rate, stirring speed, initial pH value, and biuret raw material particle size. Each factor is set with 4 to 5 levels.

[0012] In a preferred embodiment: The reaction temperature was set at five levels: 17℃, 19℃, 21℃, 23℃, and 25℃. The chlorination rate is set to four levels: 160 kg / h, 180 kg / h, 200 kg / h, and 220 kg / h. The stirring speed can be set to 5 levels: 70 rpm, 90 rpm, 110 rpm, 130 rpm, and 150 rpm. The initial pH value was set at four levels: 0.9, 1.1, 1.3, and 1.5. The particle size of the biuret raw material was set at four levels: 5μm, 10μm, 15μm, and 20μm.

[0013] In a preferred embodiment, this application employs L 25 (5 6 25 basic experiments were designed using an orthogonal array. Each experiment was repeated 3 times, and the average value was taken as the effective data. A total of 75 basic samples were obtained as the basic dataset.

[0014] Based on the basic dataset, this application also conducts supplementary experiments for the particle size abrupt change region. Specifically, 16 supplementary experiments are set according to a particle size gradient of 0.5 μm, with each group repeated twice, to obtain 32 supplementary data sets, in order to ensure the mapping accuracy of the particle size range.

[0015] After supplementary experiments, the collected raw data were standardized to remove abnormal data that deviated from the normal range, and the effective samples were divided into training and test sets in an 8:2 ratio.

[0016] In a preferred embodiment, the model training of this application uses Gradient Boosting Decision Tree (GBDT) as the basic algorithm. The training process is as follows: Hyperparameter settings: 100 trees, maximum depth of 6, learning rate of 0.1, minimum number of splits of 5.

[0017] Loss function: The mean squared error (MSE) is used as the loss function, with the optimization objective being to minimize the particle size prediction bias.

[0018] Initial training: Use the training set data for initial training to obtain the basic model.

[0019] Cross-validation: Five-fold cross-validation was used to evaluate the model’s generalization ability, and the average particle size prediction error on the validation set was ≤ ±0.3 μm.

[0020] Test set validation: Use test set data for validation, requiring particle size prediction accuracy ≥98% and process parameter matching degree ≥95%.

[0021] After the model is put into operation, its accuracy is continuously improved through a batch iterative optimization mechanism, as follows: Automatic data collection: After each batch of production is completed, three sets of data, namely "target particle size - actual control parameters - final actual particle size", are automatically collected and stored in the historical database; Regular retraining: Every 50 sets of new data are accumulated, the model is automatically retrained to update the mapping relationship and adapt to changes in operating conditions such as raw material changes and equipment aging. Deviation warning: When the particle size prediction deviation exceeds ±0.5μm for 3 consecutive batches, the model calibration reminder will be automatically triggered, prompting the operator to adjust the parameters.

[0022] After generating initial control parameters for the oxidation reaction, this application controls the reaction according to these initial control parameters during the initial stage. Subsequently, multimodal parameters of the oxidation reaction process are acquired in real time, and based on these acquired multimodal parameters, a pre-trained ADC crystallization kinetic model is invoked to predict the crystallization nucleation rate, growth rate, and particle size change trend. In a preferred embodiment, this application uses a temperature sensor, pH meter, chlorine flow meter, stirring torque sensor, and FBRM online particle size analyzer to collect multimodal parameters of the oxidation reaction process in real time. The multimodal parameters include: Thermodynamic modes: reaction liquid temperature, jacket inlet and outlet temperatures, and pressure inside the reactor; Component modes: real-time pH value, redox potential (ORP), and residual chlorine content; Flow field modes: stirring motor torque, stirring speed, and pressure difference between the top and bottom of the vessel; Material modalities: cumulative chlorine flow rate, biurea feed rate, acid replenishment rate; Particle size modality: ADC average particle size, particle size distribution, and target particle size content.

[0023] This application characterizes the exothermic or endothermic state of a reaction through thermodynamic modes to determine the reaction kinetic process. Temperature detection uses a temperature sensor, and pressure detection uses a pressure sensor.

[0024] This application characterizes the oxidation reaction progress using component modes to determine whether chlorine is in excess or insufficient. Real-time pH value is measured using a pH electrode, oxidation-reduction potential is measured using an ORP electrode, and residual chlorine content is measured using an online residual chlorine analyzer.

[0025] This application characterizes the gas-liquid-solid three-phase mixing state through flow field modes to determine the crystallization suspension effect. A torque sensor is used to detect the stirring motor torque, a frequency converter speed sensor is used to detect the stirring speed, and a differential pressure transmitter is used to detect the pressure difference between the upper and lower parts of the vessel.

[0026] This application characterizes the cumulative amount of material input through material modal analysis for material balance calculations of reaction heat and conversion rate. A chlorine mass flow meter is used to detect the cumulative chlorine flow rate, while an electromagnetic flow meter is used to detect the biuret feed rate and acid replenishment rate.

[0027] This application characterizes the crystallization state through particle size modality to provide data for growth rate prediction. The average particle size, particle size distribution, and target particle size content of the ADC are detected using the FBRM focused beam reflectance meter.

[0028] The ADC crystallization kinetic model described in this application is a semi-empirical and semi-mechanistic model, which consists of three coupled parts: nucleation rate equation, growth rate equation, and particle size distribution prediction equation.

[0029] The nucleation rate equation is as follows: , In the formula: B Nucleation rate, the number of crystal nuclei generated per unit volume per unit time, expressed in units of nuclei / (m²). 3 ·s) k b The nucleation rate constant is temperature-dependent and is calculated using the following equation: ,in A b Forward factor, E b R is the nucleation activation energy, R is the gas constant, and T is the reaction temperature; S ADC supersaturation (actual concentration / saturated solubility); N The suspension density of the existing crystal (mass of crystal per unit volume, kg / m³) 3 ); n , m The nucleation series is determined through experimental fitting.

[0030] The growth rate equation is: In the formula: G is the linear growth rate of the crystal (μm / min); kg This is the growth rate constant; g This represents the growth stage.

[0031] The particle size distribution prediction equation is: In the formula: Δd nuc This represents the average change in grain size caused by the formation of new crystal nuclei.

[0032] All constant parameters in the model were obtained through laboratory intermittent crystallization experiments. Specifically, five sets of gradient experiments were set up within the reaction temperature range of 17–23°C. Samples were taken every 5 minutes in each set to measure the ADC concentration, average particle size, and number of crystal nuclei. The experimental data were fitted using the nonlinear least squares method to determine the values ​​of all kinetic parameters.

[0033] Subsequently, based on the deviation between the predicted particle size and the target particle size, a feedforward-feedback dual closed-loop control is adopted to coordinately adjust the chlorination rate, stirring speed, and cooling flow rate of the three control loops to achieve a dynamic balance between the nucleation rate and the growth rate. Specifically, this application adopts a two-layer structure of outer loop feedforward prediction and inner loop feedback control, with the three control loops working in tandem to stably control the supersaturation S within a predetermined range, thereby achieving a dynamic balance between the nucleation rate and the growth rate. In a preferred embodiment, the predetermined range for the supersaturation S is 1.1 ± 0.05.

[0034] When the predicted particle size deviation is too large within a predetermined timeframe, or the supersaturation S exceeds the predetermined range, the outer loop feedforward adjustment is automatically triggered. That is: When the predicted particle size growth rate is less than the set value, reduce the chlorination rate by 5% to 8% to reduce the ADC nucleation rate; reduce the stirring speed by 5 to 10 rpm to reduce crystal nucleus collision and breakage; and increase the temperature by 0.3 to 0.5℃ to reduce supersaturation and promote growth. All three are adjusted simultaneously.

[0035] When the predicted ADC nucleation rate is less than the set value, increase the chlorine flow rate by 5% to 8% to increase the ADC nucleation rate; increase the stirring speed by 5 to 10 rpm to promote crystal nucleus dispersion; and decrease the temperature by 0.3 to 0.5℃ to increase supersaturation and promote nucleation. All three are adjusted simultaneously.

[0036] The inner-loop feedback control in this application is based on the actual detected particle size deviation of the FBRM focused beam reflection measuring instrument for fine-tuning, eliminating residual errors and unpredictable disturbances after feedforward adjustment.

[0037] Specifically, when the actual detected particle size deviates from the target value by more than the set value, and the deviation direction is consistent for three consecutive detections, feedback adjustment is triggered: When the particle size growth rate is less than the set value, the cooling flow rate is reduced by 10% to maintain a stable supersaturation. When the ADC nucleation rate is less than the set value, the cooling flow rate is increased by 10% to suppress excessive crystal growth.

[0038] The three loops are not adjusted independently, but rather achieve coordinated action through a decoupling algorithm to avoid coupling interference between parameters. That is: Chlorine flow rate is adjusted first: As the main control variable for ADC generation rate, it is adjusted first during feedforward regulation, and no other parameters are adjusted within a predetermined time after adjustment.

[0039] Stirring speed adjustment: After adjusting the chlorination rate, adjust the stirring speed synchronously according to the torque change to ensure uniform crystal suspension.

[0040] Cooling flow compensation: If there is still a residual error after adjusting the first two, the temperature is finely adjusted by adjusting the cooling flow to achieve the stability of supersaturation.

[0041] Subsequently, as the reaction proceeds, based on online particle size detection data and reaction kinetic parameters, multi-parameter cross-validation is used. When the product particle size reaches the target range and meets the preset conditions, the reaction endpoint is determined, chlorination is stopped, and the control process ends.

[0042] It should be noted that the particle size mutation during the oxidation process of the ADC foaming agent mentioned above is an inherent characteristic formed by the coupling effect of crystallization kinetics and reaction kinetics, and it is also the current challenge for particle size control in the industry. The ADC oxidation reaction is a "reaction-crystallization" coupled process. When the reaction proceeds to a certain extent, two kinetic inflection points will appear, which together lead to the particle size mutation: First, there is a sudden change in supersaturation. At this stage, the consumption rate of biuret and the ADC generation rate reach a critical equilibrium point. The concentration of ADC in the solution rapidly exceeds the saturation solubility, and the supersaturation rises sharply from 1.05 to 1.3-1.4. The crystallization nucleation rate increases exponentially. Within 10 minutes, the number of newly generated crystal nuclei can reach 2-3 times the total number of the previous reaction stages. If not properly controlled, this can lead to a sudden decrease in particle size.

[0043] Secondly, there is the abrupt change in crystal form: when ADC crystals grow to a certain range, a change in crystal growth mode occurs: below the critical particle size, surface growth mode is dominant, with a stable growth rate of 0.08–0.12 μm / min; after exceeding the critical particle size, it switches to agglomeration growth mode, with the growth rate surging to 0.2–0.3 μm / min, and the particle size easily increases suddenly, exceeding the target range. Traditional control methods struggle to overcome this abrupt change, resulting in large batch-to-batch particle size deviations.

[0044] This application utilizes a pre-trained ADC crystallization kinetic model to predict abnormal increases in nucleation rate based on real-time collected multimodal parameters such as temperature, ORP, and pH. It adjusts relevant parameters before particle size mutations occur, thereby suppressing particle size changes caused by these mutations. Simultaneously, a feedforward-feedback dual closed-loop control is employed to coordinate the adjustment of three control loops: chlorination rate, stirring speed, and cooling flow rate. This stabilizes the nucleation / growth rate ratio within a preset range, preventing explosive nucleation caused by a sudden increase in supersaturation.

[0045] Furthermore, this application cross-validates online particle size detection data with kinetic model prediction results. When an abnormal increase in nucleation rate is detected, the chlorination rate is immediately and automatically adjusted to reduce the supersaturation of the system. At the same time, the stirring speed is increased to promote the dissolution of fine crystals, thereby suppressing the occurrence of abrupt particle size changes.

[0046] The system records the operating characteristics of each particle size fluctuation, automatically optimizes the crystallization kinetic model parameters, continuously improves the accuracy of predicting particle size mutations under different operating conditions, and can significantly reduce the occurrence rate of particle size mutations under long-term operation.

[0047] In a preferred embodiment, the feedforward-feedback dual closed-loop control specifically involves: when the predicted particle size deviation exceeds ±0.5μm, the chlorination rate and stirring speed are rapidly adjusted through the feedforward loop, and the cooling flow rate is finely adjusted through the feedback loop to control the response time to ≤30 seconds.

[0048] In a preferred embodiment, the adaptive endpoint determination adopts multi-parameter cross-validation and simultaneously meets three preset conditions: particle size meets the standard, redox potential is stable at 640-660mV, and residual chlorine content is ≤50mg / L.

[0049] To achieve the above control method, this application also proposes an intelligent control system for the oxidation reaction of ADC foaming agent with controllable particle size, comprising: Particle size matching parameter preset module: used to generate initial control parameters according to target particle size requirements; Multimodal parameter acquisition module: used to acquire real-time parameters of the reaction process, such as temperature, pH, chlorine flow rate, torque, and particle size distribution. Crystallization kinetics prediction module: used to predict crystallization nucleation rate, growth rate, and grain size change trend; Multi-parameter collaborative control module: used to collaboratively adjust the chlorination rate, stirring speed, and cooling flow control loop to achieve dynamic balance between ADC nucleation and growth rate; Adaptive endpoint determination module: used to determine the response endpoint through multi-parameter cross-validation; Batch Iterative Optimization Module: Used to periodically optimize model parameters to adapt to changes in operating conditions.

[0050] This system can be deployed and run without large-scale modifications. The functional positioning, input / output, and interaction logic of each module are as follows: The particle size matching parameter preset module automatically converts customer requirements into process parameters and serves as the system's front-end interactive entry point. When the target particle size of the ADC, the average particle size of biuret, and the number of mother liquor cycles are input, the particle size matching parameter preset module calls the pre-trained particle size-process parameter mapping model to automatically match and generate the initial control parameters for this batch of production, including the reaction temperature curve, the staged chlorination rate, the stirring speed range, and the pH control threshold, and outputs them to the underlying control unit for execution.

[0051] The multimodal parameter acquisition module constructs a full-dimensional reaction process sensing network, providing data support for subsequent prediction and control. The multimodal parameter acquisition module collects data from five categories of objects: Thermodynamic modal parameters: reaction liquid temperature, jacket inlet and outlet temperatures, and pressure inside the reactor; Component modal parameters: real-time pH value, redox potential (ORP), and residual chlorine content; Flow field modal parameters: stirring motor torque, stirring speed, and pressure difference between the top and bottom of the vessel; Material modal parameters: chlorine cumulative flow rate, biurea feed rate, acid replenishment rate; Particle size modality: ADC average particle size, particle size distribution, and target particle size content.

[0052] The system automatically performs outlier removal and data standardization during data acquisition, and supports multi-parameter cross-substitution calculations in the event of a single sensor failure. After data standardization, the data is transmitted to the crystallization kinetics prediction module and the multi-parameter collaborative control module.

[0053] The multi-parameter collaborative control module enables decoupled multi-parameter control, maintaining a dynamic balance between nucleation and growth. The module employs a feedforward-feedback dual closed-loop control, collaboratively adjusting three control loops: Chlorine flow rate loop: Adjustable range ±5%~10%, controls ADC generation rate; Stirring speed circuit: Adjustment range ±5~10rpm, controlling the probability of crystal nucleus suspension and collision; Cooling flow loop: Adjustment range ±10%~15%, fine-tuning reaction temperature to stabilize supersaturation.

[0054] After receiving the prediction results from the crystallization kinetic prediction module and the real-time feedback data from the multi-modal acquisition module, the multi-parameter collaborative control module directly sends control commands to the field valves, frequency converters and other actuators.

[0055] The adaptive endpoint determination module can determine the reaction endpoint and prevent insufficient reaction or excessive oxidation. The adaptive endpoint determination module uses a weighted voting mechanism for multi-parameter cross-validation. When three conditions are met simultaneously—particle size meets the target, ORP is stable, and residual chlorine is less than the set value—the reaction endpoint is automatically determined.

[0056] After receiving online particle size data, ORP potential, and residual chlorine content information from the system, the adaptive endpoint determination module cuts off the chlorine feed valve and starts the cooling program based on the determination result.

[0057] The batch iteration optimization module automatically performs incremental training on the particle size-process parameter mapping model and crystallization kinetic model after accumulating a certain number of batches of production data, updates the model parameters, and adapts to long-term operating conditions such as fluctuations in biuret raw material batches, agitator wear, and accumulation of impurities in mother liquor.

[0058] After receiving the full-process data of "target particle size - control parameters - actual particle size" for each batch, the record of changes in raw material characteristics, and the equipment operating status data, the batch iterative optimization module updates the model parameters and synchronizes them to the particle size matching module and the crystallization kinetic prediction module.

[0059] The following comparative experiments will illustrate the effectiveness of this method. It should be noted that all experiments were conducted at 10m... 3 The reaction was carried out in an industrial-grade ADC oxidation reactor. The raw material was a biuret suspension with a hydrazine hydrate concentration of 120 g / L. The reaction temperature was controlled within the range of 25–45 °C, and the chlorine pressure was 0.2–0.3 MPa. Each group was tested in parallel for 3 times, and the average value of the results was taken.

[0060] Example 1 Using the method described in this application: Input the target particle size of 12 μm, and call the particle size-process parameter mapping model to generate initial parameters: chlorination rate 180 kg / h, stirring speed 60 rpm, cooling flow rate 12 m³ / h. 3 / h; Multimodal parameters are collected in real time, and the crystallization kinetic model is called every 5 minutes to predict nucleation, growth rate and particle size changes; The three control loops are dynamically adjusted based on the predicted deviation: when the nucleation rate is too high in the early stage of the reaction, the chlorine flow rate is reduced to 160 kg / h; when the growth rate is too slow in the middle stage of the reaction, the stirring speed is increased to 65 rpm; and the nucleation / growth rate ratio is kept stable in the range of 0.8 to 1.2 throughout the process. Online particle size analysis showed that the particle size reached the range of 12μm±0.1μm and the ORP value remained stable above 650mV for 10 minutes. The reaction endpoint was then determined and chlorination was stopped.

[0061] Example 2 Using the method described in this application: Input the target particle size of 25 μm, and call the mapping model to generate initial parameters: chlorination rate 220 kg / h, stirring speed 45 rpm, cooling flow rate 10 m³ / h. 3 / h; Real-time acquisition of multimodal parameters combined with kinetic prediction allows for dynamic adjustment of the control loop, maintaining the nucleation / growth rate ratio stable within the range of 0.3 to 0.5. When the particle size reaches the range of 25μm±0.1μm and the ORP stabilizes above 650mV for 10 minutes, the reaction is stopped.

[0062] Comparative Example 1 Using existing industry-standard empirical control methods, the target particle size is 12μm: The chlorination rate was set manually at 200 kg / h and the stirring speed at 55 rpm, based on experience. The cooling flow rate was adjusted manually according to the temperature. Without kinetic prediction, particle size is detected solely by manual sampling at regular intervals, once every 1 hour, and parameters are manually adjusted based on the test results. The reaction endpoint is determined by manually observing the color of the reaction solution and combining it with experience.

[0063] Comparative Example 2 Using a conventional feedback control method with a target particle size of 12 μm: the chlorination rate is adjusted only based on the online particle size detection results, without pre-trained mapping models or crystallization kinetic model prediction steps.

[0064] The test results of the examples and comparative examples are shown in Table 1.

[0065] Table 1 In Comparative Example 1 and Comparative Example 2, the ADC yield is the yield after sorting, minus the material loss.

[0066] As can be seen from Table 1, the particle size deviations of Examples 1 and 2 of this application are 0.03 μm and 0.03 μm, respectively, which are much lower than the deviation of 3.8 μm in Comparative Example 1 and 0.3 μm in Comparative Example 2, and can meet the personalized particle size requirements of different downstream application scenarios.

[0067] The particle size distribution variation coefficients of the embodiments in this application are all ≤9.2%, which are far superior to 21.3% of Comparative Example 1 and 16.8% of Comparative Example 2. This indicates that by dynamically controlling the nucleation / growth rate, burst nucleation and particle size mutation can be effectively suppressed, and a narrow distribution ADC product can be obtained, which can be used directly without subsequent screening.

[0068] The ADC yields of the embodiments of this application are all ≥94%, which is more than 7.7 percentage points higher than the 86.3% of Comparative Example 1 and more than 3 percentage points higher than the 91.2% of Comparative Example 2. This yield is the yield of the finished product obtained directly after the reaction, without the need for a screening process. In contrast, the yield of the comparative examples is the yield of qualified products after screening, and the actual total reaction yield is even lower.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for controlling the particle size of an ADC foaming agent, characterized in that: Includes the following steps: Based on the target particle size, a pre-trained particle size-process parameter mapping model is invoked to generate initial control parameters for the oxidation reaction; Multimodal parameters of the oxidation reaction process are collected in real time, and based on the collected multimodal parameters, a pre-trained ADC crystallization kinetic model is called to predict the crystallization nucleation rate, growth rate and particle size change trend. Based on the deviation between the particle size prediction result and the target value, a feedforward-feedback dual closed-loop control is adopted to coordinate the adjustment of three control loops: chlorination rate, stirring speed, and cooling flow rate, so as to achieve a dynamic balance between the ADC nucleation rate and growth rate. Based on online particle size detection data and parameters of the ADC crystallization kinetic model, multi-parameter cross-validation is used. When the ADC particle size reaches the target range and meets the preset conditions, the reaction endpoint is determined and chlorination is stopped.

2. The particle size control method for ADC foaming agent as described in claim 1, characterized in that: The initial control parameters for the oxidation reaction include the reaction temperature range, chlorine flow rate curve, stirring speed range, and pH control range.

3. The particle size control method for ADC foaming agent as described in claim 1, characterized in that: The particle size-process parameter mapping model is generated through training on historical production data.

4. The particle size control method for ADC foaming agent as described in claim 3, characterized in that: The particle size-process parameter mapping model adopts a multi-input multi-output nonlinear mapping architecture, with input parameters being the target particle size range, the biuret feed particle size range, and the number of mother liquor circulations. The output parameters are the initial control parameters combination for the oxidation reaction.

5. The particle size control method for ADC foaming agent as described in claim 4, characterized in that: The initial control parameter combination for the oxidation reaction includes: Temperature parameters: initial reaction temperature, heating rate, and temperature control range for each stage; Chlorine parameters: initial chlorine rate, stage chlorine adjustment curve, chlorine excess coefficient; Mixing parameters: initial mixing speed, stage adjustment rate, and final mixing speed; Acidity parameters: initial pH value, acid replenishment threshold.

6. The particle size control method for the ADC foaming agent as described in claim 1, characterized in that: The multimodal parameters include: Thermodynamic modes: reaction liquid temperature, jacket inlet and outlet temperatures, and pressure inside the reactor; Component modes: real-time pH value, redox potential, and residual chlorine content; Flow field modes: stirring motor torque, stirring speed, and pressure difference between the top and bottom of the vessel; Material modalities: cumulative chlorine flow rate, biurea feed rate, acid replenishment rate; Particle size modality: ADC average particle size, particle size distribution, and target particle size content.

7. The particle size control method for ADC foaming agent as described in claim 1, characterized in that: The ADC crystallization kinetic model is a semi-empirical and semi-mechanistic model, consisting of three coupled parts: nucleation rate equation, growth rate equation, and particle size distribution prediction equation.

8. The particle size control method for ADC foaming agent as described in claim 1, characterized in that: The feedforward-feedback dual closed-loop control specifically means that when the predicted particle size deviation exceeds ±0.5μm, the chlorine flow rate and stirring speed are quickly adjusted through the feedforward loop, and the cooling flow rate is finely adjusted through the feedback loop to control the response time to ≤30 seconds.

9. The particle size control method for ADC foaming agent as described in claim 1, characterized in that: The endpoint determination adopts multi-parameter cross-validation and simultaneously meets the preset conditions, namely, the particle size meets the standard, the oxidation-reduction potential is stable at 640-660mV, and the residual chlorine content is ≤50mg / L.

10. A system for implementing the particle size control method of the ADC foaming agent according to any one of claims 1 to 9, characterized in that, include: Particle size matching parameter preset module: used to generate initial control parameters according to target particle size requirements; Multimodal parameter acquisition module: used to acquire real-time parameters of the reaction process, such as temperature, pH, chlorine flow rate, torque, and particle size distribution. Crystallization kinetics prediction module: used to predict crystallization nucleation rate, growth rate, and grain size change trend; Multi-parameter coordinated control module: used to coordinate the control loops of chlorination rate, stirring speed, and cooling flow rate to achieve dynamic balance between nucleation and growth rates; Adaptive endpoint determination module: used to determine the response endpoint through multi-parameter cross-validation; Batch Iterative Optimization Module: Used to periodically optimize model parameters to adapt to changes in operating conditions.

Images