A dynamic proportioning adjusting method and system for a mixed gas cabinet for electronic special gas

Abstract

This invention discloses a dynamic proportioning adjustment method and system for a mixing cabinet of electronic specialty gases, belonging to the field of electronic specialty gas technology. The method involves inputting electronic specialty gases into the mixing device at an initial proportion via a mass flow controller; using a tunable semiconductor laser absorption spectroscopy sensor to monitor the concentration of components in the mixed gas in real time; employing a model predictive control algorithm to predict future trends based on real-time concentration deviations and dynamically generate optimized flow setpoints, and reducing calculation frequency through a sampling verification mechanism; based on the optimized setpoints and combined with real-time collected gas flow rate, temperature, and pressure data, adjusting the opening of the mass flow controller through a compensation algorithm to compensate for proportioning deviations caused by various factors. This invention significantly improves mixing accuracy and shortens mixing time, meeting the stringent requirements of high-end processes such as semiconductor manufacturing for electronic specialty gas proportioning.

Description

Technical Field

[0001] This invention relates to the field of electronic specialty gases, and in particular to a method and system for dynamic proportioning adjustment of a mixing chamber for electronic specialty gases. Background Technology

[0002] Electronic specialty gases are indispensable key raw materials for the microelectronics industries such as semiconductors, flat panel displays, and photovoltaics, and are known as the "food" of the electronics industry. These industries have extremely stringent requirements for the cleanliness, purity, and mixing precision of electronic specialty gases. Large-scale integrated circuit manufacturing often requires gas purity of 6N (99.9999%) or higher, with impurity content below 10⁻⁻⁶. 7 Scale. In core processes of wafer manufacturing such as photolithography, chemical vapor deposition (CVD), etching, and doping, it is often necessary to use mixed gases composed of a variety of electronic specialty gases that are precisely mixed. The accuracy of the ratio directly determines the performance, yield, and production safety of integrated circuits.

[0003] Currently, commonly used gas mixing methods in industry mainly include gravimetric methods and traditional flow control methods. Gravimetric methods calculate the content of each component by filling and weighing the gas cylinders in stages. While the accuracy is acceptable, it is inefficient, cannot achieve continuous production, and suffers from proportioning deviations due to gas residue. Continuous flow mixing methods based on mass flow controllers (MFCs) are more common online mixing technologies. However, in practical applications, especially in multi-component electronic specialty gas mixing scenarios, existing technologies still face significant challenges:

[0004] 1. Diffusion differences lead to mismatched proportions: Different gas components have varying molecular weights and diffusion coefficients, making it difficult to achieve instantaneous and uniform mixing within the mixing device. Density differences cause gas stratification, while variations in diffusion coefficients result in inconsistent arrival times of the components at the monitoring point during transport, leading to dynamic proportioning deviations. This problem is particularly pronounced with low-concentration components (such as doped gases), where the actual concentration may deviate from the target value by more than 10%, making it difficult to meet the stringent requirements of advanced processes such as EUV lithography.

[0005] 2. Environmental fluctuations affect control accuracy: Temperature fluctuations in the semiconductor workshop environment and the process itself can alter gas density and volumetric flow rate, while traditional MFCs are calibrated based on constant temperature conditions. When the operating temperature changes, the MFC's measured values ​​will deviate from the actual mass flow rate. Simultaneously, fluctuations in gas source pressure or downstream pressure can also interfere with stable flow control. Existing systems lack an effective mechanism for real-time, joint compensation of operating parameters such as temperature and pressure, making it difficult to maintain stable mixing accuracy during long-term operation or environmental changes.

[0006] 3. Control Response Lag and Insufficient Intelligence: Most gas mixing cabinets employ single closed-loop feedback control, adjusting the MFC based on concentration monitoring results. However, due to delays in gas mixing, delivery, and concentration analysis, this control method inherently suffers from lag, making it difficult to quickly correct mixing ratio deviations. Furthermore, existing control strategies typically adjust only for current errors, lacking the ability to predict future system behavior (such as concentration change trends determined by diffusion processes). This prevents proactive and optimized control, resulting in longer mixing times and a higher risk of overshoot or oscillations during adjustments.

[0007] To address the aforementioned issues, while some technical solutions have proposed improving the structure of mixing devices or employing quantitative pre-storage methods, these often fail to fundamentally resolve the challenges of online real-time control arising from differences in multi-component diffusion and dynamic fluctuations in operating conditions. Therefore, there is an urgent need for a high-precision gas mixing method capable of real-time sensing of the mixing state, intelligent prediction of changing trends, and dynamic compensation for various interfering factors, in order to meet the increasingly stringent precision, efficiency, and stability requirements of high-end electronics manufacturing for specialty gas proportions. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for dynamic proportioning adjustment of mixing chambers for electronic special gases.

[0009] The objective of this invention is achieved through the following technical solution: The first aspect of this invention provides: a method for dynamic proportioning adjustment of a mixing chamber for electronic special gases, comprising the following steps:

[0010] Flow preset and gas input stage: The input flow of various electronic special gases is controlled by multiple mass flow controllers in the mixing cabinet, so that they enter the mixing device according to the initial input ratio;

[0011] Real-time spectral analysis stage: Using a tunable semiconductor laser absorption spectral sensor integrated on the mixing device, the concentration of each component of the mixed gas is monitored online in real time, and real-time concentration data is obtained;

[0012] Dynamic optimization phase: The real-time concentration data is compared with the preset target concentration value at the current time node to obtain the real-time concentration deviation. A model predictive control algorithm is used to predict the concentration deviation trend at subsequent time nodes based on the initial input ratio, real-time concentration data, and real-time concentration deviation. Then, the optimized flow rate setpoints for each component gas at subsequent time nodes are dynamically generated based on the concentration deviation trend. The optimization effect is verified by sampling at preset intervals to reduce the number of calculations. If the real-time concentration deviation at the sampling time node is less than or equal to the predicted concentration deviation at that time node in the concentration deviation trend obtained in the previous optimization, the optimized flow rate setpoints generated in the previous optimization are used. If the real-time concentration deviation at the sampling time node is greater than the predicted concentration deviation at that time node in the concentration deviation trend obtained in the previous optimization, dynamic optimization is performed again.

[0013] Flow compensation stage: Based on the optimized flow setpoint, the opening of the corresponding mass flow controller is adjusted in combination with the real-time collected gas flow rate, gas temperature and gas pressure data to compensate for the ratio deviation caused by differences in gas diffusion coefficient, temperature fluctuation and pressure fluctuation.

[0014] Preferably, each electronic special gas pipeline in the mixing cabinet is equipped with a two-stage pressure reducing valve and a check valve; the flow control of the mass flow controller is initially calibrated based on the initial input ratio and gas molecular weight, and the pressure of each pipeline is controlled by the two-stage pressure reducing valve to prevent gas backflow and pressure fluctuation.

[0015] Preferably, the inner walls of all pipes, valves, and mixing devices that come into contact with the electronic specialty gas are electropolished to prevent metal ion contamination.

[0016] Preferably, the laser beam path of the tunable semiconductor laser absorption spectroscopy sensor passes through an inert gas-purged protective window and the outlet gas flow channel of the mixing device, and the wavelength is tuned to select isolated peaks of the molecular absorption spectrum of the low-concentration component gas, thereby capturing transient changes in concentration in real time.

[0017] Preferably, the protective window and the mixing device are sealed with perfluoroether rubber to prevent outside air from entering the analytical optical path and to ensure the cleanliness of the electronic special gas.

[0018] Preferably, the model predictive control algorithm constructs a predictive model based on a gas diffusion dynamics model and a thermodynamic equation of state. Its objective function is to minimize the integral sum of squares of the concentration deviation over a future period of time. The algorithm also calculates the optimal flow rate setpoint sequence for each component gas through rolling optimization and uses the first element of the sequence as an immediate control command.

[0019] Preferably, the gas diffusion kinetics model includes a diffusion coefficient difference compensation term for each component gas. The diffusion coefficient difference compensation term is dynamically adjusted according to the residence time of the gas in the mixing device, wherein the diffusion compensation weight of the low-concentration component is higher than that of the high-concentration component.

[0020] Preferably, the traffic compensation stage further includes the following steps:

[0021] The real-time collected gas velocity, gas temperature, and gas pressure data are input into the ideal gas equation of state, and a diffusion coefficient compensation factor for the target electronic special gas is introduced to construct a flow compensation model with velocity, temperature, and gas pressure as variables. The comprehensive compensation coefficient under the current operating condition is calculated through the flow compensation model, and the optimized flow setpoint is multiplied by the comprehensive compensation coefficient to generate an actual control command to adjust the opening degree of the corresponding mass flow controller.

[0022] Preferably, for different electronic specialty gases, the diffusion coefficient compensation factor of the target electronic specialty gas under various combinations of flow rate, temperature and pressure is obtained in advance through experimental calibration and stored in the compensation factor database; when constructing the flow compensation model, the diffusion coefficient compensation factor of the target electronic specialty gas under the current operating condition is queried from the compensation factor database based on the real-time collected gas flow rate, gas temperature and pressure data.

[0023] Preferably, the mixing device includes a mixing tank with a porous jet pipe, wherein the nozzles for low-density gas are distributed in the lower part of the mixing tank and the nozzles for high-density gas are distributed in the upper part of the mixing tank, and the nozzle size is designed according to the gas density gradient to promote natural diffusion and shorten the mixing time.

[0024] Preferably, the mixing tank is equipped with a stirring element, which adaptively adjusts its working state and movement amplitude according to the density of the mixed gas to enhance the turbulent mixing effect.

[0025] A second aspect of the present invention provides: a dynamic proportioning adjustment system for a mixing chamber for electronic specialty gases, used to implement any of the above-mentioned dynamic proportioning adjustment methods for mixing chambers for electronic specialty gases, comprising:

[0026] The flow preset and gas input module is used to control the input flow of various electronic special gases through multiple mass flow controllers in the mixing cabinet, so that they enter the mixing device according to the initial input ratio.

[0027] The real-time spectral analysis module is used to monitor the concentration of each component of the mixed gas in real time and obtain real-time concentration data using a tunable semiconductor laser absorption spectral sensor integrated on the mixing device.

[0028] The dynamic optimization module compares real-time concentration data with the preset target concentration value at the current time node to obtain the real-time concentration deviation. It then uses a model predictive control algorithm to predict the concentration deviation trend for subsequent time nodes based on the initial input ratio, real-time concentration data, and real-time concentration deviation. Based on this trend, it dynamically generates optimized flow rate setpoints for each gas component at subsequent time nodes and verifies the optimization effect by sampling at preset intervals to reduce the number of calculations. If the real-time concentration deviation at the sampling time node is less than or equal to the predicted concentration deviation for that time node in the previous optimization trend, the previously generated optimized flow rate setpoint is used. If the real-time concentration deviation at the sampling time node is greater than the predicted concentration deviation for that time node in the previous optimization trend, dynamic optimization is performed again.

[0029] The flow compensation module is used to adjust the opening of the corresponding mass flow controller based on the optimized flow setpoint and the real-time collected gas flow rate, gas temperature and pressure data, in order to compensate for the ratio deviation caused by differences in gas diffusion coefficient, temperature fluctuation and pressure fluctuation.

[0030] A third aspect of the present invention provides: a dynamic proportioning adjustment device for a mixing chamber for electronic specialty gases, comprising a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement any of the above-described dynamic proportioning adjustment methods for a mixing chamber for electronic specialty gases.

[0031] A fourth aspect of the present invention provides: a computer-readable storage medium storing computer-executable instructions, wherein when the computer-executable instructions are loaded and executed by a processor, the above-described dynamic proportioning adjustment method for a mixing chamber for electronic special gases is implemented.

[0032] The fifth aspect of the present invention provides: a computer program product containing instructions, which, when run on a terminal, causes the terminal to execute any of the above-described methods for dynamic proportioning adjustment of mixing chambers for electronic special gases.

[0033] The beneficial effects of this invention are:

[0034] 1) This invention significantly improves the accuracy of gas mixing ratios, particularly effectively solving the problem of ratio deviations caused by differences in diffusion coefficients of low-concentration electronic specialty gas components. By introducing a TDLAS (Tunable Laser Absorption Spectroscopy) sensor for real-time online monitoring, the true concentration changes of each component can be captured with low latency, providing a data foundation for precise control. Through Model Predictive Control (MPC) algorithms, future trends are predicted based on real-time concentration deviations, and higher diffusion compensation weights are assigned to low-concentration components for forward-looking flow rate adjustment. This reduces the ratio error of low-concentration components, meeting the stringent gas ratio requirements of high-end processes such as EUV lithography and advanced chip manufacturing.

[0035] 2) This invention achieves a high degree of intelligence and efficiency improvement in the mixing process. The traditional "blind adjustment" and repeated testing and purging methods are replaced by a dynamic optimization and sampling verification mechanism based on MPC. The system can make autonomous decisions and respond quickly, reducing mixing time and significantly minimizing gas waste. The MPC algorithm generates the optimal flow setpoint sequence through rolling optimization and combines it with interval sampling verification, ensuring accuracy while reducing unnecessary computational overhead and improving the system's response speed and economy.

[0036] 3) Addressing the specific needs of electronic specialty gas applications, the system's adaptability and long-term reliability have been enhanced in multiple aspects. At the hardware level, all pipelines and components in contact with the gas are required to undergo electrolytic polishing and use perfluoroelastomer rubber seals, fundamentally preventing metal ion contamination and volatile organic compound release, ensuring the ultra-high cleanliness of the electronic specialty gases. At the control level, a multi-dimensional compensation algorithm based on gas flow rate, temperature, and pressure is introduced, enabling real-time correction of density and flow rate changes caused by fluctuations in operating conditions. This overcomes the measurement errors caused by traditional mass flow controllers (MFCs) under temperature and pressure variations, ensuring the extreme stability of the mixing ratio during long-term operation. Attached Figure Description

[0037] Figure 1 This is a flowchart of the method of the present invention;

[0038] Figure 2 This is a flowchart of the real-time spectral analysis process.

[0039] Figure 3 This is a system principle block diagram of the present invention. Detailed Implementation

[0040] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] See Figures 1-3 The first aspect of this invention provides: a method for dynamic proportioning adjustment of a mixing chamber for electronic special gases, comprising the following steps:

[0042] Flow preset and gas input stage: The input flow of various electronic special gases is controlled by multiple mass flow controllers in the mixing cabinet, so that they enter the mixing device according to the initial input ratio;

[0043] Real-time spectral analysis stage: Using a tunable semiconductor laser absorption spectral sensor integrated on the mixing device, the concentration of each component of the mixed gas is monitored online in real time, and real-time concentration data is obtained;

[0044] Dynamic optimization phase: The real-time concentration data is compared with the preset target concentration value at the current time node to obtain the real-time concentration deviation. A model predictive control algorithm is used to predict the concentration deviation trend at subsequent time nodes based on the initial input ratio, real-time concentration data, and real-time concentration deviation. Then, the optimized flow rate setpoints for each component gas at subsequent time nodes are dynamically generated based on the concentration deviation trend. The optimization effect is verified by sampling at preset intervals to reduce the number of calculations. If the real-time concentration deviation at the sampling time node is less than or equal to the predicted concentration deviation at that time node in the concentration deviation trend obtained in the previous optimization, the optimized flow rate setpoints generated in the previous optimization are used. If the real-time concentration deviation at the sampling time node is greater than the predicted concentration deviation at that time node in the concentration deviation trend obtained in the previous optimization, dynamic optimization is performed again.

[0045] Flow compensation stage: Based on the optimized flow setpoint, the opening of the corresponding mass flow controller is adjusted in combination with the real-time collected gas flow rate, gas temperature and gas pressure data to compensate for the ratio deviation caused by differences in gas diffusion coefficient, temperature fluctuation and pressure fluctuation.

[0046] In this embodiment, the specific implementation of the present invention is described in detail using the preparation of a neon / krypton / fluorine mixed gas for EUV lithography as an example. This mixed gas requires neon as the background gas, accounting for approximately 98%, and krypton and fluorine as low-concentration key components, accounting for approximately 1.5% and 0.5% respectively. The mixing accuracy is extremely high (error of low-concentration components ≤ ±2%), and the mixing must be completed rapidly to match the production line cycle time.

[0047] The mixing unit system mainly includes: a gas path unit with three independent electronic specialty gas supply lines, corresponding to neon, krypton, and fluorine, respectively. Each line is equipped with a primary pressure reducing valve and a secondary pressure reducing valve upstream, forming a two-stage pressure stabilization to ensure the inlet pressure remains stable at 0.7MPa ± 0.5%. After the pressure reducing valves, a check valve (to prevent gas backflow) and a high-precision mass flow controller (MFC) are installed in sequence. The MFC's range is selected according to the preset proportions of each gas and is initially calibrated based on its molecular weight.

[0048] Mixing and Detection Unit: This unit includes a vertical mixing tank. Three gas outlets extend into the mixing tank via porous jet pipes. To achieve efficient mixing, the nozzles for the low-density neon gas lines are primarily located in the lower part of the mixing tank, while the nozzles for the high-density krypton and fluorine gas lines are primarily located in the upper part. The nozzle sizes are differentiated according to the gas density. The mixing tank outlet pipe integrates a TDLAS (Tunable Semiconductor Laser Absorption Spectroscopy) sensor. The laser emitter and receiver of this sensor are connected to the mixing gas flow channels via a perfluoroether rubber-sealed inert gas purge protection window, ensuring the optical components are clean and unaffected by ambient air interference. The TDLAS sensor is wavelength-tuned for the characteristic absorption lines of krypton and fluorine gases (especially their isolated absorption peaks), with a sampling frequency set to 20Hz to capture transient concentration changes of low-concentration components in real time.

[0049] Control Unit: The core is an industrial-grade programmable logic controller (PLC) or industrial computer (IPC), which incorporates the aforementioned model predictive control (MPC) algorithm and temperature-pressure compensation algorithm. The PLC / IPC is electrically connected to all MFCs, TDLAS sensors, and temperature and pressure sensors installed inside the mixing tank.

[0050] In some embodiments, each electronic special gas pipeline of the mixing cabinet is equipped with a two-stage pressure reducing valve and a check valve; the flow control of the mass flow controller is initially calibrated based on the initial input ratio and gas molecular weight, and the pressure of each pipeline is controlled by the two-stage pressure reducing valve to prevent gas backflow and pressure fluctuation.

[0051] In some embodiments, the inner walls of all pipes, valves, and mixing devices in contact with the electronic specialty gas are electropolished to prevent metal ion contamination.

[0052] In this embodiment, flow rate preset and gas input: The operator inputs the target ratio (Ne: 98%, Kr: 1.5%, F2: 0.5%) via a human-machine interface (HMI). Based on this ratio and the MFC calibration parameters, the control system calculates the initial flow rate setpoints for each gas and sends instructions to the corresponding MFC. The MFC precisely controls the valve opening, ensuring that the three electronic specialty gases enter the mixing tank at the initially set flow rates. Two-stage pressure reducing valves and check valves on each pipeline work together to ensure stable inlet pressure and prevent pressure fluctuations and gas cross-contamination.

[0053] In some embodiments, the laser beam path of the tunable semiconductor laser absorption spectroscopy sensor passes through an inert gas-purged protective window and the outlet gas flow channel of the mixing device, and the wavelength is tuned to select isolated peaks of the molecular absorption spectrum of the low-concentration component gas, thereby capturing transient changes in concentration in real time.

[0054] In some embodiments, the protective window and the mixing device are sealed with perfluoroether rubber to prevent outside air from entering the analytical optical path while ensuring the cleanliness of the electronic special gas.

[0055] In this embodiment, the real-time spectral analysis stage further includes the following steps:

[0056] Signal acquisition and baseline correction steps: The laser beam emitted by the tunable semiconductor laser absorption spectroscopy (TDLAS) sensor passes through the airflow in the mixing device, and the sensor receiver acquires the raw spectral signal containing the characteristic absorption information of the target gas; First, the raw spectral signal is baseline corrected by selecting a specific wavelength region without the absorption peak of the target gas as the reference baseline to eliminate the influence of background noise and instrument drift on the measurement results;

[0057] Feature extraction and peak identification steps: Analyze the baseline-corrected spectral signal, identify the characteristic absorption peaks corresponding to different electronic characteristic gas components, and record the peak intensity and position information of each characteristic absorption peak;

[0058] Concentration inversion calculation steps: Input the peak intensity and position information of the characteristic absorption peak into the pre-trained multi-output least squares support vector machine regression (MLS-SVR) algorithm model. This model is trained based on standard sample spectral data of different electronic characteristic gas components and can simultaneously process the concentration and interference effects of multi-component gases. The real-time concentration data of each component gas is calculated and output through the least squares support vector machine regression (MLS-SVR) algorithm model.

[0059] In this embodiment, signal acquisition involves scanning with a low-frequency triangular wave and simultaneously modulating the wavelength with a high-frequency sine wave. The light intensity signal received by the detector is sampled by a high-precision analog-to-digital converter to obtain the original second harmonic signal containing gas absorption information. Baseline correction is performed beforehand by scanning against a pure nitrogen background without the target gas, storing a standard baseline signal. In actual measurements, the system identifies characteristic wavelength regions far from the gas absorption peak in each scan signal. For example, within the scan range of 1390 nm to 1395 nm, the intervals of 1390.5 nm to 1391 nm and 1394.5 nm to 1394.8 nm, which do not contain absorption peaks, are selected as reference baseline regions. The response of the current scan signal in these regions is fitted to the stored standard baseline using the least squares method to calculate a baseline offset and slope. This baseline is then subtracted from the original signal across the entire scan range, effectively eliminating low-frequency noise caused by optical device drift and background fluctuations.

[0060] Feature extraction and peak identification are performed on the baseline-corrected signal: the system has a built-in database of characteristic absorption peaks of electronic gases, which stores the standard absorption peak positions and shape characteristics of different gas components at specific wavelengths. The algorithm iterates through the corrected signal, calculates its first derivative, and uses a sliding window peak detection algorithm to identify inflection points in the signal, which correspond to the peak positions of the absorption peaks. For each identified peak, the system records its center wavelength and peak intensity. For example, in the detection of a neon / krypton / fluorine mixture, the system simultaneously identifies the characteristic peaks corresponding to krypton and fluorine, and records their data separately. For overlapping or adjacent absorption peaks, the algorithm uses Gaussian fitting or Lorentz fitting for peak deconvolution to accurately separate the contributions of each component.

[0061] Concentration inversion calculation is the core of this stage, and its implementation relies on a pre-trained multi-output least squares support vector machine regression model: Model training: Under laboratory conditions, using a mixture of various electronic specialty gas standard samples with known precise concentrations, a large number of spectral scans are performed under various temperature and pressure conditions to construct a spectral database containing tens of thousands of samples. Each sample data includes the peak intensity, position information, and corresponding true concentration value of each characteristic absorption peak. This database is used to train the MLS-SVR model. The model selects the radial basis function as the kernel function and optimizes the regularization parameter and kernel width through cross-validation to ensure the model's generalization ability and prediction accuracy. Real-time calculation: The extracted real-time peak intensity and position information (as a multi-dimensional feature vector) are input into the pre-trained MLS-SVR model. The advantage of this model lies in its multi-output characteristics, which can simultaneously process the concentration information of multiple gas components and effectively learn the mutual interference relationship between the absorption peaks of different components. After receiving the input vector, the model performs internal calculations and directly outputs a vector containing the concentration values ​​of each component, such as [Ne concentration, Kr concentration, F2 concentration]. This process achieves a nonlinear mapping from spectral features to concentration values, and features high computational efficiency and strong resistance to cross-interference.

[0062] In an embodiment for preparing a neon / krypton / fluorine mixed gas for EUV lithography, as the mixed gas flows through the detection area of ​​a TDLAS sensor, a laser beam of a specific wavelength emitted by the sensor passes through the gas. Krypton and fluorine molecules absorb light of specific wavelengths, causing light intensity attenuation. The TDLAS sensor detects the transmitted light intensity signal in real time and performs concentration inversion calculations based on the Beer-Lambert Law. To obtain high-precision data, the system performs real-time conditional filtering on the raw spectral signal, such as low-pass filtering to remove background noise and setting thresholds to filter out invalid spectral signals. The calculated real-time concentration data of krypton and fluorine are uploaded to the control unit at a frequency of 20 Hz.

[0063] In some embodiments, the model predictive control algorithm constructs a predictive model based on a gas diffusion dynamics model and a thermodynamic equation of state. Its objective function is to minimize the integral sum of squares of the concentration deviation over a future period of time. The algorithm calculates the optimal flow rate setpoint sequence for each component gas through rolling optimization and uses the first element of the sequence as an immediate control command.

[0064] In some embodiments, the gas diffusion kinetics model includes a diffusion coefficient difference compensation term for each component gas. The diffusion coefficient difference compensation term is dynamically adjusted according to the residence time of the gas in the mixing device, wherein the diffusion compensation weight of the low-concentration component is higher than that of the high-concentration component.

[0065] In this embodiment, dynamic optimization (MPC control) is employed: the control unit compares the real-time concentration data (e.g., Kr: 1.52%, F2: 0.48%) fed back by TDLAS with the target concentration to obtain the real-time concentration deviation. The built-in Model Predictive Control (MPC) algorithm then begins operation. This algorithm is based on a predictive model incorporating gas diffusion kinetics and thermodynamic equations of state. This model considers the dynamic mixing behavior of different gases (especially low-concentration components Kr and F2) within the mixing tank due to differences in diffusion coefficients, and assigns higher compensation weights to low-concentration components. The MPC algorithm also predicts the concentration deviation trend over a future period (the specific time is set according to the actual working environment; the longer the set time, the greater the error in the prediction results at subsequent time nodes, requiring comprehensive consideration before setting). Based on this prediction, the algorithm dynamically generates a sequence of optimized flow rate setpoints for each gas component over the future period. To reduce the computational burden on the controller, the system performs sampling verification every preset time interval (not limited to a specific time value, but also determined according to the actual situation, requiring a balance between computational resources and accuracy requirements). If the actual concentration deviation at the sampling point is less than or equal to the predicted value, the current optimized setting value is maintained; if the actual deviation is greater than the predicted value, a new round of MPC optimization calculation is immediately performed to ensure fast and accurate control.

[0066] In some embodiments, the traffic compensation phase further includes the following steps:

[0067] The real-time collected gas velocity, gas temperature, and gas pressure data are input into the ideal gas equation of state, and a diffusion coefficient compensation factor for the target electronic special gas is introduced to construct a flow compensation model with velocity, temperature, and gas pressure as variables. The comprehensive compensation coefficient under the current operating condition is calculated through the flow compensation model, and the optimized flow setpoint is multiplied by the comprehensive compensation coefficient to generate an actual control command to adjust the opening degree of the corresponding mass flow controller.

[0068] In some embodiments, for different electronic specialty gases, the diffusion coefficient compensation factor of the target electronic specialty gas under various combinations of flow rate, temperature and pressure is obtained in advance through experimental calibration and stored in the compensation factor database; when constructing the flow compensation model, the diffusion coefficient compensation factor of the target electronic specialty gas under the current operating condition is queried from the compensation factor database based on the real-time collected gas flow rate, gas temperature and pressure data.

[0069] In this embodiment, the system will first establish a compensation factor database for different electronic specialty gases through experiments. This database records the compensation value corresponding to each gas under various combinations of gas flow rate, temperature, and operating pressure.

[0070] Real-time parameter acquisition: During operation, the system monitors three key parameters in real time: the gas flow rate through the mass flow controller (MFC), the gas temperature in the mixing unit, and the gas pressure data.

[0071] Coefficient Calculation: Based on real-time collected flow rate, temperature, and pressure data, the system calculates the final comprehensive compensation coefficient (i.e., the diffusion coefficient compensation factor of the target electronic specialty gas under current operating conditions) using a weighted geometric mean model. The weighted geometric mean model assigns a weighting factor to each parameter (flow rate, temperature, pressure), and these weighting factors are related to the diffusion characteristics of the target electronic specialty gas. The calculation process essentially involves a weighted comprehensive calculation of the ratios of the real-time values ​​of each parameter to their standard reference values. Multidimensional compensation using real-time flow rate v, real-time temperature T, and real-time pressure P as variables includes the following core steps:

[0072] Compensation factor matching steps: The system has a built-in database of compensation factors for different electronic specialty gases. This database was established through prior experimental calibration and stores the compensation values ​​of the target electronic specialty gas under different combinations of flow rate, temperature, and pressure.

[0073] Calculation of comprehensive compensation coefficient: Using a compensation model based on weighted geometric mean, the comprehensive compensation coefficient K_comp under the current working condition is calculated.

[0074] K_comp = α * (v / v0) k_v + β * (T / T0) k_T + γ * (P / P0) k_P

[0075] Wherein, v0, T0, and P0 are the standard reference values ​​for flow rate, temperature, and pressure, respectively. These are preset and are usually benchmark values ​​measured under ideal and stable calibration conditions in the laboratory (such as a specific flow rate, 25°C temperature, and standard atmospheric pressure), or standard operating points specified by the process. As a benchmark for calculating the relative change ratio, the formula calculates the change ratio of the real-time value to the standard value (such as v / v0), thereby quantifying the degree to which the current operating condition deviates from the standard operating condition.

[0076] α, β, and γ are weighting factors assigned to the three compensation terms of flow rate, temperature, and pressure (α+β+γ=1). They are used to measure the relative importance of the three factors on the proportioning deviation under different physical properties (mainly diffusion coefficient) of electronic specialty gases. For example, for gases with slow diffusion rates and easily affected by flow states, the weight α of flow rate will be set higher; for gases that are sensitive to temperature, the weight β of temperature will be higher. The weighting factors are determined jointly through gas characteristic analysis and experimental calibration.

[0077] k_v, k_T, and k_P are sensitivity indices for flow rate, temperature, and pressure, determined by fitting through a compensation factor database. These indices reflect the sensitivity of changes in these parameters to the final flow rate compensation. Since the effects of flow rate, temperature, and pressure on gas flow rate and mixing behavior are not simple linear relationships, these indices characterize their nonlinear effects. For example, k_v is close to 1 (linear effect), while k_T is greater than 1, indicating that temperature changes have a more sensitive and significant impact on the mixing ratio.

[0078] Command generation: Finally, the optimized flow setpoint is multiplied by the calculated comprehensive compensation coefficient to generate the actual control command sent to the MFC, thereby achieving precise adjustment of the flow valve.

[0079] In some embodiments, the mixing device includes a mixing tank with a porous jet nozzle, wherein nozzles for low-density gas are distributed in the lower part of the mixing tank and nozzles for high-density gas are distributed in the upper part of the mixing tank, and the nozzle size is designed according to the gas density gradient to promote natural diffusion and shorten the mixing time.

[0080] In this embodiment, the core purpose of injecting high-density gas from the top of the mixing tank and low-density gas from the bottom is to utilize the natural diffusion characteristics of the gases themselves to accelerate mixing and shorten mixing time. This approach is not limited to any specific mixing tank structure; any solution employing this design concept is within the scope of protection of this invention.

[0081] In some embodiments, the mixing tank is provided with a stirring element, which adaptively adjusts its working state and movement amplitude according to the density of the mixed gas to enhance the turbulent mixing effect.

[0082] In this embodiment, for application scenarios where the mixing time is too long under certain conditions, an agitator can be optionally installed in the mixing tank to enhance the mixing effect; furthermore, the mixing effect can also be enhanced by increasing the gas flow rate of gases of different densities ejected from the nozzle. The method of increasing the gas flow rate can be used alone or in combination with the agitator, depending on the specific engineering situation.

[0083] like Figure 3 As shown, the second aspect of the present invention provides: a dynamic proportioning adjustment system for a mixing tank of electronic specialty gases, used to implement any of the above-mentioned dynamic proportioning adjustment methods for a mixing tank of electronic specialty gases, comprising:

[0084] The flow preset and gas input module is used to control the input flow of various electronic special gases through multiple mass flow controllers in the mixing cabinet, so that they enter the mixing device according to the initial input ratio.

[0085] The real-time spectral analysis module is used to monitor the concentration of each component of the mixed gas in real time and obtain real-time concentration data using a tunable semiconductor laser absorption spectral sensor integrated on the mixing device.

[0086] The dynamic optimization module compares real-time concentration data with the preset target concentration value at the current time node to obtain the real-time concentration deviation. It then uses a model predictive control algorithm to predict the concentration deviation trend for subsequent time nodes based on the initial input ratio, real-time concentration data, and real-time concentration deviation. Based on this trend, it dynamically generates optimized flow rate setpoints for each gas component at subsequent time nodes and verifies the optimization effect by sampling at preset intervals to reduce the number of calculations. If the real-time concentration deviation at the sampling time node is less than or equal to the predicted concentration deviation for that time node in the previous optimization trend, the previously generated optimized flow rate setpoint is used. If the real-time concentration deviation at the sampling time node is greater than the predicted concentration deviation for that time node in the previous optimization trend, dynamic optimization is performed again.

[0087] The flow compensation module is used to adjust the opening of the corresponding mass flow controller based on the optimized flow setpoint and the real-time collected gas flow rate, gas temperature and pressure data, in order to compensate for the ratio deviation caused by differences in gas diffusion coefficient, temperature fluctuation and pressure fluctuation.

[0088] A third aspect of the present invention provides: a dynamic proportioning adjustment device for a mixing chamber for electronic specialty gases, comprising a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement any of the above-described dynamic proportioning adjustment methods for a mixing chamber for electronic specialty gases.

[0089] A fourth aspect of the present invention provides: a computer-readable storage medium storing computer-executable instructions, wherein when the computer-executable instructions are loaded and executed by a processor, the above-described dynamic proportioning adjustment method for a mixing chamber for electronic special gases is implemented.

[0090] The fifth aspect of the present invention provides: a computer program product containing instructions, which, when run on a terminal, causes the terminal to execute any of the above-described methods for dynamic proportioning adjustment of mixing chambers for electronic special gases.

[0091] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for dynamic proportioning adjustment of a mixing chamber for electronic specialty gases, characterized in that: Includes the following steps: Flow preset and gas input stage: The input flow of various electronic special gases is controlled by multiple mass flow controllers in the mixing cabinet, so that they enter the mixing device according to the initial input ratio; Real-time spectral analysis stage: Using a tunable semiconductor laser absorption spectral sensor integrated on the mixing device, the concentration of each component of the mixed gas is monitored online in real time, and real-time concentration data is obtained; The real-time spectral analysis stage also includes the following steps: Signal acquisition and baseline correction steps: The laser beam emitted by the tunable semiconductor laser absorption spectroscopy sensor passes through the airflow in the mixing device, and the sensor receiver acquires the original spectral signal containing the characteristic absorption information of the target gas; First, the original spectral signal is baseline corrected by selecting a specific wavelength region without the absorption peak of the target gas as the reference baseline to eliminate the influence of background noise and instrument drift on the measurement results. Feature extraction and peak identification steps: Analyze the baseline-corrected spectral signal, identify the characteristic absorption peaks corresponding to different electronic characteristic gas components, and record the peak intensity and position information of each characteristic absorption peak; Concentration inversion calculation steps: The peak intensity and position information of the characteristic absorption peak are input into a pre-trained multi-output least squares support vector machine regression algorithm model. The multi-output least squares support vector machine regression algorithm model is trained based on standard sample spectral data of different electronic characteristic gas components and is used to simultaneously process the concentration and interference effects of multi-component gases. The real-time concentration data of each component gas is calculated and output through the multi-output least squares support vector machine regression algorithm model. Dynamic optimization phase: The real-time concentration data is compared with the preset target concentration value at the current time node to obtain the real-time concentration deviation. A model predictive control algorithm is used to predict the concentration deviation trend at subsequent time nodes based on the initial input ratio, real-time concentration data, and real-time concentration deviation. Then, the optimized flow rate setpoints for each component gas at subsequent time nodes are dynamically generated based on the concentration deviation trend. The optimization effect is verified by sampling at preset intervals to reduce the number of calculations. If the real-time concentration deviation at the sampling time node is less than or equal to the predicted concentration deviation at that time node in the concentration deviation trend obtained in the previous optimization, the optimized flow rate setpoints generated in the previous optimization are used. If the real-time concentration deviation at the sampling time node is greater than the predicted concentration deviation at that time node in the concentration deviation trend obtained in the previous optimization, dynamic optimization is performed again. Flow compensation stage: Based on the optimized flow setpoint, the opening of the corresponding mass flow controller is adjusted in combination with the real-time collected gas flow rate, gas temperature and gas pressure data to compensate for the ratio deviation caused by differences in gas diffusion coefficient, temperature fluctuation and pressure fluctuation. The traffic compensation phase also includes the following steps: The real-time collected gas flow rate, gas temperature, and gas pressure data are input into the ideal gas state equation, and the diffusion coefficient compensation factor of the target electronic special gas is introduced to construct a flow compensation model with flow rate, temperature, and gas pressure as variables. The comprehensive compensation coefficient under the current operating condition is calculated through the flow compensation model, and the optimized flow setpoint is multiplied by the comprehensive compensation coefficient to generate actual control commands to adjust the opening degree of the corresponding mass flow controller. For different electronic specialty gases, the diffusion coefficient compensation factor of the target electronic specialty gas under various combinations of flow rate, temperature and pressure is obtained in advance through experimental calibration and stored in the compensation factor database. When constructing the flow compensation model, the diffusion coefficient compensation factor of the target electronic specialty gas under the current operating condition is queried from the compensation factor database based on the real-time collected gas flow rate, gas temperature and pressure data.

2. The method for dynamic proportioning adjustment of a mixing chamber for electronic special gases according to claim 1, characterized in that: Each electronic special gas pipeline in the gas mixing cabinet is equipped with a two-stage pressure reducing valve and a check valve; The mass flow controller's flow control is initially calibrated based on the initial input ratio and gas molecular weight, and the pressure in each pipeline is controlled by two-stage pressure reducing valves to prevent gas backflow and pressure fluctuations.

3. The method for dynamic proportioning adjustment of a mixing chamber for electronic special gases according to claim 2, characterized in that: All pipes, valves, and mixing devices that come into contact with electronic specialty gases have their inner walls electropolished to prevent metal ion contamination.

4. The method for dynamic proportioning adjustment of a mixing chamber for electronic special gases according to claim 1, characterized in that: The laser beam path of the tunable semiconductor laser absorption spectroscopy sensor passes through an inert gas-purged protective window and the outlet gas flow channel of the mixing device, and the wavelength is tuned to select isolated peaks of the molecular absorption spectrum of the low-concentration component gas, thereby capturing transient changes in concentration in real time.

5. The method for dynamic proportioning adjustment of a mixing chamber for electronic special gases according to claim 4, characterized in that: The protective window and the mixing device are sealed with perfluoroether rubber to prevent outside air from entering the analytical optical path and to ensure the cleanliness of the electronic special gas.

6. The method for dynamic proportioning adjustment of a mixing chamber for electronic special gases according to claim 1, characterized in that: The model predictive control algorithm described above is based on a gas diffusion dynamics model and a thermodynamic equation of state to construct a predictive model. Its objective function is to minimize the integral sum of squares of the concentration deviation over a future period of time. It calculates the optimal flow setpoint sequence for each component gas through rolling optimization and uses the first element of the sequence as an immediate control command.

7. The method for dynamic proportioning adjustment of a mixing chamber for electronic special gases according to claim 6, characterized in that: The gas diffusion kinetics model includes a diffusion coefficient difference compensation term for each component gas. This diffusion coefficient difference compensation term is dynamically adjusted according to the residence time of the gas in the mixing device, wherein the diffusion compensation weight of the low-concentration component is higher than that of the high-concentration component.

8. The method for dynamic proportioning adjustment of a mixing chamber for electronic special gases according to any one of claims 1-7, characterized in that: The mixing device includes a mixing tank with a porous jet pipe, wherein the nozzles for low-density gas are distributed in the lower part of the mixing tank and the nozzles for high-density gas are distributed in the upper part of the mixing tank, and the nozzle size is designed according to the gas density gradient to promote natural diffusion and shorten the mixing time.

9. The method for dynamic proportioning adjustment of a mixing chamber for electronic special gases according to claim 8, characterized in that: The mixing tank is equipped with a stirring element, which adaptively adjusts its working state and movement amplitude according to the density of the mixed gas to enhance the turbulent mixing effect.

10. A dynamic proportioning adjustment system for a mixing chamber used for electronic special gases, characterized in that: The method for dynamically adjusting the mixing ratio of a gas mixing chamber for electronic special gases as described in any one of claims 1-9 includes: The flow preset and gas input module is used to control the input flow of various electronic special gases through multiple mass flow controllers in the mixing cabinet, so that they enter the mixing device according to the initial input ratio. The real-time spectral analysis module is used to monitor the concentration of each component of the mixed gas in real time and obtain real-time concentration data using a tunable semiconductor laser absorption spectral sensor integrated on the mixing device. The dynamic optimization module compares real-time concentration data with the preset target concentration value at the current time node to obtain the real-time concentration deviation. It then uses a model predictive control algorithm to predict the concentration deviation trend for subsequent time nodes based on the initial input ratio, real-time concentration data, and real-time concentration deviation. Based on this trend, it dynamically generates optimized flow rate setpoints for each gas component at subsequent time nodes and verifies the optimization effect by sampling at preset intervals to reduce the number of calculations. If the real-time concentration deviation at the sampling time node is less than or equal to the predicted concentration deviation for that time node in the previous optimization trend, the previously generated optimized flow rate setpoint is used. If the real-time concentration deviation at the sampling time node is greater than the predicted concentration deviation for that time node in the previous optimization trend, dynamic optimization is performed again. The flow compensation module is used to adjust the opening of the corresponding mass flow controller based on the optimized flow setpoint and the real-time collected gas flow rate, gas temperature and pressure data, in order to compensate for the ratio deviation caused by differences in gas diffusion coefficient, temperature fluctuation and pressure fluctuation.

11. A dynamic proportioning adjustment device for a mixing chamber of electronic specialty gases, comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that: When the processor executes the computer program, it implements the dynamic proportioning adjustment method for the mixing chamber of electronic special gases as described in any one of claims 1-9.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when loaded and executed by a processor, implement the dynamic proportioning adjustment method for a mixing chamber for electronic special gases as described in any one of claims 1-9.

13. A computer program product containing instructions, characterized in that: When the computer program product is run on a terminal, the terminal executes the dynamic proportioning adjustment method for a mixing chamber for electronic special gases as described in any one of claims 1-9.

Images