Adjusting chamber pressure parameter fitting algorithm based on control valve
By using a nonlinear parameter fitting algorithm based on control valves, the accuracy and adaptability issues of chamber pressure control in semiconductor wafer processing were solved, achieving high-precision automatic modeling and adaptive adjustment, thus improving control performance and response speed.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JINGJIANG JIASHENG VACUUM TECH CO LTD
- Filing Date
- 2026-01-23
- Publication Date
- 2026-05-15
AI Technical Summary
Existing technologies for chamber pressure control in semiconductor wafer processing suffer from problems such as reliance on manually tuned parameters, inaccurate linear approximation models, and low efficiency of trial-and-error optimization, making it difficult to achieve high-precision modeling and rapid convergence control.
A nonlinear parameter fitting algorithm based on control valves is adopted to achieve automatic modeling and adaptive adjustment of chamber pressure through data acquisition and preprocessing, nonlinear parameter fitting and adaptive controller design. Combined with multinomial function fitting and Levenberg-Marquardt algorithm to optimize parameters, it supports multi-mode control strategies and fault diagnosis.
It improves the accuracy and adaptability of chamber pressure control, reduces the need for manual tuning, achieves faster response speed and lower steady-state error, and is suitable for various vacuum process control fields.
Smart Images
Figure CN122043914A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of process control technology for semiconductor high vacuum equipment, and in particular to a fitting algorithm for pressure parameters of a control valve regulating chamber. Background Technology
[0002] In semiconductor wafer fabrication processes (such as etching, deposition, and ion implantation), the stability of chamber pressure directly affects process uniformity and product yield. Typically, chamber pressure is regulated through the combined action of a mechanical pump, a gas mass flow controller (MFC), and a vacuum valve.
[0003] The existing control methods mainly have the following problems:
[0004] 1. Experience-based PID control relies on manual parameter tuning, making it difficult to adapt to changes in process conditions;
[0005] 2. Linear approximation models cannot accurately describe the nonlinear pressure dynamics in actual chambers;
[0006] 3. The trial-and-error optimization method is inefficient and cannot achieve online parameter updates.
[0007] Because the dynamics of chamber pressure are nonlinear and have multiple time delays, existing methods struggle to maintain high-precision modeling and fast convergence control under different process conditions. Summary of the Invention
[0008] The purpose of this application is to provide a pressure parameter fitting algorithm based on the regulating chamber of a control valve. By combining a high-precision mathematical model with a nonlinear least squares fitting method, the pressure response curve of the chamber is automatically modeled and key dynamic parameters are extracted to achieve better pressure control.
[0009] The technical solution provided in this application is a fitting algorithm for pressure parameters of a control valve regulating chamber, which adopts the following approach:
[0010] It includes three core modules:
[0011] The data acquisition and preprocessing module is configured to acquire the input signal of the control valve and the output signal of the chamber pressure, and perform noise filtering, normalization and data segmentation processing.
[0012] It is responsible for real-time acquisition of control valve opening signals and chamber pressure sensor output signals, eliminating noise interference through digital filtering technology, normalizing the data, and dividing the data into steady-state and dynamic segments according to pressure response characteristics, so as to provide high-quality input data for subsequent modeling.
[0013] The nonlinear parameter fitting module, connected to the data acquisition and preprocessing module, is configured to automatically model the chamber pressure response curve using a high-precision mathematical model and extract key dynamic parameters using a nonlinear least squares fitting method.
[0014] A multinomial function superposition model is used to accurately describe the dynamic characteristics of chamber pressure. The improved Levenberg-Marquardt algorithm is used for parameter optimization, and key parameters such as gain coefficient and time constant are automatically extracted. The model fit evaluation is also provided.
[0015] An adaptive controller design module, connected to the nonlinear parameter fitting module, is configured to dynamically update the fitted parameters to the chamber pressure controller to achieve adaptive adjustment of the chamber pressure.
[0016] The fitted parameters are dynamically updated to the pressure controller, and the control parameters are automatically adjusted according to different process conditions to achieve precise and rapid adjustment of the chamber pressure. It also has fault diagnosis and early warning functions.
[0017] The algorithm also includes:
[0018] The real-time monitoring module is configured to continuously monitor the chamber pressure control effect and trigger a remodeling process when the control deviation exceeds a threshold.
[0019] The process parameter adaptation module is configured to store the optimal parameter combinations corresponding to different process formulas, enabling rapid parameter retrieval during process switching.
[0020] The nonlinear parameter fitting module is also configured to:
[0021] It supports multiple model structure options, including Gaussian function superposition model, exponential decay model, etc.
[0022] Provides model fit evaluation metrics and automatically selects the optimal model structure;
[0023] It supports real-time visualization of the parameter fitting process and monitoring of the convergence status.
[0024] The algorithm also includes:
[0025] The fault diagnosis module is configured to identify abnormal states such as control valve performance degradation and chamber leakage by analyzing the residual characteristics of parameter fitting.
[0026] The early warning module is configured to generate early warning information and provide maintenance suggestions when an anomaly is detected.
[0027] The adaptive controller design module is also configured to:
[0028] It supports multi-mode control strategies, including PID control and model predictive control;
[0029] Automatically select the optimal control strategy according to process requirements;
[0030] Used to provide self-tuning and optimization functions for control parameters.
[0031] In summary, this application includes at least one of the following beneficial technical effects:
[0032] 1. High modeling accuracy: The multinomial function fitting method can more accurately describe nonlinear pressure dynamics. This multinomial function fitting method has achieved a major breakthrough in modeling accuracy. The traditional linear approximation model simplifies the complex chamber pressure dynamics to a first- or second-order system with a single time constant. However, the pressure changes in the actual semiconductor process chamber involve the coupling of multiple physical processes such as gas molecule flow, valve nonlinearity, and chamber geometric effects.
[0033] 2. High adaptability: It can automatically update parameters under different process conditions, reducing the need for manual tuning. The adaptive capability of this algorithm is reflected in the multi-level parameter optimization mechanism. Traditional PID control requires manual parameter tuning for different process formulations (such as etching, deposition, ion implantation), which is time-consuming and relies on expert experience.
[0034] 3. Superior control performance: It outperforms traditional PID control in terms of response speed and steady-state error;
[0035] 4. Good versatility: The algorithm can be extended to other vacuum process control fields such as OLED, photovoltaic coating, and vacuum metallurgy. Attached Figure Description
[0036] Figure 1 This is a flowchart of the algorithm for the data acquisition and preprocessing stages of this application;
[0037] Figure 2 This is a flowchart of the algorithm for the nonlinear parameter fitting stage of this application;
[0038] Figure 3 This is a flowchart of the algorithm during the design phase of the adaptive controller in this application. Detailed Implementation
[0039] The present application will be further described in detail below with reference to the accompanying drawings.
[0040] An algorithm for fitting pressure parameters in the regulating chamber of a control valve, referring to... Figure 1 , Figure 2 and Figure 3 ,include:
[0041] The data acquisition and preprocessing module is configured to acquire the input signal of the control valve and the output signal of the chamber pressure, and perform noise filtering, normalization and data segmentation processing.
[0042] The nonlinear parameter fitting module, connected to the data acquisition and preprocessing module, is configured to automatically model the chamber pressure response curve using a high-precision mathematical model and extract key dynamic parameters using a nonlinear least squares fitting method.
[0043] The adaptive controller design module is connected to the nonlinear parameter fitting module and is configured to dynamically update the fitted parameters to the chamber pressure controller to achieve adaptive adjustment of the chamber pressure.
[0044] The key parameters obtained from the fitting are updated in real time to the Model Predictive Controller (MPC), which automatically selects the optimal control strategy based on the current process formulation. The system has a built-in process parameter knowledge base that stores the optimal parameter combinations for different gas types and flow ranges, enabling seamless transitions during process switching.
[0045] The specific configuration of the data acquisition and preprocessing module is as follows:
[0046] Real-time acquisition of control valve opening signals and chamber pressure sensor output signals;
[0047] A digital filter is used to remove noise from the acquired signal, and the data is normalized.
[0048] The data is divided into steady-state and dynamic segments based on the pressure response characteristics, providing standardized data input for modeling.
[0049] This module uses high-precision sensors to collect control valve opening signals and chamber pressure values in real time. The sampling frequency is no less than 100Hz. The collected raw data is first filtered for noise by a Butterworth low-pass filter with a cutoff frequency of 10Hz. Then, the data is normalized, mapping the pressure value to the [0,1] interval and the valve opening to the [0,100%] range.
[0050] The data processing stage employs sliding window technology to identify dynamic feature points of the pressure response in real time, automatically dividing the data into steady-state and transitional segments, providing targeted data support for modeling at different stages.
[0051] The nonlinear parameter fitting module is specifically configured as follows:
[0052] A multinomial superposition model of chamber pressure is established, and the model is expressed as:
[0053]
[0054] in, , These are the parameters to be fitted to the model;
[0055] Construct the residual function: ;
[0056] The Levenberg-Marquardt nonlinear least squares algorithm is used for parameter optimization to obtain the optimal parameter vector. .
[0057] The residual sum of squares is calculated in real time during the fitting process, and the model is considered to have converged when the goodness of fit R² reaches 0.99 or higher.
[0058] The adaptive controller design module is specifically configured as follows:
[0059] The key dynamic parameters obtained from the fitting are updated to the pressure controller parameter library in real time.
[0060] The controller parameters are automatically adjusted based on the current process conditions and gas flow rate.
[0061] By precisely adjusting the opening of the control valve, rapid and stable control of the chamber pressure can be achieved.
[0062] The key parameters obtained from the fitting are updated in real time to the Model Predictive Controller (MPC), which automatically selects the optimal control strategy based on the current process formulation. The system has a built-in process parameter knowledge base that stores the optimal parameter combinations for different gas types and flow ranges, enabling seamless transitions during process switching.
[0063] The algorithm also includes:
[0064] The real-time monitoring module is configured to continuously monitor the chamber pressure control effect and trigger a remodeling process when the control deviation exceeds a threshold.
[0065] The process parameter adaptation module is configured to store the optimal parameter combinations corresponding to different process formulas, enabling rapid parameter retrieval during process switching.
[0066] The nonlinear parameter fitting module is also configured as follows:
[0067] It supports multiple model structure options, including Gaussian function superposition model, exponential decay model, etc.
[0068] Provides model fit evaluation metrics and automatically selects the optimal model structure;
[0069] It supports real-time visualization of the parameter fitting process and monitoring of the convergence status.
[0070] The algorithm also includes:
[0071] The fault diagnosis module is configured to identify abnormal states such as control valve performance degradation and chamber leakage by analyzing the residual characteristics of parameter fitting.
[0072] The early warning module is configured to generate early warning information and provide maintenance suggestions when an anomaly is detected.
[0073] The adaptive controller design module is also configured as follows:
[0074] It supports multi-mode control strategies, including PID control and model predictive control;
[0075] Automatically select the optimal control strategy according to process requirements;
[0076] Used to provide self-tuning and optimization functions for control parameters.
Claims
1. An algorithm for fitting pressure parameters in a control valve regulating chamber, characterized in that, include: The data acquisition and preprocessing module is configured to acquire the input signal of the control valve and the output signal of the chamber pressure, and perform noise filtering, normalization and data segmentation processing. The nonlinear parameter fitting module, connected to the data acquisition and preprocessing module, is configured to automatically model the chamber pressure response curve using a high-precision mathematical model and extract key dynamic parameters using a nonlinear least squares fitting method. An adaptive controller design module, connected to the nonlinear parameter fitting module, is configured to dynamically update the fitted parameters to the chamber pressure controller, thereby achieving adaptive adjustment of the chamber pressure.
2. The algorithm for fitting pressure parameters of a control valve regulating chamber according to claim 1, characterized in that, The data acquisition and preprocessing module is specifically configured as follows: Real-time acquisition of control valve opening signals and chamber pressure sensor output signals; A digital filter is used to remove noise from the acquired signal, and the data is normalized. The data is divided into steady-state and dynamic segments based on the pressure response characteristics, providing standardized data input for modeling.
3. The algorithm for fitting pressure parameters of a control valve regulating chamber according to claim 1, characterized in that, The nonlinear parameter fitting module is specifically configured as follows: A multinomial superposition model of chamber pressure is established, and the model is expressed as: in, , These are the parameters to be fitted to the model; Construct the residual function: ; The Levenberg-Marquardt nonlinear least squares algorithm is used for parameter optimization to obtain the optimal parameter vector. .
4. The algorithm for fitting pressure parameters of a control valve regulating chamber according to claim 1, characterized in that, The adaptive controller design module is specifically configured as follows: The key dynamic parameters obtained from the fitting are updated to the pressure controller parameter library in real time. The controller parameters are automatically adjusted based on the current process conditions and gas flow rate.
5. The algorithm for fitting pressure parameters of a control valve regulating chamber according to claim 1, characterized in that, The algorithm also includes: The real-time monitoring module is configured to continuously monitor the chamber pressure control effect and trigger a remodeling process when the control deviation exceeds a threshold. The process parameter adaptation module is configured to store the optimal parameter combinations corresponding to different process formulas, enabling rapid parameter retrieval during process switching.
6. The algorithm for fitting pressure parameters of a control valve regulating chamber according to claim 1, characterized in that, The nonlinear parameter fitting module is also configured to: It supports multiple model structure options, including Gaussian function superposition model, exponential decay model, etc. Provides model fit evaluation metrics and automatically selects the optimal model structure; It supports real-time visualization of the parameter fitting process and monitoring of the convergence status.
7. The algorithm for fitting pressure parameters of a control valve regulating chamber according to claim 1, characterized in that, The algorithm also includes: The fault diagnosis module is configured to identify abnormal states such as control valve performance degradation and chamber leakage by analyzing the residual characteristics of parameter fitting. The early warning module is configured to generate early warning information and provide maintenance suggestions when an anomaly is detected.
8. The algorithm for fitting pressure parameters of a control valve regulating chamber according to claim 1, characterized in that, The adaptive controller design module is also configured to: It supports multi-mode control strategies, including PID control and model predictive control; Automatically select the optimal control strategy according to process requirements; Used to provide self-tuning and optimization functions for control parameters.