Method for adjusting reaction parameters for semiconductor grade electronic hydrofluoric acid production

Abstract

This invention relates to the field of non-electric variable regulation technology and discloses a method for regulating reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid. The method includes: acquiring temperature and mass flow rate data of the reaction system; extracting the transient phase difference between the temperature and flow rate differential signals using phase difference tracking logic to characterize the nonlinear physical resistance fluctuations of the fluid; dynamically scheduling the nonlinear coupling coefficient based on the temperature deviation to identify the system error state; correcting the reaction enthalpy change coefficient using the transient phase difference, determining the feedforward compensation control quantity and superimposing it with the feedback control quantity, and outputting control commands for the fluid regulation actuator. This invention achieves compensation for fluid resistance through high-frequency signal phase decoupling, and, combined with an error state-based adaptive switching mechanism, improves the anti-disturbance stability and steady-state convergence accuracy of the purification process.

Description

Technical Field

[0001] This invention relates to a method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid, belonging to the field of non-electrical variable adjustment technology. Background Technology

[0002] Currently, semiconductor-grade hydrofluoric acid is a key cleaning and etching medium in integrated circuit manufacturing. Its purity level directly determines the chip yield. In the oxidation and impurity removal stage of the production process, an oxidant is usually injected into the hydrofluoric acid system to cause trace impurities to undergo an exothermic reaction and generate high-boiling-point complexes. A distributed control system is used to regulate the circulation flow of the cooling medium to maintain the steady state of the reaction temperature field. As the purity requirements of the reagents in the process have entered the extremely small scale, the reaction temperature control accuracy needs to meet the requirements of millisecond-level response and ultra-low fluctuation range. This type of oxidation process has strong exothermic characteristics. The diffusion of heat from the reaction center to the temperature sensor has an intrinsic physical time delay. The existing regulation strategy is based on the static assumption that the heat transfer path parameters are constant. As the depth of the complexation reaction evolves, the microscopic viscosity inside the fluid undergoes nonlinear drift, resulting in dynamic fluctuations in the thermal conduction resistance. This pure hysteresis variation caused by the evolution of the medium properties causes the control action to lag behind the transient of heat excitation, resulting in a phase misalignment between the control command and the physical phenomenon.

[0003] To address the heat transfer hysteresis problem, increasing the sampling frequency or introducing feedforward compensation with a fixed operator are common improvement paths. However, high-frequency acquisition is prone to signal noise interference, inducing high-frequency mechanical oscillations in the actuator. Static feedforward models cannot track conduction hindrance caused by fluid property variations, resulting in systematic deviations during periods of drastic reaction fluctuations, ultimately inducing temperature rise overshoot. Specifically, in existing technologies, the response accuracy of the regulating mechanism is limited by the spatial layout of the sensor placement points, making it difficult to eliminate the intrinsic physical delay of the heat transfer path simply by increasing the hardware sampling frequency. In addition to the aforementioned hardware limitations, the software-level control algorithm also has shortcomings in handling complex fluid property evolution. For example, Chinese invention patent application with publication number CN117046140A... A device for the distillation of electronic-grade hydrofluoric acid was developed, using a deep learning algorithm to establish a mapping relationship between hydrogen fluoride gas velocity and liquid temperature. However, such feature information mined based on artificial intelligence is essentially a black-box mapping driven by data statistics, lacking a causal description of the evolution of underlying physical mechanisms such as fluid viscosity and enthalpy change during the dynamic process of chemical reaction. In the precision oxidation and impurity removal process of semiconductor-grade electronic hydrofluoric acid, the property drift has extremely strong nonlinearity and transient suddenness. When faced with abnormal operating condition fluctuations outside the training set, the black-box model has difficulty accurately inverting the real-time phase shift caused by the variation of thermal conduction hindrance. This results in an unmatchable time axis misalignment between the control output and the physical exothermic pulse, which cannot meet the almost demanding stability requirements of the temperature field for the ppt-level purification process.

[0004] Therefore, the technical problem to be solved by this invention is how to achieve real-time offsetting between the control phase and the excitation of heat load under the condition of dynamic evolution of fluid properties and significant hysteresis in heat transfer, and how to eliminate the influence of heat transfer hysteresis variation on temperature control accuracy. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid, comprising the following steps: Step 101: Using a platinum resistance sensor installed in the reaction containment space and a mass flow meter installed on the fluid delivery pipeline, real-time temperature data characterizing the state of the reaction system in the reaction containment space and instantaneous mass flow data corresponding to the input flow of the reaction medium are obtained. Step 102: Extract the first-order differential signal of the real-time temperature data and the first-order differential signal of the instantaneous mass flow rate data, and input the two first-order differential signals into the phase difference tracking logic. By identifying the extreme points of the jumps of the two first-order differential signals within the sampling period, obtain the transient phase difference characterization value between the two first-order differential signals, and use the transient phase difference characterization value to define the fluid nonlinear physical resistance fluctuation caused by the evolution of physical properties in the reaction system. Step 103: Calculate the absolute difference between the real-time temperature data and the preset target temperature. Based on the mapping relationship between the absolute difference and the preset error distribution model, dynamically schedule the nonlinear coupling coefficient used to adjust the feedforward channel access ratio in order to identify the error state of the reaction system. Step 104: Establish the mapping ratio between the transient phase difference characterization value and the standard thermal diffusion constant, and based on the mapping ratio and the proportional coefficient of the preset fluid rheology model, synchronously and inversely correct the reaction enthalpy change coefficient adapted to the current control cycle to achieve compensation for the nonlinear physical resistance of the fluid. Step 105: Based on the error state and the corrected reaction enthalpy change coefficient, a feedforward compensation control quantity is generated. The feedforward compensation control quantity and the feedback control quantity are superimposed and processed to output the opening control command for the fluid regulating actuator, so as to accurately adjust the real-time addition amount of the reaction medium.

[0006] Preferably, step 103 includes the following sub-steps: step 1031, determining whether the absolute difference is within the preset steady-state dead zone threshold range; step 1032, if the absolute difference exceeds the steady-state dead zone threshold range, determining that the reaction system is subjected to input flow abrupt disturbance, setting the nonlinear coupling coefficient to 1.0, and realizing full access of the feedforward compensation control quantity; step 1033, if the absolute difference returns to the steady-state dead zone threshold range, reducing the nonlinear coupling coefficient to 0 according to the preset exponential decay rule, suppressing the disturbance injection of the feedforward channel during the steady-state fine-tuning period.

[0007] Preferably, step 102 includes the following sub-steps: step 1021, within each sampling period, the extreme points of the jumps of the first-order differential signal of the real-time temperature data and the extreme points of the jumps of the first-order differential signal of the instantaneous mass flow rate data are tracked in parallel; step 1022, the time axis offset between the two extreme points is calculated, and the time axis offset is converted into the signal phase difference time reflecting the fluid viscosity fluctuation, as a transient phase difference characterization value.

[0008] Preferably, before outputting the opening control command for the fluid regulating actuator in step 105, the method further includes: importing the opening control command into first-order inertial filtering logic, performing frequency domain attenuation processing on the pulse signal with a frequency higher than 2Hz in the opening control command, and suppressing the high-frequency fluctuations of the opening control command from being transmitted to the fluid regulating actuator.

[0009] Preferably, the method further includes the following long-term self-calibration steps: Step 601, using 24h as the statistical period, calculate the algebraic mean deviation of the basic feedback control quantity; Step 602, extract the adjustment trend characteristics within the statistical period, reverse the model static weights of the feedforward compensation control quantity, and eliminate the model static error of the feedforward compensation control quantity.

[0010] Preferably, step 104 includes the following sub-steps: step 1041, obtaining the activation energy benchmark of the oxidation reaction of the controlled object in the reaction system; step 1042, performing nonlinear modulation on the enthalpy change coefficient of the reaction according to the mapping ratio to compensate for the hysteresis deviation caused by the change of heat transfer path during the ppt-level purification process of the reaction system.

[0011] Preferably, the filtering constant of the first-order inertial filtering logic is corrected in real time based on the mechanical wear life model of the fluid regulating actuator, thereby improving the operating conditions of the fluid regulating actuator in a highly corrosive environment while ensuring control accuracy.

[0012] Preferably, in step 101, the mass flow rate data is obtained by acquiring the instantaneous flow rate data of the reaction medium entering the reaction container space in real time through a mass flow rate acquisition unit installed on the fluid delivery pipeline.

[0013] Preferably, the method is applied to the purification process of the oxidation reaction of electronic hydrofluoric acid. After the opening control command is output in step 105, the dropping rate of the reaction medium is controlled by the opening control command to convert the low boiling point impurities in the reaction system into high boiling point complexes.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. In the production of semiconductor-grade electronic hydrofluoric acid, a dynamic adjustment mechanism based on phase lead compensation is constructed by synergistically extracting the first derivative of the oxidant feed mass flow rate and the second derivative of the real-time temperature of the reaction zone. The derivative peak of the feed flow rate is used as the logical starting point for heat injection, and the extreme value of the response of the second derivative of temperature is used as the physical inflection point of heat conduction. By calculating the transient phase difference between the two signals on the time axis, the intrinsic heat conduction hysteresis during the impurity removal process of high-purity hydrofluoric acid is directly stripped away. This processing method, which aligns the trend characteristics of the feedforward signal with the acceleration characteristics of the feedback signal in phase, ensures that the action command of the cooling actuator leads the deviation trend of the overall temperature in the physical time dimension, eliminates the spatiotemporal misalignment between the execution action and the physical micro-change caused by the pure hysteresis of fluid heat transfer, and ensures that the steady-state temperature fluctuation of the reaction zone is clamped within ±0.05℃ of the set value.

[0015] 2. By utilizing the mapping ratio between the real-time tracked transient phase difference characterization value and the standard thermal diffusion constant, a self-excited compensation logic for the reaction enthalpy change coefficient for drastic changes in fluid microviscosity is established. Since the instantaneous formation of high-boiling-point complexes in oxidation reactions leads to nonlinear jumps in the local viscosity of the reaction system, which in turn causes dynamic drift in heat transfer resistance, this method directly inverts the change in this microscopic physical resistance through real-time phase difference monitoring and simultaneously corrects the reaction enthalpy change coefficient in the feedforward channel online. The originally imperceptible fluctuations in medium properties are reduced in dimension and mapped to measurable signal time differences, so that the calculated feedforward control quantity can always evolve synchronously with the actual transient heat load inside the fluid, avoiding the overreaction or low-frequency oscillation of the system induced by the staticization of preset parameters in traditional control models.

[0016] 3. Based on the absolute difference between the real-time temperature and the target set temperature, the nonlinear coupling coefficient is dynamically adjusted to construct an adaptive constraint execution flow within the control law. When the system suffers strong disturbances such as sudden changes in feed, causing the absolute difference to exceed the dead zone threshold, the nonlinear coupling coefficient converges to 1.0 to achieve full access of the feedforward control quantity and forcibly offset external energy impacts. When the system returns to the steady-state dead zone range, the coefficient decays to zero to suppress the disturbance injection of the feedforward channel into the steady-state fine-tuning period. This switching mode based on error state recognition enables the system to have high-strength anti-disturbance stability while also ensuring high convergence accuracy of steady-state control, achieving precise coverage of the ppt-level semiconductor reagent preparation process window without adding any hardware detection sensors. Attached Figure Description

[0017] Figure 1 This is a flowchart of the adaptive phase compensation adjustment of reaction parameters in this invention; Figure 2 This is a diagram of the feedforward-feedback coupling control architecture under the property drift condition of this invention.

[0018] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] 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.

[0020] A method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid includes the following steps: Step 101: Using a platinum resistance sensor installed in the reaction containment space and a mass flow meter installed on the fluid delivery pipeline, real-time temperature data characterizing the state of the reaction system in the reaction containment space and instantaneous mass flow data corresponding to the input flow of the reaction medium are obtained. Step 102: Extract the first-order differential signal of the real-time temperature data and the first-order differential signal of the instantaneous mass flow rate data, and input the two first-order differential signals into the phase difference tracking logic. By identifying the extreme points of the jumps of the two first-order differential signals within the sampling period, obtain the transient phase difference characterization value between the two first-order differential signals, and use the transient phase difference characterization value to define the fluid nonlinear physical resistance fluctuation caused by the evolution of physical properties in the reaction system. Step 103: Calculate the absolute difference between the real-time temperature data and the preset target temperature. Based on the mapping relationship between the absolute difference and the preset error distribution model, dynamically schedule the nonlinear coupling coefficient used to adjust the feedforward channel access ratio in order to identify the error state of the reaction system. Step 104: Establish the mapping ratio between the transient phase difference characterization value and the standard thermal diffusion constant, and based on the mapping ratio and the proportional coefficient of the preset fluid rheology model, synchronously and inversely correct the reaction enthalpy change coefficient adapted to the current control cycle to achieve compensation for the nonlinear physical resistance of the fluid. Step 105: Based on the error state and the corrected reaction enthalpy change coefficient, a feedforward compensation control quantity is generated. The feedforward compensation control quantity and the feedback control quantity are superimposed and processed to output the opening control command for the fluid regulating actuator, so as to accurately adjust the real-time addition amount of the reaction medium.

[0021] Preferably, step 103 includes the following sub-steps: step 1031, determining whether the absolute difference is within the preset steady-state dead zone threshold range; step 1032, if the absolute difference exceeds the steady-state dead zone threshold range, determining that the reaction system is subjected to input flow abrupt disturbance, setting the nonlinear coupling coefficient to 1.0, and realizing full access of the feedforward compensation control quantity; step 1033, if the absolute difference returns to the steady-state dead zone threshold range, reducing the nonlinear coupling coefficient to 0 according to the preset exponential decay rule, suppressing the disturbance injection of the feedforward channel during the steady-state fine-tuning period.

[0022] Preferably, the dynamic scheduling of nonlinear coupling coefficients in step 103 follows the following calculation logic: ,in, For nonlinear coupling coefficients, The gain constant is used to characterize the thermal inertia of the reaction system. For real-time temperature data, The preset target temperature.

[0023] Preferably, step 102 includes the following sub-steps: step 1021, within each sampling period, the extreme points of the jumps of the first-order differential signal of the real-time temperature data and the extreme points of the jumps of the first-order differential signal of the instantaneous mass flow rate data are tracked in parallel; step 1022, the time axis offset between the two extreme points is calculated, and the time axis offset is converted into the signal phase difference time reflecting the fluid viscosity fluctuation, as a transient phase difference characterization value.

[0024] Preferably, before outputting the opening control command for the fluid regulating actuator in step 105, the method further includes: importing the opening control command into first-order inertial filtering logic, performing frequency domain attenuation processing on the pulse signal with a frequency higher than 2Hz in the opening control command, and suppressing the high-frequency fluctuations of the opening control command from being transmitted to the fluid regulating actuator.

[0025] Preferably, the method further includes the following long-term self-calibration steps: Step 601, using 24h as the statistical period, calculate the algebraic mean deviation of the basic feedback control quantity; Step 602, extract the adjustment trend characteristics within the statistical period, reverse the model static weights of the feedforward compensation control quantity, and eliminate the model static error of the feedforward compensation control quantity.

[0026] Preferably, step 104 includes the following sub-steps: step 1041, obtaining the activation energy benchmark of the oxidation reaction of the controlled object in the reaction system; step 1042, performing nonlinear modulation on the enthalpy change coefficient of the reaction according to the mapping ratio to compensate for the hysteresis deviation caused by the change of heat transfer path during the ppt-level purification process of the reaction system.

[0027] Preferably, the filtering constant of the first-order inertial filtering logic is corrected in real time based on the mechanical wear life model of the fluid regulating actuator, thereby improving the operating conditions of the fluid regulating actuator in a highly corrosive environment while ensuring control accuracy.

[0028] Preferably, in step 101, the mass flow rate data is obtained by acquiring the instantaneous flow rate data of the reaction medium entering the reaction container space in real time through a mass flow rate acquisition unit installed on the fluid delivery pipeline.

[0029] Preferably, the method is applied to the purification process of the oxidation reaction of electronic hydrofluoric acid. After the opening control command is output in step 105, the dropping rate of the reaction medium is controlled by the opening control command to convert the low boiling point impurities in the reaction system into high boiling point complexes.

[0030] Example 1: In a continuous operation scenario of the oxidation and impurity removal section in the production of hydrofluoric acid for semiconductor-grade electronics, the process equipment quantitatively adds oxidant to a circulating hydrofluoric acid system to convert low-boiling-point impurities. The instantaneous exothermic reaction and the formation of high-boiling-point complexes cause a nonlinear jump in the local fluid viscosity of the reaction system. This evolution of the medium's physical state constitutes the physical resistance to heat diffusion in the fluid, causing the physical hysteresis time between the start of the heat release and the capture of the corresponding temperature jump signal by the platinum resistance sensor to evolve into an unsteady parameter that fluctuates drastically with the reaction depth. Conventional control architectures relying on constant heat transfer parameters cannot withstand the heat with pure spatial diffusion hysteresis. After collecting the data, a cooling execution command deviating from the actual transient heat load is output downstream, causing the reaction fluid to oscillate between the supercooled and superheated boundaries, resulting in a decline in impurity conversion efficiency within a specific process window. To address the control deviation caused by the aforementioned physical obstruction, real-time temperature data characterizing the state of the reaction system within the reaction chamber and instantaneous mass flow rate data corresponding to the input flow of the reaction medium are acquired using a platinum resistance sensor installed in the reaction chamber and a mass flow meter installed on the fluid delivery pipeline. The first-order differential signals of the real-time temperature data and the instantaneous mass flow rate data are extracted, and these two first-order differential signals are input into a phase difference tracking logic. By identifying the two first-order differential signals... The transient phase difference between two first-order differential signals is obtained by identifying the extreme points of the abrupt changes in the first-order differential signal within the sampling period. This transient phase difference is then used to define the nonlinear physical resistance fluctuations of the fluid caused by the evolution of physical properties within the reaction system. A mapping ratio between the transient phase difference and the standard thermal diffusivity constant is established. Based on this mapping ratio and the proportionality coefficient of the preset fluid rheological model, the reaction enthalpy coefficient adapted to the current control period is synchronously and inversely corrected. The preceding flow rate abrupt extreme points and the subsequent temperature response extreme points form a correlation constraint on the time axis. The flow rate calculus information provides a perturbation injection benchmark for the recalculation of the reaction enthalpy coefficient, while the temperature calculus information provides physical confirmation of heat arrival. The two work together to establish the dynamic compensation for the nonlinear heat transfer hindrance of the system. At the hardware execution level of this compensation system, the platinum resistance sensor, due to the intrinsic thermal inertia brought about by its encapsulation sleeve, does not directly sense the high-frequency details of the microscopic viscosity change itself. Instead, it acts as a heat flux integration component at the flow field boundary, measuring the overall wavefront advancement hysteresis phenomenon of the heat transfer flow field caused by the cross-linking effect of microscopic molecular chains in the fluid on the overall scale. The control logic intercepts this overall heat transfer delay difference that appears on the second-level time axis and uses inverse mapping to deduce the equivalent scale of the microscopic local resistance change. Thus, at the logic level, it bridges the physical scale gap between the low-pass characteristics of the temperature probe and the microscopic transients of the fluid.

[0031] When outputting fluid regulation actions, the absolute difference between real-time temperature data and the preset target temperature is calculated. Based on the mapping relationship between the absolute difference and the preset error distribution model, the nonlinear coupling coefficient used to adjust the feedforward channel access ratio is dynamically scheduled. When the absolute difference exceeds the preset steady-state dead zone threshold range, it is determined that the reaction system has suffered a sudden disturbance of the input flow. The nonlinear coupling coefficient is set to 1.0, and the feedforward compensation control quantity is fully connected. When the absolute difference returns to the steady-state dead zone threshold range, the nonlinear coupling coefficient is reduced to 0 according to the preset exponential decay rule to suppress the disturbance injection of the feedforward channel. Based on the identified error state and the corrected reaction enthalpy change coefficient, a feedforward compensation control quantity is generated. The feedforward compensation control quantity and the feedback control quantity are superimposed to output the opening control command for the fluid regulation actuator, which precisely adjusts the real-time addition of the reaction medium. The opening control command drives the regulating valve in the initial stage of the distortion of the internal thermodynamic state of the fluid. The system compensates and regulates fluid flux, eliminating spatiotemporal misalignment caused by variations in heat transfer resistance. This ensures that all thermodynamic parameters within the closed fluid network converge to the target range. When the system adjusts the enthalpy change coefficient to match the fluid physical resistance fluctuations, the computation module extracts the transient phase difference characterization value and the standard thermal diffusivity fixed in the storage unit. It calculates the quotient of the transient phase difference characterization value divided by the standard thermal diffusivity as the mapping ratio. Based on the Arrhenius chemical reaction kinetics, the sensitivity of the fluid micro-reaction rate to the system's thermal fluctuations is rigidly constrained by the inherent physical activation energy of the medium. The computation module simultaneously acquires the activation energy benchmark of the controlled object's oxidation reaction in the reaction system and introduces a correlation term containing the activation energy benchmark index to establish a nonlinear modulation base. In the instruction computation step, the proportional coefficient of the fluid rheological model and the modulation amount containing the activation energy benchmark index are extracted and multiplied together to obtain the dynamic compensation factor. The specific dynamic compensation factor K is continuously calculated using the relational formula. Symbols in mathematical constraint logic To calculate and obtain the mapping ratio, the symbol... To invoke the proportional coefficient output of the fluid rheology model, the symbol E represents the activation energy benchmark for the extracted oxidation reaction and is specified to have a value range in the positive real number interval. The constant R is determined to be the ideal gas constant. In the actual operational architecture of the controller, to satisfy the dimensional consistency of the underlying physical logic, the exponent denominator in the above operational relationship is implicitly multiplied by the benchmark steady-state thermodynamic temperature constant under the current operating conditions of the system at the compilation layer. This transforms the exponent into a dimensionless algebraic term that conforms to the classical kinetic laws, converting the activation energy benchmark into the inherent static bias ratio of the system. The transient fluctuation characteristics of the fluid micro-reaction rate are decoupled in real time by multiplying the mapping ratio and the proportional coefficient. The proportional coefficient of the fluid rheology model is extracted, and the dynamic compensation factor is obtained by multiplying the mapping ratio and the proportional coefficient. The initial enthalpy change coefficient of the reaction system is extracted, and the quotient of the initial enthalpy change coefficient divided by the dynamic compensation factor is calculated. This quotient is output as the corrected enthalpy change coefficient of the reaction. This operational process transforms the evolution of fluid viscosity properties into algebraic modulation parameters within the feedforward control loop, constituting the underlying algorithm basis for the resistance compensation process.

[0032] When the control system performs the superposition of feedforward compensation control and feedback control, the arithmetic module obtains the corrected enthalpy change coefficient and the first-order differential signal of the instantaneous mass flow rate data, calculates their product to obtain the basic feedforward reference quantity, extracts the nonlinear coupling coefficient based on the absolute difference, multiplies the basic feedforward reference quantity and the nonlinear coupling coefficient to generate the feedforward compensation control quantity, extracts the basic feedback adjustment quantity output by the proportional-integral-derivative control model based on the absolute difference between the real-time temperature data and the preset target temperature, extracts the physical opening value of the fluid regulating actuator in the previous sampling period, performs an algebraic summation operation on the feedforward compensation control quantity, the basic feedback adjustment quantity, and the physical opening value, generates the opening control command for the fluid regulating actuator in the current control period, and outputs the drive current to control the valve body action. Before the valve body action, the controller imports the opening control command into the first-order inertial filtering logic. Based on the fatigue cumulative damage physical model, the relative wear degree of the mechanical surface of the valve body is positively correlated with the cumulative value of the valve core reciprocating stroke. The control system runs the fluid regulation in parallel in the background. The mechanical wear life model of the actuator is used. In each sampling cycle, the computing unit extracts the absolute value of the displacement difference between the current opening control command and the historical command of the previous cycle and performs cumulative summation to continuously output the cumulative action stroke of the valve core. The system corrects the first-order inertial filter logic constant in real time by capturing the cumulative action stroke of the valve core. The specific numerical calculation constraint follows the formula F=C+k×S. In the algebraic relationship, the symbol F represents the dynamic filter constant output in the current control cycle, the symbol C indicates the basic filter constant when the actuator is in the initial assembly state with zero wear, and the symbol S is the real-time tracking and statistical valve core cumulative action stroke. The coefficient k is set as the wear penalty gain corresponding to a specific strong corrosive process medium environment and is limited to a real value greater than zero. Relying on the valve core wear stroke proportionally amplifying the filter constant algorithm logic to forcibly attenuate the high-frequency pulse electrical signal branch with a frequency higher than 2Hz in the command array, invalid command fluctuations are truncated and transmitted to the physical execution end of the fluorine-related special control valve. This summation path transforms the multi-dimensional disturbance compensation requirements into a single physical execution input, establishing a command closed loop for fluid flux regulation.

[0033] Example 2: When the system faces continuous disturbance in the high-purity hydrofluoric acid oxidation and impurity removal section, the heat transfer hysteresis induced by the sudden change in fluid kinematic viscosity causes the distillation column reboiler temperature to deviate from the set value. A physical verification platform with a closed-loop pipeline network and a jacketed reactor is constructed. The platform integrates a Coriolis mass flow meter with a range of 0 to 500 kg / h and an accuracy of 0.1%, and introduces an armored platinum resistance sensor with a response time of less than 0.5 s. The basic input flow data waveform is extracted from the historical operating database, and the temperature control accuracy of the physical test environment is set to ±0.1℃. The input flow reference information is then sent to the system. The signal is actively superimposed with Gaussian white noise with a signal-to-noise ratio of 15dB to simulate fluid pulse interference caused by the start-up and shutdown of high-power pumps and valves. The engineering consideration for setting the sampling period is to balance the real-time performance of high-frequency phase difference capture with the computational load of the controller. When the spectral bandwidth of the extracted first-order differential signal is in a high-dynamic-range change, a sampling period of 10ms is selected to prevent aliasing at small jump extreme points. This value ensures that the capture accuracy of the transient phase difference characterization value meets the hysteresis compensation requirement. The oxidant metering pump is started, and the initial system target temperature is set to 65.0℃. This invention is not loaded. When the conventional feedback comparison sample of the Ming control method encountered a 10% input flow mutation, the original temperature sensing data showed heat transfer hysteresis. The peak temperature in the reaction chamber reached 68.3℃ 45s after the disturbance, deviating from the target temperature by 3.3℃, ​​and the system exhibited a divergent oscillation. Running the sample of this invention, the first-order differential signals of real-time temperature data and instantaneous mass flow rate data were extracted. For the input waveform with a signal-to-noise ratio of 15dB, the system identified the extreme points of the flow differential jump and the extreme points of the temperature differential jump. The calculated transient phase difference characterization value changed from the initial stage of the disturbance to the current value of the flow differential jump. The temperature climbs from 2.1s to 5.8s, establishing the mapping ratio between the transient phase difference characterization value and the standard thermal diffusivity constant, and inversely correcting the reaction enthalpy change coefficient. The absolute difference between the real-time temperature data and the preset target temperature is calculated. If this absolute difference exceeds the steady-state dead zone threshold range of 0.5℃ within 1.2s, the system adjusts the nonlinear coupling coefficient to 1.0, fully connects the feedforward compensation control quantity, and outputs the opening control command to adjust the real-time addition amount of the reaction medium. The measured data shows that the temperature peak of the sample group under disturbance converges to 65.4℃, and the steady-state recovery time is shortened to 12s.

[0034] A partially missing control group was set up, cutting off the correction loop between the transient phase difference characterization value and the enthalpy change coefficient of the reaction, while retaining the nonlinear coupling coefficient scheduling logic based on the absolute difference. Tests showed that this sample group, when rapidly and fully connected to the feedforward compensation control quantity, experienced a shift in the feedforward reference due to the lack of dynamic compensation for fluid nonlinear physical resistance fluctuations, resulting in a temperature peak of 66.5℃ and a second overshoot. This verifies that the correlation constraint between flow rate calculus information and temperature calculus information on the time axis provides a disturbance injection reference, forming a causal coupling synergistic effect with the error scheduling mechanism. An out-of-range control group with a steady-state dead zone threshold was also set. When the steady-state dead zone threshold was set to 0.1℃ below the lower limit, white noise frequently triggered system state jumps, and the nonlinear coupling coefficient switched frequently between 0 and 1.0, exacerbating wear on the regulating valve and causing the system temperature to exhibit continuous high-frequency flutter with a standard deviation of 0.6℃. When the steady-state dead zone threshold was set to 2.0℃ above the upper limit, the system lost sensitivity to sudden changes in the actual input flow, and temperature deviation occurred. After reaching 1.8℃, full feedforward access was initiated, missing the compensation window. The data exhibited nonlinear inflection point characteristics, confirming that 0.5℃ to 0.8℃ is the optimal working window that balances noise resistance stability and adjustment agility. A problem intensity gradient control system was set up, with the input flow disturbance amplitude set to three levels: low, medium, and high, namely 5%, 15%, and 25%. The measured maximum temperature dynamic deviation values ​​corresponding to the sample group of this invention were 0.2℃, 0.6℃, and 1.1℃, respectively. As the disturbance intensity increased, the maximum temperature deviation showed a convergent linear correlation change, and the system remained stable without diverging. The experimental data confirmed that by extracting the extreme points of the jumps in the first-order differential signals of temperature and flow to obtain the transient phase difference characterization value, and dynamically scheduling the nonlinear coupling coefficient, the control loop achieves adaptive compensation of parameters under complex fluid property evolution and noise interference conditions. This compensation mechanism eliminates the spatiotemporal misalignment of control caused by the nonlinear heat transfer hindrance of the fluid, and stably limits the thermodynamic state parameters in the closed fluid pipeline network within the target range.

[0035] Example 3: This example combines Figures 1 to 2 The method for adjusting reaction parameters for the production of semiconductor-grade electronic hydrofluoric acid is explained, such as... Figure 1The diagram illustrates the specific flow of a reaction parameter adjustment method for semiconductor-grade electronic hydrofluoric acid production. The method begins in step 101, where a platinum resistance sensor installed in the reaction chamber and a mass flow meter installed on the fluid delivery pipeline acquire real-time temperature data characterizing the state of the reaction system within the reaction chamber, as well as instantaneous mass flow data corresponding to the input flow of the reaction medium. Step 102 involves extracting the first-order differential signals of the real-time temperature data and the instantaneous mass flow data, and inputting these two signals into a phase difference tracking logic. By identifying the extreme points of the jumps in the two first-order differential signals within the sampling period, the transient phase difference characterization value between the two signals is obtained. This transient phase difference characterization value is used to define the nonlinear physical resistance fluctuations of the fluid caused by the evolution of physical properties within the reaction system. Then, in step 103… The absolute difference between the real-time temperature data and the preset target temperature is calculated. Based on the mapping relationship between the absolute difference and the preset error distribution model, the nonlinear coupling coefficient used to adjust the feedforward channel access ratio is dynamically scheduled to identify the error state of the reaction system. Then, step 104 is executed to establish the mapping ratio between the transient phase difference characterization value and the standard thermal diffusion constant. Based on the mapping ratio and the proportional coefficient of the preset fluid rheology model, the reaction enthalpy change coefficient adapted to the current control cycle is synchronously and inversely corrected to compensate for the nonlinear physical resistance of the fluid. Finally, in step 105, based on the error state and the corrected reaction enthalpy change coefficient, a feedforward compensation control quantity is generated. The feedforward compensation control quantity and the feedback control quantity are superimposed and output as an opening control command for the fluid regulation actuator to accurately adjust the real-time addition amount of the reaction medium.

[0036] like Figure 2 As shown, the system architecture and data flow of this regulation logic are as follows: the reaction system, as the basic unit of the production process, generates raw physical signals that are captured by the real-time operating condition parameter acquisition module. The signal output by this module is processed in two ways: the first way enters the dynamic adaptation nonlinear mapping module and is then fed into the temperature difference error state establishment module; the second way enters the reaction enthalpy change feedback correction module, which simultaneously participates in the verification and identification of the temperature difference error state while adjusting the calculation coefficients required for feedforward control in real time. The process controller is responsible for integrating the calculation results of the above modules and converging them into the feedforward superposition control action module. This control logic generates a phase lead to offset the fluid heat transfer resistance and finally outputs drive commands to the regulation actuator, thereby achieving precise closed-loop control of the reaction system state.

[0037] Example 4: When the semiconductor-grade electronic hydrofluoric acid production equipment exhibits heat exchanger surface scaling and hydrodynamic boundary drift during continuous operation, the preset constant thermal inertia parameters and rheological coefficients deviate from the current physical state of the equipment, causing overcompensation or undercompensation imbalance in the feedforward compensation channel. A physical calibration procedure is set for the gain constant α characterizing the thermal inertia of the reaction system and the proportional coefficient of the preset fluid rheological model. The oxidant metering pump is controlled to apply a step change signal with an amplitude of 5% at the reference mass flow rate. A platinum resistance sensor collects the real-time temperature data response sequence within the reaction chamber. The time hysteresis span corresponding to when the temperature change in the real-time temperature data response sequence reaches 63.2% of the steady-state deviation from the extreme value is extracted, and the time hysteresis is calculated. The reciprocal of the span is used to determine the basic thermal inertia characteristic value. The preset equivalent heat capacity value of the reaction containment space and the reference flow rate of the reaction medium output by the mass flow meter are extracted. The product of the preset equivalent heat capacity value and the reference flow rate of the reaction medium is calculated. The basic thermal inertia characteristic value is weighted using this product, and the gain constant adapted to the current physical state of the equipment is output. When the control system performs the above weighted calculation, the internal calculation unit simultaneously extracts the product of the equipment heat transfer area constant and the reciprocal of the reference heat transfer coefficient preset in the register. The additional dimensions brought in by the equivalent heat capacity and mass flow rate are canceled by the dimensional normalization operation, ensuring that the final output gain constant has only the physical dimension of the reciprocal of the thermodynamic temperature, so that it can achieve dimensional consistency with the absolute difference in subsequent calculations.

[0038] Within the bypass measurement module of the fluid transport pipeline, shear stress data of the reaction medium at a set temperature and different flow rates are extracted. The slope of the change between the logarithm of shear stress and the logarithm of flow rate is calculated to determine the power-law exponent of the reaction medium. Its product with the standard thermal diffusivity constant is calculated, and the proportional coefficient of the fluid rheological model is output. This calibration procedure transforms the static parameters at the bottom layer of the control system into quantitative measured values ​​calculated based on field sensor physical data. The system dynamically schedules the nonlinear coupling coefficient according to the updated gain constant α, and corrects the reaction enthalpy change coefficient adapted to the current control cycle in combination with the updated proportional coefficient. The feedforward compensation control quantity synchronously maps the current thermal conduction hindrance and fluid viscosity state of the equipment, limiting the deviation of thermodynamic state parameters in the closed fluid pipeline network during the equipment operation cycle.

[0039] Example 5: When a newly built semiconductor-grade electronic hydrofluoric acid production line is in the physical baseline calibration condition before commissioning, there are engineering deviations between the heat exchange boundary conditions of the fluid pipeline network and the theoretical settings. Before the process fluid is injected into the reaction containment space in batches, the control system initiates an on-site offline calibration procedure targeting the preset error distribution model and steady-state dead zone threshold. It controls the oxidant metering pump to inject a constant power standard heat source pulse into the circulating medium in a stepwise manner within the safe test temperature range. The temperature response sequence is continuously collected using a platinum resistance sensor, and the peak value of the natural temperature drift envelope when the system reaches thermal equilibrium is calculated. The shift envelope peak value is set as the boundary reference for the steady-state dead zone threshold. The power injection amplitude of the standard heat source pulse is increased incrementally. The absolute difference between the real-time temperature data and the preset target temperature is recorded. The critical feedforward compensation opening ratio required by the system to suppress divergent oscillations at each absolute difference node is recorded synchronously. The absolute difference sequence and the critical feedforward compensation opening ratio sequence are mapped and combined to generate a discretized error mapping matrix. The controller writes the discretized error mapping matrix into the storage space as the underlying addressing library of the preset error distribution model. The output nonlinear coupling coefficient scheduling command is synchronously associated with the measured heat conduction parameters of the physical device.

[0040] When setting the exponential decay rule parameters for the nonlinear coupling coefficient to decay towards 0, the limiting heat removal rate of the cooling jacket of the reaction containment space under the maximum refrigerant flow state is extracted. Combined with the equivalent heat capacity of the reaction medium obtained by calibration, the natural temperature fall-off time constant after removing the feedforward compensation control is derived. The reciprocal of the natural temperature fall-off time constant is written into the calculation module of the exponential decay rule as the decay damping variable. When the absolute difference returns to the steady-state dead zone threshold range, the calculation module containing the decay damping variable is called to drive the nonlinear coupling coefficient to decrease at the corresponding rate. The decay damping variable constrains the withdrawal rate of the feedforward channel command to be synchronized with the physical relaxation period of the cooling mechanism, suppressing the secondary fluid pulse disturbance caused by the step closure of the regulating valve opening. The allocation action of the feedforward channel and the feedback channel is associated with the heat transfer hysteresis boundary conditions measured on site. The thermodynamic state parameters of the closed-loop pipe network system under the input flow disturbance are stably converged within the target range.

[0041] Example 6: When initializing the physical parameters of the control system of a semiconductor-grade electronic hydrofluoric acid production equipment before deployment, the preset underlying time constant in the controller lacks a quantitative benchmark to adapt to the boundary drift of the on-site fluid dynamics. Therefore, the system establishes an offline optimization calibration procedure for the sampling period required to extract the first-order differential signal. A fluid disturbance injection generator is connected to the physical testing section. The fluid disturbance injection generator is controlled to continuously sweep the frequency and output transient flow pulse signals to the circulating reaction medium within the frequency range of 200Hz to 500Hz. The corresponding response is synchronously acquired by a mass flow meter and a platinum resistance sensor. The waveform sequence is processed by applying a Fast Fourier Transform algorithm to the response waveform sequence to extract the upper limit physical quantity of the dominant frequency band, which includes more than 95% of the total signal energy. The reciprocal of the upper limit physical quantity of the dominant frequency band is divided by a preset constant of 5 to generate the sampling period time parameter required by the control module. The sampling period time parameter is written into the hardware clock register. This parameter extraction mechanism establishes a mapping relationship between the hardware sensing rate and the fluctuation of the nonlinear heat transfer resistance of the fluid based on the physical signal sampling law, suppresses the signal aliasing interference encountered by the extreme points of small jumps in the high dynamic change range, and provides the underlying computing loop with distortion-free discrete data samples.

[0042] When processing two first-order differential signals to obtain transient phase difference characterization values, the preset comparison logic faces the risk of extreme value misjudgment caused by high-frequency random fluctuations of discrete data points. The control module loads the first-order differential signals of the acquired real-time temperature data and the first-order differential signals of the instantaneous mass flow rate data into a sliding data window spanning a preset number of sampling periods. Within the sliding data window, the slope reversal flag of adjacent data points before and after each discrete data point is calculated. Specific discrete data points that simultaneously satisfy the condition of the slope changing from positive to negative and the amplitude being greater than twice the local background noise baseline are extracted as jump extreme points. The first timestamp of the jump extreme point in the instantaneous mass flow rate data is extracted, and the second timestamp of the corresponding jump extreme point in the real-time temperature data is extracted. The difference data is obtained by subtracting the second timestamp from the first timestamp and is assigned as the transient phase difference characterization value. This operation process transforms the abstract phase difference calculation into a deterministic algebraic subtraction action of timestamps based on the hardware clock sequence. The output transient phase difference characterization value directly quantifies the physical hysteresis time span caused by the thermal diffusion of the reaction medium.

[0043] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0044] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid, characterized in that, Includes the following steps: Step 101: Using a platinum resistance sensor installed in the reaction containment space and a mass flow meter installed on the fluid delivery pipeline, real-time temperature data characterizing the state of the reaction system in the reaction containment space and instantaneous mass flow data corresponding to the input flow of the reaction medium are obtained. Step 102: Extract the first-order differential signal of the real-time temperature data and the first-order differential signal of the instantaneous mass flow rate data, and input the two first-order differential signals into the phase difference tracking logic. By identifying the extreme points of the jumps of the two first-order differential signals within the sampling period, obtain the transient phase difference characterization value between the two first-order differential signals, and use the transient phase difference characterization value to define the fluid nonlinear physical resistance fluctuation caused by the evolution of physical properties in the reaction system. Step 103: Calculate the absolute difference between the real-time temperature data and the preset target temperature. Based on the mapping relationship between the absolute difference and the preset error distribution model, dynamically schedule the nonlinear coupling coefficient used to adjust the feedforward channel access ratio in order to identify the error state of the reaction system. Step 104: Establish the mapping ratio between the transient phase difference characterization value and the standard thermal diffusion constant, and based on the mapping ratio and the proportional coefficient of the preset fluid rheology model, synchronously and inversely correct the reaction enthalpy change coefficient adapted to the current control cycle to achieve compensation for the nonlinear physical resistance of the fluid. Step 105: Based on the error state and the corrected reaction enthalpy change coefficient, a feedforward compensation control quantity is generated. The feedforward compensation control quantity and the feedback control quantity are superimposed and processed to output the opening control command for the fluid regulating actuator, so as to accurately adjust the real-time addition amount of the reaction medium.

2. The method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid according to claim 1, characterized in that, Step 103 includes the following sub-steps: Step 1031, determine whether the absolute difference is within the preset steady-state dead zone threshold range; Step 1032, if the absolute difference exceeds the steady-state dead zone threshold range, determine that the reaction system is subjected to sudden input flow disturbance, set the nonlinear coupling coefficient to 1.0, and realize the full access of the feedforward compensation control quantity; Step 1033, if the absolute difference returns to the steady-state dead zone threshold range, reduce the nonlinear coupling coefficient to 0 according to the preset exponential decay rule, and suppress the disturbance injection of the feedforward channel during the steady-state fine-tuning period.

3. The method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid according to claim 1, characterized in that, Step 102 includes the following sub-steps: Step 1021, within each sampling period, track the extreme points of the first-order differential signal of the real-time temperature data and the extreme points of the first-order differential signal of the instantaneous mass flow rate data in parallel; Step 1022, calculate the time axis offset between the two extreme points, and convert the time axis offset into the signal phase difference time reflecting the fluid viscosity fluctuation, as a transient phase difference characterization value.

4. The method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid according to claim 1, characterized in that, Before outputting the opening control command for the fluid regulating actuator in step 105, the method further includes: importing the opening control command into first-order inertial filtering logic, performing frequency domain attenuation processing on the pulse signal with a frequency higher than 2Hz in the opening control command, and suppressing the high-frequency fluctuations of the opening control command from being transmitted to the fluid regulating actuator.

5. The method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid according to claim 1, characterized in that, The method also includes the following long-term self-calibration steps: Step 601, using a 24-hour statistical period, calculate the algebraic mean deviation of the basic feedback control quantity; Step 602, extract the adjustment trend characteristics within the statistical period, reverse the model static weights of the feedforward compensation control quantity, and eliminate the model static error of the feedforward compensation control quantity.

6. The method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid according to claim 1, characterized in that, Step 104 includes the following sub-steps: Step 1041, obtaining the activation energy benchmark for the oxidation reaction of the controlled object in the reaction system; Step 1042, performing nonlinear modulation on the enthalpy change coefficient of the reaction according to the mapping ratio to compensate for the hysteresis deviation caused by the change in heat transfer path during the ppt-level purification process of the reaction system.

7. The method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid according to claim 4, characterized in that, The filtering constant of the first-order inertial filtering logic is corrected in real time based on the mechanical wear life model of the fluid regulating actuator, thereby improving the operating conditions of the fluid regulating actuator in a highly corrosive environment while ensuring control accuracy.

8. The method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid according to claim 1, characterized in that, In step 101, the mass flow rate data is obtained by acquiring the instantaneous flow rate data of the reaction medium entering the reaction container space in real time through a mass flow rate acquisition unit installed on the fluid delivery pipeline.

9. The method for adjusting reaction parameters in the production of semiconductor-grade electronic hydrofluoric acid according to claim 1, characterized in that, The method is applied to the purification process of the oxidation reaction of electronic hydrofluoric acid. After the opening control command is output in step 105, the dropping rate of the reaction medium is controlled by the opening control command to convert low-boiling-point impurities in the reaction system into high-boiling-point complexes.

Images