A control method and system for improving hydrazine carbonate yield

Abstract

This application provides a control method for improving the yield of biuret, comprising the following steps: collecting the characteristic spectrum of the reaction solution and obtaining real-time concentration data of hydrazine hydrate and biuret through near-infrared spectroscopy analysis; employing a multi-parameter decoupled dual-closed-loop control architecture, with the inner loop adjusting the feed flow ratio of hydrochloric acid, urea, and hydrazine hydrate in real time to control the pH value of the reaction solution; the outer loop correcting the reaction temperature setpoint based on the real-time concentration data of hydrazine hydrate and biuret; adjusting the stirring rate based on the rheological characteristics feedback of the reaction solution to maintain stable biuret crystal growth; and determining the reaction endpoint by integrating the multi-parameter analysis results of real-time concentration, cumulative heat of reaction, and pH value change trends. When the conversion rate of hydrazine hydrate reaches or exceeds the set value, heating is stopped and gradient cooling is initiated. By using real-time component detection and pH-temperature dual-closed-loop control, the problems of offline detection lag and parameter coupling are solved, and the final average product yield is increased from 91.7% to 95.1%.

Description

Technical Field

[0001] This application relates to the field of biuret preparation, and more particularly to a method and system for controlling the yield of biuret. Background Technology

[0002] Biuret (N,N'-dicarbamoylhydrazine) is a core intermediate in the production of ADC blowing agents (azodicarbonamide), and the stability of its preparation process directly determines the quality and production cost of downstream blowing agent products. Currently, most small and medium-sized ADC manufacturers in China adopt an acidic condensation process. This process uses hydrazine hydrate, urea, and hydrochloric acid as main raw materials, and generates biuret through a condensation reaction under acidic conditions. The main reaction equation is: N2H4·H2O+2HCl+2CO(NH2)2→NH2CONHNHCONH2+2NH4Cl+H2O This process has been widely used in the industry due to its fast reaction rate and low equipment investment. However, problems such as poor production stability and low resource utilization have become increasingly prominent during long-term operation. Currently, the yield of biuret in this process is generally only 85% to 92%, with batch-to-batch yield fluctuations reaching ±5%, and some companies with insufficient management even experiencing fluctuations exceeding 8%. At the same time, the process generates a high proportion of byproducts such as biuret, aminourea, and nitrogen-containing organic impurities, which not only increases the difficulty of subsequent product refining but also leads to a significant increase in wastewater content.

[0003] The main reason for the instability of existing process production indicators lies in the limitations of process control methods. Currently, most enterprises still use manual experience-based control or single-loop PID control modes, which are difficult to adapt to the nonlinear characteristics of the strong coupling of multiple factors in condensation reactions. During the reaction process, parameters such as material ratio, pH value, reaction temperature, reaction time, and stirring mixing effect have complex interactive effects: for example, a fluctuation of ±2℃ in reaction temperature can change the condensation reaction rate by more than 15%, and at the same time promote the deamination of urea molecules to form biuret byproducts; when the pH value deviates from the predetermined range, it will not only lead to abnormal protonation of hydrazine hydrate, reducing the selectivity of the main reaction, but also accelerate the formation of toxic byproducts such as aminourea. Under manual control mode, operators need to sample and analyze parameters every 30 to 60 minutes. Adjusting the feed rate or heating power based on experience has obvious control lag, and the difference in judgment standards among different operators directly leads to poor consistency of process parameters between batches. Traditional single-loop PID control can only adjust the setpoint of a single parameter and cannot perceive the synergistic relationship between multiple parameters. When disturbances such as fluctuations in feed concentration or unstable stirring occur, the control loops of each parameter will interfere with each other, further aggravating the oscillation of process parameters. For example, when the temperature control loop adjusts the heating power, it will indirectly affect the measurement accuracy of the pH value of the reaction solution. The change in the hydrochloric acid feed rate during pH adjustment will cause fluctuations in the reaction heat load, forming a control coupling dilemma. Ultimately, this will lead to problems such as unstable conversion rate of the main reaction and large fluctuations in the amount of by-products generated. Summary of the Invention

[0004] In view of this, this application proposes a control method to improve the yield of biuret. It adopts a dual closed-loop control architecture, combines a mechanism model and a data-driven method to achieve dynamic collaborative optimization of multiple parameters, and improves the yield of biuret while maintaining the stability of the yield between batches.

[0005] A method for controlling the yield of biuret includes the following steps: The characteristic spectrum of the reaction solution was collected, and the real-time concentration data of hydrazine hydrate and biuret were obtained by near-infrared spectroscopy analysis. A multi-parameter decoupled dual-closed-loop control architecture is adopted. The inner loop adjusts the feed flow ratio of hydrochloric acid, urea and hydrazine hydrate in real time to control the pH value of the reaction solution. The outer loop corrects the reaction temperature setpoint based on the real-time concentration data of hydrazine hydrate and biuret. Based on the feedback of the rheological properties of the reaction solution, the stirring rate is adjusted to maintain the stability of biurea crystal growth; By integrating the results of multi-parameter analysis of real-time concentration, cumulative heat of reaction, and pH value change trends, the reaction endpoint is determined. When the conversion rate of hydrazine hydrate reaches or exceeds the set value, heating is stopped and gradient cooling is initiated.

[0006] A control system for improving the yield of biuret includes: The data acquisition unit includes a mass flow meter, a density meter, an online pH analyzer, a multi-point temperature sensor, an online rheometer, and a near-infrared spectrometer, which are used to collect production process parameters in real time. The control unit, which is communicatively connected to the data acquisition unit, includes a programmable logic controller and an industrial computer, and is equipped with the control method for generating control commands. The execution unit, which is communicatively connected to the control unit, includes a feed regulating valve, a heat transfer oil regulating valve, a variable frequency stirring motor, and a discharge control valve, and is used to execute control commands; The data storage and analysis unit is communicatively connected to the control unit and the data acquisition unit, and is used to store historical production data and periodically update the prediction model parameters.

[0007] The technical advantages of this application are as follows: This solution reduces process parameter fluctuations by more than 60% through real-time component detection and pH-temperature dual closed-loop control, solving the problems of offline detection lag and parameter coupling; the combined effect of feedforward compensation and dynamic parameter matching improves the main reaction selectivity to 95.2% and reduces by-product generation by 83%; the synergy of accurate endpoint determination and gradient crystallization control improves product particle size uniformity by 60%, reduces filtration loss by 84%, and ultimately increases the average product yield from 91.7% to 95.1%. 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] The control methods for improving the yield of biuret include the following steps: Step 1: Collect the characteristic spectrum of the reaction solution and obtain real-time concentration data of hydrazine hydrate and biuret through near-infrared spectroscopy analysis. Near-infrared spectroscopy can identify the overtone and combination frequency vibrations of chemical bonds such as CH, NH, and OH in molecules. Significant differences exist in the characteristic absorption peaks of the NH bond in hydrazine hydrate and the amide group in biuret. Through near-infrared spectroscopy analysis, the real-time acquired characteristic spectra can be converted into the concentration values ​​of the corresponding components. This reduces the response delay of traditional manual sampling and detection from 30–60 minutes to less than 10 seconds, and controls the concentration detection error to below 0.5%. This avoids the control lag caused by offline detection and provides an accurate data basis for subsequent reaction process control and endpoint determination, eliminating batch-to-batch parameter differences caused by manual analysis errors at the source.

[0010] Step 2: A multi-parameter decoupled dual-closed-loop control architecture is adopted. The inner loop adjusts the feed flow ratio of hydrochloric acid, urea, and hydrazine hydrate in real time to control the pH value of the reaction solution; the outer loop corrects the reaction temperature setpoint based on real-time concentration data of hydrazine hydrate and biuret. This application employs a layered control logic of inner and outer loops: the inner loop is a rapid pH control loop, which can offset the influence of feed concentration fluctuations and reaction consumption on the pH value by adjusting the feed flow ratio of hydrochloric acid, urea, and hydrazine hydrate in real time, thus stabilizing the pH value within the optimal range; the outer loop is a reaction conversion rate control loop, which calculates the reaction rate based on real-time concentration data and dynamically corrects the reaction temperature setpoint, matching the kinetic requirements of the main reaction by controlling the rate of heat generation. The decoupling algorithm between the two loops can eliminate the mutual interference between temperature changes on pH detection and pH adjustment on reaction heat load, achieving independent and stable control of the two parameters. In this way, the pH fluctuation of the reaction process can be controlled within ±0.2 and the temperature fluctuation within ±1℃, the selectivity of the main reaction can be increased to more than 95%, and the amount of by-products biuret and aminourea can be reduced by more than 40%. This solves the problems of high raw material consumption and many by-products at the reaction process level, and provides high-purity reaction mother liquor for the subsequent crystallization process.

[0011] In a preferred embodiment, the inner loop response time is less than 1 second; the outer loop response time is less than 10 seconds.

[0012] Step 3: Based on the rheological feedback of the reaction solution, adjust the stirring rate to maintain stable biuret crystal growth. The rheological properties of the reaction solution, such as viscosity and shear stress, are directly related to the concentration and particle size distribution of biuret crystals: In the early stage of the reaction, the reaction solution is a homogeneous solution with low viscosity. High-speed stirring can enhance material mixing and eliminate local concentration inhomogeneities. In the middle and later stages of the reaction, biuret crystals gradually precipitate, and the system viscosity increases with the increase of crystal concentration. At this time, reducing the stirring rate can avoid excessive shear force that could break the crystals, while ensuring uniform system temperature and concentration to prevent excessive local supersaturation and the formation of impurity crystals. By monitoring the viscosity changes of the reaction solution in real time using an online rheometer, the crystal growth stage can be accurately determined, thus enabling the stirring rate to match the crystallization process. In this way, the particle size distribution variation coefficient of the biuret product can be reduced from 35% in the traditional process to less than 15%, and the amount of impurities encapsulated during crystallization can be reduced by 60%. This avoids subsequent filtration losses caused by excessively fine crystals, reduces the difficulty of crude product refining, and can increase the product yield by 2-3 percentage points.

[0013] Step 4: Integrate the multi-parameter analysis results of real-time concentration, cumulative heat of reaction, and pH trend to determine the reaction endpoint. When the hydrazine hydrate conversion rate reaches or exceeds the set value, stop heating and initiate gradient cooling. The endpoint is determined by the hydrazine hydrate conversion rate reaching or exceeding the set value. Single concentration detection may introduce random errors; therefore, three parameters are integrated for comprehensive determination: real-time concentration data directly reflects component changes; cumulative heat of reaction verifies the degree of conversion through energy conservation; and pH trend helps determine the raw material consumption status. The endpoint is determined when the consistency of the three parameters reaches a threshold. Subsequently, the temperature is gradually reduced using a gradient cooling rate to avoid impurity co-precipitation caused by rapid cooling, ensuring crystal purity.

[0014] In a preferred embodiment, when the conversion rate of hydrazine hydrate is less than 70%, the reaction temperature is controlled at 108°C to 112°C; when the conversion rate of hydrazine hydrate reaches 70% to 90%, the reaction temperature is controlled at 112°C to 116°C; and when the conversion rate of hydrazine hydrate is greater than 90%, the reaction temperature is controlled at 105°C to 108°C.

[0015] Traditional single-loop pH control only adjusts the feed after the pH deviates from the set value, which is prone to overshoot oscillation. However, the outer loop in step 2 can predict reaction rate changes in advance through real-time concentration, and fine-tune the temperature setpoint before pH fluctuations occur to match reaction kinetics requirements. Combined with the rapid adjustment of the feed ratio in the inner loop, the two work together to reduce pH fluctuations from ±0.5 to ±0.2 and temperature fluctuations from ±3℃ to ±1℃. The selectivity of the main reaction increases from 92% to over 95%, and the amount of by-products is reduced by 40%. At the same time, the stable control in step 2 reduces temperature and impurity interference in the reaction system, reducing the near-infrared spectroscopy detection error in step 1 from 1.2% to less than 0.5%, forming a positive cycle.

[0016] If only step 3 is used to adjust the stirring, a large amount of biuret will be generated in a short time if the control in step 2 becomes unstable, leading to excessively high local supersaturation. No matter how the stirring is adjusted, a large number of fine and impurity crystals will appear. However, the stable control in step 2 ensures a uniform biuret formation rate, and the rheological feedback in step 3 can transmit the crystal growth state back to step 2. For example, if rheological data shows that crystal precipitation is too fast, step 2 will slightly reduce the reaction temperature to slow down the biuret formation rate and avoid local supersaturation. With the synergy of both, the particle size distribution variation coefficient of biuret crystals is reduced from 35% in the traditional process to less than 15%, the amount of impurities encapsulated in the crystals is reduced by 60%, the energy consumption of subsequent crude product refining is reduced by 30%, and the product loss in the filtration process is reduced to about 1%. Meanwhile, there is also a two-way synergistic relationship between steps 1 and 3: the near-infrared spectroscopy in step 1 can detect the concentration changes of hydrazine hydrate and biuret in the reaction solution in real time. In the early stage of the reaction before crystallization, the near-infrared spectral concentration data can be used to indirectly infer whether the reaction rate is uniform. When the hydrazine hydrate consumption rate fluctuates abnormally, it prompts step 3 to promptly increase the stirring rate to eliminate local concentration unevenness. In the crystallization precipitation stage, the biuret concentration detected in step 1 and the solid content of the crystals predicted by the rheological model in step 3 can be cross-validated. Specifically, if the deviation between the two exceeds the set threshold, it indicates that the near-infrared spectroscopy is biased due to crystal scattering, or the rheological model parameters need to be updated. At this time, model correction or equipment inspection alarm is triggered. This two-way verification mechanism keeps the concentration detection accuracy in the crystallization stage below 0.5% and the solid content prediction error of the rheological model within 2%, ensuring that the stirring rate adjustment is always based on accurate crystallization state information. If the crystallization stage is determined solely by the rheological data from step 3, fluctuations in the feed concentration can cause a deviation between the actual crystallization process and the rheological model's expectations. This could lead to the stirring rate being switched at the wrong time, resulting in either excessively low rotation speed during the nucleation stage causing crystal agglomeration or excessively high rotation speed during the growth stage breaking the crystals. However, the real-time concentration data from step 1 provides a second criterion for step 3, independent of rheological properties. Combining these two criteria increases the accuracy of the stirring rate switching timing during the crystallization stage from 92% with a single rheological assessment to over 98%, significantly improving the consistency of crystal particle size distribution between batches.

[0017] The endpoint determination in step 4 does not rely solely on the concentration data from step 1, but also combines the cumulative heat of reaction from step 2 (calculated using temperature and feed rate) and the rheological property changes from step 3 (viscosity reaches a stable threshold near the endpoint). This cross-validation of three parameters effectively avoids misjudgments of the endpoint caused by fluctuations in a single detection data point. Simultaneously, the batch-to-batch reaction time difference is reduced from the traditional 20-30 minutes to less than 5 minutes. The gradient cooling process initiated after endpoint determination also adjusts the cooling rate by referencing the residual hydrazine hydrate concentration from step 1 and the crystallization particle size data from step 3: if the residual trace hydrazine hydrate is slightly high, the cooling rate is appropriately slowed to avoid impurity co-precipitation; if the crystallization particle size is too small, the high-temperature holding time is extended to allow the crystals to continue growing. Ultimately, the batch-to-batch product yield fluctuation is reduced from the traditional ±5% to within ±2%.

[0018] In a preferred embodiment, the cumulative heat of reaction in step 4 is calculated using temperature and feed rate, as follows: To simplify calculations and ensure accuracy for industrial applications, the following reasonable assumptions are made first: The reactor has good thermal insulation performance, and the heat dissipation to the environment can be obtained by offline calibration with a fixed coefficient and an error of ≤5%. The specific heat capacity of the reaction solution changes very little with temperature and concentration, so the average specific heat capacity is used for calculation. No material overflows during the reaction process, and the total mass of the feed is equal to the sum of the total mass of the reaction liquid and the mass of the discharged material.

[0019] The calculation parameters are defined as follows:

[0020] The cumulative heat of reaction is calculated in real time using the energy balance method. That is, the total heat released by the reaction is equal to the sum of the enthalpy change of the system temperature rise, the heat carried away by the reaction, and the heat lost to the environment. The calculation cycle is synchronized with the outer loop control cycle.

[0021] Step 1: Calculate the enthalpy change of the reaction system The reaction system from the initial temperature T 0 Rise to the current temperature T mix The heat absorbed is

[0022] The formula consists of two parts: one is the enthalpy change of the overall temperature rise of the reaction liquid, and the other is the heat absorbed by the feed at room temperature as it is heated to the reaction temperature.

[0023] Step 2: Calculate the heat transfer of the jacket medium. During the heating stage, the jacketed heat transfer oil releases heat to the system; during the cooling stage, the cooling water removes heat from the system. The calculations are performed using the following formula:

[0024] Where τ is the reaction time. Δt The calculation period is defined as follows: a positive result indicates that the jacket supplies heat to the system, while a negative result indicates that the jacket removes heat from the system.

[0025] Step 3: Calculate the cumulative heat of reaction According to the law of conservation of energy, the actual heat released by the reaction is equal to the enthalpy change of the system minus the heat supplied by the jacket, plus the heat dissipated to the environment:

[0026] To eliminate detection noise, a sliding window average is used to process all real-time parameters during the calculation, avoiding calculation errors caused by instantaneous fluctuations.

[0027] Step 4: Conversion Rate Verification The calculated cumulative heat of reaction can be compared with the theoretical heat of reaction to verify the conversion rate of hydrazine hydrate.

[0028] in This represents the total amount of hydrazine hydrate in the feed. The conversion rate estimate based on the heat of reaction is:

[0029] The deviation between this value and the conversion rate obtained from near-infrared spectroscopy should be ≤2%; otherwise, it is considered an abnormal detection.

[0030] In a preferred embodiment, step 2 is performed as follows.

[0031] First, complete the hardware deployment and data acquisition required for control to support subsequent decoupled control. The specific process is as follows: A mass flow meter is installed on the feed pipeline of the condensation reactor to achieve real-time measurement of the feed flow rates of hydrochloric acid, urea, and hydrazine hydrate; a pH electrode and a platinum resistance temperature sensor are installed inside the reactor. Near-infrared spectral real-time concentration data and process parameter data are integrated into the existing DCS (Distributed Control System), and an independent control server, such as an APC server, is configured to achieve bidirectional data communication with the DCS system, enabling control commands to be quickly issued to the actuators; We continuously collect full-process data from more than 10 normal production batches, covering operating conditions under different raw material concentrations and loads. The data includes key indicators such as feed flow rate, pH value, temperature, concentration, and by-product content, providing a data foundation for the subsequent development of decoupling models.

[0032] Subsequently, through a combination of mechanistic analysis and data modeling, the coupling relationship between pH and temperature was clarified, and a decoupling control method was established. The specific process is as follows: The coupling degree between the two control loops is calculated using the relative gain matrix method, i.e.: With other parameters fixed, the first amplification factor of the change in hydrochloric acid feed rate on pH value and the second amplification factor when the temperature control loop is closed were tested respectively, and the relative gain of the pH control loop was calculated. The first amplification factor of the heating power change on temperature and the second amplification factor when the pH control loop is closed are tested to obtain the relative gain of the temperature control loop. If the relative gains of the two loops both deviate from 1, it indicates that there is a significant coupling effect, and the interference must be eliminated through a decoupling algorithm. A feedforward compensation decoupling scheme is adopted, and a two-loop feedforward compensation model is constructed based on the reaction kinetics mechanism and historical data of biuret: The feedforward compensation model for pH detection temperature is as follows:

[0033] in, pH comp This represents the effective pH value at the k-th sampling time after temperature compensation. pH raw This represents the raw measurement value of the pH electrode output at the k-th sampling time (without temperature compensation). K p1 This represents the static temperature compensation coefficient. T ( k () represents the measured temperature of the reaction liquid inside the reactor at the k-th sampling time. T ref This indicates the reference calibration temperature of the pH electrode. K d1 It is a dynamic compensation coefficient that can be adjusted according to the actual electrode characteristics. After compensation, the pH detection error is ≤0.03.

[0034] When the temperature control loop adjusts the heating power, it pre-compensates the pH detection value based on the effect of temperature change on the pH electrode detection value, thus avoiding malfunction of the pH control loop.

[0035] The feedforward compensation model for temperature-controlled hydrochloric acid feed is as follows:

[0036] in, ΔP ( k () represents the feedforward correction amount of the heating power of the heat transfer oil at the k-th sampling time. K p2 This represents the static feedforward gain of the hydrochloric acid feed heat effect. ΔF ( k () represents the instantaneous change in hydrochloric acid feed flow rate at the k-th sampling time. K d2 This represents the dynamic feedforward gain of the hydrochloric acid feed heat effect. ΔF ( k -1) represents the change in hydrochloric acid feed flow rate at the (k-1)th sampling time, α represents the first-order autoregressive coefficient, and β represents the second-order autoregressive coefficient. ΔP ( k -1) represents the feedforward correction amount of the heating power at the (k-1)th sampling time. ΔP ( k -2) represents the feedforward correction of heating power at the (k-2)th sampling time.

[0037] When the pH control loop adjusts the hydrochloric acid feed rate, the impact on the heat load of the reaction system is calculated based on the heat of reaction brought in by the hydrochloric acid and the feed temperature. The heating power of the temperature control loop is then fed forward to compensate for the temperature fluctuations caused by the change in feed rate.

[0038] Finally, a hierarchical control logic is adopted, with the inner loop providing rapid stabilization and the outer loop providing dynamic optimization, to enable the two loops to operate in coordination: Inner loop: pH rapid control loop: The deviation between the real-time pH value and the set value is input into the PID controller, and the output signal simultaneously adjusts the feed flow regulating valves of hydrochloric acid, urea, and hydrazine hydrate to maintain the predetermined ratio of the three. The loop prioritizes pH stability. When the pH deviation exceeds the set value, the temperature adjustment of the outer loop is paused, and the feed ratio is adjusted first to eliminate pH fluctuations and avoid a surge in side reactions.

[0039] Outer loop: Reaction conversion optimization loop: Based on the real-time concentrations of hydrazine hydrate and biurea detected by near-infrared spectroscopy, the deviation between the current reaction rate and the target reaction rate is calculated, and the optimal temperature setpoint is output in combination with the reaction kinetic model. By adjusting the flow rate of the heating / cooling medium, the reaction temperature is stabilized at the optimal set value, and the deviation is controlled to not exceed the set value. This ensures that the reaction rate matches the biuret formation rate and avoids the formation of impurities caused by local oversaturation.

[0040] In a preferred embodiment, during the raw material feeding stage, the molar ratio of hydrazine hydrate to urea is controlled at 1:2.3 to 2.8, and the pH value is controlled at 1 to 2.

[0041] In a preferred embodiment, step 3 is performed as follows.

[0042] First, we need to clarify the quantitative correspondence between the rheological parameters of the reaction solution and the crystallization state of biurea, so as to provide a basis for adjusting the stirring rate: The apparent viscosity and shear stress of the reaction solution are two essential monitoring parameters: In the early stage of the reaction, before biuret precipitates, the system is a homogeneous aqueous solution with a stable viscosity of 0.8–1.2 mPa·s; as the reaction proceeds, biuret concentration exceeds its solubility and crystals begin to precipitate, and the system becomes a solid-liquid suspension. The viscosity increases exponentially with the increase of the solid content of the crystals. When the solid content reaches 15%–20%, the viscosity can rise to 3–5 mPa·s. Moreover, the shear stress is directly related to the crystal particle size distribution, that is, the higher the content of fine crystals, the greater the shear stress at the same solid content.

[0043] Samples were collected at different reaction stages under actual process conditions, and rheological parameters, solid content, and particle size distribution data were measured offline. A correlation model between rheological parameters and crystallization state was constructed using a nonlinear regression method. Real-time rheological data can accurately predict the current solid content and average particle size, providing a quantitative basis for adjusting the stirring rate.

[0044] Based on the reaction process and rheological data, the stirring rate adjustment is divided into three stages, which operate in conjunction with the reaction process: In the early stage of the reaction, i.e., the homogeneous reaction stage: at this time, biuret has not yet precipitated, and the system viscosity is stable at around 1 mPa·s. The stirring speed is controlled at 200–300 rpm to enhance the mixing effect of the materials, eliminate local exothermic reactions of the hydrochloric acid and hydrazine hydrate neutralization reaction, avoid local pH and temperature deviations from the target range, and ensure the uniformity of the condensation reaction. Rheological data at this stage are mainly used to verify the homogeneity of the system mixing. When viscosity fluctuations exceed ±0.2 mPa·s, the stirring speed is appropriately increased to eliminate local concentration inhomogeneities.

[0045] In the mid-reaction stage, i.e., the crystallization nucleation stage: when near-infrared spectroscopy detects that the biuret concentration has reached the saturation threshold of 12%–15%, and rheological data shows that the viscosity begins to rise rapidly, it is determined that crystallization has begun. The stirring speed is then gradually reduced to 150–200 rpm. During this stage, sufficient shear force must be maintained to prevent crystal nuclei from agglomerating, while also preventing excessively high stirring speeds from breaking the nuclei. The nucleation rate is monitored through rheological data: if the viscosity increase rate exceeds 0.3 mPa·s / h, it indicates that the nucleation rate is too fast, and the stirring speed can be appropriately reduced. Simultaneously, feedback is sent to the dual closed-loop control loop to slightly lower the reaction temperature, slowing down the biuret formation rate and preventing the formation of a large number of fine crystals in a short period.

[0046] The later stage of the reaction, i.e., the crystal growth stage: When rheological data shows that the system viscosity rises above 3 mPa·s and the solid content exceeds 15%, the crystal growth stage begins, and the stirring rate is further reduced to 60–120 rpm. During this stage, the low stirring speed reduces crystal breakage while ensuring uniform system temperature and concentration, avoiding excessive local supersaturation and the formation of impurities. The crystal growth status is monitored through rheological data: if the shear stress abnormally increases, it indicates that the fine crystal content is too high. In this case, the stirring speed can be temporarily increased to break up the agglomerated small crystals, and then the low stirring speed should be resumed to promote crystal growth.

[0047] In a preferred embodiment, step 4 is performed as follows.

[0048] Measurement and calibration of basic parameters and determination of thresholds First, the calibration of each detection parameter and the determination of the process threshold are completed to provide a reliable basis for endpoint determination: Calibration of detection equipment: Perform metrological calibration on the near-infrared spectral concentration detection system, pH electrode, temperature sensor, and heating / cooling flow meter to prevent excessive errors.

[0049] Process threshold calibration: Based on historical data from more than 10 standard batches, calibrate the core threshold of the reaction endpoint. hydrazine hydrate conversion rate ≥95%; The cumulative heat of reaction reaches more than 98% of the theoretical heat of reaction; The rate of pH increase is less than 0.01 / minute for 5 consecutive minutes.

[0050] The endpoint determination logic of this application for multi-parameter fusion adopts a two-level determination rule of single-parameter anomaly filtering and multi-parameter consistency verification to avoid misjudgment caused by single data fluctuations. Level 1: Outlier Filtering: The real-time data of the three types of parameters are smoothed by a sliding window to remove single abnormal jump data: If the detection value of a certain parameter exceeds the normal process range for 3 consecutive times, it is automatically judged as a detection anomaly, the judgment weight of the parameter is temporarily removed, and the equipment inspection alarm is triggered at the same time.

[0051] Level 2: Multi-parameter consistency determination: A weighted voting mechanism is adopted, with the following weights for the three parameters: real-time concentration, cumulative heat of reaction, and pH trend. The reaction is considered to be at its endpoint when any of the following conditions are met: All three parameters reach the endpoint threshold simultaneously; Two parameters reach the threshold, and the deviation of the third parameter is less than 10% of the threshold and there is no continuous deviation trend.

[0052] This rule can effectively avoid random errors in near-infrared spectroscopy caused by crystal adhesion and calculation deviations in heat accumulation caused by heat dissipation fluctuations, thereby improving the accuracy of judgment.

[0053] In a preferred embodiment, the weights of the real-time concentration, cumulative heat of reaction, and pH change trend are 50%, 30%, and 20%, respectively.

[0054] Once the reaction endpoint is determined, the shutdown procedure is executed according to the following logic: Immediately close the steam feed valve of the reactor and stop heating. At the same time, record the total reaction time, final hydrazine hydrate conversion rate, predicted by-product content, and other data, and store them in the batch production file. Pause the feed adjustment of the dual closed-loop control loop, maintain the stirring rate at the current set value, and avoid crystallization and sedimentation; If only two parameters meet the standard during the judgment, and the third parameter has a slight deviation, an offline sampling verification will be automatically triggered. After the offline titration result confirms that the hydrazine hydrate conversion rate is ≥95% within 10 minutes, the subsequent procedures will be executed to avoid misjudgment.

[0055] After confirming the reaction endpoint, initiate a gradient cooling procedure to reduce impurity precipitation during the cooling process. Segmented cooling rate control: A segmented variable-rate cooling strategy is adopted, with an average cooling rate of approximately 0.5℃ / minute, executed in three stages: First stage (105℃→85℃): The cooling rate is set to 0.6℃ / min to rapidly reduce the supersaturation of the system, prevent biuret from continuing to generate byproducts, and ensure uniform temperature. Second stage (85℃→60℃): The cooling rate is reduced to 0.4℃ / min. The slow cooling promotes crystal growth and reduces the formation of fine crystals. At the same time, the stirring rate is adjusted in real time according to the rheological data to avoid crystal sedimentation. The third stage (60℃→40℃): The cooling rate is restored to 0.5℃ / minute, and the material discharge temperature is rapidly reduced, shortening the production cycle.

[0056] Cooling process linkage control: Real-time monitoring of reaction liquid temperature deviation during cooling process. When the actual temperature deviates from the set value by more than ±1℃, the opening of the cooling water valve is automatically adjusted, while avoiding scaling on the inner wall of the reactor caused by excessively fast cooling rate.

[0057] The following is a specific embodiment of this application.

[0058] Preliminary preparation stage Near-infrared spectral model calibration Sample set preparation: A total of 200 representative reaction solution samples were collected from different production batches and different reaction stages, covering the full concentration range of hydrazine hydrate (0-40%) and biuret (0-25%). Near-infrared spectroscopy and laboratory offline chromatographic analysis data were collected for each sample as calibration benchmarks. Spectral preprocessing: The acquired raw spectra are smoothed at 9 points, subjected to standard normal variable transformation (SNV) and first derivative processing to eliminate interference factors such as baseline drift and particle scattering; Model training: Partial least squares (PLS) was used to establish quantitative analysis models for hydrazine hydrate concentration and biuret concentration, with 150 samples as the calibration set and 50 samples as the validation set. On-site verification: The trained model was deployed to a near-infrared spectroscopy analyzer and ran continuously for 72 hours. Every 2 hours, manual sampling was performed to compare the online analysis results with the offline detection values, and the deviation data was recorded. Acceptance criteria: Model prediction correlation coefficient R 2 ≥0.99, cross-validation error RMSECV≤0.3%, average deviation of field validation ≤0.5%, detection response time ≤10 seconds, meeting the requirements for online detection.

[0059] Feedforward compensator parameter tuning Open-loop characteristic test: The pH control loop and temperature control loop were switched to manual mode, while other process parameters remained stable under normal production conditions. Two step temperature changes of ±2℃ and ±5℃ were applied respectively, and the response curves of the temperature detection value and the original pH detection value were recorded simultaneously. The sampling frequency was 1Hz, and the test duration was ≥30 minutes. Then, ±0.5m... 3 / h, ±1.0 m 3 The step flow rate of hydrochloric acid at two gradients was measured, and the response curves of hydrochloric acid flow rate, reactor temperature and heating power were recorded simultaneously. The sampling frequency was 1Hz and the test duration was ≥60 minutes. Step response fitting: For pH temperature compensation test data, the parameters of the first-order difference equation were fitted using the least squares method. Kp and Kd The optimization objective was to minimize the deviation between the compensated pH value and the offline laboratory test value at the baseline of 25℃. For the feedforward test data of hydrochloric acid, the parameters a, b, c, d, and e of the second-order difference equation were fitted using the system identification method, with the optimization objective being to minimize temperature fluctuation. The fitted parameters were then substituted into the control model, and offline simulation was performed using historical test data to verify the compensation effect. Closed-loop operation verification: The feedforward compensator parameters were written into the control system, the loop was switched to automatic mode, and a continuous 72-hour industrial trial run was conducted. During the operation, disturbance tests were applied with temperature step change and hydrochloric acid flow step change, and the dynamic response process of pH control deviation and temperature control deviation was recorded. During continuous operation, the product quality indicators (purity, yield, and particle size distribution of biuret) were compared daily to evaluate the control effect. Acceptance criteria: pH control deviation ≤ ±0.1℃ and temperature control deviation ≤ ±0.5℃ under normal operating conditions; pH recovery time ≤ 10s and temperature recovery time ≤ 30s under disturbed operating conditions; product yield ≥ 95% and particle size distribution variation coefficient ≤ 15%, meeting the production process requirements. Control loop threshold setting pH control loop settings: Set the pH control range to 1.0~2.0, the alarm threshold to 0.8 (low alarm) and 2.2 (high alarm), and the interlock threshold to 0.5 (low interlock) and 2.5 (high interlock); when the pH deviation exceeds 0.3, automatically pause the outer loop temperature adjustment and prioritize adjusting the feed ratio; Temperature control loop settings: Temperature control ranges are set in stages: 108-112℃ when conversion rate is <70%, 112-116℃ when conversion rate is 70%-90%, and 105-108℃ when conversion rate is >90%; Alarm thresholds are set to ±2℃ for each range, and interlock thresholds are set to ±5℃ for each range. Stirring rate control threshold settings: when the viscosity is 0.8 to 1.2 mPa·s, the stirring rate is 200 to 300 rpm; when the biuret concentration is 12% to 15% and the viscosity rise rate is >0.5 mPa·s / 10min, the stirring rate is reduced to 150 to 200 rpm; when the viscosity is >3 mPa·s and the solid content is >15%, the stirring rate is reduced to 60 to 120 rpm. Acceptance criteria: All threshold settings are written into the PLC control program, and the trigger logic is continuously simulated and tested 100 times with 100% accuracy and no false triggers or missed triggers.

[0060] The trained system will be in 10m 3 The test data, obtained from 30 consecutive batches of tests conducted in an industrial reactor and a 10,000-ton-per-year biuret unit, and compared with traditional processes, are as follows:

[0061] In this embodiment, the main reaction selectivity, total by-product content, particle size distribution coefficient of variation, filtration loss rate, and product yield of 30 consecutive batches conformed to a normal distribution. Statistical analysis showed that the 95% confidence interval for product yield was 94.9%~95.3%, proving that this process can stably achieve yield improvement with small batch-to-batch fluctuations, and the technical effect is reproducible. To avoid disclosing too many production process details, only the statistical analysis results are disclosed.

[0062] The comparative example used in this test is the conventional production process commonly used by domestic biuret manufacturers. The specific control process and parameters are as follows: Detection method: Manual offline sampling and titration analysis is adopted. Samples are taken every 30 minutes and sent to the central laboratory for titration to detect the concentration of hydrazine hydrate and pH value. The single detection cycle is about 45 minutes. The detection results are manually entered into the DCS system. Control method: Single-loop PID control is adopted, and only the reaction temperature is controlled at a fixed value. The pH value is manually adjusted by the operator based on the offline detection results. There is no automatic decoupling control or feedforward compensation mechanism, and no coordinated control logic for pH and temperature. Endpoint determination: A determination method combining fixed reaction time and offline concentration detection is adopted. The preset reaction time is 10 hours. After the time is reached, the concentration of hydrazine hydrate is sampled and detected. If the concentration is ≤0.5%, the endpoint is determined. Otherwise, the reaction continues and the concentration is retested every 15 minutes until the target is reached. Crystallization control: After the reaction is completed, circulating water is directly introduced for rapid cooling without any control over the cooling rate. The stirring rate is fixed at 120 rpm throughout the process. After cooling to room temperature, the material is directly discharged for centrifugation and filtration without any particle size control measures.

[0063] This process is a traditional production scheme commonly used in the biuret industry. In this test, the 30 batches in the control group were strictly operated according to the above process without any additional process optimization measures. All operation records and test data were collected synchronously by the DCS system on the production site and the LIMS system in the laboratory.

[0064] The data in the table above shows that... This solution replaces manual offline sampling with near-infrared online detection, reducing the concentration detection error from 1.2% to 0.42% and the detection response delay from 45 minutes to 10 seconds, eliminating the lag of offline detection. Combined with dual closed-loop decoupling control and feedforward compensation mechanism, the pH fluctuation is narrowed from ±0.5 to ±0.18 and the temperature fluctuation is narrowed from ±3.2℃ to ±0.8℃.

[0065] Test data quantitatively verified the direct correlation between process parameter stability and reaction selectivity: for every 0.1 decrease in pH fluctuation, the main reaction selectivity increased by 0.8 percentage points; for every 1°C decrease in temperature fluctuation, the main reaction selectivity increased by 0.7 percentage points. This scheme, by stabilizing pH and temperature, increased the main reaction selectivity from 91.8% to 95.2%, and reduced the total formation of byproducts biuret and aminourea from 0.82% to 0.14%, directly reducing the ineffective consumption of raw materials.

[0066] The effects of process control optimization in this application are transmitted to the post-processing stage. By controlling gradient cooling and matching stirring rate, the product particle size distribution variation coefficient is reduced from 34.7% to 13.8%, significantly reducing the fine crystal content and decreasing the filtration loss rate from 3.6% to 0.9%. The main contributors to the yield improvement are: improved main reaction selectivity reducing side reaction losses and improved particle size uniformity reducing filtration losses, resulting in a total yield improvement of 3.4 percentage points.

[0067] This application also proposes a control system for improving the yield of biuret to achieve the above control method, comprising: The data acquisition unit includes a mass flow meter, a density meter, an online pH analyzer, a multi-point temperature sensor, an online rheometer, and a near-infrared spectrometer, which are used to collect production process parameters in real time. In a preferred embodiment, the mass flow meter is installed on the hydrochloric acid, urea, and hydrazine hydrate feed lines; The densitometer is installed on the circulation pipeline at the outlet of the reactor, preferably a vibrating densitometer; The pH online analyzer uses an insertable PTFE-sheathed electrode, which is inserted 30 cm below the liquid surface in the reaction vessel. It has a built-in temperature compensation function and a measurement range of 0–14. The multi-point temperature sensors are respectively arranged in the upper, middle and lower layers of the reactor, the jacket inlet and outlet, and the feed pipeline. PT100 platinum resistance thermometers are preferred. The online rheometer is installed on the output shaft of the stirring motor, with a measurement range of 0 to 500 N·m, and can reflect the changes in stirring load in real time. The near-infrared spectrometer uses an insertable sapphire window fiber optic probe, which is inserted 40-50 cm below the liquid surface in the reaction vessel, avoiding the impact area of ​​the agitator.

[0068] The data acquisition unit is periodically cleaned with high-pressure pure water to prevent crystal adhesion.

[0069] The control unit, which is communicatively connected to the data acquisition unit, includes a programmable logic controller and an industrial computer, and is equipped with the control method for generating control commands. In a preferred embodiment, the programmable logic controller is a Siemens S7-400H redundant PLC or a product with equivalent performance, and is configured with 16 analog input modules, 8 analog output modules, and 32 digital input / output modules to meet the requirements of all acquisition and control points. The industrial computer (IPC) is rack-mounted and equipped with an i7 processor, 16GB of memory, and 1TB of solid-state storage. It also deploys advanced algorithms such as multi-parameter decoupling control and endpoint determination.

[0070] Analog signals from all field acquisition devices are connected to the PLC's AI module, and digital signals are connected to the DI module. After A / D conversion, they are stored in the PLC's input image area. The PLC and the industrial computer are connected via an industrial Ethernet switch and communicate using the Modbus TCP protocol. The PLC uploads real-time acquired data to the IPC every 100ms, and the IPC sends control parameter settings to the PLC every 1 second. The PLC's AO module outputs a 4-20mA control signal to the feed regulating valve and the heat transfer oil regulating valve, while the DO module outputs a switch signal to the discharge valve. It communicates with the inverter of the variable frequency stirring motor via the Profibus-DP bus to achieve speed regulation.

[0071] The execution unit, which is communicatively connected to the control unit, includes a feed regulating valve, a heat transfer oil regulating valve, a variable frequency stirring motor, and a discharge control valve, and is used to execute control commands; In a preferred embodiment, the feed regulating valve is installed on the raw material feed pipeline and is a highly corrosion-resistant pneumatic regulating valve; The heat transfer oil regulating valve is installed on the heat transfer oil inlet and outlet pipelines of the reactor jacket and is electrically operated. The variable frequency stirring motor is equipped with a vector frequency converter; The discharge control valve is installed on the discharge pipeline at the bottom of the reactor. It is a pneumatic switch valve with valve position feedback signal.

[0072] The data storage and analysis unit is communicatively connected to the control unit and the data acquisition unit, and is used to store historical production data and periodically update the prediction model parameters.

[0073] 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 yield of biuret, characterized in that, Includes the following steps: The characteristic spectrum of the reaction solution was collected, and the real-time concentration data of hydrazine hydrate and biuret were obtained by near-infrared spectroscopy analysis. A multi-parameter decoupled dual-closed-loop control architecture is adopted. The inner loop adjusts the feed flow ratio of hydrochloric acid, urea and hydrazine hydrate in real time to control the pH value of the reaction solution. The outer loop corrects the reaction temperature setpoint based on the real-time concentration data of hydrazine hydrate and biuret. Based on the feedback of the rheological properties of the reaction solution, the stirring rate is adjusted to maintain the stability of biurea crystal growth; By integrating the results of multi-parameter analysis of real-time concentration, cumulative heat of reaction, and pH value change trends, the reaction endpoint is determined. When the conversion rate of hydrazine hydrate reaches or exceeds the set value, heating is stopped and gradient cooling is initiated.

2. The method for controlling the yield of biuret as described in claim 1, characterized in that: In the dual closed-loop control architecture, the inner loop response time is less than 1 second, and the outer loop response time is less than 10 seconds.

3. The method for controlling the yield of biuret as described in claim 1, characterized in that: The reaction temperature control follows these rules: When the conversion rate of hydrazine hydrate is less than 70%, the reaction temperature is controlled at 108℃~112℃; When the conversion rate of hydrazine hydrate reaches 70% to 90%, the reaction temperature is controlled at 112℃ to 116℃. When the conversion rate of hydrazine hydrate is greater than 90%, the reaction temperature is controlled at 105℃~108℃.

4. The method for controlling the yield of biuret as described in claim 1, characterized in that: The multi-parameter decoupling is performed in the following manner: The coupling degree between the pH control loop and the temperature control loop is quantitatively calculated using the relative gain matrix method. When the relative gains of both loops deviate from 1, a feedforward compensation decoupling scheme is adopted, i.e.: When adjusting the temperature, the pH detection value is pre-compensated for temperature based on the effect of temperature changes on the pH electrode detection value; When adjusting the hydrochloric acid feed rate, the heating power of the temperature control loop is feedforward based on the reaction heat brought in by the hydrochloric acid to offset the temperature fluctuations caused by the feed rate change.

5. The method for controlling the yield of biuret as described in claim 1, characterized in that: The adjustment of the stirring rate based on the rheological characteristics of the reaction solution is as follows: when the viscosity of the reaction solution is between 0.8 and 1.2 mPa·s, the stirring rate is controlled at 200 to 300 rpm; when the concentration of biuret reaches 12% to 15%, the stirring rate is reduced to 150 to 200 rpm; when the viscosity of the reaction solution rises to above 3 mPa·s and the solid content exceeds 15%, the stirring rate is reduced to 60 to 120 rpm.

6. The method for controlling the yield of biuret as described in claim 1, characterized in that: When the rheological data shows that the viscosity rise rate exceeds 0.3 mPa·s / h, the feedback is sent to the dual closed-loop control loop to reduce the reaction temperature and slow down the biuret formation rate.

7. The method for controlling the yield of biuret as described in claim 1, characterized in that: The multi-parameter endpoint determination adopts a two-stage determination method: Level 1: Outlier filtering. The real-time data of the three types of parameters are smoothed by a sliding window. Single abnormal jump data is removed. Parameters that exceed the process range three times in a row are temporarily removed from the judgment weight. The second level adopts a weighted voting mechanism, with the weights of the three parameters being: real-time concentration, cumulative heat of reaction, and pH change trend. When all three parameters reach the endpoint threshold simultaneously, or when two parameters reach the threshold and the deviation of the third parameter is less than 10% of the threshold, the reaction is determined to be at its endpoint.

8. The method for controlling the yield of biuret as described in claim 7, characterized in that: The endpoint threshold includes: hydrazine hydrate conversion rate ≥95%; The cumulative heat of reaction reaches more than 98% of the theoretical heat of reaction; The rate of pH increase is less than 0.01 / minute for 5 consecutive minutes.

9. The method for controlling the yield of biuret as described in claim 1, characterized in that: The gradient cooling employs segmented control: 105℃→85℃: Cooling rate is 0.6℃ / minute; 85℃→60℃: Cooling rate is 0.4℃ / min, and the stirring rate is adjusted in real time according to rheological data; 60℃→40℃: The cooling rate is 0.5℃ / minute, rapidly reducing the material discharge temperature.

10. A control system for improving the yield of biuret by implementing the control method of any one of claims 1 to 9, characterized in that, include: The data acquisition unit includes a mass flow meter, a density meter, an online pH analyzer, a multi-point temperature sensor, an online rheometer, and a near-infrared spectrometer, which are used to collect production process parameters in real time. The control unit, which is communicatively connected to the data acquisition unit, includes a programmable logic controller and an industrial computer, and is equipped with the control method for generating control commands. The execution unit, which is communicatively connected to the control unit, includes a feed regulating valve, a heat transfer oil regulating valve, a variable frequency stirring motor, and a discharge control valve, and is used to execute control commands; The data storage and analysis unit is communicatively connected to the control unit and the data acquisition unit, and is used to store historical production data and periodically update the prediction model parameters.

Images