Novel intelligent lithium deposition control method for preparing lithium carbonate

Abstract

The invention relates to the technical field of lithium compounds, in particular to a novel intelligent lithium deposition control method for preparing lithium carbonate, which comprises the following steps: mixing lithium sulfate and sodium carbonate in a reaction kettle at 95 DEG C, monitoring the pH value to 9.0-10.5, adjusting the pulse width of the sodium carbonate if the deviation exceeds 0.1, obtaining the granularity and crystal morphology, calculating the growth rate and morphological characteristics, and recording the pulse width adjustment. The pH deviation frequency is calculated, the secondary nucleation risk is judged in combination with the growth rate and the particle size parameter n, and if the risk exists, the stirring rotating speed is lowered to 100 r / min, and the pH upper limit is adjusted to 10.0. According to the invention, the pH value of the reaction environment is monitored in real time to carry out micro-scale dropwise addition adjustment so as to stabilize the nucleation environment, obtain crystal growth and morphology characteristics, establish the association between process parameters and crystal quality, perspectively identify the secondary nucleation risk, dynamically optimize the stirring intensity and the reaction endpoint according to the risk judgment result, effectively inhibit the disordered nucleation process, and improve the nucleation efficiency. Crystals are guided to grow uniformly, and the particle size uniformity and the crystal form regularity of a lithium carbonate product are improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of lithium compounds, in particular to a novel intelligent lithium precipitation control method for preparing lithium carbonate. BACKGROUND

[0002] The technical field of lithium compounds includes the extraction, purification and preparation of alkali metal compounds, especially lithium carbonate. This technical field generally involves the separation and enrichment of lithium ions from salt lake brine or ore leaching solution through chemical reactions, and finally precipitates lithium carbonate products. Among them, the traditional lithium precipitation control method refers to adding a precipitant such as sodium carbonate to the lithium-rich solution during the preparation of lithium carbonate. This control usually relies on manual monitoring and manual adjustment of key process parameters, such as batch addition of sodium carbonate solution according to preset empirical values, and detection of the pH value or lithium ion concentration at the reaction endpoint at regular intervals to determine whether the precipitation is complete.

[0003] The traditional lithium precipitation control method relies on manual experience and regular detection, and it is difficult to accurately control the reaction process. Batch addition of sodium carbonate can easily lead to excessive local concentration in the reactor, causing a sharp fluctuation in pH value. This unstable reaction environment can easily induce secondary nucleation, resulting in a large number of fine grains, which can destroy the normal growth of the crystal, ultimately leading to uneven particle size distribution and poor crystal form, affecting the filtration performance and consistency between product batches, increasing the difficulty of subsequent processing and reducing the overall production efficiency. SUMMARY

[0004] The purpose of the present application is to solve the problems existing in the prior art, and a novel intelligent lithium precipitation control method for preparing lithium carbonate is proposed.

[0005] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows: a novel intelligent lithium precipitation control method for preparing lithium carbonate, comprising the following steps: S1: passing lithium sulfate solution and sodium carbonate solution into a lithium carbonate reactor, setting the jacket temperature of the lithium carbonate reactor to 95 DEG C, monitoring the real-time pH value, setting the target pH range to 9.0-10.5, setting the pH deviation threshold to 0.1, and determining whether the real-time pH value exceeds the target pH range and the absolute value of the difference between the target pH range boundary is greater than the pH deviation threshold. If the answer is yes, adjust the sodium carbonate solution feed pump drop pulse width to 50 milliseconds; S2: obtaining real-time particle size distribution data, calculating the crystal growth rate, collecting crystal morphology images, identifying the crystal particles in the crystal morphology images, calculating the aspect ratio and circularity, and generating a crystal morphology feature vector composed of the aspect ratio and the circularity; S3: record the number of times of adjusting the dropwise adding pulse width of the sodium carbonate solution feeding pump, calculate the pH deviation event frequency, obtain the crystal growth rate, input the real-time particle size distribution data into the Rosin-Rammler distribution model fitting, and calculate the particle size distribution characteristic parameters , based on the pH deviation event frequency, the crystal growth rate, and the particle size distribution characteristic parameters , determine whether the reaction process is in a secondary nucleation risk state; S4: based on the determination result that the reaction process is in a secondary nucleation risk state, adjust the stirring speed of the lithium carbonate reaction kettle from 150 rpm to 100 rpm, and adjust the upper limit of the target pH interval from 10.5 to 10.0.

[0006] As a further scheme of the present application, the crystal growth rate includes a D50 growth rate and a D90 growth rate, the pH deviation event frequency specifically refers to the total number of pH adjustments within a unit time window, and the particle size distribution characteristic parameters specifically refer to the homogeneity index in the Rosin-Rammler distribution model, and the determination result specifically refers to a Boolean value indicating whether the reaction process is in a secondary nucleation risk state.

[0007] As a further scheme of the present application, the step of adjusting the dropwise adding pulse width of the sodium carbonate solution feeding pump specifically includes: calculating the absolute difference between the real-time pH value and the nearest boundary of the target pH interval to obtain a pH deviation value; based on the pH deviation value and the pH deviation threshold value, using a proportional-integral-derivative control algorithm to calculate a pulse width adjustment amount, which comprehensively considers the current deviation, historical deviation accumulation, and deviation change rate; the parameters of the proportional-integral-derivative control algorithm are adaptively set according to reaction data of at least 100 batches collected online; the calculated pulse width adjustment amount is applied to the current dropwise adding pulse width of the sodium carbonate solution feeding pump to generate an updated dropwise adding pulse width.

[0008] As a further scheme of the present application, the step of identifying the crystal particles in the crystal morphology image and calculating the aspect ratio and the circularity specifically includes: performing Gaussian filtering on the collected crystal morphology image, and then using the Otsu algorithm for global threshold segmentation to convert the image into a binary image containing only crystals and background; using morphological opening operation to eliminate small noise points in the binary image and smooth the crystal particle boundaries, and then using a watershed segmentation algorithm to accurately separate the mutually adhered crystal particles to obtain the pixel area of each independent crystal particle; For each individual crystal grain pixel region, a minimum bounding rectangle is fitted to extract its principal axis length and minor axis length. The ratio of the principal axis length to the minor axis length is calculated to obtain the aspect ratio. At the same time, the actual area of ​​the pixel region is calculated. With perimeter Through formula Once the circularity is obtained, a feature dataset containing all crystal grain morphology parameters is established.

[0009] As a further aspect of the present invention, the step of determining whether the reaction process is in a state of secondary nucleation risk specifically comprises: Based on the pH deviation event frequency, the D50 growth rate, the D90 growth rate, and the particle size distribution characteristic parameters Real-time risk index is calculated using a secondary nucleation risk assessment model. ; The secondary nucleation risk assessment model is as follows: ; in, Represents a real-time risk index. This represents the frequency of the pH deviation events. This represents the growth rate of the D90. This represents the growth rate of the D50. The particle size distribution characteristic parameter n represents the particle size distribution characteristic parameter. The weighting coefficients representing the frequency of the pH deviation events. A weighting coefficient representing the ratio of the D90 growth rate to the D50 growth rate. The weighting coefficient representing the deviation of the particle size distribution characteristic parameter n, and , , The sum of the three is 1. The maximum permissible pH deviation event frequency, The target homogeneity index; The value range is from 0.4 to 0.6. The value range is from 0.2 to 0.4. The value range is from 0.1 to 0.3; Set a predetermined risk threshold The calculated real-time risk index is 0.8. Exceeding the predetermined risk threshold When the reaction process is in a state of secondary nucleation risk, the Boolean value is a true result.

[0010] As a further aspect of the present invention, the maximum permissible pH deviation event frequency The value ranges from 6 to 10 times per minute; The target homogeneity index The value range is from 3.0 to 4.0.

[0011] As a further aspect of the present invention, the specific duration of the unit time window is set to 60 seconds; The calculation process for the D50 growth rate and the D90 growth rate is as follows: The real-time granularity distribution data is collected once at the beginning and end of each unit time window to obtain the initial D50 value, initial D90 value, end D50 value and end D90 value; The difference between the ending D50 value and the initial D50 value is divided by the duration of the unit time window to obtain the D50 growth rate. The difference between the final D90 value and the initial D90 value is divided by the duration of the unit time window to obtain the D90 growth rate.

[0012] As a further aspect of the present invention, the generated crystal morphology feature vector also includes eccentricity and convexity; The eccentricity is calculated by fitting an ellipse to the pixel region of each individual crystal particle and then calculating the ratio of the focal distance to the major axis length of the fitted ellipse. The convexity is calculated by obtaining the pixel area of ​​each independent crystal particle, calculating its convex hull area, and using the ratio of the actual area of ​​the crystal particle to the convex hull area as the convexity. A four-dimensional crystal morphology feature vector is then generated, consisting of the aspect ratio, the circularity, the eccentricity, and the convexity.

[0013] As a further aspect of the present invention, the step of adjusting the stirring speed of the lithium carbonate reactor and the upper limit of the target pH range specifically comprises: Once the system receives a judgment that the reaction process is at risk of secondary nucleation, it immediately executes a linear rate decay program to uniformly reduce the stirring speed of the lithium carbonate reactor from 150 rpm to 100 rpm within 60 seconds. Simultaneously, the control system smoothly adjusts the upper limit setting of the target pH range from 10.5 to 10.0. The adjustment process uses a first-order hysteresis filtering algorithm to generate the adjusted process parameter setting values. After the stirring speed and the upper limit of the target pH range are adjusted, the risk status of secondary nucleation is continuously monitored. If the risk status is resolved within 3 consecutive monitoring cycles, the process parameters are restored to the initial settings.

[0014] As a further aspect of the present invention, the concentration of the lithium sulfate solution is 1.5 mol / L to 2.5 mol / L, and the feed flow rate is 10 L / min to 20 L / min; The concentration of the sodium carbonate solution is from 1.0 mol / L to 1.8 mol / L; When adjusting the dripping pulse width of the sodium carbonate solution feed pump, the adjustment step size of the dripping pulse width is 5% to 10% of the current pulse width, and the maximum adjustment amount in a single instance does not exceed 100 milliseconds.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the nucleation environment is stably controlled by real-time monitoring of the pH value of the reaction environment and micro-quantitative dropwise adjustments. At the same time, the crystal growth rate and morphology characteristics are acquired online, establishing an intrinsic correlation between process parameters and crystal quality. This allows for the proactive identification of secondary nucleation risks, and the dynamic optimization of stirring intensity and reaction endpoint based on the risk assessment results. This effectively suppresses disordered nucleation processes, guides crystal growth towards a uniform direction, and significantly improves the particle size uniformity and crystal regularity of lithium carbonate products, thereby ensuring stable product quality. Attached Figure Description

[0016] Figure 1 This is a flowchart of the novel intelligent lithium carbonate deposition control preparation method of the present invention; Figure 2 This is a flowchart illustrating the adjustment of the dripping pulse width of the sodium carbonate solution feed pump according to the present invention. Figure 3 This is a flowchart of the crystal particle feature vector generation process of the present invention; Figure 4 This is a flowchart for determining the risk status of secondary nucleation in this invention; Figure 5 This is a flowchart of the secondary nucleation risk handling and parameter recovery process of this invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the technical solutions of this invention and do not constitute a limitation on the scope of protection.

[0018] In the description of this invention, the process flow relationships or material and energy transfer paths indicated by terms such as "unit," "step," "equipment," "pipeline," "material flow," and "process parameters" are defined based on the process flow diagram or equipment structure diagram corresponding to the embodiments. This way of expression is only used to clearly illustrate the logical relationship between the elements in the technical solution, and not to limit the specific equipment connection method or physical layout. The term "multiple" includes two or more technical units, including but not limited to multiple reactors, pumps, valves, separation units, or detection instruments and other expandable elements. The specific number is determined according to specific process requirements or production scale and needs to be specifically stated.

[0019] Example 1 In this embodiment, the concentration of the sodium carbonate solution is set to its lower limit of 1.0 mol / L; when adjusting the dripping pulse width of the sodium carbonate solution feed pump, the adjustment step size of the dripping pulse width is set to its lower limit, i.e., 5% of the current pulse width; in the secondary nucleation risk assessment model, the weighting coefficients... , and A set of constraints ( The effective combination values ​​within their respective ranges are set to 0.4, 0.4, and 0.2, respectively; the maximum permissible pH deviation event frequency. Use its lower limit of 6 times / minute; target homogeneity index The lower limit of the range was adopted as 3.0; the concentration of lithium sulfate solution was adopted as the lower limit of the range as 1.5 mol / L, and the feed flow rate was adopted as the lower limit of the range as 10 L / min.

[0020] Please see Figure 1 and Figure 2 A novel intelligent lithium carbonate deposition control method for preparing lithium carbonate, based on the above-mentioned intelligent lithium carbonate deposition control, includes the following steps: S1: Introduce lithium sulfate solution and sodium carbonate solution into the lithium carbonate reactor. Set the jacket temperature of the lithium carbonate reactor to 95℃. Monitor the real-time pH value. Set the target pH range to 9.0-10.5 and the pH deviation threshold to 0.1. If the real-time pH value exceeds the target pH range and the absolute value of the difference between the real-time pH value and the boundary of the target pH range is greater than the pH deviation threshold, adjust the drip pulse width of the sodium carbonate solution feed pump to 50 milliseconds.

[0021] The specific steps for adjusting the drip pulse width of the sodium carbonate solution feed pump are as follows: Calculate the absolute difference between the real-time pH value and the nearest boundary of the target pH range to obtain the pH deviation value; Based on the pH deviation value and pH deviation threshold, a proportional-integral-derivative control algorithm is used to calculate the pulse width adjustment amount, which comprehensively considers the current deviation, historical deviation accumulation, and deviation change rate. The parameters of the proportional-integral-derivative control algorithm are adaptively tuned based on reaction data from at least 100 batches collected online. The calculated pulse width adjustment is applied to the current dripping pulse width of the sodium carbonate solution feed pump to generate the updated dripping pulse width.

[0022] The concentration of the sodium carbonate solution is 1.0 mol / L to 1.8 mol / L; When adjusting the dripping pulse width of the sodium carbonate solution feed pump, the adjustment step size of the dripping pulse width is 5% to 10% of the current pulse width, and the maximum adjustment amount in a single step does not exceed 100 milliseconds.

[0023] A 1.5 mol / L lithium sulfate solution and a 1.0 mol / L sodium carbonate solution were pumped into a 500 L jacketed lithium carbonate reactor via a separate feed line at a flow rate of 10 L / min. Heat transfer oil was circulated within the reactor jacket, and the jacket temperature was precisely controlled at 95 °C using an external temperature control unit. A Mettler Toledo InPro3253i high-temperature composite pH electrode was installed in the middle of the liquid phase zone of the reactor. This electrode was connected to a Siemens S7-1500 series programmable logic controller (PLC) via a 4-20 mA signal. The PLC continuously collected real-time pH values ​​at a sampling period of 200 milliseconds. The target pH range set within the control system was 9.0 to 10.5, with a pH deviation threshold of 0.1.

[0024] Within a specific control cycle, the PLC collects a real-time pH value of 8.88 at 10:05:30.200. The system first determines whether this value is outside the target range [9.0, 10.5], and the result is yes. Subsequently, the system calculates the absolute difference between this real-time pH value and the nearest boundary (9.0) of the target pH range, i.e. Next, the system compares this difference with the preset pH deviation threshold of 0.1. Since 0.12 is greater than 0.1, the system triggers the pulse width adjustment program of the sodium carbonate solution feed pump. At this time, the current dripping pulse width of the sodium carbonate solution feed pump is 400 milliseconds.

[0025] The pulse width adjustment is calculated based on the proportional-integral-derivative (PID) control algorithm; the parameters of this algorithm (proportional coefficient) Integral coefficient Differential coefficients The values ​​were determined based on regression analysis and optimization of production data from the previous 150 batches, with the following set values: , , The proportional term (P) is calculated as follows: The integral term (I) is calculated as follows: The PLC has accumulated the pH deviation value over the last 60 seconds, and its time integral value is 1.85. The differential term (D) is calculated as: current deviation ( The deviation from 200 milliseconds ago (corresponding to pH 8.89, the deviation is...) The difference is ,but The calculated total pulse width adjustment is: millisecond.

[0026] The system performs constraint checks on the adjustment amount; firstly, the minimum adjustment step size is 5% of the current pulse width, i.e. The maximum adjustment step size is 100 milliseconds; secondly, the maximum adjustment step size is 100 milliseconds; the calculated adjustment step size of 39.1 milliseconds is greater than the minimum adjustment step size of 20 milliseconds and less than the maximum adjustment step size of 100 milliseconds, so the adjustment step size is valid; therefore, this adjustment step size is applied to the current pulse width to generate the updated dripping pulse width: Milliseconds; the PLC writes this new pulse width value into a control instruction and sends it to the sodium carbonate solution feed pump driver, completing one closed-loop control adjustment.

[0027] To verify the rationality of the pH deviation threshold (0.1), multiple parallel experiments were conducted. Under the condition that other process parameters remained the same, only the pH deviation threshold was changed to investigate its impact on the stability of the reaction process and the quality of the final product. The experimental data are recorded in Table 1.

[0028] Table 1. Impact of pH Deviation Threshold Setting on Reaction Process and Product Performance

[0029] As shown in Table 1, when the pH deviation threshold is set to 0.1, the pH fluctuation of the reaction system remains at a low level, the consumption of sodium carbonate solution is most economical, and the content of the key impurity L2 in the final product is the lowest, indicating that this threshold is set to the optimal value. If the threshold is too low (0.05), the control system will respond too frequently, resulting in over-adjustment of the actuator (feed pump) and waste of raw materials. If the threshold is too high (0.20 and 0.50), the pH will deviate from the target range for too long, affecting the microenvironment for crystal growth and reducing the quality of the final product.

[0030] Please see Figure 1 and Figure 3S2: Acquire real-time particle size distribution data, calculate crystal growth rate, collect crystal morphology images, identify crystal particles in crystal morphology images, calculate aspect ratio and roundness, and generate a crystal morphology feature vector composed of aspect ratio and roundness.

[0031] Crystal growth rates include D50 growth rate and D90 growth rate.

[0032] The specific steps for identifying crystal grains in a crystal morphology image and calculating the aspect ratio and roundness are as follows: The acquired crystal morphology image was processed by Gaussian filtering, and then the Otsu algorithm was used for global threshold segmentation to convert the image into a binary image containing only the crystal and the background. Morphological opening operations are used to eliminate tiny noise in the binarized image and smooth the boundaries of crystal particles. Then, the watershed segmentation algorithm is used to accurately separate the mutually adhered crystal particles and obtain the pixel region of each independent crystal particle. For each individual crystal grain pixel region, a minimum bounding rectangle is fitted to extract its principal axis length and minor axis length. The ratio of the principal axis length to the minor axis length is calculated to obtain the aspect ratio. At the same time, the actual area of ​​the pixel region is calculated. With perimeter Through formula Obtain the roundness and establish a feature dataset containing all crystal grain morphology parameters.

[0033] The generated crystal morphology feature vector also includes eccentricity and convexity; The eccentricity is calculated by fitting an ellipse to the pixel region of each individual crystal particle and then calculating the ratio of the focal distance to the major axis length of the fitted ellipse. The convexity is calculated by obtaining the pixel region of each independent crystal particle, calculating its convex hull area, and using the ratio of the actual area of ​​the crystal particle to the convex hull area as the convexity. This generates a four-dimensional crystal morphology feature vector composed of aspect ratio, circularity, eccentricity, and convexity.

[0034] The specific duration of the unit time window is set to 60 seconds; The calculation process for D50 growth rate and D90 growth rate is as follows: Real-time granularity distribution data were collected once at the beginning and end of each unit time window to obtain the initial D50 value, initial D90 value, end D50 value, and end D90 value. The D50 growth rate is obtained by dividing the difference between the final D50 value and the initial D50 value by the duration of the unit time window. The D90 growth rate is obtained by dividing the difference between the final D90 value and the initial D90 value by the duration of the unit time window.

[0035] During the reaction, the reaction slurry is pumped into the online monitoring system through a circulation pipeline connected to the bottom of the reactor. This system integrates an Insatec laser particle size analyzer from Malvern Panaco and a PVMV819i particle imaging measurement system from Mettler Toledo, Switzerland. The laser particle size analyzer outputs complete particle size distribution data every 30 seconds, including characteristic particle size values ​​such as D10, D50, and D90. The PVM particle imaging measurement system continuously acquires grayscale images of crystal morphology with a resolution of 1280x960 pixels every 10 seconds.

[0036] At the beginning of a unit time window (set to 60 seconds) (e.g., 10:10:00), the control system records the initial particle size distribution data measured by the laser particle size analyzer, where the initial D50 value is 52.5 μm and the initial D90 value is 95.8 μm; at the end of the time window (10:11:00), data is collected again, yielding an ending D50 value of 55.7 μm and an ending D90 value of 101.3 μm; the D50 growth rate is calculated as follows: the difference between the ending D50 value and the initial D50 value (… Dividing the μm by the duration of the unit time window (60 seconds) yields the D50 growth rate. μm / s; The calculation process for the D90 growth rate is as follows: the difference between the final D90 value and the initial D90 value ( Dividing the growth rate of D90 by the unit time window duration (60 seconds) yields the growth rate of D90. μm / s.

[0037] Simultaneously, the image processing unit analyzes the crystal morphology images acquired by the PVM system. First, a 5x5 Gaussian kernel is applied to the original image acquired at 10:10:35 for convolution filtering to smooth the image and suppress random noise. After processing, Otsu's method is used to calculate a global threshold of 138. Pixels with a value greater than 138 are set to 255 (white, representing crystals), and those less than or equal to 138 are set to 0 (black, representing the background), generating a binarized image. Subsequently, a morphological opening operation (erosion followed by dilation) of a 3x3 structuring element is performed to eliminate isolated small white noise points in the image and to smooth the boundary contours of the crystal particles. For the phenomenon of multiple crystal particles adhering together in the image, a watershed segmentation algorithm based on distance transform is adopted. First, the distance from all white pixels to the nearest black background pixel in the binarized image is calculated to generate a distance transform map. Then, local maxima are found on the distance transform map as seed points, ultimately achieving effective segmentation of the adhering crystal clusters and obtaining 157 independent crystal particle pixel regions.

[0038] For one successfully segmented individual crystal particle, its pixel region contains 3240 pixels; by fitting this region using the minimum bounding rectangle algorithm, its principal axis length is found to be 88 pixels and its minor axis length is 50 pixels; the aspect ratio is calculated as follows: The actual area of ​​this pixel region 3240 pixels Its perimeter is calculated using a boundary tracing algorithm. It is 235 pixels; the calculated value for roundness is... Next, an ellipse was fitted to the pixel region, resulting in a focal distance of 45 pixels and a major axis length of 87 pixels. The eccentricity was then calculated. Finally, the convex hull of this pixel region is calculated, and its area is 3450 pixels. Then the convexity is These four parameters are combined to generate the morphological feature vector of the crystal [1.76, 0.735, 0.517, 0.939]. This process is repeated for all 157 independent crystal particles, and finally a feature dataset containing all crystal morphological parameters is generated for subsequent analysis.

[0039] Please see Figure 1 and Figure 4 S3: Record the number of times the sodium carbonate solution feed pump pulse width is adjusted, calculate the pH deviation event frequency, obtain the crystal growth rate, input the real-time particle size distribution data into the Rosing-Lammler distribution model for fitting, calculate the particle size distribution characteristic parameter n, and determine whether the reaction process is in a state of secondary nucleation risk based on the pH deviation event frequency, crystal growth rate, and particle size distribution characteristic parameter n.

[0040] pH deviation event frequency specifically refers to the total number of pH adjustments within a unit time window.

[0041] The particle size distribution characteristic parameter n specifically refers to the homogeneity index in the Rossin-Lammler distribution model.

[0042] The judgment result is a Boolean value indicating whether the reaction process is in a state of secondary nucleation risk.

[0043] The specific steps for determining whether the reaction process is at risk of secondary nucleation are as follows: Based on the frequency of pH deviation events, D50 growth rate, D90 growth rate, and particle size distribution parameter n, a real-time risk index is calculated using a secondary nucleation risk assessment model. ; The secondary nucleation risk assessment model is as follows ; in, Represents a real-time risk index. Represents the frequency of pH deviation events. Represents the growth rate of D90. Represents the D50 growth rate. The particle size distribution characteristic parameter n, Weighting coefficients representing the frequency of pH deviation events. The weighting coefficient representing the ratio of the growth rate of D90 to the growth rate of D50. The weighting coefficient representing the deviation of the particle size distribution characteristic parameter n, and , , The sum of the three is 1. The maximum permissible pH deviation event frequency, The target homogeneity index; The value range is from 0.4 to 0.6. The value range is from 0.2 to 0.4. The value range is from 0.1 to 0.3; Set a predetermined risk threshold The calculated real-time risk index is 0.8. Exceeding the predetermined risk threshold When the reaction process is in a state of secondary nucleation risk, the Boolean value is a true result.

[0044] Maximum permissible pH deviation event frequency The value ranges from 6 to 10 times per minute; Target homogeneity index The value range is from 3.0 to 4.0.

[0045] Within the same time window of S2 (10:10:00 to 10:11:00), the control system counts the pH adjustment actions in S1; the log shows that within this 60-second period, due to pH fluctuations exceeding the threshold, a total of 5 adjustments were made to the dripping pulse width of the sodium carbonate solution feed pump; therefore, the frequency of pH deviation events is... Recorded at 5 times / minute; simultaneously, the crystal growth rate and D50 growth rate within this time window are obtained from S2. The growth rate of D90 is 0.053 μm / s. It is 0.092 μm / s.

[0046] At the end of this time window (10:11:00), the complete particle size distribution data (a set of data pairs containing the volume percentage of particles in different particle size ranges) measured by the laser particle size analyzer in S2 is input into the preset Rosin-Rammler distribution model; the homogeneity index of the distribution model, i.e. the particle size distribution characteristic parameter n, is calculated by fitting with the nonlinear least squares method, and its value is 2.65.

[0047] The Rosin-Rammler distribution model described above is a mathematical model used to describe the particle size distribution of a particle population. Its cumulative distribution function is usually expressed as: ;in, Indicates particle size is smaller than The volume fraction of particles, It is the characteristic particle size, and This refers to the uniformity index in this embodiment, which reflects the width of the particle size distribution. The larger the value, the narrower the particle size distribution and the better the uniformity.

[0048] Next, the system invokes the secondary nucleation risk assessment model to determine whether the reaction process is in a state of secondary nucleation risk; this model is a linear weighted model: The physical meanings and values ​​of the parameters in the formula are as follows: The secondary nucleation risk index represents real-time calculation; This is the frequency of pH deviation events in the current window, with a measured value of 5 times / minute; and These are the crystal growth rates for D90 and D50, respectively, with measured values ​​of 0.092 μm / s and 0.053 μm / s. It is the homogeneity index of the Rosing-Lammer distribution, with a measured value of 2.65; the weighting coefficient , and These represent the contributions of pH deviation event frequency, growth rate ratio, and homogeneity index deviation to the total risk, respectively. Based on historical data analysis, pH stability has the most direct impact on secondary nucleation, followed by abnormal crystal growth behavior; therefore, the following values ​​are set: , , ; This is the maximum permissible pH deviation event frequency, set at 6 times / minute. This value is based on experimental verification: in multiple batches of experiments, when the pH adjustment frequency exceeded 6 times per minute, the micro powder content in the product increased significantly, indicating that crystal breakage or secondary nucleation phenomena intensified. The target uniformity index is set to 3.0. The basis for this value is that when the uniformity index is 3.0, the product has a narrow particle size distribution and good flowability and compaction performance, which is the ideal product state.

[0049] The aforementioned secondary nucleation risk assessment model is a comprehensive index model used to quantify the risk of generating fine crystals (micropowder) during the crystallization process; this model integrates three key process indicators: (Frequency of pH deviation events) reflects the stability of the chemical environment; a high frequency indicates drastic fluctuations in supersaturation. The growth rate ratio reflects the consistency of crystal growth. When the growth rate of large particles (D90) is significantly faster than that of medium particles (D50), it may indicate abnormal growth or agglomeration and breakage. (Uniformity index deviation) reflects the degree of deviation between the actual particle size distribution and the target distribution; the larger the deviation, the worse the product uniformity. This is achieved through linear weighting (…). Combining these three normalized indicators yields a comprehensive risk index. .

[0050] Substitute the above parameter values ​​into the risk assessment model for calculation: ; ; ; ; .

[0051] Preset risk threshold The threshold is 0.8; the process for determining this threshold is as follows: A retrospective analysis of production data from 100 historical batches is performed, and the average value of each batch during the mid-reaction period is calculated. The value was then correlated with the micron content (the percentage of particles smaller than 10 μm) of the final product in that batch, and a correlation analysis was performed.

[0052] Table 2. Correlation Analysis between Risk Index and Product Micron Powder Content

[0053] As shown in Table 2, when When the value is below 0.8, the product's micron powder content remains stable at a low level (5%); while when When the value exceeds 0.8, the micron content increases sharply; therefore, the risk threshold is set... Set to 0.8.

[0054] The calculated real-time risk index With risk threshold Comparison; due to The system generates a Boolean value of "true" to indicate that the current reaction process is in a state of secondary nucleation risk, and transmits this signal to the next process parameter adjustment module. The calculation results show that the current growth rate ratio of D90 to D50 is too high, which is the main reason for the risk index exceeding the standard.

[0055] Please see Figure 1 and Figure 5 S4: Based on the judgment that the reaction process is in a state of secondary nucleation risk, adjust the stirring speed of the lithium carbonate reactor from 150 rpm to 100 rpm, and adjust the upper limit of the target pH range from 10.5 to 10.0.

[0056] The specific steps for adjusting the stirring speed and upper limit of the target pH range in the lithium carbonate reactor are as follows: Once the system receives a judgment that the reaction process is at risk of secondary nucleation, it immediately executes a linear rate decay program to uniformly reduce the stirring speed of the lithium carbonate reactor from 150 rpm to 100 rpm within 60 seconds. Meanwhile, the control system smoothly adjusts the upper limit setting of the target pH range from 10.5 to 10.0. The adjustment process uses a first-order hysteresis filtering algorithm to generate the adjusted process parameter setting values. After adjusting the stirring speed and the upper limit of the target pH range, continuously monitor the risk status of secondary nucleation. If the risk status is resolved within 3 consecutive monitoring cycles, restore the process parameters to the initial settings.

[0057] The concentration of the lithium sulfate solution is 1.5 mol / L to 2.5 mol / L, and the feed flow rate is 10 L / min to 20 L / min.

[0058] When the control system receives a Boolean signal of "true" generated by S3 indicating that the reaction process is in a state of secondary nucleation risk at 10:11:01, it immediately starts the preset process parameter adjustment program. This program first sends a command to the frequency converter (VFD) controlling the agitator of the lithium carbonate reactor to execute a linear rate decay. The setpoint for the agitator speed will decrease uniformly over 60 seconds, starting from the current 150 rpm. The specific setpoint function is as follows: ,in This is a time variable ranging from 0 to 60 seconds; for example, the RPM setpoint sent to the VFD at the 30th second after program startup is... At 10:12:01, the stirring speed stabilized at 100 rpm.

[0059] Simultaneously, the control system adjusts the target pH range of the pH control loop in S1; the upper limit of the target range is smoothly adjusted from 10.5 to 10.0, while the lower limit remains unchanged at 9.0; to avoid system oscillation caused by sudden changes in the setpoint, a first-order hysteresis filtering algorithm is used in the adjustment process, the formula of which is: ,in The sampling period is 200ms. The filter time constant is set to 10 seconds. This algorithm causes the upper limit pH setting to gradually change from 10.5 to 10.0 over approximately 30 seconds, rather than jumping instantaneously. After the adjustment is complete, the control logic of S1 will operate according to the new target pH range [9.0, 10.0].

[0060] After the process parameters are adjusted, the system enters the continuous monitoring and evaluation phase; during the next four monitoring cycles (60 seconds each), the secondary nucleation risk assessment model of S3 runs continuously: *- First monitoring cycle (10:12:01-10:13:01): Calculated real-time risk index Decreased to 0.91; *- Second monitoring period (10:13:01-10:14:01): Calculated real-time risk index The risk index dropped to 0.78, indicating it was below the risk threshold of 0.8, and the risk status was lifted. *- Third monitoring period (10:14:01-10:15:01): Calculated real-time risk index =0.75; *- Fourth monitoring period (10:15:01-10:16:01): Calculated real-time risk index It is 0.72.

[0061] Because the risk status was in a de-escalation state for three consecutive monitoring periods (the second, third, and fourth periods) (i.e. At 10:16:01, the control system triggered the parameter recovery mechanism; the stirring speed and the upper limit of the target pH range gradually recovered to the initial set value within 120 seconds in the same smooth and linear manner as when it decreased, that is, the stirring speed recovered to 150 rpm and the target pH range recovered to [9.0, 10.5].

[0062] Table 3 Comparison of Product Performance in Example 1

[0063] Example 2 In this embodiment, the concentration of the sodium carbonate solution is set to its upper limit of 1.8 mol / L; when adjusting the dripping pulse width of the sodium carbonate solution feed pump, the adjustment step size of the dripping pulse width is set to its upper limit, i.e., 10% of the current pulse width; in the secondary nucleation risk assessment model, the weighting coefficients... , and Given a set of combinations that are within their respective ranges and whose sum is 1, the preferred choice is... and The upper limits are set to 0.6, 0.3, and 0.1 respectively; the maximum permissible pH deviation event frequency. Use its upper limit of 10 times / minute; target homogeneity index The upper limit of the range of 4.0 is adopted; the concentration of lithium sulfate solution is adopted as the upper limit of the range of 2.5 mol / L, and the feed flow rate is adopted as the upper limit of the range of 20 L / min.

[0064] Please see Figure 1 and Figure 2 A novel intelligent lithium carbonate deposition control method for preparing lithium carbonate, based on the above-mentioned intelligent lithium carbonate deposition control, includes the following steps: S1: Introduce lithium sulfate solution and sodium carbonate solution into the lithium carbonate reactor. Set the jacket temperature of the lithium carbonate reactor to 95℃. Monitor the real-time pH value. Set the target pH range to 9.0-10.5 and the pH deviation threshold to 0.1. If the real-time pH value exceeds the target pH range and the absolute value of the difference between the real-time pH value and the boundary of the target pH range is greater than the pH deviation threshold, adjust the drip pulse width of the sodium carbonate solution feed pump to 50 milliseconds.

[0065] The specific steps for adjusting the drip pulse width of the sodium carbonate solution feed pump are as follows: Calculate the absolute difference between the real-time pH value and the nearest boundary of the target pH range to obtain the pH deviation value; Based on the pH deviation value and pH deviation threshold, a proportional-integral-derivative control algorithm is used to calculate the pulse width adjustment amount, which comprehensively considers the current deviation, historical deviation accumulation, and deviation change rate. The parameters of the proportional-integral-derivative control algorithm are adaptively tuned based on reaction data from at least 100 batches collected online. The calculated pulse width adjustment is applied to the current dripping pulse width of the sodium carbonate solution feed pump to generate the updated dripping pulse width.

[0066] The concentration of the sodium carbonate solution is 1.0 mol / L to 1.8 mol / L; When adjusting the dripping pulse width of the sodium carbonate solution feed pump, the adjustment step size of the dripping pulse width is 5% to 10% of the current pulse width, and the maximum adjustment amount in a single step does not exceed 100 milliseconds.

[0067] A 2.5 mol / L lithium sulfate solution and a 1.8 mol / L sodium carbonate solution were precisely injected into a 1000L continuous crystallization reactor with a jacket temperature control via metering pumps at a flow rate of 20 L / min. Saturated water vapor at 95°C was circulated within the reactor jacket to maintain a constant temperature of the materials inside the reactor. An online pH monitoring system (Endress+Hauser Orbisint CPS11D) installed inside the reactor transmitted the real-time measured pH signal to a central distributed control system (DCS) via a PROFIBUS bus. The target pH operating range set in the DCS was 9.0 to 10.5, and a pH deviation threshold of 0.1 was defined.

[0068] In a specific control scenario, the DCS collected a real-time pH value of 10.65 at 14:20:15.500. The system program first compared this value with the target range [9.0, 10.5] and determined that it exceeded the upper limit. Subsequently, the program calculated the absolute difference between this real-time value and the upper boundary of the target range, i.e. This deviation value is determined to be greater than the preset pH deviation threshold of 0.1, so the system immediately initiates the pulse width adjustment logic for the sodium carbonate solution feed pump; at this moment, the drop pulse width setting of the sodium carbonate solution feed pump is 550 milliseconds.

[0069] The pulse width adjustment is calculated by the PID function block built into the DCS; the parameters of this function block (proportional gain) Integral time Differential time The formula (P) is derived from data from over 200 historical batches, optimized using the process response curve method and the Ziegler-Nichols tuning rule; the proportional term (P) is calculated as follows: The integral term (I) is calculated as follows: the cumulative historical deviation of the DCS internal integrator is -2.5 pH·s, then... The differential term (D) is calculated as: current deviation ( The deviation from 500 milliseconds ago (corresponding to pH 10.63, the deviation is...) The difference is ,but The calculated total pulse width adjustment is: millisecond.

[0070] The system performs an output limit check on the adjustment amount; the upper limit of the adjustment step size is 10% of the current pulse width, i.e. Milliseconds; the maximum adjustment amount per pulse is 100 milliseconds; the calculated absolute value of the adjustment amount is 87.375 milliseconds, which is greater than the upper limit of the adjustment step size of 55 milliseconds and less than the maximum adjustment amount per pulse of 100 milliseconds, therefore the adjustment amount is valid; thus, the system applies this adjustment amount to the current pulse width to generate the updated drip pulse width: Milliseconds; This setting is sent by the DCS system to the frequency converter of the sodium carbonate feed pump, thereby reducing the injection volume of sodium carbonate solution and causing the pH value to return to the target range.

[0071] To determine the rationality of the 10% upper limit of the drop pulse width adjustment step, a series of comparative experiments were conducted, and the results are shown in Table 4.

[0072] Table 4. Impact of the Upper Limit of Pulse Width Adjustment Step Size on Control Performance

[0073] As can be seen from the data in Table 4, when the upper limit of the adjustment step size is set to 10%, the system response speed is the fastest (the adjustment time is the shortest), and the pH overshoot is the smallest. This is conducive to maintaining the stability of the reaction environment and finally obtaining the product crystals with the largest particle size. Too small a step size (5% or 8%) leads to a slow response, while too large a step size (12%) causes system oscillation, which is not conducive to control.

[0074] Please see Figure 1 and Figure 3 S2: Acquire real-time particle size distribution data, calculate crystal growth rate, collect crystal morphology images, identify crystal particles in crystal morphology images, calculate aspect ratio and roundness, and generate a crystal morphology feature vector composed of aspect ratio and roundness.

[0075] Crystal growth rates include D50 growth rate and D90 growth rate.

[0076] The specific steps for identifying crystal grains in a crystal morphology image and calculating the aspect ratio and roundness are as follows: The acquired crystal morphology image was processed by Gaussian filtering, and then the Otsu algorithm was used for global threshold segmentation to convert the image into a binary image containing only the crystal and the background. Morphological opening operations are used to eliminate tiny noise in the binarized image and smooth the boundaries of crystal particles. Then, the watershed segmentation algorithm is used to accurately separate the mutually adhered crystal particles and obtain the pixel region of each independent crystal particle. For each individual crystal grain pixel region, a minimum bounding rectangle is fitted to extract its principal axis length and minor axis length. The ratio of the principal axis length to the minor axis length is calculated to obtain the aspect ratio. At the same time, the actual area of ​​the pixel region is calculated. With perimeter Through formula Obtain the roundness and establish a feature dataset containing all crystal grain morphology parameters.

[0077] The generated crystal morphology feature vector also includes eccentricity and convexity; The eccentricity is calculated by fitting an ellipse to the pixel region of each individual crystal particle and then calculating the ratio of the focal distance to the major axis length of the fitted ellipse. The convexity is calculated by obtaining the pixel region of each independent crystal particle, calculating its convex hull area, and using the ratio of the actual area of ​​the crystal particle to the convex hull area as the convexity. This generates a four-dimensional crystal morphology feature vector composed of aspect ratio, circularity, eccentricity, and convexity.

[0078] The specific duration of the unit time window is set to 60 seconds; The calculation process for D50 growth rate and D90 growth rate is as follows: Real-time granularity distribution data were collected once at the beginning and end of each unit time window to obtain the initial D50 value, initial D90 value, end D50 value, and end D90 value. The D50 growth rate is obtained by dividing the difference between the final D50 value and the initial D50 value by the duration of the unit time window. The D90 growth rate is obtained by dividing the difference between the final D90 value and the initial D90 value by the duration of the unit time window.

[0079] The circulating slurry from the reactor is fed into an online particle characteristic analysis unit at a flow rate of 2 L / min. In this unit, a SympatecHELOS / KR laser diffractometer and a QICPIC dynamic image analyzer work together. The HELOS / KR provides a high-precision particle size distribution curve and characteristic particle size parameters once per minute, while the QICPIC captures images of crystal particles passing through the flow cell at a rate of 30 frames per second and calculates their morphological parameters in real time.

[0080] Within a specific time window (set to 60 seconds), from 14:30:00 to 14:31:00, the control system recorded the following data: At the start of the window, the initial D50 value measured by HELOS / KR was 68.0 μm, and the initial D90 value was 120.5 μm; at the end of the window, the final D50 value was 71.8 μm, and the final D90 value was 128.0 μm; the D50 growth rate was calculated as follows: μm / s; the growth rate of D90 is calculated as follows: μm / s.

[0081] During this period, the QICPIC image analysis software processed thousands of crystal images. The processing flow for one frame of the image is as follows: First, a Gaussian low-pass filter (kernel size 7x7, standard deviation σ=1.5) was applied to preprocess the original 2048x2048 pixel image to eliminate background noise and uneven illumination. Then, an adaptive threshold segmentation algorithm based on local entropy was used to convert the image into a binary image. This algorithm is more robust than the global Otsu algorithm for cases where crystal edges are blurred or overlaps exist. After that, a morphological closing operation (dilation followed by erosion, using 5x5 disk-shaped structural elements) was performed to fill any small holes that may exist inside the crystal and to connect fracture boundaries. Finally, a marker-controlled watershed algorithm was applied, using pre-identified crystal nuclei centroids as markers to accurately segment crystals that are in contact with each other or slightly aggregated. A total of 213 independent crystal particles were identified in this processing.

[0082] Parameter calculations were performed on a typical independent crystal particle. The pixel region of this particle, statistically analyzed, contains 8560 pixels. Its minimum bounding rectangle was calculated using the Rotating Calipers algorithm, yielding a principal axis length of 155 pixels and a minor axis length of 75 pixels. The aspect ratio was then calculated to be... The perimeter of a particle is calculated by tracing its outline using Freeman chaincode. It has 410 pixels, but its actual area is... 8560 pixels The calculated value for roundness is... Furthermore, an ellipse is fitted to this pixel region. The distance from the focal point to the center of the ellipse is 58 pixels, and the length of the major semi-axis is 77 pixels. The eccentricity is then calculated. Finally, the convex hull area of ​​this pixel region is calculated to be 9150 pixels. Then the convexity is These four parameters [2.07, 0.641, 0.753, 0.935] constitute the four-dimensional morphological feature vector of the crystal, and are stored in the database along with the feature vectors of the other 212 crystals for subsequent population equilibrium model analysis.

[0083] Please see Figure 1 and Figure 4 S3: Record the number of times the sodium carbonate solution feed pump pulse width is adjusted, calculate the pH deviation event frequency, obtain the crystal growth rate, input the real-time particle size distribution data into the Rosing-Lammler distribution model for fitting, calculate the particle size distribution characteristic parameter n, and determine whether the reaction process is in a state of secondary nucleation risk based on the pH deviation event frequency, crystal growth rate, and particle size distribution characteristic parameter n.

[0084] pH deviation event frequency specifically refers to the total number of pH adjustments within a unit time window.

[0085] The particle size distribution characteristic parameter n specifically refers to the homogeneity index in the Rossin-Lammler distribution model.

[0086] The judgment result is a Boolean value indicating whether the reaction process is in a state of secondary nucleation risk.

[0087] The specific steps for determining whether the reaction process is at risk of secondary nucleation are as follows: Based on the frequency of pH deviation events, D50 growth rate, D90 growth rate, and particle size distribution parameter n, a real-time risk index is calculated using a secondary nucleation risk assessment model. ; The secondary nucleation risk assessment model is as follows ; in, Represents a real-time risk index. Represents the frequency of pH deviation events. Represents the growth rate of D90. Represents the D50 growth rate. The particle size distribution characteristic parameter n, Weighting coefficients representing the frequency of pH deviation events. The weighting coefficient representing the ratio of the growth rate of D90 to the growth rate of D50. The weighting coefficient representing the deviation of the particle size distribution characteristic parameter n, and , , The sum of the three is 1. The maximum permissible pH deviation event frequency, The target homogeneity index; The value range is from 0.4 to 0.6. The value range is from 0.2 to 0.4. The value range is from 0.1 to 0.3; Set a predetermined risk threshold The calculated real-time risk index is 0.8. Exceeding the predetermined risk threshold When the reaction process is in a state of secondary nucleation risk, the Boolean value is a true result.

[0088] Maximum permissible pH deviation event frequency The value ranges from 6 to 10 times per minute; Target homogeneity index The value range is from 3.0 to 4.0.

[0089] Within the time window from 14:30:00 to 14:31:00, the DCS event log shows that due to pH fluctuations reaching the control boundary, the system performed adjustments to the dripping pulse width of the sodium carbonate solution feed pump a total of 8 times; accordingly, the frequency of pH deviation events within this time window is... The rate was determined to be 8 times per minute; the D50 growth rate was obtained from the calculations of S2. μm / s and D90 growth rate μm / s.

[0090] At the end of the window at 14:31:00, the volumetric distribution data (128 particle size channels) output by the HELOS / KR particle size analyzer was imported into the Rosing-Lammler distribution model for fitting; the Levenberg-Marquardt algorithm was used for iterative calculation to determine the goodness of fit. The homogeneity index, i.e. the particle size distribution characteristic parameter n, is 3.30 when the value reaches 0.998.

[0091] The system then initiates real-time calculations of the secondary nucleation risk assessment model, which is as follows: The parameters in the formula are defined and assigned values ​​as follows: The real-time risk index to be calculated; This represents the frequency of pH deviation events, currently at 8 times per minute. and These represent crystal growth rates, currently 0.125 μm / s and 0.0633 μm / s, respectively. This is the uniformity index of granularity distribution, currently valued at 3.30; weighting coefficient. , , Based on process experience, at high concentrations and high flow rates, pH fluctuations and abnormal growth rates have a stronger inducing effect on secondary nucleation; therefore, it is set as follows: , , ; The maximum permissible pH deviation event frequency was set to 10 times / minute; this value was determined based on experimental data: when the reaction load was increased to the level of this embodiment, if the pH adjustment frequency exceeded 10 times / minute, it would cause a large number of tiny pits to appear on the crystal surface, which is a sign of secondary nucleation germination; The target uniformity index is set to 4.0; this value is set to pursue a more uniform and narrower particle size distribution in order to meet the stringent requirements of high-end battery materials for powder consistency.

[0092] Substitute the above values ​​into the model for calculation: ; ; ; ; .

[0093] Preset risk threshold The value remains at 0.8; the applicability of this threshold has been verified under high-load production conditions, as shown in Table 5.

[0094] Table 5. Impact of Target Homogeneity Index Setting on Product Performance

[0095] As shown in Table 5, the target homogeneity index Setting it to 4.0 yields the product with the best uniformity (highest actual n value), the narrowest particle size distribution (smallest D90), and the highest compaction density of the prepared battery electrode sheets. Although setting it to 4.5 slightly improves performance, it significantly increases the difficulty of process control and cost. Therefore, 4.0 is considered the optimal target value.

[0096] The calculated real-time risk index With risk threshold Compare; because Significantly greater than The system determines that the current reaction process is in a state of high risk of secondary nucleation and immediately generates an alarm and control trigger signal with a Boolean value of "true". The result indicates that the high frequency of pH fluctuations and the excessively large D90 / D50 growth rate ratio jointly caused the risk index to exceed the limit.

[0097] Please see Figure 1 and Figure 5 S4: Based on the judgment that the reaction process is in a state of secondary nucleation risk, adjust the stirring speed of the lithium carbonate reactor from 150 rpm to 100 rpm, and adjust the upper limit of the target pH range from 10.5 to 10.0.

[0098] The specific steps for adjusting the stirring speed and upper limit of the target pH range in the lithium carbonate reactor are as follows: Once the system receives a judgment that the reaction process is at risk of secondary nucleation, it immediately executes a linear rate decay program to uniformly reduce the stirring speed of the lithium carbonate reactor from 150 rpm to 100 rpm within 60 seconds. Meanwhile, the control system smoothly adjusts the upper limit setting of the target pH range from 10.5 to 10.0. The adjustment process uses a first-order hysteresis filtering algorithm to generate the adjusted process parameter setting values. After adjusting the stirring speed and the upper limit of the target pH range, continuously monitor the risk status of secondary nucleation. If the risk status is resolved within 3 consecutive monitoring cycles, restore the process parameters to the initial settings.

[0099] The concentration of the lithium sulfate solution is 1.5 mol / L to 2.5 mol / L, and the feed flow rate is 10 L / min to 20 L / min.

[0100] After receiving the trigger signal from S3 indicating that the "secondary nucleation risk status" was "true" at 14:31:02, the DCS immediately executed two adjustment operations in parallel. First, it sent a command to the servo motor controller controlling the magnetic stirrer at the bottom of the reactor to initiate a linear ramp-down program for the stirring speed. The speed setpoint decreased linearly from 150 rpm to 100 rpm within 60 seconds. The DCS refreshed the speed setpoint every second; for example, at 14:31:32 (30 seconds after program startup), the speed setpoint was... 100 revolutions per minute; by 14:32:02, the stirring speed had precisely stabilized at 100 revolutions per minute.

[0101] Simultaneously, the upper limit setting of the target pH range in the pH control loop of the DCS also begins to adjust; through a first-order hysteresis filter function block with dead zone and change rate limit, the upper limit is smoothly transitioned from 10.5 to 10.0; the filter time constant is set to 15 seconds, and the change rate limit is 0.05 pH / s; this makes the adjustment process of the setting smooth and shock-free, and after about 45 seconds, the new target pH range [9.0, 10.0] is fully effective.

[0102] After 14:32:02, the process parameters were adjusted, and the system entered the continuous monitoring phase of risk status; the DCS calculated and recorded the real-time risk index every 60 seconds. * - First monitoring period (14:32:02-14:33:02): Dropped to 0.95; *- Second monitoring period (14:33:02-14:34:02): The value dropped to 0.79, at which point the risk status was lifted; *- Third monitoring period (14:34:02-14:35:02): Stable at 0.76; *- Fourth monitoring period (14:35:02-14:36:02): It further decreased to 0.73.

[0103] The system detected that for three consecutive monitoring cycles (i.e., from 14:33:02 to 14:36:02), starting from the second monitoring cycle... The value was below the risk threshold of 0.8; therefore, at 14:36:02, the system automatically triggered the parameter recovery program; the stirring speed and the upper limit of the target pH range were smoothly restored to the initial 150 rpm and [9.0, 10.5] range within 120 seconds using the same ramp and filtering algorithm as when it was falling.

[0104] Table 6 Comparison of Product Performance in Example 2

[0105] The above embodiments illustrate preferred embodiments of the present invention. Any equivalent adjustments to the technical solution based on chemical engineering methods are within the scope of protection, including but not limited to: using different chemical reaction processes to achieve technical effects, optimizing the production process flow, adjusting the raw material ratio scheme, improving reactor design, and improving energy efficiency. Any implementation scheme derived from reasonable modifications to the production process, raw material utilization, equipment configuration, or system integration level without departing from the core technology of the present invention should be considered within the scope of protection defined by the claims of the present invention.

Claims

1. A novel intelligent lithium deposition control method for preparing lithium carbonate, characterized in that, Includes the following steps: S1: Introduce lithium sulfate solution and sodium carbonate solution into a lithium carbonate reactor. Set the jacket temperature of the lithium carbonate reactor to 95°C. Monitor the real-time pH value. Set the target pH range to 9.0-10.5 and the pH deviation threshold to 0.

1. Determine if the real-time pH value exceeds the target pH range and the absolute value of the difference between the real-time pH value and the boundary of the target pH range is greater than the pH deviation threshold. If the determination is yes, adjust the drip pulse width of the sodium carbonate solution feed pump to 50 milliseconds. S2: Acquire real-time particle size distribution data, calculate crystal growth rate, collect crystal morphology images, identify crystal particles in the crystal morphology images, calculate aspect ratio and roundness, and generate a crystal morphology feature vector composed of aspect ratio and roundness. S3: Record the number of times the sodium carbonate solution feed pump pulse width is adjusted, calculate the pH deviation event frequency, obtain the crystal growth rate, input the real-time particle size distribution data into the Rossin-Lammler distribution model for fitting, calculate the particle size distribution characteristic parameter n, and determine whether the reaction process is in a state of secondary nucleation risk based on the pH deviation event frequency, the crystal growth rate, and the particle size distribution characteristic parameter n. S4: Based on the judgment that the reaction process is in a state of secondary nucleation risk, adjust the stirring speed of the lithium carbonate reactor from 150 rpm to 100 rpm, and adjust the upper limit of the target pH range from 10.5 to 10.

0.

2. The novel intelligent lithium deposition control method for preparing lithium carbonate according to claim 1, characterized in that, The crystal growth rate includes the D50 growth rate and the D90 growth rate; the pH deviation event frequency specifically refers to the total number of pH adjustments within a unit time window; the particle size distribution characteristic parameter n specifically refers to the homogeneity index in the Rossin-Lammler distribution model; and the judgment result is specifically a Boolean value indicating whether the reaction process is in a state of secondary nucleation risk.

3. The novel intelligent lithium deposition control method for preparing lithium carbonate according to claim 1, characterized in that, The specific steps for adjusting the drip pulse width of the sodium carbonate solution feed pump are as follows: The absolute difference between the real-time pH value and the nearest boundary of the target pH range is calculated to obtain the pH deviation value; Based on the pH deviation value and the pH deviation threshold, a proportional-integral-derivative control algorithm is used to calculate the pulse width adjustment amount, which comprehensively considers the current deviation, historical deviation accumulation, and deviation change rate. The parameters of the proportional-integral-derivative control algorithm are adaptively tuned based on reaction data collected online from at least 100 batches. The calculated pulse width adjustment is applied to the current dripping pulse width of the sodium carbonate solution feed pump to generate an updated dripping pulse width.

4. The novel intelligent lithium deposition control method for preparing lithium carbonate according to claim 1, characterized in that, The specific steps of identifying crystal grains in the crystal morphology image and calculating the aspect ratio and the roundness are as follows: The acquired crystal morphology image is processed by Gaussian filtering, and then global threshold segmentation is performed using the Otsu algorithm to convert the image into a binary image containing only the crystal and the background. The tiny noise in the binarized image is eliminated by morphological opening operation and the crystal particle boundaries are smoothed. Then, the watershed segmentation algorithm is used to accurately separate the mutually sticky crystal particles and obtain the pixel region of each independent crystal particle. For each individual crystal grain pixel region, a minimum bounding rectangle is fitted to extract its principal axis length and minor axis length. The ratio of the principal axis length to the minor axis length is calculated to obtain the aspect ratio. At the same time, the actual area of ​​the pixel region is calculated. With perimeter Through formula Once the circularity is obtained, a feature dataset containing all crystal grain morphology parameters is established.

5. The novel intelligent lithium deposition control method for preparing lithium carbonate according to claim 2, characterized in that, The specific steps for determining whether the reaction process is in a state of secondary nucleation risk are as follows: Based on the pH deviation event frequency, the D50 growth rate, the D90 growth rate, and the particle size distribution characteristic parameter n, a real-time risk index is calculated using a secondary nucleation risk assessment model. ; The secondary nucleation risk assessment model is as follows: ; in, Represents a real-time risk index. This represents the frequency of the pH deviation events. This represents the growth rate of the D90. This represents the growth rate of the D50. The particle size distribution characteristic parameter n represents the particle size distribution characteristic parameter. The weighting coefficients representing the frequency of the pH deviation events. A weighting coefficient representing the ratio of the D90 growth rate to the D50 growth rate. The weighting coefficient representing the deviation of the particle size distribution characteristic parameter n, and , , The sum of the three is 1. The maximum permissible pH deviation event frequency, The target homogeneity index; The value range is from 0.4 to 0.

6. The value range is from 0.2 to 0.

4. The value range is from 0.1 to 0.3; Set a predetermined risk threshold The calculated real-time risk index is 0.

8. Exceeding the predetermined risk threshold When the reaction process is in a state of secondary nucleation risk, the Boolean value is a true result.

6. The novel intelligent lithium deposition control method for preparing lithium carbonate according to claim 5, characterized in that, The maximum permissible pH deviation event frequency The value ranges from 6 to 10 times per minute; The target homogeneity index The value range is from 3.0 to 4.

0.

7. The novel intelligent lithium deposition control method for preparing lithium carbonate according to claim 2, characterized in that, The specific duration of the unit time window is set to 60 seconds; The calculation process for the D50 growth rate and the D90 growth rate is as follows: The real-time granularity distribution data is collected once at the beginning and end of each unit time window to obtain the initial D50 value, initial D90 value, end D50 value and end D90 value; The difference between the ending D50 value and the initial D50 value is divided by the duration of the unit time window to obtain the D50 growth rate. The difference between the final D90 value and the initial D90 value is divided by the duration of the unit time window to obtain the D90 growth rate.

8. The novel intelligent lithium deposition control method for preparing lithium carbonate according to claim 1, characterized in that, The generated crystal morphology feature vector also includes eccentricity and convexity; The eccentricity is calculated by fitting an ellipse to the pixel region of each individual crystal particle and then calculating the ratio of the focal distance to the major axis length of the fitted ellipse. The convexity is calculated by obtaining the pixel area of ​​each independent crystal particle, calculating its convex hull area, and using the ratio of the actual area of ​​the crystal particle to the convex hull area as the convexity. A four-dimensional crystal morphology feature vector is then generated, consisting of the aspect ratio, the circularity, the eccentricity, and the convexity.

9. The novel intelligent lithium deposition control method for preparing lithium carbonate according to claim 4, characterized in that, The specific steps for adjusting the stirring speed of the lithium carbonate reactor and the upper limit of the target pH range are as follows: Once the system receives a judgment that the reaction process is at risk of secondary nucleation, it immediately executes a linear rate decay program to uniformly reduce the stirring speed of the lithium carbonate reactor from 150 rpm to 100 rpm within 60 seconds. Simultaneously, the control system smoothly adjusts the upper limit setting of the target pH range from 10.5 to 10.

0. The adjustment process uses a first-order hysteresis filtering algorithm to generate the adjusted process parameter setting values. After the stirring speed and the upper limit of the target pH range are adjusted, the risk status of secondary nucleation is continuously monitored. If the risk status is resolved within 3 consecutive monitoring cycles, the process parameters are restored to the initial settings.

10. The novel intelligent lithium deposition control method for preparing lithium carbonate according to claim 1, characterized in that, The concentration of the lithium sulfate solution is 1.5 mol / L to 2.5 mol / L, and the feed flow rate is 10 L / min to 20 L / min; The concentration of the sodium carbonate solution is from 1.0 mol / L to 1.8 mol / L; When adjusting the dripping pulse width of the sodium carbonate solution feed pump, the adjustment step size of the dripping pulse width is 5% to 10% of the current pulse width, and the maximum adjustment amount in a single instance does not exceed 100 milliseconds.

Images