Chemical liquid proportioning control method

Abstract

The application discloses a chemical liquid proportioning control method and relates to the technical field of semiconductor manufacturing processes. The method comprises the following steps: performing real-time filtering processing on a liquid concentration observation signal in a circulating pipeline; triggering an intermittent compensation action when the concentration estimation value deviates from a target concentration and exceeds a set dead zone threshold; obtaining a single optimal liquid supplement amount through a correction algorithm based on a current concentration deviation and a historical compensation effect; executing the single optimal liquid supplement amount according to the deviation polarity of the concentration estimation value; starting a forced waiting time after the single optimal liquid supplement amount, and maintaining the closed-loop circulation of the liquid in a mixing tank through the circulating pipeline to eliminate the local concentration gradient in space. By using the above scheme, in the closed-loop control, the sampling distortion caused by the measurement noise and the local concentration unevenness is overcome, and the misjudgment caused by the false data is avoided.

Description

Technical Field

[0001] This invention relates to the field of semiconductor manufacturing process technology, and in particular to a method for controlling the proportion of chemical solutions. Background Technology

[0002] Currently, in processes such as semiconductor wet cleaning and etching, diluted hydrofluoric acid (DHF) or other chemical solutions of specific concentrations are required as working solutions. Existing technologies typically employ online mixing and replenishment systems, which generally consist of a mixing tank for storing the working solution, multiple inlet pipes connecting the high-concentration chemical solution source and the diluent source, a circulation pipeline located outside the mixing tank, and a concentration meter for online monitoring of the chemical solution concentration. During operation, the system triggers a replenishment action based on the drop in liquid level caused by the consumption of liquid by the equipment using the solution. It calculates the required volume of high-concentration chemical solution and diluent according to the law of conservation of mass and replenishes it to the mixing tank through the inlet pipes. A circulation pump then drives the chemical solution to flow within the circulation pipeline to promote uniform mixing, thereby maintaining the basic level and concentration of the chemical solution.

[0003] To further improve mixing accuracy and reduce concentration fluctuations, existing technologies propose an intermittent replenishment method based on concentration feedback. For example, in addition to level-based replenishment, a concentration compensation mechanism dependent on online concentration meter monitoring is added: when the real-time concentration detected online deviates from the target concentration, the system initiates a concentration correction action, calculates the theoretical compensation volume based on the current concentration deviation, and injects the corresponding drug solution or diluent in one go, aiming to quickly bring the drug concentration in the mixing tank back to the target range. This method incorporates the concentration signal into closed-loop control, which, compared to open-loop mixing methods that rely solely on level triggers, improves the steady-state control capability of the concentration to a certain extent.

[0004] However, in practical applications, the defects in each step of the above-mentioned improvement scheme are not isolated, making the system prone to severe oscillations during concentration adjustment. Specifically, when a high-concentration drug solution is added to the system, due to the uneven flow field distribution within the mixing tank, a local high-concentration peak region far exceeding the target concentration is instantaneously formed at the sampling point in the circulation pipeline. Simultaneously, the concentration sensor itself is affected by random noise interference caused by fluid turbulence and bubbles, resulting in a distorted data signal that does not reflect the true concentration but is superimposed with significant random oscillation spikes. If this extreme abnormal transient pseudo-data, which combines local unmixed peaks and random noise spikes, is directly captured, it can lead to a misjudgment that the system has experienced severe concentration overshoot, resulting in the calculation of a compensation amount that deviates excessively from the actual requirement. Therefore, how to overcome the sampling distortion caused by measurement noise and local concentration unevenness in closed-loop control, and avoid misjudgments caused by pseudo-data, is a pressing technical challenge that needs to be addressed in this field. Summary of the Invention

[0005] The purpose of this invention is to provide a chemical solution ratio control method to solve the technical problem in the prior art of how to overcome sampling distortion caused by measurement noise and local concentration unevenness in closed-loop control, and avoid misjudgment caused by false data.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows: A method for controlling the proportioning of chemical solutions, applied to a proportioning system including a mixing tank, a circulation pipeline, and at least one inlet pipeline for replenishing the solution, comprising the following steps: S1: Perform real-time filtering on the concentration observation signal of the drug solution in the circulation pipeline to filter out random disturbance noise of the fluid and output the concentration estimate for the current period. S2: When the estimated concentration deviates from the target concentration and exceeds the set dead zone threshold, an intermittent compensation action is triggered; based on the current concentration deviation and historical compensation effects, the optimal replenishment volume for a single operation is obtained through a correction algorithm; S3: Based on the deviation polarity of the concentration estimate, select the corresponding inlet pipeline and execute the single optimal replenishment volume through closed-loop control of the flow rate; S4: After the single optimal replenishment volume is injected, a preset forced waiting time is initiated, and during the forced waiting time, the closed-loop circulation of the drug solution in the mixing tank is maintained through the circulation pipeline to eliminate the spatial local concentration gradient; after the forced waiting time ends, the process returns to step S1 until the concentration estimate stabilizes within the tolerance range of the target concentration.

[0007] Furthermore, the real-time filtering of the drug concentration observation signal, using the Kalman filtering algorithm, includes the following steps: Based on the state estimate and process noise covariance of the previous cycle, state prediction and covariance prediction are performed. Calculate the Kalman gain based on the measurement noise covariance; Using the concentration observation signal acquired in the current period and the Kalman gain, the state is updated, and the concentration estimate for the current period is output. Update the covariance matrix for use in the next iteration.

[0008] Furthermore, the step of determining the optimal single-time fluid replacement volume based on the current concentration deviation and historical compensation effects using a correction algorithm specifically includes the following steps: Extract the effective volume of the drug solution in the current system and the estimated concentration value, and calculate the static compensation volume based on the law of conservation of mass; Extract the actual fluid volume replenished in the previous intermittent compensation action and the actual change in concentration it caused. Combine this with the target concentration change in the current case to calculate the estimated compensation volume. The static compensation volume and the estimated compensation volume are weighted and summed to obtain the optimal single-time replenishment volume.

[0009] Furthermore, in the weighted summation, the weighting coefficient α of the static compensation volume and the weighting coefficient β of the estimated compensation volume satisfy α+β=1, and the weighting coefficients α and β are dynamically adjusted according to the system operating status or historical estimated error rate.

[0010] Furthermore, the dynamic adjustment based on the system operating status or historical prediction error rate includes: Record the estimated compensation volume and actual concentration change for each intermittent compensation action, and calculate the estimation error rate. When the estimated error rate exceeds the preset upper limit, increase α and decrease β; When the estimated error rate is lower than the preset lower limit, decrease α and increase β.

[0011] Furthermore, the proportioning system includes: A mixing tank is used to hold and store liquid medicine. A circulation pipeline connects the outlet and return port of the mixing tank to form a closed-loop circulation circuit. A circulation pump and a concentration meter are installed on the circulation pipeline. A first liquid inlet pipeline is connected to a first liquid source and the mixing tank. A first valve and a first flow meter are installed on the first liquid inlet pipeline. The second liquid inlet pipeline connects the second liquid source to the mixing tank, and the second liquid inlet pipeline is equipped with a second valve and a second flow meter; The control module is electrically connected to the concentration meter, the first valve, the second valve, the first flow meter, and the second flow meter, respectively, and is used to execute steps S1 to S4.

[0012] Furthermore, by selecting the corresponding inlet pipeline and implementing closed-loop flow control, the single optimal replenishment volume includes: When the estimated concentration is lower than the target concentration, the first inlet pipeline is selected and a high-concentration first drug solution is added. When the estimated concentration is higher than the target concentration, select the second inlet pipeline and add a low-concentration second drug solution; During the replenishment process, the instantaneous flow rate is fed back in real time through the flow meter, and the opening degree of the corresponding valve is adjusted by the closed-loop control algorithm so that the actual flow rate tracks the target flow rate.

[0013] Furthermore, the first valve is a high-precision miniature needle valve with a flow control resolution of not less than 0.1 mL / min, used to achieve precise replenishment of trace amounts of drug solution; the second valve is a proportional valve, used to achieve stable replenishment of large flow diluent.

[0014] Furthermore, the closed-loop control algorithm is a PID control algorithm, and different PID parameter groups are configured for the response characteristics of the first valve and the second valve, respectively.

[0015] Furthermore, the duration of the forced waiting time is dynamically determined based on the ratio of the single optimal replenishment volume to the total volume of the mixing tank.

[0016] Furthermore, the dynamic determination method for the forced waiting time is as follows: ,in, To enforce waiting time, The optimal fluid replacement volume for a single treatment is [value missing]. The total volume of the mixing tank. The base waiting time is K, where K is the proportionality coefficient.

[0017] Furthermore, the dead zone threshold in step S2 is greater than the tolerance range of the target concentration in step S4, in order to avoid frequently triggering intermittent compensation actions near the target value.

[0018] Furthermore, the concentration meter is a conductivity type concentration meter, and the control module stores a three-dimensional calibration relationship table of drug concentration-conductivity-temperature; The mixing system also includes a temperature sensor for real-time detection of the liquid temperature. The control module performs temperature compensation on the concentration measurement value based on the feedback from the temperature sensor and the three-dimensional calibration relationship table, and then performs real-time filtering processing in step S1.

[0019] Furthermore, the control module is also used to perform the following liquid level maintenance and replenishment process: When the liquid level in the mixing tank is detected to be lower than the preset low liquid level threshold, a liquid level replenishment action is triggered. Based on the current liquid level, the volume of the liquid, the current estimated concentration, and the target concentration, the type of liquid to be replenished and the total amount of liquid to be replenished are calculated based on the law of conservation of mass. Control the valve on the corresponding liquid inlet pipeline to open and execute the total liquid replenishment; The liquid level replenishment action and the intermittent compensation action in step S2 are triggered independently. Liquid level replenishment is responsible for maintaining the liquid level within a preset range, while intermittent compensation is responsible for precise concentration adjustment. When the liquid level replenishment action and the intermittent compensation action are triggered simultaneously, the control module prioritizes executing the action with the larger concentration deviation, or combines the replenishment amounts of both actions in one replenishment action.

[0020] Furthermore, the iteration termination condition until the concentration estimate stabilizes within the tolerance range of the target concentration stops the intermittent compensation operation when any of the following conditions are met: Condition 1: The estimated concentration value falls within the tolerance range of the target concentration, i.e. ,in, For tolerance range, This is a concentration estimate. The target concentration; Condition 2: After N consecutive intermittent compensation actions, the actual change in concentration caused by two adjacent actions is less than the preset change threshold, indicating that the system has become stable. Wherein, N is an integer greater than or equal to 3, and the tolerance range is less than the dead zone threshold in step S2.

[0021] Furthermore, it also includes an adaptive learning unit, which records the amount of fluid replenishment, the estimated concentration before replenishment, and the estimated concentration after replenishment for each intermittent compensation action, and continuously optimizes the prediction accuracy of the estimated compensation volume through machine learning methods.

[0022] The beneficial effects of the chemical solution ratio control method provided by this invention are as follows: By real-time filtering of the observed concentration signal of the solution in the circulation pipeline, the filtered concentration estimate is used to replace the original observed signal as the basis for subsequent control judgment. This separates the random high-frequency noise introduced by fluid turbulence and bubbles from the actual concentration change trend, avoiding the influence of single-point spikes on control, and making subsequent judgments based on a smoother signal that is closer to the actual distribution. Furthermore, a dead zone threshold is set; intermittent compensation is only triggered when the concentration estimate deviates from the target concentration and exceeds this threshold. This prevents small random fluctuations near the target value caused by residual noise from triggering system responses, avoiding meaningless frequent adjustments. The correction algorithm simultaneously introduces two variables—the current concentration deviation and the historical compensation effect—to calculate the optimal replenishment volume for a single operation. The actual observed concentration response after historical compensation operations is used to correct purely theoretical calculations, preventing the static formula from outputting a replenishment volume that deviates from actual needs under abnormal data. The corresponding inlet pipeline is selected based on the deviation polarity of the concentration estimate, i.e., high-concentration drug solution or diluent is added to the mixing tank. The optimal replenishment volume obtained in step S2 is executed through flow closed-loop control. The flow closed-loop control ensures that the compensation volume is injected with controlled precision, avoiding over-execution caused by uncontrolled discharge from ordinary switching valves, and making the actual injection volume close to the ideal compensation value output by the algorithm. After a single replenishment is completed, a forced waiting time is set, and the closed-loop operation of the circulation pipeline is maintained during this period to drive sufficient exchange of drug solution in the mixing tank. Its function is to actively eliminate the spatial local concentration gradient caused by replenishment injection before the next concentration sampling of the system, ensuring that the circulating drug solution sampled in the next cycle step S1 has approached the overall mean rather than a local spurious peak. The above steps proceed sequentially, filtering out noise at the signal source, avoiding false data-driven misjudgments in the decision-making stage, matching and compensating for accuracy in the execution stage, and forcibly mixing and eliminating local gradients during the sampling interval. Each step has a clear division of labor, and together they enable the system to gradually converge from the initial deviation state to the target concentration tolerance range, eliminating the positive feedback oscillation phenomenon caused by false data inducing overshoot and overshoot generating false data again. Attached Figure Description

[0023] Figure 1 This is a flowchart of a chemical solution ratio control method according to an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art. The terms "comprising" and similar expressions used herein mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, but does not exclude other elements or objects. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] Reference Figure 1 In some embodiments of the present invention, a method for controlling the proportion of a chemical drug solution is provided, which is applied to a chemical drug solution proportioning system. The proportioning system includes a mixing tank, a circulation pipeline, and at least one inlet pipeline for replenishing the drug solution.

[0026] Specifically, the mixing tank is used to contain and store the drug solution. A circulation pipeline connects the outlet and return port of the mixing tank, forming a closed-loop circulation circuit. A circulation pump and a concentration meter are installed on the circulation pipeline. The circulation pump drives the continuous flow of the drug solution within the circulation pipeline, and the concentration meter detects the concentration of the drug solution flowing through the circulation pipeline online and outputs a corresponding observation signal. An inlet pipeline connects the drug solution source to the mixing tank, used to replenish the mixing tank with drug solutions of different concentrations. A valve and a flow meter are installed on the inlet pipeline. The valve controls the on / off state and flow rate of the replenishment, and the flow meter provides real-time feedback of the instantaneous flow rate. The proportioning system also includes a control module, which is electrically connected to the concentration meter, valve, and flow meter, respectively, and is used to execute the steps of the control method described below.

[0027] In this embodiment, a conductivity-type concentration meter is used, and the control module stores a three-dimensional calibration table of drug concentration, conductivity, and temperature. The mixing system also includes a temperature sensor installed on the circulation pipeline or in the mixing tank. The temperature sensor is electrically connected to the control module and is used to detect the drug temperature in real time and feed the temperature signal back to the control module. After reading the original measurement value from the concentration meter, the control module performs temperature compensation on the concentration measurement value according to the feedback from the temperature sensor and the three-dimensional calibration table to obtain a temperature-corrected concentration observation signal. This concentration observation signal is then sent to step S1 for subsequent real-time filtering. Through the temperature compensation step, the influence of drug temperature changes on the conductivity measurement results is eliminated, ensuring that the concentration observation signal reflects the true concentration of the drug under different operating temperature conditions.

[0028] In some embodiments of the present invention, the control module performs the following steps S1 to S4 to achieve precise control of the concentration of the chemical solution.

[0029] Step S1: Perform real-time filtering on the drug concentration observation signal in the circulation pipeline to remove random disturbance noise of the fluid and output the concentration estimate for the current period.

[0030] In this embodiment, real-time filtering is implemented using the Kalman filter algorithm, and its specific execution process is as follows: First, state prediction and covariance prediction are performed based on the state estimate and process noise covariance of the previous cycle; second, the Kalman gain is calculated based on the measurement noise covariance; then, the state is updated using the concentration observation signal acquired in the current cycle and the Kalman gain, and the concentration estimate for the current cycle is output; finally, the covariance matrix is ​​updated for use in the next cycle iteration. When the concentration observation signal is measured while flowing through the circulation pipeline, it is affected by random disturbances such as fluid turbulence and bubbles, resulting in the superposition of irregular noise spikes in the original observation signal. The Kalman filter algorithm uses the process noise covariance to characterize the slow change in drug concentration and the measurement noise covariance to characterize the intensity of random measurement noise from the concentration meter. It optimizes the choice between trusting the state prediction model and trusting the current observation value through the Kalman gain: when measurement noise is high, the Kalman gain automatically decreases, and the filtered output tends to maintain the state estimate of the previous cycle, thus suppressing the impact of noise spikes on the filtering result; when process noise is relatively high, the Kalman gain automatically increases, and the filtered output tracks the actual concentration change trend more quickly. Therefore, the filtered concentration estimate effectively removes random fluid disturbance noise without introducing significant phase lag, providing a smooth and nearly true concentration signal for subsequent compensation decisions.

[0031] In some specific embodiments of the present invention, taking the preparation of a diluted hydrofluoric acid working solution with a target concentration of 0.49% (mass percentage concentration) as an example, the total volume of the mixing tank is 100L, and the concentration meter continuously collects concentration data with a sampling period of 100ms. During the replenishment and circulation process, due to the influence of fluid turbulence and bubbles, the original measured signal of the concentration meter has random noise fluctuations of approximately ±0.03%. The control module performs Kalman filtering on the original concentration observation signal, setting the process noise covariance Q to 0.0001 and the measurement noise covariance R to 0.001. These values ​​are determined based on the calibration accuracy of the concentration meter and the actual operating noise level. After Kalman filtering, the standard deviation of the concentration estimate is reduced from ±0.03% of the original signal to within ±0.005%, and the noise amplitude is compressed to less than one-sixth of the original level. In actual operation, if the original measured signal in a certain sampling period exhibits a noise spike deviating from the true value by 0.04% due to bubble interference, this spike, without filtering, is sufficient to trigger the intermittent compensation action in subsequent step S2, causing an erroneous liquid replenishment. However, after Kalman filtering, this spike is significantly attenuated in the filtered output, and the filtered concentration estimate remains smooth and is not significantly affected by single-point spikes, thus avoiding erroneous actions. The filtered concentration curve can smoothly track the true trend of drug concentration changes while removing noise, avoiding erroneous compensation triggers caused by single-point noise, and also avoiding response lag caused by excessive smoothing of the filter.

[0032] Step S2: When the estimated concentration deviates from the target concentration and exceeds the set dead zone threshold, an intermittent compensation action is triggered; based on the current concentration deviation and historical compensation effect, the optimal liquid replenishment volume for a single operation is obtained through a correction algorithm.

[0033] The purpose of the dead zone threshold is to prevent the system from triggering any compensation action when the concentration estimate fluctuates within a preset small range near the target concentration, thus avoiding frequent ineffective adjustments caused by residual noise or minor concentration fluctuations. The system only determines that concentration correction is needed when the concentration estimate clearly deviates from the target concentration and exceeds the dead zone threshold. The correction algorithm integrates two information dimensions: the current concentration deviation and historical compensation effects. The purpose is that the current concentration deviation provides a static compensation requirement based on the real-time state, but this requirement is calculated on the premise that the drug solution is uniformly mixed and the concentration estimate fully represents the overall concentration. However, this premise is not necessarily strictly true in actual dynamic processes. The historical compensation effect records the correspondence between the actual injected volume and the actual change in concentration during a past intermittent compensation action, reflecting the true transmission characteristics of the injection volume and concentration response under the current system state. The correction algorithm combines both to derive the optimal injection volume for a single operation. This absorbs the immediacy of the current deviation and corrects the theoretical calculation using historical response data, making the single injection volume closer to the actual compensation value required by the system and avoiding excessive compensation commands driven by pseudo-data.

[0034] In some specific embodiments of the present invention, the above-mentioned correction algorithm for deriving the optimal replenishment volume for a single operation based on the current concentration deviation and historical compensation effects specifically includes the following steps: First, extract the effective volume of the drug solution in the current system and the estimated concentration value, and calculate the static compensation volume based on the law of conservation of mass; second, extract the actual replenishment volume of the previous intermittent compensation action and the actual change in concentration it caused, and calculate the estimated compensation volume in combination with the current target concentration change; finally, perform a weighted summation of the static compensation volume and the estimated compensation volume to obtain the optimal replenishment volume for a single operation. The calculation of the static compensation volume is based on the assumption that the current concentration estimate is completely accurate and that the drug solution in the tank is uniformly mixed, and its calculation result is a theoretical compensation value under ideal conditions; the estimated compensation volume bypasses the assumptions of uniform mixing and accurate sampling, and directly uses the proportional relationship between the previous actual replenishment volume and the measured concentration response to extrapolate the replenishment volume required to achieve the target change this time. The weighted summation of the two is equivalent to a compromise between theoretical calculation and empirical derivation. When historical compensation results are good and the prediction model is reliable, the weight of the predicted compensation volume can be increased; when operating conditions change significantly and the reference value of historical experience decreases, the weight of the static compensation volume can be increased.

[0035] Furthermore, in the above-mentioned scheme of weighted summation of static compensation volume and estimated compensation volume, the weighting coefficient α of static compensation volume and the weighting coefficient β of estimated compensation volume satisfy α + β = 1, and the weighting coefficients α and β are dynamically adjusted according to the system operating state or historical prediction error rate. The weighting coefficients are no longer fixed values, but adaptively change with the system operating state. The basis for their dynamic adjustment includes, but is not limited to: when the system is in the initial preparation stage or the recovery stage after experiencing a large disturbance, the concentration deviation is large and the historical compensation record may not be fully established. At this time, the static compensation volume can be given a higher weight for rapid response; after the system tends to a steady state, the historical compensation record is rich, and the accuracy of the prediction model improves, the estimated compensation volume can be given a higher weight to improve the smoothness and accuracy of compensation. Dynamic adjustment enables the correction algorithm to maintain a suitable balance in different operating stages.

[0036] For example, a specific method for dynamic adjustment based on historical prediction error rates is as follows: The control module records the compensation effect of several intermittent compensation actions in the past and calculates the prediction error rate E of the predicted compensation volume, which is defined as |actual concentration change - predicted concentration change| / predicted concentration change. When the prediction error rate E exceeds a preset upper limit (e.g., 20%), it indicates that the prediction accuracy of the recent prediction model is poor. The system automatically increases α and decreases β, making the control more dependent on the static compensation volume based on the current concentration deviation. When the prediction error rate E is lower than a preset lower limit (e.g., 5%), it indicates that the recent prediction model is accurate and reliable. The system automatically decreases α and increases β, making the control more dependent on historical experience extrapolation. The dynamic adjustment of the weighting coefficients can use a linear mapping formula, for example, α = 0.5 + 0.3 × (E - 0.05) / 0.25, and limit α to the range of 0.2 to 0.8, β = 1 - α. When E=5%, α=0.5, β=0.5, with static calculation and historical prediction each accounting for half the weight; when E rises to 20%, α=0.68, β=0.32, and the control focus shifts towards static calculation; when E drops to near zero, α approaches the lower limit of 0.2, and β approaches 0.8, with historical experience becoming dominant. Through the above feedback adjustment based on the prediction error rate, the weighting coefficients continuously converge towards the optimal value under the current operating conditions. It should be noted that the above-mentioned preset upper limit of 20%, preset lower limit of 5%, and the range of α from 0.2 to 0.8 are all exemplary values. In practical applications, they can be adjusted according to the system scale, drug characteristics, and process requirements, which does not affect the control logic of this invention.

[0037] Step S3: Based on the deviation polarity of the concentration estimate, select the corresponding inlet pipeline and execute the single optimal replenishment volume through closed-loop flow control. Deviation polarity refers to whether the concentration estimate is too high or too low relative to the target concentration. When the concentration estimate is lower than the target concentration, a high-concentration drug solution needs to be added to the mixing tank to increase the concentration; in this case, an inlet pipeline connected to a high-concentration drug solution source is selected. When the concentration estimate is higher than the target concentration, a low-concentration drug solution or diluent needs to be added to the mixing tank to decrease the concentration; in this case, an inlet pipeline connected to a low-concentration drug solution source or diluent source is selected. During the replenishment process, the flow closed-loop control continuously adjusts the opening of the corresponding valve based on the deviation between the instantaneous flow rate fed back by the flow meter and the target flow rate, ensuring that the actual injected instantaneous flow rate accurately tracks the preset target flow rate curve until the cumulative replenishment volume reaches the single optimal replenishment volume calculated in step S2, at which point the valve is closed. By controlling the flow rate in a closed loop, the optimal replenishment volume output in step S2 can be executed precisely, avoiding the replenishment volume deviation caused by factors such as nonlinearity between valve opening and flow rate and pipeline pressure fluctuations. This ensures that the actual replenishment volume injected under control is highly consistent with the ideal compensation value expected by the algorithm, thus guaranteeing the accuracy of concentration correction from the execution end.

[0038] In some embodiments of the present invention, a circulation pipeline is connected between the outlet and return port of the mixing tank of the proportioning system, and a circulation pump and a concentration meter are sequentially arranged on the circulation pipeline. A first inlet pipeline connects a first drug source to the mixing tank, and a first valve and a first flow meter are arranged on the first inlet pipeline; a second inlet pipeline connects a second drug source to the mixing tank, and a second valve and a second flow meter are arranged on the second inlet pipeline. The control module is electrically connected to the concentration meter, the first valve, the second valve, the first flow meter, and the second flow meter, respectively, receives measurement signals from the concentration meter and the flow meter, and sends control commands to each valve. Preferably, the circulation pump is a magnetically driven pump, characterized by zero leakage and low pulsation, and the circulation flow rate is 3 to 10 times the volume of the mixing tank per hour. In this embodiment, the first drug source provides a high-concentration first drug solution, such as a 49% hydrofluoric acid stock solution or an intermediate solution that still has a high concentration after dilution; the second drug source provides a low-concentration second drug solution or a pure diluent, such as deionized water. The specific rule for selecting the inlet pipeline based on the deviation polarity of the concentration estimate is as follows: when the concentration estimate is lower than the target concentration, the first inlet pipeline connected to the first drug source is selected and a high-concentration first drug solution is added; when the concentration estimate is higher than the target concentration, the second inlet pipeline connected to the second drug source is selected and a low-concentration second drug solution is added. During the replenishment process, the instantaneous flow rate is fed back in real time by the flow meter on the corresponding pipeline, and the control module uses a closed-loop control algorithm to adjust the opening of the corresponding valve so that the actual flow rate tracks the target flow rate. Through the above structural configuration, the system has independent and precise compensation execution channels in both directions of concentration deviation (high and low).

[0039] In some embodiments of the present invention, the first valve and the second valve employ valve components with different characteristics to match their respective flow control requirements. The first valve is a high-precision miniature needle valve with a flow control resolution of not less than 0.1 mL / min, used to achieve precise micro-volume replenishment of high-concentration drug solutions; the second valve is a proportional valve, used to achieve stable large-flow replenishment of diluent; specifically, the proportional valve is an electro-pneumatic proportional valve or an electric proportional valve, with a response time of no more than 100 ms. When the concentration deviation is small and the required single optimal replenishment volume is only a few milliliters or even less than one milliliter, the needle valve, with its extremely high flow control resolution, can accurately inject such a small amount of high-concentration drug solution into the mixing tank without over-injection due to the limitation of the valve's minimum controllable flow rate. The proportional valve, when a larger volume of diluent needs to be replenished, can complete the replenishment action at a stable flow rate, avoiding pipeline impact and flow fluctuations. The differentiated selection of the first and second valves enables the system to maintain precise execution capability over a wide range from micro-volume to large flow rates. Furthermore, the closed-loop control algorithm employs a PID control algorithm, and different PID parameter sets are configured for the response characteristics of the first and second valves, respectively. Specifically, due to its relatively slow response caused by its mechanical structure, the needle valve's PID parameter can be set to proportional gain. =1.8, Integral Gain =0.4, Differential gain =0.6, appropriately increasing the derivative gain to improve the dynamic response speed; the proportional valve has a fast response and a large flow rate, and the parameters of its PID controller can be optimally set: ; It emphasizes stability and resistance to integral saturation. By configuring PID parameters independently for different valves, the flow closed-loop control of each actuator channel can achieve good dynamic response and steady-state accuracy.

[0040] Step S4: After the optimal replenishment volume is injected in a single operation, a preset forced waiting time is initiated. During this forced waiting time, the liquid in the mixing tank is kept in a closed loop through the circulation pipeline to eliminate local concentration gradients. After the forced waiting time ends, the process returns to step S1 until the estimated concentration stabilizes within the tolerance range of the target concentration. When a single replenishment operation is completed and a high-concentration or low-concentration liquid is injected into a local area of ​​the mixing tank, uneven spatial concentration distribution is inevitable: the concentration in the area near the injection point has not yet been fully exchanged with the surrounding liquid. If the liquid in the circulation pipeline is sampled immediately at this time, the concentration meter is likely to measure a false peak in a local concentration that has not yet been mixed. Step S4 sets a forced waiting time and keeps the circulation pump running continuously during this period to drive sufficient convection and exchange of the liquid in the tank, so that the high-concentration or low-concentration area at the injection point is fully diffused and diluted to the entire tank before the next sampling cycle. The length of the forced waiting time ensures that when the system executes step S1 in the next cycle, the liquid flowing through the concentration meter in the circulation pipeline represents the overall average concentration of the mixing tank, rather than spurious data from locally unevenly mixed areas. This cuts off the positive feedback loop from spurious data to misjudgment, and from misjudgment to reverse overshoot. After the forced waiting time ends, the system returns to step S1 to re-evaluate the concentration estimate. If the concentration estimate is still outside the tolerance range, intermittent compensation is triggered again; if it is within the tolerance range, compensation stops. Through the iterative cycles of steps S1 to S4, the system concentration gradually approaches the target concentration from its initial deviation and remains stable within the tolerance range.

[0041] In some embodiments of the present invention, the duration of the forced waiting time is not a fixed value, but is dynamically determined based on the ratio of the optimal single replenishment volume to the total volume of the mixing tank. The larger the replenishment volume relative to the tank volume, the longer it takes for the newly injected liquid to mix evenly throughout the entire tank, and the required forced waiting time increases accordingly. When the replenishment volume is very small, for example, only one ten-thousandth of the total volume, it can quickly disperse evenly under strong convection conditions, and the forced waiting time can be shortened accordingly. By dynamically linking the waiting time to the replenishment volume / volume ratio, the drawbacks of using a fixed waiting time—insufficient mixing for large replenishment volumes and excessively long waiting times for small replenishment volumes—are avoided.

[0042] For example, one way to dynamically determine the forced waiting time is as follows: ,in, To enforce waiting time, The optimal fluid replacement volume for a single treatment is [value missing]. The total volume of the mixing tank. The base waiting time is K, and the proportionality coefficient is K. The base waiting time and the proportionality coefficient K can be determined experimentally based on parameters such as the geometry of the mixing tank, the circulation flow rate of the circulating pump, and the viscosity of the solution. For example, when... It is 2.3L. With a base wait time of 60 seconds and K set to 1.0, the forced wait time is... Approximately 1.4 seconds. It should be noted that the above values ​​are merely examples; the actual parameters should be selected and calibrated based on the mixing efficiency of the specific system, which does not conflict with the control logic framework of this invention. In other embodiments of this invention, the forced waiting time can also be directly determined as a fixed value based on the ratio of the circulation pump's flow rate to the total volume of the mixing tank (i.e., the circulation ratio). For example, when the circulation ratio is 3 to 10 times / hour, the forced waiting time can be a fixed value within the range of 10 to 60 seconds. This simplified method can also meet the requirement of eliminating local concentration gradients in systems with sufficiently strong circulation mixing capabilities.

[0043] In some embodiments of the present invention, the dead zone threshold in step S2 is greater than the tolerance range of the target concentration in step S4. The dead zone threshold determines whether intermittent compensation is triggered, and the tolerance range determines when the system stops intermittent compensation. Setting the dead zone threshold to be greater than the tolerance range means that the concentration estimate must deviate to a certain extent before compensation is initiated. Once compensation is initiated, the system will continuously iterate until the concentration estimate enters a tolerance window that is more stringent than the trigger threshold. This two-layer threshold design creates a hysteresis interval near the target value: compensation is triggered at the outer dead zone threshold and stopped at the inner tolerance range. The gap between the two layers effectively absorbs the small overshoot and residual fluctuations caused by the compensation action, preventing the system from repeatedly switching between compensation and stop due to small fluctuations around the target value after convergence, thus improving the stability after convergence. For example, the target concentration is 0.49%, and the dead zone threshold is... The threshold can be set to 0.005%, and the tolerance range δ_accept can be set to 0.002%. In this case, the dead zone is 0.49% ± 0.005% of the target concentration, i.e., 0.485% to 0.495%, and the tolerance range is 0.49% ± 0.002% of the target concentration, i.e., 0.488% to 0.492%. The concentration estimate must exceed the 0.485% to 0.495% range to trigger compensation, and the compensation stops when the concentration estimate enters the 0.488% to 0.492% range, with a hysteresis gap of 0.003% between the two.

[0044] In some embodiments of the present invention, the intermittent compensation operation is stopped when the iteration termination condition of steps S1 to S4 meets any of the following conditions: Condition 1: The estimated concentration value falls within the tolerance range of the target concentration, i.e. ,in, For tolerance range, This is a concentration estimate. The target concentration; Condition 2: After N consecutive intermittent compensation actions, the actual concentration change caused by two adjacent actions is less than a preset change threshold, indicating that the system has stabilized. Here, N is an integer greater than or equal to 3. For example, N can be 3, 4, 5, or 6.

[0045] The purpose of setting condition two is that, under certain operating conditions, due to inherent measurement noise or minor environmental disturbances, the concentration estimate may consistently fail to fall within the set tolerance range, even though the actual concentration fluctuation is minimal and the system is in a substantial steady state. In such cases, relying solely on condition one might cause the system to continuously trigger compensation actions and fail to exit the iteration. Condition two determines whether the system has substantially converged by monitoring the actual effect of adjacent compensations. Even if the concentration estimate does not yet meet condition one, as long as the concentration change resulting from multiple consecutive compensations is negligible, the system determines that there is no need to continue compensation and terminates the iteration. Thus, the two conditions complement each other, taking into account reasonable exit scenarios for both concentration target achievement and substantial convergence.

[0046] In some specific embodiments of the present invention, the complete iterative process of steps S1 to S4 is taken as an example for further illustration. Assuming the system is in a stable operating phase, the target concentration... The dead zone threshold is 0.49%. The concentration is 0.005%, the total volume of the mixing tank is 100L, and the effective volume corresponding to the current liquid level is approximately 45L. After Kalman filtering in step S1, the estimated concentration for the current period is... The concentration was 0.51%, which exceeded the dead zone range of 0.49% ± 0.005%, or 0.485% to 0.495%, triggering intermittent compensation. Since the estimated concentration was higher than the target concentration, the concentration was determined to be too high, requiring the addition of deionized water for dilution; therefore, a second inlet line was selected.

[0047] In step S2, the control module first calculates the static compensation volume. : =45×(0.51%-0.49%) / 0.49%≈1.837L. Next, read the historical record of the previous intermittent compensation: after the last addition of 1.5L of deionized water, the concentration decreased from 0.505% to 0.495%, an actual change of 0.01%. The target change is 0.02% (from 0.51% to 0.49%), therefore the estimated compensation volume is... =1.5×0.02% / 0.01%=3.0L.

[0048] The weighting coefficients are set to α=0.6 and β=0.4. These values ​​correspond to the system being in a dynamic adjustment phase, with slightly higher weights in static calculations. The weighted calculation yields... =0.6×1.837+0.4×3.0=2.302L. In step S3, the control module opens the proportional valve on the second inlet pipeline and performs replenishment with PID closed-loop control at a target flow rate of 1L / min until the cumulative flow of the second flow meter reaches 2.302L, at which point the proportional valve closes. In step S4, after replenishment is completed, a forced waiting time of 60s is initiated, and the circulating pump runs continuously at a flow rate of 5m³ / h to ensure that the newly added 2.3L of deionized water is fully mixed with the approximately 45L of drug solution in the tank. After the 60s waiting time ends, the system returns to step S1, rereads the concentration meter signal, and performs Kalman filtering to obtain an updated concentration estimate of 0.495%, which is exactly at the dead zone boundary (the upper limit of the dead zone is 0.495%). The system then triggers intermittent compensation again. In the second round of compensation, =47.3×(0.495%-0.49%) / 0.49%≈0.483L, Based on the actual effect of the previous round of compensation (the concentration decreased from 0.51% to 0.495% after adding 2.302L, with an actual change of 0.015%) and the updated calculation of the target change of 0.005%, the weighted average was calculated as follows: The initial concentration was approximately 0.52 L. After performing the second round of replenishment and waiting for mixing, the concentration estimate obtained by filtering again was 0.491%, which fell within the tolerance range of 0.49% ± 0.002% (0.488% to 0.492%), satisfying condition one, and the intermittent compensation action terminated. Through the above two rounds of iteration, the system concentration converged from 0.51% to 0.491%, and no overshoot or oscillation occurred throughout the process.

[0049] In some embodiments of the present invention, the control module is also used to execute a liquid level maintenance replenishment process. Specifically, when the liquid level in the mixing tank is detected to be lower than a preset low liquid level threshold, a liquid level replenishment action is triggered; based on the current liquid level corresponding to the liquid volume, the current estimated concentration, and the target concentration, the type of liquid to be replenished and the total replenishment amount are calculated based on the law of conservation of mass; the valve on the corresponding inlet pipeline is controlled to open, and the total replenishment amount is executed. The liquid level replenishment action and the intermittent compensation action described in step S2 are triggered independently: liquid level replenishment is responsible for replenishing the liquid level to the normal operating range when continuous liquid consumption causes the liquid level to drop to the low limit, and its calculation is based on the total liquid volume to be replenished and the target concentration; intermittent compensation is responsible for precisely fine-tuning the concentration when the liquid level is within the normal range but the concentration deviates. When the liquid level replenishment action and the intermittent compensation action are triggered simultaneously, the control module prioritizes executing the action with the larger concentration deviation, or combines the replenishment amounts of both actions in one replenishment action. By decoupling the functions of liquid level maintenance and concentration regulation, the system will not have logical conflicts when facing both liquid level drop and concentration shift at the same time, and the calculation and execution of liquid replenishment for each function can remain clear and independent.

[0050] In a specific application example of this invention, a system with a total mixing tank volume of 100L and a target concentration of 0.49% is used to illustrate the independent triggering of liquid level replenishment and concentration compensation. During normal system operation, the liquid-using equipment continuously draws liquid from the mixing tank, and the liquid level gradually decreases. When the liquid level sensor detects that the liquid level has dropped to 20% of the preset low liquid level threshold, corresponding to a volume of approximately 20L, the control module triggers the liquid level replenishment procedure. The control module reads the effective volume V0 = 20L corresponding to the current liquid level, and simultaneously reads the current concentration estimate after Kalman filtering. =0.47%, lower than the target concentration of 0.49%. Based on the law of conservation of mass, the volume of 49% hydrofluoric acid that needs to be added is calculated as follows: =20×(0.49%-0.47%) / (49%-0.49%)≈0.0082L=8.2mL. The control module sets the target flow rate of the needle valve on the first inlet line to 5mL / min, and starts PID closed-loop control to perform liquid replenishment. The first flow meter provides real-time feedback of the instantaneous flow rate, and the actual flow rate stabilizes within the range of 4.8mL / min to 5.2mL / min. When the cumulative flow rate of the first flow meter reaches 8.2mL, the control module closes the needle valve, and the single liquid replenishment action is completed. At this time, the total volume in the tank is approximately 20.008L, and the liquid level is still below the low liquid level threshold. After the liquid replenishment is completed, the circulation pump continues to run. After waiting for 30 seconds of forced mixing time, the concentration estimate is reread as 0.488%, which is still lower than the target concentration, triggering the second liquid replenishment, adding approximately 1.6mL of hydrofluoric acid. After multiple replenishment iterations, the concentration gradually approached the target value, with a final cumulative replenishment volume of approximately 50 mL and a total tank volume of approximately 20.05 L. The liquid level had not yet returned to the normal range, but the concentration had reached the tolerance range of 0.49% ± 0.01%. This example demonstrates that the liquid level replenishment action and the intermittent concentration compensation action are functionally independent: after a single trigger, liquid level replenishment may only add a small amount of liquid to compensate for concentration correction, and the recovery of the liquid level requires the accumulation of multiple liquid level replenishments. However, fine-tuning of the concentration can be completed simultaneously during the liquid level replenishment process, and the two will not conflict due to triggering conditions.

[0051] In some embodiments of the present invention, the above-mentioned correction algorithm further includes an adaptive learning unit. This adaptive learning unit records the replenishment volume, the estimated concentration before replenishment, and the estimated concentration after replenishment for each intermittent compensation action, and continuously optimizes the prediction accuracy of the estimated compensation volume using machine learning methods. The adaptive learning unit operates within the data storage unit of the control module. Its working principle is as follows: as the system's operating cycle accumulates, the number of historical compensation record samples continuously increases. The adaptive learning unit uses these samples to train and update the parameters of the prediction model online, gradually reducing the deviation between the predicted value of the estimated compensation volume and the actual required compensation value. As the accuracy of the prediction model improves, the weight β of the estimated compensation volume during weighted summation in step S2 can be gradually increased, improving the smoothness of compensation and the success rate on the first attempt, and reducing the number of iterations for intermittent compensation actions. It should be noted that the introduction of the adaptive learning unit does not change the main control framework of steps S1 to S4, but rather continuously optimizes the calculation accuracy of the estimated compensation volume within the correction algorithm of step S2, thus enhancing the performance of the correction algorithm.

[0052] In some other embodiments of the present invention, some of the algorithms or execution mechanisms in the above scheme can be implemented in an alternative manner.

[0053] For example, in addition to the Kalman filter algorithm, the real-time filtering process in step S1 can also employ moving average filtering, exponentially weighted moving average filtering, or an adaptive denoising algorithm based on wavelet transform. The advantage of Kalman filtering lies in its ability to perform optimal recursive estimation based on noise statistics, enabling it to effectively track the dynamic changes in the true concentration while filtering out noise. When the system has limited computational resources or the concentration changes extremely slowly, the simpler exponentially weighted moving average filtering can also achieve the function of step S1 within acceptable accuracy; its smoothness is controlled by adjusting the forgetting factor.

[0054] For example, in step S3, the flow closed-loop control can employ model predictive control or fuzzy control in addition to the PID control algorithm. The PID control algorithm, due to its simple structure and mature parameter tuning methods, has wide applicability in flow control applications. Model predictive control may perform better in situations where valve response lag is significant and pipeline delays are long, but it involves greater computation and places higher demands on controller performance. Regardless of the closed-loop control algorithm used, as long as the valve opening can be adjusted in real time based on flowmeter feedback to track the target flow rate, the precise execution of the replenishment amount required in step S3 can be achieved.

[0055] For example, in addition to adopting a priority execution or combined execution strategy, the triggering and coordination of the liquid level maintenance replenishment process and the concentration intermittent compensation action can also be set to a mutual exclusion execution strategy. That is, if a replenishment action is completed and a forced waiting time has elapsed, regardless of whether it is triggered by liquid level or concentration, another replenishment action will not be triggered, thereby further reducing the mutual interference that may occur when the two replenishment logics work at the same time.

[0056] It should be noted that the order of description of the embodiments in this application is not intended to limit the priority of the embodiments. Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the specification of this application is for the purpose of describing specific embodiments only and is not intended to limit this application. The embodiments of this application have been described above with reference to the accompanying drawings, but this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many forms under the guidance of this application without departing from the spirit and scope of protection of the claims. All equivalent transformations made under the inventive concept of this application, using the content of the specification and drawings of this application, or direct / indirect applications in other related technical fields are included within the scope of patent protection of this application.

Claims

1. A method for controlling the proportioning of chemical solutions, applied to a proportioning system comprising a mixing tank, a circulation pipeline, and at least one inlet pipeline for replenishing the chemical solution, characterized in that, Includes the following steps: S1: Perform real-time filtering on the concentration observation signal of the drug solution in the circulation pipeline to filter out random disturbance noise of the fluid and output the concentration estimate for the current period. S2: When the estimated concentration deviates from the target concentration and exceeds the set dead zone threshold, an intermittent compensation action is triggered; based on the current concentration deviation and historical compensation effects, the optimal replenishment volume for a single operation is obtained through a correction algorithm; S3: Based on the deviation polarity of the concentration estimate, select the corresponding inlet pipeline and execute the single optimal replenishment volume through closed-loop control of the flow rate; S4: After the single optimal replenishment volume is injected, a preset forced waiting time is initiated, and during the forced waiting time, the closed-loop circulation of the drug solution in the mixing tank is maintained through the circulation pipeline to eliminate the spatial local concentration gradient; after the forced waiting time ends, the process returns to step S1 until the concentration estimate stabilizes within the tolerance range of the target concentration.

2. The method for controlling the proportion of chemical solutions according to claim 1, characterized in that, The real-time filtering of the drug concentration observation signal, using the Kalman filter algorithm, includes the following steps: Based on the state estimate and process noise covariance of the previous cycle, state prediction and covariance prediction are performed. Calculate the Kalman gain based on the measurement noise covariance; Using the concentration observation signal acquired in the current period and the Kalman gain, the state is updated, and the concentration estimate for the current period is output. Update the covariance matrix for use in the next iteration.

3. The method for controlling the proportion of chemical solutions according to claim 1, characterized in that, The process of determining the optimal fluid replacement volume for a single treatment based on the current concentration deviation and historical compensation effects using a correction algorithm includes the following steps: Extract the effective volume of the drug solution in the current system and the estimated concentration value, and calculate the static compensation volume based on the law of conservation of mass; Extract the actual fluid volume replenished in the previous intermittent compensation action and the actual change in concentration it caused. Combine this with the target concentration change in the current case to calculate the estimated compensation volume. The static compensation volume and the estimated compensation volume are weighted and summed to obtain the optimal single-time replenishment volume.

4. The method for controlling the proportion of chemical solutions according to claim 3, characterized in that, In the weighted summation, the weighting coefficient α of the static compensation volume and the weighting coefficient β of the estimated compensation volume satisfy α+β=1, and the weighting coefficients α and β are dynamically adjusted according to the system operating status or historical estimated error rate.

5. The method for controlling the proportion of a chemical drug solution according to claim 4, characterized in that, The dynamic adjustment based on system operating status or historical prediction error rate includes: Record the estimated compensation volume and actual concentration change for each intermittent compensation action, and calculate the estimation error rate. When the estimated error rate exceeds the preset upper limit, increase α and decrease β; When the estimated error rate is lower than the preset lower limit, decrease α and increase β.

6. The method for controlling the proportion of chemical solutions according to claim 1, characterized in that, The proportioning system includes: A mixing tank is used to hold and store liquid medicine. A circulation pipeline connects the outlet and return port of the mixing tank to form a closed-loop circulation circuit. A circulation pump and a concentration meter are installed on the circulation pipeline. A first liquid inlet pipeline is connected to a first liquid source and the mixing tank. A first valve and a first flow meter are installed on the first liquid inlet pipeline. The second liquid inlet pipeline connects the second liquid source to the mixing tank, and the second liquid inlet pipeline is equipped with a second valve and a second flow meter; The control module is electrically connected to the concentration meter, the first valve, the second valve, the first flow meter, and the second flow meter, respectively, and is used to execute steps S1 to S4.

7. The method for controlling the proportion of a chemical drug solution according to claim 6, characterized in that, Select the corresponding inlet pipeline and execute the single optimal replenishment volume through closed-loop flow control, including: When the estimated concentration is lower than the target concentration, the first inlet pipeline is selected and a high-concentration first drug solution is added. When the estimated concentration is higher than the target concentration, select the second inlet pipeline and add a low-concentration second drug solution; During the replenishment process, the instantaneous flow rate is fed back in real time through the flow meter, and the opening degree of the corresponding valve is adjusted by the closed-loop control algorithm so that the actual flow rate tracks the target flow rate.

8. The method for controlling the proportion of a chemical drug solution according to claim 6, characterized in that, The first valve is a high-precision miniature needle valve with a flow control resolution of not less than 0.1 mL / min, used to achieve precise replenishment of trace amounts of drug solution; the second valve is a proportional valve, used to achieve stable replenishment of large flow diluent.

9. The method for controlling the proportion of a chemical drug solution according to claim 7, characterized in that, The closed-loop control algorithm is a PID control algorithm, and different PID parameter groups are configured for the response characteristics of the first valve and the second valve, respectively.

10. The method for controlling the proportion of a chemical drug solution according to claim 1, characterized in that, The duration of the forced waiting time is dynamically determined based on the ratio of the single optimal replenishment volume to the total volume of the mixing tank.

11. The method for controlling the proportion of a chemical drug solution according to claim 10, characterized in that, The dynamic determination method for the mandatory waiting time is as follows: ,in, To enforce waiting time, The optimal fluid replacement volume for a single treatment is [value missing]. The total volume of the mixing tank. The base waiting time is K, where K is the proportionality coefficient.

12. The method for controlling the proportion of chemical solutions according to claim 1, characterized in that, The dead zone threshold mentioned in step S2 is greater than the tolerance range of the target concentration mentioned in step S4, in order to avoid frequently triggering intermittent compensation actions near the target value.

13. The method for controlling the proportion of a chemical drug solution according to claim 6, characterized in that, The concentration meter is a conductivity type concentration meter, and the control module stores a three-dimensional calibration relationship table of drug concentration-conductivity-temperature. The mixing system also includes a temperature sensor for real-time detection of the liquid temperature. The control module performs temperature compensation on the concentration measurement value based on the feedback from the temperature sensor and the three-dimensional calibration relationship table, and then performs real-time filtering processing in step S1.

14. The method for controlling the proportion of a chemical drug solution according to claim 6, characterized in that, The control module is also used to perform the following liquid level maintenance and replenishment process: When the liquid level in the mixing tank is detected to be lower than the preset low liquid level threshold, a liquid level replenishment action is triggered. Based on the current liquid level, the volume of the liquid, the current estimated concentration, and the target concentration, the type of liquid to be replenished and the total amount of liquid to be replenished are calculated based on the law of conservation of mass. Control the valve on the corresponding liquid inlet pipeline to open and execute the total liquid replenishment; The liquid level replenishment action and the intermittent compensation action in step S2 are triggered independently. Liquid level replenishment is responsible for maintaining the liquid level within a preset range, while intermittent compensation is responsible for precise concentration adjustment. When the liquid level replenishment action and the intermittent compensation action are triggered simultaneously, the control module prioritizes executing the action with the larger concentration deviation, or combines the replenishment amounts of both actions in one replenishment action.

15. The method for controlling the proportion of a chemical drug solution according to claim 1, characterized in that, The iteration termination condition until the concentration estimate stabilizes within the tolerance range of the target concentration is met, and the intermittent compensation action stops when any of the following conditions are met: Condition 1: The estimated concentration value falls within the tolerance range of the target concentration, i.e. ,in, For tolerance range, This is a concentration estimate. The target concentration; Condition 2: After N consecutive intermittent compensation actions, the actual change in concentration caused by two adjacent actions is less than the preset change threshold, indicating that the system has become stable. Wherein, N is an integer greater than or equal to 3, and the tolerance range is less than the dead zone threshold in step S2.

16. The method for controlling the proportion of a chemical drug solution according to claim 3, characterized in that, It also includes an adaptive learning unit, which records the amount of fluid replenishment, the estimated concentration before replenishment, and the estimated concentration after replenishment for each intermittent compensation action, and continuously optimizes the prediction accuracy of the estimated compensation volume through machine learning methods.

Images