Method for automatic adjustment of reaction pressure for deep dearsenication of semiconductor-grade electronic hydrofluoric acid

Abstract

The application relates to the technical field of non-electric variable adjustment, and discloses a reaction pressure automatic adjustment method for deep dearsenification of semiconductor-grade electronic hydrogen fluoride acid, which comprises the following steps: obtaining a dearsenification agent dosing flow signal and a tank pressure feedback signal with a signal delay constant; extracting historical state parameters in a time sequence data buffer based on the signal delay constant; calculating an asynchronous deviation of the pressure feedback signal and the historical state parameters, and determining a pressure adjustment instruction through an adjustment gain model; and driving a pressure relief adjustment mechanism to produce displacement pre-adjustment before a pressure mutation is sensed in a signal feedback loop by using the pressure adjustment instruction. Through construction of an asynchronous control closed loop based on a physical delay constant, the application eliminates system parasitic oscillation caused by chasing of lagging signals in the adjustment action, and ensures microcosmic dynamic balance of an internal pressure field in a dearsenification reaction process under the premise of maintaining the strength of physical isolation protection.

Description

Technical Field

[0001] This invention belongs to the field of non-electric variable regulation technology, and particularly relates to an automatic regulation method for reaction pressure of deep arsenic removal from semiconductor-grade electronic hydrofluoric acid. Background Technology

[0002] Currently, in the deep arsenic removal process of semiconductor-grade electronic hydrofluoric acid, the reaction system typically involves adding an oxidizing arsenic removal agent within a closed container, accompanied by the generation of trace gases. Maintaining stable pressure within the vessel is essential for ensuring the direction of the arsenic removal reaction and the equilibrium of the microscopic interface. Existing technologies often employ pressure sensors to acquire signals and utilize controllers for closed-loop regulation. Due to the highly corrosive nature of hydrofluoric acid, the sensing end of the pressure sensor is usually covered with a thickened polytetrafluoroethylene (PTFE) membrane. This membrane exhibits viscoelastic physical properties, generating a physical filtering effect during signal transmission. This results in a time lag when the pressure change signal reaches the sensor's sensitive components, causing the pressure feedback parameters acquired by the system to deviate in the time dimension. The compatibility of the control scheme with physical laws also restricts the regulation accuracy. For example, publicly available... Chinese invention patent application CN120459897A discloses a pressure stabilization structure for the reactor environment in anhydrous hydrogen fluoride arsenic removal process. This structure achieves physical compensation of environmental pressure by setting rubber septa and using a lower support to change the volume of the reaction chamber. However, this approach has limitations in semiconductor-grade ultra-high purity environments: the adjustment logic relies on the instantaneous nature of sensor feedback; physical isolation barriers in extreme corrosive environments induce signal delays; if mechanical compensation is only performed within a synchronous feedback framework, the adjustment action lags behind the actual pressure evolution within the reactor, leading to a phase mismatch between the control command and the pressure peak, triggering parasitic oscillations in the system, shortening the lifespan of the actuator due to adjustment loop jitter, causing local disturbances in the reaction liquid, and resulting in the re-suspension of precipitated impurity particles that penetrate the filter medium. This ultimately fails to meet the microscopic dynamic pressure balance requirements of semiconductor processes.

[0003] When the control logic determines the control output based on this feedback signal with phase deviation, the control action becomes mismatched with the actual physical evolution process inside the vessel. If the control system uses the feedback parameter as the actual state at the current moment for deviation calculation, the control command lags behind the actual pressure peak, inducing oscillation in the control loop, causing the pressure relief valve to reciprocate, which in turn causes local disturbance in the reaction liquid, resulting in the resuspension of precipitated impurity particles and their penetration through the filter medium. To solve this problem, the industry usually tries to select a more sensitive sensor or set a smaller control dead zone. However, the physical properties of the isolation membrane limit the improvement of signal transmission speed, and an excessively small dead zone setting can easily amplify the measurement noise at the sensor end, leading to a decrease in system reliability.

[0004] Therefore, the technical problem to be solved by this invention is to construct a regulation mechanism that can align control parameters with physical processes in time to address the physical sensing hysteresis caused by the isolation medium in a highly corrosive environment, thereby eliminating pressure pulsation and impurity suspension caused by control disorder. Summary of the Invention

[0005] This invention provides an automatic pressure regulation method for deep arsenic removal from semiconductor-grade electronic hydrofluoric acid, comprising the following steps: Step S101: Obtain the real-time arsenic removal agent dosing flow signal fed back by the arsenic removal agent dosing drive unit; Step S102: Obtain the real-time pressure signal inside the vessel fed back by the sensing branch. The inner wall of the sensing branch is lined with a polytetrafluoroethylene isolation membrane. Due to the physical isolation characteristics of the polytetrafluoroethylene isolation membrane, the real-time pressure signal has a signal delay constant relative to the actual state of the physical pressure field inside the vessel. Step S103: Based on the signal delay constant, retrieve and extract the historical state parameters that are in the same physical time slice as the real-time pressure signal in the time series data buffer by timestamp. The historical state parameters characterize the expected value of the pressure evolution of the pressure field inside the vessel at historical moments. Step S104: Calculate the algebraic difference between the real-time pressure signal and the historical state parameters to obtain the asynchronous deviation, and input the asynchronous deviation into the adjustment gain model to generate a pressure adjustment command. The pressure adjustment command is used to eliminate the phase interference caused by the signal delay constant to the control loop. Step S105: The pressure regulation command is sent to the pressure relief regulation mechanism, which drives the pressure relief regulation mechanism to generate a preset opening displacement before the physical pressure field inside the vessel fluctuates. This completes the preset opening of the pressure relief regulation mechanism before the real-time pressure signal is fed back to the control loop, thereby achieving stable suppression of micro-fluctuations in the pressure field inside the vessel.

[0006] Preferably, in step S104, the adjustment gain model establishes a nonlinear correlation between the feedforward component of the real-time injection flow signal and the attenuation coefficient based on the cumulative injection amount, thereby mapping the evolution trend of gas generation rate caused by substrate consumption in real time. The adjustment gain model dynamically adjusts the proportional regulation intensity according to the evolution trend of gas generation rate, so that the pressure regulation command maintains adaptive convergence of gain throughout the entire reaction cycle, preventing excessive pressure relief caused by the decrease in gas generation due to the decrease in substrate concentration in the later stages of the reaction.

[0007] Preferably, step S103 specifically includes the following sub-steps: step S1031, establishing a timestamp cache sequence for recording the pressure field state inside the vessel; step S1032, based on the signal delay constant determined by the physical thickness of the polytetrafluoroethylene isolation membrane, performing reverse addressing in the timestamp cache sequence to locate the physical real node that matches the current signal acquisition time, and extracting the historical state parameters under that node.

[0008] Preferably, step S105 specifically includes the following sub-steps: step S1051, obtaining the current valve position feedback signal of the pressure relief regulating mechanism, and calculating the target opening displacement according to the pressure regulating command; step S1052, driving the pressure relief regulating mechanism to move towards the target opening displacement position within the sampling period defined by the signal delay constant.

[0009] Preferably, the method further includes the following steps: using historical operating data, extracting the pressure fluctuation residual within the signal delay constant, and online correcting the proportional coefficient of the adjustment gain model based on the pressure fluctuation residual to compensate for the signal response characteristic drift caused by the elastic deformation of the polytetrafluoroethylene isolation membrane.

[0010] Preferably, in step S101, the acquisition period of the real-time added flow signal is less than the signal delay constant, and the real-time added flow signal is processed by a first-order low-pass filter to filter out the transient pulse noise generated by the added pump, thereby ensuring that the data input to the adjustment gain model has physical stability.

[0011] Preferably, in step S102, the real-time pressure signal is acquired by a piezoresistive sensor located at the end of the sensing branch and transmitted to the analog-to-digital conversion module via a shielded cable, and mapped into a digital physical quantity for asynchronous deviation calculation.

[0012] Preferably, the opening preset displacement generated by the pressure relief regulating mechanism forms a complementary dynamic response topology with the asynchronous deviation. By superimposing the advance amount of the regulating algorithm with the preset amount of the mechanical opening, closed-loop suppression of control loop oscillation caused by physical isolation protection is achieved.

[0013] Preferably, the method constructs an asynchronous control closed loop based on the physical delay constant, which maintains the physical barrier effect of the polytetrafluoroethylene isolation membrane while eliminating the parasitic oscillations of the system caused by the adjustment action chasing the lag feedback signal, thus ensuring the microscopic dynamic balance of the pressure field inside the reactor during the arsenic removal reaction.

[0014] Compared with existing technologies, the automatic pressure adjustment method for deep arsenic removal from semiconductor-grade electronic hydrofluoric acid of the present invention has the following advantages: 1. In the deep arsenic removal of semiconductor-grade electronic hydrofluoric acid, a data alignment architecture with non-homogeneous temporal states is established by introducing a state caching mechanism with timestamps into the regulation logic. This effectively eliminates the phase interference of the control loop caused by the physical sensing lag generated by the thickened polytetrafluoroethylene isolation film. The system ensures that the deviation calculation is based on an absolutely aligned physical causal chain by backtracking and extracting historical expected parameters that are in the same physical time slice as the current feedback signal. This avoids the phase distortion caused by the regulation command chasing the lag feedback and improves the underlying stability of the regulation loop in the face of sensing lag under extreme corrosive conditions.

[0015] 2. By deeply coupling the feedforward capture of real-time dosing flow rate with the kinetic mapping matrix based on cumulative dosing, the system can map the nonlinear evolution of gas generation rate caused by the decrease in substrate concentration in real time. This synergistic gain between the feedforward logic and the inherent decay trend of reaction kinetics enables the adjustment command to maintain adaptive convergence of gain throughout the entire reaction cycle, preventing the risk of excessive pressure relief due to excessive feedforward gain in the middle and late stages of the reaction, and achieving stable suppression of pressure field fluctuations inside the reactor under complex and variable reaction conditions.

[0016] 3. The established asynchronous deviation backtracking logic and the pre-displacement deflection mechanism of the actuator form a complementary dynamic response topology. Utilizing the objective physical law that the injection action must precede the pressure fluctuation, the basic displacement preset of the pressure relief mechanism is completed before the sensing components detect the effective signal change. Combined with model drift calibration based on historical data from the same source, this scheme solves the inherent contradiction between physical isolation layer protection and real-time signal acquisition from the perspective of temporal topology reconstruction within the control domain. It eliminates parasitic oscillations in the control system caused by signal feedback phase lag and ensures the microscopic dynamic balance of non-electrical physical quantities during the arsenic removal process. Attached Figure Description

[0017] Figure 1 This is an asynchronous control flowchart for automatic pressure adjustment in semiconductor-level arsenic removal reaction according to the present invention; Figure 2 This is the timing sequence interaction diagram for calibrating the signal delay constant of the pressure step excitation of this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0019] It should be noted that all directional and positional terms used in this invention, such as: up, down, left, right, front, back, vertical, horizontal, inner, outer, top, bottom, transverse, longitudinal, center, etc., are only used to explain the relative positional relationship and connection between components in a specific state (as shown in the accompanying drawings). They are only for the convenience of describing this invention and do not require that this invention be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention. In addition, the descriptions of "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated.

[0020] In the description of this invention, unless otherwise explicitly specified and limited, the terms installation, connection, and linking should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections; they can refer to direct connections or indirect connections through an intermediate medium; they can refer to the internal connection of two components. For those skilled in the art, the specific meaning of the above terms in this invention can be understood according to the specific circumstances.

[0021] In the description of this specification, references to the terms "an embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example, and the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] An automatic pressure regulation method for deep arsenic removal from semiconductor-grade electronic hydrofluoric acid includes the following steps: Step S101: Obtain the real-time arsenic removal agent dosing flow signal fed back by the arsenic removal agent dosing drive unit; Step S102: Obtain the real-time pressure signal inside the vessel fed back by the sensing branch. The inner wall of the sensing branch is lined with a polytetrafluoroethylene isolation membrane. Due to the physical isolation characteristics of the polytetrafluoroethylene isolation membrane, the real-time pressure signal has a signal delay constant relative to the actual state of the physical pressure field inside the vessel. Step S103: Based on the signal delay constant, retrieve and extract the historical state parameters that are in the same physical time slice as the real-time pressure signal in the time series data buffer by timestamp. The historical state parameters characterize the expected value of the pressure evolution of the pressure field inside the vessel at historical moments. Step S104: Calculate the algebraic difference between the real-time pressure signal and the historical state parameters to obtain the asynchronous deviation, and input the asynchronous deviation into the adjustment gain model to generate a pressure adjustment command. The pressure adjustment command is used to eliminate the phase interference caused by the signal delay constant to the control loop. Step S105: The pressure regulation command is sent to the pressure relief regulation mechanism, which drives the pressure relief regulation mechanism to generate a preset opening displacement before the physical pressure field inside the vessel fluctuates. This completes the preset opening of the pressure relief regulation mechanism before the real-time pressure signal is fed back to the control loop, thereby achieving stable suppression of micro-fluctuations in the pressure field inside the vessel.

[0023] Preferably, in step S104, the adjustment gain model establishes a nonlinear correlation between the feedforward component of the real-time injection flow signal and the attenuation coefficient based on the cumulative injection amount, thereby mapping the evolution trend of gas generation rate caused by substrate consumption in real time. The adjustment gain model dynamically adjusts the proportional regulation intensity according to the evolution trend of gas generation rate, so that the pressure regulation command maintains adaptive convergence of gain throughout the entire reaction cycle, preventing excessive pressure relief caused by the decrease in gas generation due to the decrease in substrate concentration in the later stages of the reaction.

[0024] Preferably, step S103 specifically includes the following sub-steps: step S1031, establishing a timestamp cache sequence for recording the pressure field state inside the vessel; step S1032, based on the signal delay constant determined by the physical thickness of the polytetrafluoroethylene isolation membrane, performing reverse addressing in the timestamp cache sequence to locate the physical real node that matches the current signal acquisition time, and extracting the historical state parameters under that node.

[0025] Preferably, step S105 specifically includes the following sub-steps: step S1051, obtaining the current valve position feedback signal of the pressure relief regulating mechanism, and calculating the target opening displacement according to the pressure regulating command; step S1052, driving the pressure relief regulating mechanism to move towards the target opening displacement position within the sampling period defined by the signal delay constant.

[0026] Preferably, the method further includes the following steps: using historical operating data, extracting the pressure fluctuation residual within the signal delay constant, and online correcting the proportional coefficient of the adjustment gain model based on the pressure fluctuation residual to compensate for the signal response characteristic drift caused by the elastic deformation of the polytetrafluoroethylene isolation membrane.

[0027] Preferably, in step S101, the acquisition period of the real-time added flow signal is less than the signal delay constant, and the real-time added flow signal is processed by a first-order low-pass filter to filter out the transient pulse noise generated by the added pump, thereby ensuring that the data input to the adjustment gain model has physical stability.

[0028] Preferably, in step S102, the real-time pressure signal is acquired by a piezoresistive sensor located at the end of the sensing branch and transmitted to the analog-to-digital conversion module via a shielded cable, and mapped into a digital physical quantity for asynchronous deviation calculation.

[0029] Preferably, in step S104, the pressure regulation command is calculated using the following quantization rule: ,in, The control output of the pressure regulation command; This is the preset feedback adjustment coefficient; The real-time pressure signal obtained in step S102; The data extracted in step S103 after time offset Processed historical state parameters; The signal delay constant is determined by the polytetrafluoroethylene (PTFE) separator. To be based on real-time flow signal The calculated feedforward compensation component.

[0030] Preferably, the opening preset displacement generated by the pressure relief regulating mechanism forms a complementary dynamic response topology with the asynchronous deviation. By superimposing the advance amount of the regulating algorithm with the preset amount of the mechanical opening, closed-loop suppression of control loop oscillation caused by physical isolation protection is achieved.

[0031] Preferably, the method constructs an asynchronous control closed loop based on the physical delay constant, which maintains the physical barrier effect of the polytetrafluoroethylene isolation membrane while eliminating the parasitic oscillations of the system caused by the adjustment action chasing the lag feedback signal, thus ensuring the microscopic dynamic balance of the pressure field inside the reactor during the arsenic removal reaction.

[0032] Example 1: An automatic pressure regulation method for deep arsenic removal from semiconductor-grade electronic hydrofluoric acid is applied to a 10nm process ultrapure electronic hydrofluoric acid large-scale deep arsenic removal purification production line. In the sealed reactor of this production line, the continuous addition of the arsenic removal agent induces a chemical phase change in the reaction substrate, accompanied by the release of trace amounts of gas. The dynamic equilibrium of the transient pressure field within the reactor constitutes the physical boundary conditions that maintain the unidirectional arsenic removal reaction. A piezoresistive sensor, located at the end of the sensing branch, has a polytetrafluoroethylene (PTFE) membrane of a predetermined thickness as its inner wall. The viscoelastic properties of this PTFE membrane constitute a mechanical low-pass filter, allowing changes in physical pressure within the reactor to be transmitted to the sensing components. The physical filtering effect and time lag cause the real-time pressure signal output by the analog-to-digital converter module to have a signal delay constant relative to the physical pressure field state inside the reactor. Conventional closed-loop control logic collects the real-time pressure signal with phase deviation and subtracts it from the set reference value to drive the downstream actuator. The control unit sets the mechanical adjustment opening at the current moment according to the physical state of the reaction system's historical time slice. The timing misalignment between this sensor feedback and physical reality causes the pressure relief adjustment mechanism to generate reciprocating vibration and parasitic oscillations in the system, inducing local micro-boiling disturbances in the reaction liquid, causing the arsenic complex impurity particles that have settled at the bottom of the reactor to be resuspended and penetrate the subsequent filter medium.

[0033] The controller acquires the real-time arsenic removal agent flow rate signal from the dosing pump via the arsenic removal agent dosing drive unit. After passing through a first-order low-pass filter, this signal is used as the feedforward disturbance input to the adjustment gain model. Based on the mapping matrix between the feedforward disturbance input and a preset value representing the nonlinear decay of the cumulative dosing amount, the transient expected pressure increment for the current control cycle is calculated. The processor, using the ideal gas law and the stoichiometric ratio of the arsenic removal reaction, multiplies the acquired real-time arsenic removal agent flow rate signal by the molar gas generation conversion constant of the arsenic removal reaction, converting it into the standard gas generation volume per unit control cycle. This volume is then divided by the current gas phase residual volume parameter of the reactor to obtain the theoretical pressure rise bias caused by the purely chemical reaction within this physical control cycle. The processor calculates and acquires the transient expected pressure increment, carrying the current system timestamp, and sequentially writes it into a preset timestamp cache sequence. Within the same physical control cycle, the signal acquisition branch acquires the real-time pressure signal transmitted from the vessel via the PTFE isolation membrane. To eliminate feedback phase interference caused by the physical isolation medium, the controller initiates reverse addressing logic in the timestamp cache sequence based on the signal delay constant calibrated by the physical thickness and elastic deformation of the PTFE isolation membrane. This locates and extracts historical state parameters that are physically aligned with the currently acquired real-time pressure signal. The processor calculates the dynamic asynchronous deviation between the real-time pressure signal and the corresponding historical state parameters based on the extracted physical alignment parameters, according to the formula... Generate pressure regulation commands, wherein, This is the control output of the pressure regulation command. The preset feedback adjustment coefficient, To obtain real-time pressure signals, Extracted elapsed time offset Processed historical state parameters The signal delay constant is determined by the polytetrafluoroethylene (PTFE) separator. To be based on real-time flow signal Under this solution framework, although the real-time pressure signal obtained has experienced the physical delay of the isolation membrane, the expected parameters extracted by the timestamp backtracking are also placed on the same historical hysteresis time coordinate axis, and the two are forcibly aligned in physical temporality. The algebraic difference between the two essentially strips away the known theoretical pressure fluctuation benchmark caused by the normal feedforward injection reaction, and extracts the unpredictable abnormal pressure residual caused by environmental temperature disturbance, valve mechanical nonlinear wear, or local micro-boiling of reaction byproducts. This unknown disturbance residual serves as the only real physical error source to drive the convergence of the feedback regulation control loop, thereby mathematically completing an effective and stable error closed loop under the isolation delay condition. This asynchronous solution logic applies phase superposition to the feedforward component of the real-time injection flow signal and the model drift calibration based on the same source physical historical data, uses the objective fluid dynamics law that the injection action is earlier than the pressure generation to provide control advance, and provides closed-loop error compensation based on the dynamic asynchronous deviation stripped of the physical sensing hysteresis time deviation, thus constructing a complementary dynamic response network between the front and back ends.

[0034] The controller will have a feedforward compensation component. Pressure regulation command with dynamic correction data The electrical circuit sends the signal to the pressure relief regulating mechanism, which moves towards the target opening displacement calculated based on the pressure regulation command before the piezoresistive sensor detects a sudden pressure change and feeds it back to the control circuit. This advance positioning capability is based on the difference in the propagation speed of the physical causal chain perceived by the control system. Since the mechanical injection action of the arsenic removal agent dosing pump must precede the chemical reaction of the substrate in the reactor and the subsequent gas phase expansion, the system acquires the dosing flow signal in real time through the arsenic removal agent dosing drive unit. This millisecond-level electrical signal transmission and calculation speed is offset against the second-level fluid dynamic diffusion speed of the reaction phase change and bubble escape, thereby achieving the desired effect in the actual reactor. Before the overall pressure peak is formed, the feedforward compensation component is used to drive the valve to open ahead of the sensor lag. When the physical pressure field inside the sealed reactor expands due to the gas generation caused by the consumption of the arsenic removal reaction substrate, the valve core of the pressure relief regulating mechanism has already deflected to the basic relief opening corresponding to the expected pressure increment. This advance mechanical opening displacement releases the local gas phase increment generated during the reaction kinetic evolution process in advance, reduces the absolute pressure overshoot in the transient response process inside the reactor, maintains the dynamic balance of the pressure field inside the reactor throughout the reaction cycle, suppresses the local disturbance phenomenon of the liquid inside the reactor excited by the feedback oscillation of the control loop, and keeps the solid precipitate generated by the arsenic removal reaction in a stable deposition state at the bottom of the highly corrosive medium system.

[0035] Example 2: This example constructs a 500L continuous flow arsenic removal reaction test platform, including a sealed reactor with a temperature control accuracy of ±0.5℃ and a piezoresistive sensor with a range of 0 to 0.5MPa and a resolution of 0.1kPa. The inner wall of the piezoresistive sensor's sensing branch is lined with a polytetrafluoroethylene (PTFE) isolation membrane. Gaussian white noise with a signal-to-noise ratio of 20dB is superimposed on the input circuit of the arsenic removal agent dosing drive unit. Simultaneously, 50Hz harmonic interference from fluid mechanical vibration is introduced. The feedback adjustment coefficient K is set to balance the pressure regulation response speed with the mechanical fatigue wear of the pressure relief mechanism. The decision rule is based on the real-time dosing flow rate. The time derivative of the signal establishes a nonlinear mapping. When the absolute value of the time derivative of the real-time applied flow signal is in the range greater than the preset flow mutation threshold, the feedback adjustment coefficient K tends to the lower limit of the value range, suppressing the overshoot oscillation of the valve core of the pressure relief regulating mechanism. The value range of the feedback adjustment coefficient K is set to 0.8 to 1.25. When the absolute value of the time derivative exceeds the preset flow mutation threshold, the linear attenuation compensation circuit is activated. The ratio of the overshoot of the current sampling period derivative to the set derivative limit value is used as the attenuation rate. The feedback adjustment coefficient K is gradually reduced according to the sampling clock step size until it is locked at the lower limit value of 0.8. According to this rule, the basic value of the feedback adjustment coefficient K is set to 1.25.

[0036] The arsenic removal agent dosing procedure was initiated. The acquisition module obtained the real-time pressure signal superimposed with the aforementioned power frequency interference and noise. The initial readings fluctuated randomly within the range of 105.3 kPa to 112.8 kPa. Three parallel independent test groups were established. Control group 1 adopted a traditional synchronous feedback control architecture. Control group 2 disconnected the asynchronous historical state comparison logic but retained the feedforward compensation link. The test group deployed complete asynchronous solution control logic. At the 10.0s mark, the real-time pressure signal inside the vessel obtained by the acquisition branch was recorded. The instantaneous value was 115.2 kPa. The signal delay constant τ, calibrated by the PTFE isolation membrane, was 1.6 s. The controller initiated reverse addressing in the timestamp buffer sequence and extracted the historical state parameter with a physical timestamp of 8.4 s. Its value is 113.8 kPa. Based on the above physical alignment parameters, the processor calculates the dynamic asynchronous deviation after removing the phase deviation to be 1.4 kPa. At the same time, the real-time arsenic removal agent dosing flow signal generates a feedforward compensation component with a corresponding equivalent of 0.8 kPa. The controller is based on the formula The above parameters are combined to generate a pressure regulation command, which drives the pressure relief regulating mechanism to output an initial valve position deflection action.

[0037] By varying the thickness of the polytetrafluoroethylene (PTFE) separator, the signal delay constant can be constructed. Five gradient test sequences of 0.5s, 0.8s, 1.6s, 2.5s, and 3.5s were used. The test data showed gradient differentiation when the signal delay constant... At 0.5s, the absolute pressure overshoot in the reactor of control group 1 was 5.2kPa, and the absolute pressure overshoot in the experimental group was 1.1kPa. When the signal delay constant... When the set lower limit of 0.8s was reached, the absolute pressure overshoot of the test group remained at 1.2 kPa. When the signal delay constant τ extended to the median of 1.6s, the absolute pressure overshoot of the control group surged to 14.7 kPa due to parasitic oscillations in the control system caused by sensor misalignment, triggering the mechanical safety relief valve. The absolute pressure overshoot of the test group remained in the steady-state range of 1.3 kPa. When the set upper limit of 2.5s was reached, the absolute pressure overshoot of the test group was 1.5kPa. When the signal delay constant τ increased to 3.5s, which exceeded the range, the excessive thickening of the PTFE separator caused the low-pass filtering effect to cross the nonlinear physical inflection point. The high-frequency components of the historical state parameters underwent irreversible attenuation, and the drift rate of the asynchronous solution increased to 15.6%. The absolute pressure overshoot of the test group subsequently climbed to 6.8kPa. The gradient data established that the optimal working window for the PTFE separator thickness to maintain the system control accuracy corresponds to the signal delay constant. Between 0.8s and 2.5s, the data from the aforementioned continuous flow arsenic removal reaction test platform show that the application of a non-electric variable control architecture with a timestamp offset addressing mechanism suppresses the negative impact of measurement loop hardware delay on the accuracy of control commands. The control output strategy, which combines feedforward compensation components with kinetic asynchronous deviation, limits the overshoot of transient pressure fluctuations within the vessel and maintains the steady-state deposition of sediment at the bottom of the reaction vessel. This architecture suppresses pressure regulation oscillations caused by the hysteresis of the isolation medium.

[0038] Example 3: In the continuous operation of a large-scale deep arsenic removal and purification production line for semiconductor-grade electronic hydrofluoric acid, as the arsenic removal agent is continuously added, the concentration of the reaction substrate in the sealed reactor gradually decreases. The gas generation rate caused by the arsenic removal agent per unit volume exhibits a non-linear decay trend. The feedforward compensation logic with fixed parameters cannot match this dynamic decay characteristic, resulting in additional valve position deflection displacement of the pressure relief regulating mechanism in the later stages of the reaction. The controller is configured with dynamic calibration and update logic for the image matrix. During the offline benchmark calibration stage, the arsenic removal reaction cycle under standard operating conditions is selected, and the discrete test gradient vector of the arsenic removal agent addition flow rate is set. At different cumulative addition nodes, the corresponding test gradient is injected. The processor measures the pressure response step peak value in the sealed reactor for the arsenic removal agent. Based on the ratio of the pressure response step peak value to the corresponding test gradient at each cumulative dosage node, the processor generates a reference gain coefficient. Based on the reference gain coefficient, it constructs the initial state mapping matrix. During the online adaptive adjustment phase, the integral calculation unit in the controller integrates the real-time arsenic removal agent dosing flow signal along the time dimension and outputs the current cumulative dosage parameter. The processor inputs the cumulative dosage parameter into a preset first-order decay function to calculate the current dynamic decay factor. The processor calculates the product of the initial state mapping matrix and the dynamic decay factor to generate an updated real-time mapping matrix.

[0039] Based on the first-order kinetics of substrate consumption, the first-order decay function is set as an exponential formula including the reaction rate constant determined by offline titration and the gas compatibility volume constant of the reactor. The cumulative dosage is extracted to calculate the gas-phase generation conversion rate, and the dimensionless decay coefficient is output. The product of the initial state mapping matrix and the dynamic decay factor is calculated to generate an updated real-time mapping matrix. The controller then transmits the real-time arsenic removal agent dosing flow signal. The input is given to the real-time mapping matrix, and the feedforward compensation component is calculated using a linear interpolation algorithm. The controller integrates feedforward compensation components. Based on the feedback component generated by the asynchronous deviation of the dynamics, a pressure regulation command is generated and output to the pressure relief regulation mechanism. By applying the dynamic calibration and update logic of the mapping matrix, an adaptive convergence path of the feedforward compensation component as the reaction substrate is consumed is constructed. This control architecture offsets the expected incremental deviation caused by the decrease in gas generation rate in the middle and late stages of the reaction, suppresses the ineffective mechanical action of the pressure relief regulation mechanism, and maintains the dynamic balance of the pressure field inside the vessel throughout the entire cycle of the arsenic removal reaction.

[0040] Example 4: When the system faces the initial deployment of the purification production line and the replacement of the PTFE separator membrane, the test module connects the sensing branch of the sealed reactor to a standard step pressure generator, applying an amplitude of [value missing] to the piezoresistive sensor covered with the PTFE separator membrane. The pressure step excitation; the timing unit inside the controller records the initial timestamp of the applied pressure step excitation. Simultaneously, the analog-to-digital conversion module acquires the real-time pressure signal output by the piezoresistive sensor at a sampling rate 10 times higher than the system's operating frequency band; the controller monitors the amplitude trajectory of this real-time pressure signal, and when the real-time pressure signal reaches 63.2% of the pressure step excitation amplitude ΔP, it extracts the corresponding response timestamp. The processor follows the formula. Calculate the signal delay constant of the current physical isolation membrane under assembly stress. ,in, Let be the signal delay constant. In response to timestamps, This is the initial timestamp; the controller will use this signal delay constant. Write the reverse addressing configuration register of the timestamp cache sequence to set the base time span for timestamp offset addressing.

[0041] When the system faces long-term operation of the arsenic removal agent dosing drive unit and the pressure relief regulating mechanism, the test module injects discrete dosing flow time derivative test vectors into the arsenic removal agent dosing drive unit. The vibration meter records the peak-to-peak values ​​of the mechanical oscillations of the valve stem of the pressure relief regulating mechanism under the drive of each test vector. A low-pass filtering algorithm with a cutoff frequency of half the valve stem's natural resonant frequency is used to preprocess and record the mechanical oscillation waveforms. The amplitude deviation signal caused by the sudden change in the regulating command is separated. The system has a pre-programmed SN fatigue life curve database characterizing the mechanical properties of the pressure relief valve body material of this model. The processor inputs the peak-to-peak values ​​of the mechanical oscillations extracted by the vibration meter within a short period as the basic alternating stress amplitude into the database model, combined with the Miner linear fatigue damage accumulation theory formula. The equivalent structural stress level under tens of millions of cycles is calculated by extrapolation. Once the calculated equivalent structural stress level crosses the upper limit of the material elastic deformation given in the database, it is determined that the microscopic failure critical point corresponding to the fatigue yield limit of the valve metal has been indirectly reached. The processor extracts the critical value of the added flow derivative at the point where the peak value of the mechanical oscillation exceeds the critical point corresponding to the fatigue yield limit of the valve metal, multiplies it by a safety margin factor of 0.8, and calculates the target flow mutation threshold. The controller writes the target flow mutation threshold into the nonlinear image criterion library of the adjustment gain model. When the absolute value of the time derivative of the real-time added flow signal crosses the target flow mutation threshold, the adjustment gain model calculates and outputs the feedback adjustment coefficient. By using control parameters that converge to the lower limit, the boundary of the underlying basic operating parameters of the non-electric variable regulation system is established.

[0042] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. 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 other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A method for automatically adjusting the reaction pressure of deep arsenic removal from semiconductor-grade electronic hydrofluoric acid, characterized in that, Includes the following steps: Step S101: Obtain the real-time arsenic removal agent dosing flow signal fed back by the arsenic removal agent dosing drive unit; Step S102: Obtain the real-time pressure signal inside the vessel fed back by the sensing branch. The inner wall of the sensing branch is lined with a polytetrafluoroethylene isolation membrane. Due to the physical isolation characteristics of the polytetrafluoroethylene isolation membrane, the real-time pressure signal has a signal delay constant relative to the actual state of the physical pressure field inside the vessel. Step S103: Based on the signal delay constant, retrieve and extract the historical state parameters that are in the same physical time slice as the real-time pressure signal in the time series data buffer by timestamp. The historical state parameters characterize the expected value of the pressure evolution of the pressure field inside the vessel at historical moments. Step S104: Calculate the algebraic difference between the real-time pressure signal and the historical state parameters to obtain the asynchronous deviation, and input the asynchronous deviation into the adjustment gain model to generate a pressure regulation command. The pressure regulation command is used to eliminate the phase interference caused by the signal delay constant to the control loop; the calculation is performed according to quantization rules. ,in, The control output of the pressure regulation command; This is the preset feedback adjustment coefficient; To acquire real-time pressure signals; Extracted elapsed time offset Processed historical state parameters; The signal delay constant is determined by the polytetrafluoroethylene (PTFE) separator. To be based on real-time flow signal The calculated feedforward compensation component; Step S105: The pressure regulation command is sent to the pressure relief regulation mechanism. Taking advantage of the objective physical law that the addition action must precede the pressure fluctuation, the pressure relief regulation mechanism is driven to generate a preset opening displacement before the physical pressure field inside the vessel fluctuates. Thus, the opening preset of the pressure relief regulation mechanism is completed before the real-time pressure signal is fed back to the control loop, thereby achieving stable suppression of micro-fluctuations in the pressure field inside the vessel.

2. The method for automatically adjusting the reaction pressure of semiconductor-grade electronic hydrofluoric acid deep arsenic removal according to claim 1, characterized in that, In step S104, the adjustment gain model establishes a nonlinear relationship between the feedforward component of the real-time injection flow signal and the attenuation coefficient based on the cumulative injection amount, thereby mapping the evolution trend of gas generation rate caused by substrate consumption in real time. The gain adjustment model dynamically adjusts the proportional regulation intensity according to the evolution trend of gas generation rate, so that the pressure regulation command maintains adaptive convergence of gain throughout the entire reaction cycle, preventing excessive pressure relief caused by the decrease in gas generation due to the decrease in substrate concentration in the middle and late stages of the reaction.

3. The method for automatically adjusting the reaction pressure of semiconductor-grade electronic hydrofluoric acid deep arsenic removal according to claim 1, characterized in that, Step S103 specifically includes the following sub-steps: Step S1031, establish a timestamp cache sequence for recording the pressure field state inside the vessel; Step S1032, based on the signal delay constant determined by the physical thickness of the polytetrafluoroethylene isolation membrane, perform reverse addressing in the timestamp cache sequence to locate the physical real node that matches the current signal acquisition time, and extract the historical state parameters under that node.

4. The method for automatically adjusting the reaction pressure of semiconductor-grade electronic hydrofluoric acid deep arsenic removal according to claim 1, characterized in that, Step S105 specifically includes the following sub-steps: Step S1051, obtain the current valve position feedback signal of the pressure relief regulating mechanism, and calculate the target opening displacement according to the pressure regulation command; Step S1052, within the sampling period defined by the signal delay constant, drive the pressure relief regulating mechanism to move towards the target opening displacement position.

5. The method for automatically adjusting the reaction pressure of semiconductor-grade electronic hydrofluoric acid deep arsenic removal according to claim 3, characterized in that, It also includes the following steps: By utilizing historical operating data, the pressure fluctuation residual within the signal delay constant is extracted, and the proportional coefficient of the adjustment gain model is corrected online based on the pressure fluctuation residual to compensate for the signal response characteristic drift caused by the elastic deformation of the polytetrafluoroethylene isolation membrane.

6. The method for automatically adjusting the reaction pressure of semiconductor-grade electronic hydrofluoric acid deep arsenic removal according to claim 1, characterized in that, In step S101, the acquisition period of the real-time added flow signal is less than the signal delay constant, and the real-time added flow signal is processed by a first-order low-pass filter to filter out the transient pulse noise generated by the added pump, thereby ensuring that the data input to the adjustment gain model has physical stability.

7. The method for automatically adjusting the reaction pressure of semiconductor-grade electronic hydrofluoric acid deep arsenic removal according to claim 1, characterized in that, In step S102, the real-time pressure signal is acquired by a piezoresistive sensor located at the end of the sensing branch and transmitted to the analog-to-digital conversion module via a shielded cable, and mapped into a digital physical quantity for asynchronous deviation calculation.

8. The method for automatically adjusting the reaction pressure of semiconductor-grade electronic hydrofluoric acid deep arsenic removal according to claim 1, characterized in that, The pressure relief regulating mechanism generates a preset opening displacement and an asynchronous deviation to form a complementary dynamic response topology. By superimposing the advance amount of the regulating algorithm with the preset mechanical opening amount, closed-loop suppression of control loop oscillations caused by physical isolation protection is achieved.

9. The method for automatically adjusting the reaction pressure of semiconductor-grade electronic hydrofluoric acid deep arsenic removal according to claim 1, characterized in that, The method constructs an asynchronous control closed loop based on the physical delay constant, which maintains the physical barrier effect of the polytetrafluoroethylene isolation membrane while eliminating the parasitic oscillations caused by the adjustment action chasing the lag feedback signal, thus ensuring the microscopic dynamic balance of the pressure field inside the reactor during the arsenic removal reaction.

Images