A method and system for controlling the inhibition of impurities in an o-aminobenzoate esterification reaction

Abstract

The application discloses a kind of o-aminobenzyl ester esterification reaction impurity inhibition control method and system, belong to industrial process control field, it includes: by micro potential sensing probe obtains the micro zone potential signal generated by key intermediate dynamic behavior;Local transfer function is obtained by using pseudo-random sequence tracer identification;Dynamic bifurcation probability is calculated and three-dimensional risk heat map is generated;Predicting impurity evolution path to generate reverse-time pre-compensation command containing pre-compensation amplitude and advance;Coordinated stirrer frequency modulation and feed pump pulse generate standing wave interference in risk area;Tracking entropy acceleration rate dynamically adjusts modulation amplitude;After reaction, controller parameters are corrected by cause-effect path backtracking algorithm.The application adopts closed-loop control logic based on physical field intervention and information entropy feedback, can realize early prediction and accurate target inhibition to impurity generation risk, improve product purity and production process stability.

Description

Technical Field

[0001] This invention relates to the field of industrial process control, and in particular to a method and system for impurity suppression and control in the esterification reaction of o-aminobenzoic acid ester. Background Technology

[0002] o-Aminobenzoate is an important fine chemical intermediate, and its esterification reaction is a key step in the synthesis process. During the esterification reaction, due to factors such as reaction kinetics and uneven heat and mass transfer, side reactions are prone to occur, generating impurities that affect the purity and yield of the final product. Therefore, precise control of the esterification process to suppress impurity formation at the source is of paramount importance for improving product quality and reducing production costs.

[0003] Currently, in the industrial production of o-aminobenzoic acid ester esterification, process control mainly relies on the monitoring and adjustment of macroscopic and average parameters within the reactor. For example, sensors such as thermometers and pressure gauges installed at specific locations within the reactor acquire overall temperature and pressure information, and traditional proportional-integral-derivative (PID) control algorithms are used to adjust the flow rate of heating or cooling media, agitator speed, etc., to maintain these macroscopic parameters near their set values. Some advanced control systems may incorporate process analysis techniques such as spectral analysis to monitor the overall component concentration of the materials within the reactor online.

[0004] However, the aforementioned existing technical solutions have significant limitations. First, their control relies on the overall or local macroscopic average parameters of the reactor, failing to detect and characterize localized high-temperature "hot spots" or "dead zones" of insufficient material mixing caused by uneven flow fields within the reactor. These microscopic regions are precisely the main sources of impurity formation. Second, this type of control is inherently lagging; the control system only responds when side reactions have already occurred and have an observable impact on macroscopic parameters or overall concentration. By this time, impurities have already formed and are difficult to reverse. Furthermore, control measures are usually global, such as uniformly increasing the stirring speed, which not only consumes a lot of energy but may also cause unnecessary disturbances to other parts of the reaction system. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides an impurity suppression and control method and system for the esterification reaction of o-aminobenzoic acid ester. This method employs a program control strategy that combines micro-region potential signal sensing, dynamic bifurcation probability risk assessment, reverse-time pre-compensation standing wave intervention, and cross-batch causal backtracking learning. This approach enables predictive suppression of impurity generation risk and adaptive optimization of control parameters.

[0006] To achieve the above objectives, this application adopts the following technical solution: In a first aspect, a method for controlling impurities in the esterification reaction of o-aminobenzoic acid ester is provided, comprising: The dynamic behavior of key intermediates collected by micro potential sensing probes at several spatial locations inside the reactor is obtained, and micro-region potential signals are generated. An inert tracer controlled by a pseudo-random sequence is injected into the reactor, and the diffusion response of each micro-region to the inert tracer is monitored according to the potential signal of the micro-region. The local transfer function is obtained by identification. Based on the local transfer function and the benchmark transfer function used to characterize the ideal reaction state, the dynamic bifurcation probability of each microregion is calculated, and a three-dimensional risk heat map is generated based on the dynamic bifurcation probability. Based on the evolution path of the impurity generation probability predicted by the three-dimensional risk heat map, a reverse-time pre-compensation command containing the pre-compensation magnitude and the pre-compensation lead amount is generated. According to the reverse time pre-compensation command, the speed of the stirrer is coordinated with the pulse flow of the micro-feeding pump in frequency modulation mode with a phase difference to form the interference of mechanical shear wave and concentration wave, generating standing waves in the high-risk areas marked by the three-dimensional risk heat map. Track the entropy acceleration rate of the high-risk area and dynamically adjust the modulation amplitude of the frequency modulation mode according to the entropy acceleration rate; After the reaction is completed, the actual impurity distribution of the final product is obtained, and the actual impurity distribution is correlated with the local transfer function, the three-dimensional risk heat map, the reverse time pre-compensation command and the entropy acceleration rate recorded in this batch of reaction. The controller parameters are then corrected using a causal path backtracking algorithm.

[0007] Based on the above technical solution, in the impurity suppression and control method for the esterification reaction of o-aminobenzoic acid ester provided in this application, a program control strategy combining micro-region potential signal sensing, dynamic bifurcation probability risk assessment, reverse time pre-compensation standing wave intervention, and cross-batch causal backtracking learning is adopted, which can achieve predictive suppression of impurity generation risk and adaptive optimization of control parameters.

[0008] In conjunction with the first aspect above, in one possible implementation, the dynamic behavior of the key intermediate acquired by the micro-potential sensing probes at several spatial locations within the reactor includes: Acquire the raw electrical signal output by the miniature potential sensing probe array; The original electrical signal is digitally filtered and normalized to obtain a preprocessed signal; The preprocessed signal is input into an intermediate migration model used to describe the relationship between electrical signals and molecular concentrations, and the concentration gradient and migration velocity of the key intermediate are inverted. The concentration gradient and the migration velocity are then combined into a micro-region potential signal.

[0009] In conjunction with the first aspect above, in one possible implementation, calculating the dynamic bifurcation probability of each microregion based on the local transfer function and the benchmark transfer function used to characterize the ideal reaction state, and generating a three-dimensional risk heatmap based on the dynamic bifurcation probability includes: The local transfer function is compared with the reference transfer function to calculate the deviation matrix containing the deviation information of each micro-region; The micro-regions corresponding to the elements in the deviation matrix whose values ​​change abruptly are extracted as candidate abnormal micro-regions; The spatial location of the candidate abnormal micro-region is weighted and fused with the dynamic bifurcation probability to generate a three-dimensional risk heat map.

[0010] In conjunction with the first aspect described above, in one possible implementation, the method further includes: Density clustering analysis was performed on the three-dimensional risk heatmap to identify high-risk clusters; Extract the geometric centroid coordinates and circumscribed rectangular bounding box of the high-risk cluster to generate the effective range parameters of the standing wave.

[0011] In conjunction with the first aspect above, in one possible implementation, generating a reverse-time pre-compensation command that includes the pre-compensation magnitude and pre-compensation lead amount, based on the evolution path of the impurity generation probability predicted by the three-dimensional risk heatmap, includes: Based on the standing wave's effective range parameters, the prediction target area is located; Within the predicted target area, the evolution path of the impurity generation probability is predicted using a Kalman filter to obtain the risk score peak and its occurrence time; Based on the peak risk score and the time of its occurrence, the initial pre-compensation magnitude and pre-compensation advance are calculated, and the two are combined into a reverse-time pre-compensation command.

[0012] In conjunction with the first aspect above, in one possible implementation, according to the reverse-time pre-compensation command, the rotational speed of the stirrer is coordinated with the pulse flow rate of the micro-feed pump in a frequency modulation mode with a phase difference to form interference between mechanical shear waves and concentration waves, generating standing waves in the high-risk areas identified in the three-dimensional risk heat map, including: The pre-compensation magnitude, pre-compensation lead, and the peak time of the risk score are parsed from the reverse-time pre-compensation command. The compensation trigger time is calculated based on the peak occurrence time of the risk score and the pre-compensation lead time. The pulse frequency of the micro-feeding pump is set to be the same as the frequency of the frequency modulation mode, and the phase difference between the two is set. When the compensation trigger time is reached, control signals for the stirrer and the micro-feeding pump are simultaneously sent to generate a standing wave in the space defined by the standing wave range parameter, wherein the modulation amplitude of the frequency modulation mode is set according to the pre-compensation amplitude.

[0013] In conjunction with the first aspect above, in one possible implementation, tracking the entropy acceleration rate of the high-risk region and dynamically adjusting the modulation amplitude of the frequency modulation mode based on the entropy acceleration rate includes: Based on the time series distribution of the micro-region potential signal or the statistical characteristics of the dynamic bifurcation probability, calculate the state disorder index of each micro-region; The instantaneous entropy rate is obtained by calculating the first derivative of the state disorder index of each micro-region with respect to time, and the entropy acceleration rate is obtained by calculating the second derivative of the instantaneous entropy rate with respect to time. Based on the positive and negative direction and magnitude of the entropy acceleration, the modulation amplitude of the frequency modulation mode is linearly adjusted so that the entropy acceleration returns to the stable range.

[0014] In conjunction with the first aspect above, in one possible implementation, the correction of controller parameters via a causal path backtracking algorithm includes: Obtain the reaction data for this batch, and perform correlation analysis between the actual impurity distribution and the reaction data to obtain the correlation coefficients between each parameter and the impurity rate; The parameters involved in the analysis are sorted from largest to smallest according to the absolute value of the correlation coefficient, and the top-ranked parameters are extracted as key parameters and a parameter influence weight table is generated. Based on the parameter influence weight table, the default values ​​of the key parameters with the highest weights are corrected to generate corrected controller default values.

[0015] In conjunction with the first aspect above, in one possible implementation, the method further includes: using the corrected controller default value as the initial execution parameter for the next batch of reactions, thereby realizing cross-batch iterative evolution of controller parameters.

[0016] Secondly, an impurity suppression and control system for the esterification reaction of o-aminobenzoic acid ester is provided, comprising: The data sensing and signal generation module is used to acquire the dynamic behavior of key intermediates collected by micro potential sensing probes at several spatial locations inside the reactor, and generate micro-area potential signals. The tracer response and transfer function identification module is used to inject an inert tracer controlled by a pseudo-random sequence into the reactor, monitor the diffusion response of each micro-region to the inert tracer according to the micro-region potential signal, and obtain the local transfer function through identification. The state assessment and risk heat map construction module is used to calculate the dynamic bifurcation probability of each micro-region based on the local transfer function and the benchmark transfer function used to characterize the ideal reaction state, and generate a three-dimensional risk heat map based on the dynamic bifurcation probability. The path prediction and reverse time command generation module is used to predict the evolution path of impurity generation probability based on the three-dimensional risk heat map and generate a reverse time pre-compensation command including pre-compensation magnitude and pre-compensation lead. The intervention execution and standing wave formation module is used to coordinate the rotation speed of the stirrer in frequency modulation mode and the pulse flow of the micro-feed pump with phase difference according to the reverse time pre-compensation command, so as to form the interference of mechanical shear wave and concentration wave, and generate standing wave in the high-risk area marked by the three-dimensional risk heat map. An entropy increase monitoring and modulation adaptive module is used to track the entropy acceleration rate of the high-risk region and dynamically adjust the modulation amplitude of the frequency modulation mode according to the entropy acceleration rate. The causal backtracking and parameter correction module is used to obtain the actual impurity distribution of the final product after the reaction is completed, and associate the actual impurity distribution with the local transfer function, the three-dimensional risk heat map, the reverse time pre-compensation command and the entropy acceleration rate recorded in this batch of reaction, and correct the controller parameters through the causal path backtracking algorithm.

[0017] Compared with the prior art, the present invention has the following advantages: This invention constructs a closed-loop control system integrating dynamic sensing, prediction, precise intervention, and self-learning functions, achieving predictive suppression of impurity generation risk and adaptive optimization of control parameters. Unlike traditional hysteresis control based on macroscopic parameters, this invention utilizes a micro-sensor array to capture the microscopic state inside the reactor in real time, enabling early identification and quantification of the risk of local impurity generation. This achieves a shift from passive response to proactive prevention in control mode, intervening before impurities form.

[0018] The intervention method proposed in this invention possesses high spatial targeting and energy efficiency. By coordinating the phase of the mechanical shear wave from the stirrer with the concentration wave from the feed pump, interferometric standing waves are generated within the high-risk region precisely identified by the three-dimensional risk heat map, achieving "surgical" precision treatment of the problem micro-region. This strategy of locally enhanced mixing effectively disrupts the microenvironment for impurity generation, while avoiding the enormous energy consumption and unnecessary disturbance to system stability caused by vigorous stirring of the entire reactor.

[0019] This invention establishes a complete adaptive and self-learning mechanism to ensure the long-term robustness and continuous evolution capability of the control strategy. In a single reaction, the intervention intensity is dynamically adjusted in real time by tracking the entropy acceleration rate, achieving online adaptive control. Between batches, a causal path backtracking algorithm is used to analyze the correlation between final product quality and process data, automatically correcting and optimizing key controller parameters. This allows the system to continuously learn from production experience, gradually approaching the optimal control strategy and effectively coping with changes in operating conditions such as raw material fluctuations and equipment aging.

[0020] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims, and drawings. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This application provides a structural architecture diagram of an impurity suppression control system for the esterification reaction of o-aminobenzoic acid ester. Figure 2 A schematic flowchart illustrating an impurity suppression and control method for the esterification reaction of o-aminobenzoic acid ester provided in this application embodiment; Figure 3 This is a three-dimensional risk heat map and a schematic diagram of the spatial distribution of risk clusters provided in the embodiments of this application.

[0023] Figure 4 This is an evolution curve of the state disorder index and entropy acceleration rate as adjusted by feedback, provided in the embodiments of this application. Detailed Implementation

[0024] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0025] The impurity suppression and control method for the esterification reaction of o-aminobenzoic acid ester provided in this application embodiment can be applied to, for example... Figure 1In an impurity suppression control system for the esterification reaction of o-aminobenzoic acid esters, as shown, Figure 1 As shown, the system includes: The data sensing and signal generation module is used to acquire the dynamic behavior of key intermediates collected by micro potential sensing probes at several spatial locations inside the reactor, and generate micro-area potential signals. The tracer response and transfer function identification module is used to inject an inert tracer controlled by a pseudo-random sequence into the reactor, monitor the diffusion response of each micro-region to the inert tracer according to the micro-region potential signal, and obtain the local transfer function through identification. The state assessment and risk heat map construction module is used to calculate the dynamic bifurcation probability of each micro-region based on the local transfer function and the benchmark transfer function used to characterize the ideal reaction state, and generate a three-dimensional risk heat map based on the dynamic bifurcation probability. The path prediction and reverse time command generation module is used to predict the evolution path of impurity generation probability based on the three-dimensional risk heat map and generate a reverse time pre-compensation command including pre-compensation magnitude and pre-compensation lead. The intervention execution and standing wave formation module is used to coordinate the rotation speed of the stirrer in frequency modulation mode and the pulse flow of the micro-feed pump with phase difference according to the reverse time pre-compensation command, so as to form the interference of mechanical shear wave and concentration wave, and generate standing wave in the high-risk area marked by the three-dimensional risk heat map. An entropy increase monitoring and modulation adaptive module is used to track the entropy acceleration rate of the high-risk region and dynamically adjust the modulation amplitude of the frequency modulation mode according to the entropy acceleration rate. The causal backtracking and parameter correction module is used to obtain the actual impurity distribution of the final product after the reaction is completed, and associate the actual impurity distribution with the local transfer function, the three-dimensional risk heat map, the reverse time pre-compensation command and the entropy acceleration rate recorded in this batch of reaction, and correct the controller parameters through the causal path backtracking algorithm.

[0026] like Figure 2 As shown in the embodiments of this application, an impurity suppression and control method for the esterification reaction of o-aminobenzoic acid ester is provided, comprising: The dynamic behavior of key intermediates collected by micro potential sensing probes at several spatial locations inside the reactor is obtained, and micro-region potential signals are generated. An inert tracer controlled by a pseudo-random sequence is injected into the reactor, and the diffusion response of each micro-region to the inert tracer is monitored according to the potential signal of the micro-region. The local transfer function is obtained by identification. Based on the local transfer function and the benchmark transfer function used to characterize the ideal reaction state, the dynamic bifurcation probability of each microregion is calculated, and a three-dimensional risk heat map is generated based on the dynamic bifurcation probability. Based on the evolution path of the impurity generation probability predicted by the three-dimensional risk heat map, a reverse-time pre-compensation command containing the pre-compensation magnitude and the pre-compensation lead amount is generated. According to the reverse time pre-compensation command, the speed of the stirrer is coordinated with the pulse flow of the micro-feeding pump in frequency modulation mode with a phase difference to form the interference of mechanical shear wave and concentration wave, generating standing waves in the high-risk areas marked by the three-dimensional risk heat map. Track the entropy acceleration rate of the high-risk area and dynamically adjust the modulation amplitude of the frequency modulation mode according to the entropy acceleration rate; After the reaction is completed, the actual impurity distribution of the final product is obtained, and the actual impurity distribution is correlated with the local transfer function, three-dimensional risk heat map, reverse time pre-compensation command and entropy acceleration recorded in this batch of reaction. The controller parameters are then corrected using a causal path backtracking algorithm.

[0027] It should be noted that the technical principle of this invention lies in constructing a complete closed-loop control system from microscopic perception to macroscopic learning. Its core is to utilize a micro-potential sensor array to achieve precise spatiotemporal monitoring of the internal state of the reactor, and to identify the system through active injection of tracers, thereby transforming the invisible dynamic behavior of local flow and concentration fields into quantified local transfer functions. Based on this, the actual reaction state is compared with an ideal benchmark, and the risk of impurity generation is quantified through dynamic bifurcation probability, visualized in the form of a three-dimensional risk heat map. The technical innovation lies in its forward-looking predictive capability; it goes beyond simply identifying current risks, generating reverse-time pre-compensation commands by predicting the evolution path of risks, and proactively deploying intervention measures. During intervention, a novel wave interference technique is employed. By coordinating the frequency of the stirrer rotation speed with the phase of the feed pump pulse flow, mechanical shear waves and concentration waves are generated. These waves interfere in a designated high-risk area to form a standing wave, achieving precise spatiotemporal focusing of energy and efficiently disrupting the microenvironment for impurity generation. The entire intervention process employs an adaptive adjustment based on entropy acceleration rate to ensure that the intervention intensity is just right, while the causal path backtracking algorithm after the reaction ends end gives the system the ability to learn and evolve across batches.

[0028] In one possible implementation of the embodiments of this application, combined with Figure 2 The dynamic behavior of the key intermediates acquired by the micro-potential sensing probes at several spatial locations within the reactor includes: Acquire the raw electrical signal output by the miniature potential sensing probe array; The original electrical signal is digitally filtered and normalized to obtain a preprocessed signal; The preprocessed signal is input into an intermediate migration model used to describe the relationship between electrical signals and molecular concentrations, and the concentration gradient and migration velocity of the key intermediate are inverted. The concentration gradient and the migration velocity are then combined into a micro-region potential signal.

[0029] In some implementations, a micro-potential sensing probe array arranged on the inner wall of the reactor and around the stirring shaft captures the raw electrical signal in real time. These micro-potential sensing probes employ ion-selective electrodes (ISEs), specifically responding to the anthranilic acid cation, a key intermediate in the anthranilic esterification reaction. The acquired raw electrical signal undergoes digital filtering for noise reduction and normalization. Digital filtering uses a Butterworth low-pass filter with a cutoff frequency set between 10Hz and 50Hz to remove mechanical vibration noise generated by stirring. Normalization involves dividing the real-time electrical signal by the average potential value over the first 5 minutes of the batch reaction. Subsequently, the pre-processed signal is input into a pre-established intermediate migration model, based on the Nernst-Planck equation. The calculation formula is as follows: ; in, The molar flux of the key intermediate is represented by the value obtained from the preprocessed signal, i.e., the local potential. time step The rate of change within is calculated by differentiation, which conforms to the linear mapping relationship between charge transfer and potential fluctuation; The diffusion coefficient of the intermediate in the reaction medium is experimentally measured and stored in the database; The instantaneous molar concentration of the key intermediate; The intermediate charge number is 1, where the value of the anthraquinone cation is 1. Let Faraday's constant be denoted as , and its value be . ; Let be the ideal gas constant, and take the value of . ; The absolute temperature value is collected in real time by a local temperature sensor inside the reactor. This represents the local potential gradient measured by a miniature potential sensing probe. The goal is to solve for two unknowns in this equation. and To address the decoupling problem, this embodiment introduces a pre-calculated velocity field lookup table from computational fluid dynamics (CFD), i.e., the local fluid velocity vector. Derived from a pre-established value related to stirring speed Relevant empirical correlations: ; Among them, coefficient The tracer flow field was determined through calibration experiments conducted under no-load conditions in the reactor. This was achieved by using known... Substituting these values ​​into the Nernst-Planck equation, the concentration gradient of the key intermediate in each microregion can be obtained by inversion. With migration speed: ; Concentration gradient Migration speed The data is fused with the corresponding spatial coordinates to generate a micro-region potential signal data packet containing spatial vector information and chemical component evolution characteristics, which serves as the basic data for characterizing the micro-reaction state.

[0030] For example, in the industrial production of o-aminobenzoic acid ester, the reactor volume is 3000L, and 32 sets of ion-selective electrode probes are arranged in a three-dimensional space with equal spacing. The following values ​​are derived from pilot-scale tests in a 50L laboratory reactor and typical values ​​derived according to the proportional scale-up criterion. When the reaction enters the core esterification stage, a certain probe measures the local absolute temperature. The local potential gradient is 353K. for The diffusion coefficient of the anthranilic acid cation in the current medium is known. for The instantaneous molar concentration of the intermediate is obtained by inverting the normalized level signal. for At this time, the stirring speed for The local fluid velocity calculated according to the corresponding empirical correlation. for Substituting into the Nernst-Planck formula above: Based on the molar flux determined within a continuous sampling period. Solve the equation to obtain the concentration gradient of the micro-region. Approximately migration speed Approximately The controller encapsulates these parameters into micro-region potential signal data packets, which serve as the basis for subsequent identification of local transfer functions and construction of three-dimensional risk heat maps, thereby accurately locating the inducing sources of impurity generation.

[0031] In one possible implementation of the embodiments of this application, combined with Figure 2 The process of calculating the dynamic bifurcation probability of each microregion based on the local transfer function and the baseline transfer function used to characterize the ideal reaction state, and generating a three-dimensional risk heatmap based on the dynamic bifurcation probability, includes: The local transfer function is compared with the reference transfer function to calculate the deviation matrix containing the deviation information of each micro-region; The micro-regions corresponding to the elements in the deviation matrix whose values ​​change abruptly are extracted as candidate abnormal micro-regions; The spatial location of the candidate abnormal micro-region is weighted and fused with the dynamic bifurcation probability to generate a three-dimensional risk heat map.

[0032] In some implementations, the baseline transfer function is obtained by computational fluid dynamics (CFD) simulation or by pre-determining the system response of the reactor at steady state and optimal yield under ideal laboratory conditions. The identified local transfer functions of each micro-region With reference transfer function By comparison, a deviation matrix is ​​constructed that includes the degree of deviation of the dynamic characteristics of each micro-region. Extracting elements in the deviation matrix whose values ​​undergo abrupt changes specifically refers to calculating... and In the complex frequency domain Norm difference, capturing spatial points where the deviation exceeds the normal disturbance threshold; the normal disturbance threshold is based on the deviation distribution in historical fault-free batches. The principle is adaptively set to eliminate inherent noise interference from the system. This is three standard deviations. To further quantify the critical risk of impurity generation caused by these mutation points, this embodiment introduces dynamic bifurcation probability. The evaluation will be conducted using the following calculation formula: ; in, It represents the dynamic bifurcation probability, which is used to measure the possibility of the micro-region reaction state transitioning from a steady state to a nonlinear impurity generation state; The sensitivity coefficient is obtained by calibrating the correlation between impurity outbreak points and deviation values ​​in historical production data; express The norm is used to extract the maximum amplitude deviation of the transfer function across the entire frequency band. When the calculated norm is... When the value is significantly higher than the average of the surrounding micro-regions or exceeds the preset warning line, the micro-region is determined to be a candidate abnormal micro-region. The warning line is obtained by simulating the critical point of impurity outbreak in a small-scale experiment and statistically analyzing the corresponding probability value, and is generally set to 0.6. Subsequently, the Kriging interpolation algorithm is used to interpolate the values ​​at each discrete probe. Values ​​extended to the three-dimensional spatial coordinates of the entire reactor. This generates a 3D risk heatmap where color intensity represents risk values. In this heatmap, the value of each spatial grid represents the probability distribution of side reaction bifurcation at that location due to deviation from the dynamic path, providing precise navigation for subsequent targeted standing wave intervention.

[0033] For example, during the esterification reaction of o-aminobenzoic acid ester, at a certain sampling time, the local transfer function of the 12th microregion was identified as follows: The reference transfer function under this operating condition is known to be... Calculate the two Norm difference, i.e., deviation matrix Corresponding element in It is approximately 0.25. The sensitivity coefficient in this reaction system is known, based on historical high-quality batch statistics. The value is 4.8. Substitute this value into the formula to calculate the dynamic bifurcation probability of this microregion: Since this value significantly exceeds the impurity critical warning line of 0.6 determined through small-scale experiments, the micro-region is automatically identified as a candidate abnormal micro-region, and its spatial coordinates are assigned. and The data is correlated. Through real-time interpolation calculation of 32 data points across the entire reactor, a three-dimensional risk heat map covering the 3000L reactor capacity is generated, which intuitively identifies high-risk clusters located near the stirring dead zone, providing a precise triggering basis for activating the reverse time pre-compensation command.

[0034] In one possible implementation of the embodiments of this application, combined with Figure 2 The method further includes: Density clustering analysis was performed on the three-dimensional risk heatmap to identify high-risk clusters; Extract the geometric centroid coordinates and circumscribed rectangular bounding box of the high-risk cluster to generate the effective range parameters of the standing wave.

[0035] In some implementations, the dynamic bifurcation probability of each spatial grid in the 3D risk heatmap is obtained. Set a probability threshold. All of the heatmap The grid points were extracted as a high-risk sample point set to be clustered. A density-based clustering method with noise (DBSCAN) was used to spatially cluster the sample point set. The DBSCAN algorithm includes two core parameters: neighborhood radius. and minimum number of contained points .in, Based on the grid resolution setting of the 3D risk heat map, the preferred value is 1.5 to 2 times the grid side length to ensure the capture of spatial continuity features; Based on the characteristic dimensions of the reactor volume and the physical micro-region, ensure that each high-risk cluster contains at least 10-20 grid points sufficient to characterize an independent anomalous region. By calculating the Euclidean distance between sample points, regions with dense spatial distribution and continuously evolving risk values ​​are identified as independent high-risk clusters. For each identified high-risk cluster, extract the spatial coordinates of all grid points it contains. Calculate the geometric centroid coordinates of the high-risk cluster. The calculation formula is: ; in, This represents the number of points within the cluster. Then, the extreme coordinates within the cluster are traversed to determine its bounding rectangle in space. The bounding box is used to define a spatially closed region that completely covers all high-risk grid points. In other embodiments, a minimum rotation bounding box can also be used to accommodate non-axially distributed anomalous flow channels. The geometric centroid coordinates combined with the circumscribed rectangular bounding box are encapsulated as the effective range parameters of the standing wave, serving as the geometric reference for subsequent adjustment of the interference position and envelope range of the mechanical shear wave and concentration wave. Figure 3 As shown, the spatial distribution of dynamic bifurcation probability in the micro-regions inside the reactor is displayed. The black high-value areas mark the centroids of the risk clusters for impurity generation, providing accurate spatial coordinate references for subsequent standing wave interferometry.

[0036] For example, during the esterification reaction of o-aminobenzoic acid ester, a three-dimensional risk heat map showed a large-area risk warning at the bearing seal at the bottom of the reactor. A probability threshold was set. And set the neighborhood radius according to the grid edge length. It is 1.8 times the grid side length. The value was set to 12. Density clustering analysis successfully identified a high-risk cluster with a volume of approximately 0.4L. The calculated geometric centroid coordinates of this cluster were... The corresponding bounding box is The centroid coordinates and boundary range are defined as the effective range parameters of the standing wave.

[0037] In one possible implementation of the embodiments of this application, combined with Figure 2 Based on the evolution path of the impurity generation probability predicted by the three-dimensional risk heatmap, a reverse-time pre-compensation command including the pre-compensation magnitude and pre-compensation lead time is generated, including: Based on the standing wave's effective range parameters, the prediction target area is located; Within the predicted target area, the evolution path of the impurity generation probability is predicted using a Kalman filter to obtain the risk score peak and its occurrence time; Based on the peak risk score and the time of its occurrence, the initial pre-compensation magnitude and pre-compensation advance are calculated, and the two are combined into a reverse-time pre-compensation command.

[0038] In some implementations, the effective range parameter of the standing wave is determined by the geometric centroid coordinates of the high-risk cluster. The spatial range defined by the circumscribed rectangular bounding box is used to identify spatially continuous high-risk clusters as the prediction target region in the 3D risk heatmap. Within the prediction target region, the impurity generation probability is defined. Dynamic bifurcation probability for all grid points within this region Spatial weighted average. In order to achieve... Accurate prediction of the evolution path is achieved by establishing a state-space model based on a Kalman filter. Its state equations are defined as follows: ; in, express The state vector representing the probability of impurity generation at time t; The system state transition matrix is ​​obtained by offline parameter identification of dynamic evolution data under the same working conditions in historical high-quality production batches, such as fitting using the least squares method. This is the control input for the current stirring speed and the pulse flow rate of the feed pump; For control gain matrix; This represents process noise, which follows a Gaussian distribution. Its observation equation is defined as follows: ; in, The observation vector is obtained by real-time acquisition and inversion through a miniature potential sensing probe array; The observation matrix; To measure noise. Predict the future through recursive iteration. The probability trajectory over a period of time identifies the peak risk score within the trajectory. and the corresponding time of occurrence Subsequently, based on the peak risk score... Calculate the initial pre-compensation magnitude: ; in, The pre-compensation amplitude corresponds to the modulation deviation of the stirring speed; The preset compensation gain coefficient was obtained by offline gradient experiments to calibrate the impurity suppression effect under different risk levels. The preset impurity risk threshold is set, such as 0.4. Simultaneously, the pre-compensation lead is calculated based on the system's physical inertia and transmission lag characteristics. ; in, To compensate for advance time; This is the characteristic distance from the center point of the predicted target area to the core region of the interferometric standing wave; The migration rate of key intermediates in this region; This is to account for the hardware response delay of the controller and actuators. Ultimately, the pre-compensation amplitude will be... , advance compensation When the risk score peaks Encapsulated as a reverse-time pre-compensation command, used to guide the actuator in... Intervention actions are initiated in advance at all times to achieve reverse time suppression of impurity generation pathways.

[0039] For example, in the prediction stage of the o-aminobenzoic acid ester esterification reaction, the prediction target region located near the bearing at the bottom of the reactor was identified, and this region was determined by the centroid coordinates. The bounding box of the micro-area is determined. At this point, the Kalman filter, based on the trend of the micro-area potential signal changes over the previous 10 minutes, and combined with the system state transition matrix identified from historical batch data, is used. The probability of impurity generation in this region is predicted to be [value] in the next 120 seconds. It will reach the maximum value in the predicted trajectory, i.e., the peak of the risk score. If the preset impurity risk threshold is... The compensation gain coefficient is 0.4. After offline calibration to 15.0, the pre-compensation amplitude is calculated. This value corresponds to the required increase in frequency tuning depth for the stirrer. Simultaneously, the migration speed of the intermediate in this region is detected. for Feature distance for Controller hardware response delay for Then calculate the pre-compensation lead time. The generated reverse-time pre-compensation command indicates that the risk score peak occurs at the specified time. 4.5 seconds before the 120th second after the current time, that is, 115.5 seconds after the current time, the actuator starts an interference pulse of preset amplitude to ensure that the standing wave field completes spatial construction before the impurity generation path reaches its peak, thereby disrupting the kinetic conditions for the generation of side reactions in advance.

[0040] In one possible implementation of the embodiments of this application, combined with Figure 2 According to the reverse-time pre-compensation command, the agitator speed is coordinated with the pulse flow rate of the micro-feed pump in frequency modulation mode with a phase difference, forming an interference between mechanical shear wave and concentration wave, generating standing waves in the high-risk areas marked on the three-dimensional risk heat map, including: The pre-compensation magnitude, pre-compensation lead, and the peak time of the risk score are parsed from the reverse-time pre-compensation command. The compensation trigger time is calculated based on the peak occurrence time of the risk score and the pre-compensation lead time. The pulse frequency of the micro-feeding pump is set to be the same as the frequency of the frequency modulation mode, and the phase difference between the two is set. When the compensation trigger time is reached, control signals for the stirrer and the micro-feeding pump are simultaneously sent to generate a standing wave in the space defined by the standing wave range parameter, wherein the modulation amplitude of the frequency modulation mode is set according to the pre-compensation amplitude.

[0041] In some implementations, the controller parses the pre-compensation amplitude from the received counter-time pre-compensation command. , advance compensation and the moment when the risk score peaks. Calculate the compensation trigger time: ; In order to generate wave interference at a specific point in space, the carrier frequency of the agitator's frequency modulation signal and the pulse frequency of the micro-feed pump are both set to characteristic frequencies. Characteristic frequencies The system was calibrated by injecting a broadband pressure noise signal into the reactor and analyzing the spectral peak values ​​of the potential response in each micro-region using Fourier transform, selecting the resonant point with the largest amplitude-frequency response. The output speed of the stirrer... Follow the frequency modulation mode: ; in, Based on the base speed, For modulation amplitude, and , Sensitivity mapping gain; feed pump pulse flow rate Follow the pulse pattern: ; in, Based on the basic feed flow rate, This refers to the pulse amplitude. The phase difference between the two sets of signals is set. This causes the mechanical shear stress wave and the material concentration wave to superimpose in space. According to the principle of wave interference, the composite wave at a certain point in space... amplitude The distribution follows: ; in, and These are the equivalent amplitudes of the shear wave and the concentration wave, respectively. The equivalent propagation velocity of the shear wave at the current viscosity of the reaction liquid is given by the equivalent wave velocity. Since the concentration wave undergoes phase pre-compensation through the initial phase angle of the feed pulse, its effective phase at the interference point is uniformly characterized by this equivalent wave velocity. For target point Distance from the center of the agitator With feed pipe outlet The process difference, its mathematical expression is: ; By adjusting the phase difference This allows the centroid coordinates within the range of action of the standing wave to be defined by the parameters. The phase term is satisfied Conditions for strengthening interference, The value is an integer, thus forming a standing wave field with concentrated energy in this local region. This standing wave, through periodic high-frequency shearing and local concentration oscillations, forcibly disrupts the nonlinear kinetic feedback path required for side reactions, thereby achieving impurity suppression.

[0042] For example, in the intervention stage of the anthranilate esterification reaction, the pre-compensation amplitude is analyzed. Pre-compensation lead time Peak time Calculate the compensation trigger time at 600 seconds after the start of the reaction. The characteristic frequencies identified by spectral analysis are known. 5Hz, sensitivity gain Set the mixer modulation amplitude to 1.2. For high-risk cluster centroids The center of the agitator is known. and feed inlet The process difference was calculated. This is an example of a symmetrical distribution. If the equivalent wave velocity... The required phase difference At 595.5s, a control command is simultaneously issued, causing the stirrer speed to maintain 5Hz at the base speed. The modulation was performed, and the feed pump started pulse feeding at the same frequency and phase. Experimental observations showed that, under the synergy of these parameters, the local turbulent kinetic energy in the high-risk region was increased by 45%, successfully suppressing the dynamic bifurcation probability of this micro-region from 0.85 to 0.32, effectively preventing the formation of dark impurity clusters in the esterification reaction.

[0043] In one possible implementation of the embodiments of this application, combined with Figure 2 Tracking the entropy acceleration rate of the high-risk region and dynamically adjusting the modulation amplitude of the frequency modulation mode based on the entropy acceleration rate includes: Based on the time series distribution of the micro-region potential signal or the statistical characteristics of the dynamic bifurcation probability, calculate the state disorder index of each micro-region; The instantaneous entropy rate is obtained by calculating the first derivative of the state disorder index of each micro-region with respect to time, and the entropy acceleration rate is obtained by calculating the second derivative of the instantaneous entropy rate with respect to time. Based on the positive and negative direction and magnitude of the entropy acceleration, the modulation amplitude of the frequency modulation mode is linearly adjusted so that the entropy acceleration returns to the stable range.

[0044] In some implementations, the micro-area potential signals of each micro-region within a high-risk area are acquired and observed within the observation time window. Amplitude discretization is performed within the micro-region. The dynamic range of the amplitude of the potential signal in the entire reaction process is divided into... There are equidistant energy level intervals, among which The value of is set according to the signal-to-noise ratio, and is preferably an integer between 10 and 20 to balance calculation accuracy and statistical stationarity. The probability distribution of the statistical signal falling within each interval. And calculate the state disorder index of each microregion based on the probability distribution: ; State disorder index This is used to characterize the degree of random fluctuation in the mixing uniformity and chemical potential of local reactants. Subsequently, real-time monitoring is performed. The dynamic evolution process, calculating its time... The first derivative yields the instantaneous entropy rate. Further calculations The derivative with respect to time yields the entropy acceleration. This is used to characterize the intensity of the disordering trend. The controller is based on the entropy acceleration rate. The sign and magnitude of the value affect the modulation amplitude of the stirrer. Perform linear adaptive adjustment: ; in, The adjusted modulation amplitude; The modulation amplitude at the current moment; This is the sampling control period. Adjust the gain coefficient. Set as the disorder index under steady-state conditions The reciprocal of the standard deviation of the fluctuation is used to normalize the dimensions of the feedback compensation amount and the inherent statistical fluctuations of the system, ensuring that the control sensitivity is adapted to the system characteristics. When this occurs, it indicates that the disordering trend is accelerating, and the controller linearly increases the modulation amplitude to enhance the standing wave interference strength; when And disorder index When the trend is downward, appropriately reduce the modulation amplitude. Through the above closed-loop adjustment, the entropy acceleration rate is increased. Return to the preset stable range Inside, among which Based on historical high-quality batches The stability threshold determined by the standard deviation of the fluctuation is used to suppress instability in local reaction kinetics. For example... Figure 4 As shown, the changing trend of the system state disorder index at the intervention time and the process of the entropy acceleration returning to the stable range under the closed-loop feedback are demonstrated, reflecting the control system's ability to suppress the system's instability trend.

[0045] For example, during the intervention process of the o-aminobenzoic acid ester esterification reaction, high-risk clusters located near the reactor bearings are continuously monitored. The micro-region potential signal is divided into... The disorder index of the current state is calculated from each energy level interval. Calculations were performed using three consecutive sets of sampling points, revealing the instantaneous entropy rate. from Rapidly rise to During the sampling period The entropy acceleration rate is calculated below. The current modulation amplitude of the stirrer is known. for If in steady state If the standard deviation of the fluctuation is 0.22, then the adjustment gain coefficient... Taking its reciprocal, approximately 4.5, the new modulation amplitude is calculated. Subsequently, the stirrer enhanced the stirring at this characteristic frequency. The reduced rotational speed oscillation depth, through stronger local standing wave energy, dispersed the locally accumulated material concentration deviation. After a 2-second feedback adjustment, Reduce to It has entered the preset stable range. This effectively avoids the risk of impurity outbreaks caused by local overheating or concentration imbalance.

[0046] In one possible implementation of the embodiments of this application, combined with Figure 2 The step of correcting the controller parameters using the causal path backtracking algorithm includes: Obtain the reaction data for this batch, and perform correlation analysis between the actual impurity distribution and the reaction data to obtain the correlation coefficients between each parameter and the impurity rate; The parameters involved in the analysis are sorted from largest to smallest according to the absolute value of the correlation coefficient, and the top-ranked parameters are extracted as key parameters and a parameter influence weight table is generated. Based on the parameter influence weight table, the default values ​​of the key parameters with the highest weights are corrected to generate corrected controller default values.

[0047] In some implementations, after the esterification reaction of a single batch of o-aminobenzoic acid ester is completed, complete reaction data for that batch is acquired. This data includes a time series of micro-region potential signals recorded by a micro-potential sensing probe array, three-dimensional risk heatmap data during the process, and recorded values ​​of various controller execution variables. The final actual impurity distribution at the product end is then detected using high-performance liquid chromatography (HPLC) to obtain the actual impurity rate corresponding to each spatial micro-region. The actual impurity rate Historical characteristic variables in reaction data Correlation analysis was performed on parameters such as the average dynamic bifurcation probability, average state disorder index, and modulation amplitude record value of each microregion, and the Pearson correlation coefficient between the two was calculated. ; in, For the first The correlation coefficients between the parameters and the impurity rate This is the average time value of this parameter within this batch. This represents the average of the actual impurity rates at each sampling point. It is calculated based on the absolute value of the correlation coefficient. The parameters are sorted from largest to smallest, and the top-ranked parameters are extracted as key parameters affecting impurity generation. A parameter influence weight table is then generated, showing the weight of each parameter in the table. The normalized correlation coefficient is used as the definition: ; Based on the parameter impact weight table, identify the key parameters with the highest weights and assign them default values ​​to the corresponding controllers, such as pre-compensation lead time. Default baseline value, characteristic frequency The center value or adjustment gain coefficient The initial value is corrected. The correction formula is: ; in, This is the corrected default value for the controller. This is the default value for the controller used in the current batch. The preset learning rate step size, ranging from 0.05 to 0.15, was determined through convergence experiments using historical data. The parameter affects the maximum weight value in the weight table. The target impurity rate is set as the baseline. Through this iterative correction based on causal backtracking, the control logic for the next batch can automatically adapt to the evolution of system characteristics caused by scaling on the inner wall of the reactor or catalyst activity drift.

[0048] For example, after completing the production of a batch of o-aminobenzoic acid ester, HPLC analysis showed that the average impurity rate of the product was... The figure was 0.12%, exceeding the target benchmark. The causal path backtracking algorithm was activated to perform correlation analysis between the impurity distribution of 50 micro-area sampling points and the control records of this batch. The pre-compensation lead time was calculated. Correlation coefficient with impurity rate This indicates that insufficient lead time leads to increased impurities, and the gain coefficient needs to be adjusted. correlation coefficient After sorting, The key parameter identified as having the highest weight has its normalized weight. The default value for pre-compensation lead time used in this batch is known. Set the learning rate The corrected default value Here, multiplying by 100 represents the dimensionless mapping constant. The database is automatically updated, and the controller will use 4.98s as the initial lead time benchmark for prediction compensation when the next batch starts.

[0049] In one possible implementation, combining Figure 2 The method further includes: using the corrected controller default value as the initial execution parameter for the next batch of reactions, thereby realizing cross-batch iterative evolution of controller parameters.

[0050] In some implementations, the corrected controller default value generated for each batch of settlements is used. Stored in the global configuration database and associated with the physical status identifier of the current reactor. Before starting the next batch, i.e., the [number missing]th batch... Before the batch reaction, the controller initialization program reads the previous batch, the first... Stored at the end of the batch The initial execution parameters for the next batch are calculated by smoothing the data based on the inertia weights from multiple batches of evolution. The formula for its calculation is: ; in, For the first Initial execution parameters for the batch; For the first The default value of the controller after batch correction is calculated from the correction formula of the aforementioned causal path backtracking algorithm; This is a dynamic evolution factor, ranging from 0.6 to 0.8, used to balance the ratio of the current batch correction amount to the historical trend. Its value is calibrated according to the reactor cleaning frequency. The sliding window length represents the number of historical batches involved in the average calculation, preferably 3 to 5 batches. The initial execution parameters obtained through calculation include the initial pre-compensation lead amount. Initial characteristic frequency and initial adjustment gain coefficient This cross-batch iterative mechanism can effectively offset the deviations caused by sensor drift or actuator wear due to long-term operation of the equipment, ensuring that the controller is in a near-optimal state at the beginning of each batch, thus shortening the adaptive adjustment convergence time of the closed-loop system.

[0051] For example, in the continuous batch production of the esterification reaction of o-aminobenzoic acid ester, after the 10th batch of reaction is completed, the corrected pre-compensation lead amount is calculated using a causal path backtracking algorithm. ,Right now It took 4.98 seconds. Historical database access revealed that the first three batches... The corrected default value average is 4.80s. Set the dynamic evolution factor. The initial pre-compensation lead time for the start-up of the 11th batch is calculated as follows: After the 11th batch of production orders was issued, the controller automatically loaded 4.926 seconds as the baseline lead time for the prediction model. Instead of using the original factory default of 4.5s, this cross-batch iteration strategy improved the success rate of building standing wave fields after identifying risk clusters in the early stages of the reaction by 18%, and reduced the batch-to-batch standard deviation of product impurity rate by 25%, achieving deep synergistic evolution between the control system and the production process.

[0052] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0053] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for controlling impurities in the esterification reaction of o-aminobenzoic acid ester, characterized in that, The method includes: The dynamic behavior of key intermediates collected by micro potential sensing probes at several spatial locations inside the reactor is obtained, and micro-region potential signals are generated. An inert tracer controlled by a pseudo-random sequence is injected into the reactor, and the diffusion response of each micro-region to the inert tracer is monitored according to the potential signal of the micro-region. The local transfer function is obtained by identification. Based on the local transfer function and the benchmark transfer function used to characterize the ideal reaction state, the dynamic bifurcation probability of each microregion is calculated, and a three-dimensional risk heat map is generated based on the dynamic bifurcation probability. Based on the evolution path of the impurity generation probability predicted by the three-dimensional risk heat map, a reverse-time pre-compensation command containing the pre-compensation magnitude and the pre-compensation lead amount is generated. According to the reverse time pre-compensation command, the speed of the stirrer is coordinated with the pulse flow of the micro-feeding pump in frequency modulation mode with a phase difference to form the interference of mechanical shear wave and concentration wave, generating standing waves in the high-risk areas marked by the three-dimensional risk heat map. Track the entropy acceleration rate of the high-risk area and dynamically adjust the modulation amplitude of the frequency modulation mode according to the entropy acceleration rate; After the reaction is completed, the actual impurity distribution of the final product is obtained, and the actual impurity distribution is correlated with the local transfer function, the three-dimensional risk heat map, the reverse time pre-compensation command and the entropy acceleration rate recorded in this batch of reaction. The controller parameters are then corrected using a causal path backtracking algorithm.

2. The method for impurity suppression and control in the esterification reaction of o-aminobenzoic acid ester according to claim 1, characterized in that, The dynamic behavior of the key intermediates acquired by the micro potential sensing probes at several spatial locations within the reactor includes: Acquire the raw electrical signal output by the miniature potential sensing probe array; The original electrical signal is digitally filtered and normalized to obtain a preprocessed signal; The preprocessed signal is input into an intermediate migration model used to describe the relationship between electrical signals and molecular concentrations, and the concentration gradient and migration velocity of the key intermediate are inverted. The concentration gradient and the migration velocity are then combined into a micro-region potential signal.

3. The method for impurity suppression and control in the esterification reaction of o-aminobenzoic acid ester according to claim 1, characterized in that, Based on the local transfer function and the baseline transfer function used to characterize the ideal reaction state, the dynamic bifurcation probability of each microregion is calculated, and a three-dimensional risk heatmap is generated based on the dynamic bifurcation probability, including: The local transfer function is compared with the reference transfer function to calculate the deviation matrix containing the deviation information of each micro-region; The micro-regions corresponding to the elements in the deviation matrix whose values ​​change abruptly are extracted as candidate abnormal micro-regions; The spatial location of the candidate abnormal micro-region is weighted and fused with the dynamic bifurcation probability to generate a three-dimensional risk heat map.

4. The method for impurity suppression and control in the esterification reaction of o-aminobenzoic acid ester according to claim 3, characterized in that, The method further includes: Density clustering analysis was performed on the three-dimensional risk heatmap to identify high-risk clusters; Extract the geometric centroid coordinates and circumscribed rectangular bounding box of the high-risk cluster to generate the effective range parameters of the standing wave.

5. The method for impurity suppression and control in the esterification reaction of o-aminobenzoic acid ester according to claim 4, characterized in that, Based on the evolution path of the impurity generation probability predicted by the three-dimensional risk heatmap, a reverse-time pre-compensation command including the pre-compensation magnitude and pre-compensation lead amount is generated, including: Based on the standing wave's effective range parameters, the prediction target area is located; Within the predicted target area, the evolution path of the impurity generation probability is predicted using a Kalman filter to obtain the risk score peak and its occurrence time; Based on the peak risk score and the time of its occurrence, the initial pre-compensation magnitude and pre-compensation advance are calculated, and the two are combined into a reverse-time pre-compensation command.

6. The method for impurity suppression and control in the esterification reaction of o-aminobenzoic acid ester according to claim 5, characterized in that, According to the reverse-time pre-compensation command, the agitator speed is coordinated with the pulse flow rate of the micro-feed pump in frequency modulation mode with a phase difference, forming an interference between mechanical shear wave and concentration wave, generating standing waves in the high-risk areas marked on the three-dimensional risk heat map, including: The pre-compensation magnitude, pre-compensation lead, and the peak time of the risk score are parsed from the reverse-time pre-compensation command. The compensation trigger time is calculated based on the peak occurrence time of the risk score and the pre-compensation lead time. The pulse frequency of the micro-feeding pump is set to be the same as the frequency of the frequency modulation mode, and the phase difference between the two is set. When the compensation trigger time is reached, control signals for the stirrer and the micro-feeding pump are simultaneously sent to generate a standing wave in the space defined by the standing wave range parameter, wherein the modulation amplitude of the frequency modulation mode is set according to the pre-compensation amplitude.

7. The method for impurity suppression and control in the esterification reaction of o-aminobenzoic acid ester according to claim 1, characterized in that, Tracking the entropy acceleration rate of the high-risk region and dynamically adjusting the modulation amplitude of the frequency modulation mode based on the entropy acceleration rate includes: Based on the time series distribution of the micro-region potential signal or the statistical characteristics of the dynamic bifurcation probability, calculate the state disorder index of each micro-region; The instantaneous entropy rate is obtained by calculating the first derivative of the state disorder index of each micro-region with respect to time, and the entropy acceleration rate is obtained by calculating the second derivative of the instantaneous entropy rate with respect to time. Based on the positive and negative direction and magnitude of the entropy acceleration, the modulation amplitude of the frequency modulation mode is linearly adjusted so that the entropy acceleration returns to the stable range.

8. The method for impurity suppression and control in the esterification reaction of o-aminobenzoic acid ester according to claim 1, characterized in that, The step of correcting the controller parameters using the causal path backtracking algorithm includes: Obtain the reaction data for this batch, and perform correlation analysis between the actual impurity distribution and the reaction data to obtain the correlation coefficients between each parameter and the impurity rate; The parameters involved in the analysis are sorted from largest to smallest according to the absolute value of the correlation coefficient, and the top-ranked parameters are extracted as key parameters and a parameter influence weight table is generated. Based on the parameter influence weight table, the default values ​​of the key parameters with the highest weights are corrected to generate corrected controller default values.

9. The method for impurity suppression and control in the esterification reaction of o-aminobenzoic acid ester according to claim 8, characterized in that, The method further includes: using the corrected controller default value as the initial execution parameter for the next batch of reactions, thereby realizing cross-batch iterative evolution of controller parameters.

10. An impurity suppression control system for the esterification reaction of o-aminobenzoic acid ester, characterized in that, The system is used for impurity suppression and control in the esterification reaction of o-aminobenzoic acid ester as described in any one of claims 1-9, the system comprising: The data sensing and signal generation module is used to acquire the dynamic behavior of key intermediates collected by micro potential sensing probes at several spatial locations inside the reactor, and generate micro-area potential signals. The tracer response and transfer function identification module is used to inject an inert tracer controlled by a pseudo-random sequence into the reactor, monitor the diffusion response of each micro-region to the inert tracer according to the micro-region potential signal, and obtain the local transfer function through identification. The state assessment and risk heat map construction module is used to calculate the dynamic bifurcation probability of each micro-region based on the local transfer function and the benchmark transfer function used to characterize the ideal reaction state, and generate a three-dimensional risk heat map based on the dynamic bifurcation probability. The path prediction and reverse time command generation module is used to predict the evolution path of impurity generation probability based on the three-dimensional risk heat map and generate a reverse time pre-compensation command including pre-compensation magnitude and pre-compensation lead. The intervention execution and standing wave formation module is used to coordinate the rotation speed of the stirrer in frequency modulation mode and the pulse flow of the micro-feed pump with phase difference according to the reverse time pre-compensation command, so as to form the interference of mechanical shear wave and concentration wave, and generate standing wave in the high-risk area marked by the three-dimensional risk heat map. An entropy increase monitoring and modulation adaptive module is used to track the entropy acceleration rate of the high-risk region and dynamically adjust the modulation amplitude of the frequency modulation mode according to the entropy acceleration rate. The causal backtracking and parameter correction module is used to obtain the actual impurity distribution of the final product after the reaction is completed, and associate the actual impurity distribution with the local transfer function, the three-dimensional risk heat map, the reverse time pre-compensation command and the entropy acceleration rate recorded in this batch of reaction, and correct the controller parameters through the causal path backtracking algorithm.

Images