Self-adaptive regulation and control method and system for radio frequency power of remote plasma source

Abstract

The invention discloses a self-adaptive regulation and control method and system for radio frequency power of a remote plasma source, and relates to the technical field of self-adaptive regulation and control, and the method comprises the steps: deploying a multi-parameter sensor network, and collecting and obtaining a discharge state parameter flow; presetting a target power curve, identifying and analyzing the discharge state parameter flow and the target power curve, and outputting a load state trend parameter; determining a radio frequency power regulation and control evaluation rule; constructing a radio frequency power adaptive regulation and control channel; and performing regulation and control calculation on the target power curve and the load state trend parameter by adopting a radio frequency power self-adaptive regulation and control channel, determining a radio frequency power correction amount, and performing power self-adaptive regulation and control on the remote plasma source system based on the radio frequency power correction amount. The technical problems that in the prior art, the radio frequency power regulation and control precision is insufficient, and load changes cannot be adapted in real time are solved, and the technical effects of improving the radio frequency power regulation and control precision and stability and improving the plasma source operation efficiency are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of adaptive control, and in particular to a method and system for adaptively controlling radio frequency power of a remote plasma source. Background Art

[0002] The power control accuracy of an RF plasma source is crucial to its stable operation and efficiency. Traditional RF power regulation methods rely on preset fixed parameters or simple feedback mechanisms, which are unable to accurately adjust the power in response to load fluctuations. When the discharge state fluctuates, traditional methods have difficulty dynamically adjusting the RF power according to actual load changes. This results in low power utilization efficiency and an inability to respond promptly to load changes caused by different process requirements, thus affecting overall efficiency. Summary of the Invention

[0003] The present application provides a method and system for adaptively controlling the radio frequency power of a remote plasma source, which is used to solve the technical problems in the prior art of insufficient radio frequency power control accuracy and inability to adapt to load changes in real time.

[0004] In view of the above problems, the present application provides a method and system for adaptively controlling the radio frequency power of a remote plasma source.

[0005] In a first aspect of the present application, a method for adaptively controlling radio frequency power of a remote plasma source is provided, the method comprising: A multi-parameter sensor network is deployed in a remote plasma source system, and a discharge state parameter stream is collected and acquired through the multi-parameter sensor network; a target power curve is preset according to target process requirements, the discharge state parameter stream and the target power curve are identified and analyzed, and a load state trend parameter is output; hierarchical indicator decomposition and evaluation factor fitting are performed on the radio frequency power control target to determine a radio frequency power control evaluation rule; the radio frequency power control evaluation rule is introduced to perform power control learning based on the radio frequency power historical working data set of the remote plasma source system, and a radio frequency power adaptive control channel is constructed; the radio frequency power adaptive control channel is used to perform control calculations on the target power curve and the load state trend parameter, determine a radio frequency power correction amount, and perform power adaptive control on the remote plasma source system based on the radio frequency power correction amount.

[0006] A second aspect of the present application provides an adaptive control system for radio frequency power of a remote plasma source, the system comprising: A parameter acquisition module is used to deploy a multi-parameter sensor network in the remote plasma source system and acquire a discharge state parameter stream through the multi-parameter sensor network; an identification and analysis module is used to preset a target power curve according to target process requirements, identify and analyze the discharge state parameter stream and the target power curve, and output a load state trend parameter; a rule determination module is used to perform hierarchical indicator decomposition and evaluation factor fitting on the RF power control target to determine the RF power control evaluation rule; a control channel construction module is used to introduce the RF power control evaluation rule to perform power control learning based on the RF power historical working data set of the remote plasma source system and construct an RF power adaptive control channel; an adaptive control module is used to use the RF power adaptive control channel to perform control calculations on the target power curve and the load state trend parameter, determine the RF power correction amount, and perform power adaptive control on the remote plasma source system based on the RF power correction amount.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: The present application deploys a multi-parameter sensor network in a remote plasma source system, collects and obtains a discharge state parameter stream through the multi-parameter sensor network; presets a target power curve according to target process requirements, identifies and analyzes the discharge state parameter stream and the target power curve, and outputs a load state trend parameter; performs hierarchical indicator decomposition and evaluation factor fitting on the RF power control target to determine an RF power control evaluation rule; introduces the RF power control evaluation rule to perform power control learning based on the RF power historical working data set of the remote plasma source system, and constructs an RF power adaptive control channel; uses the RF power adaptive control channel to perform control calculation on the target power curve and the load state trend parameter, determines an RF power correction amount, and performs power adaptive control on the remote plasma source system based on the RF power correction amount. The present invention solves the technical problems of insufficient RF power control accuracy and inability to adapt to load changes in the prior art. By using a multi-parameter sensor network to collect discharge state parameters in real time and dynamically controlling them in combination with the target power curve, the present invention achieves the technical effect of improving RF power control accuracy and stability and improving plasma source operation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1A flow chart of a method for adaptively controlling radio frequency power of a remote plasma source provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of an adaptive control system for remote plasma source radio frequency power provided in an embodiment of the present application.

[0010] Explanation of the accompanying symbols: parameter acquisition module 11, identification and analysis module 12, rule determination module 13, control channel construction module 14, adaptive control module 15. DETAILED DESCRIPTION

[0011] The present application provides a method and system for adaptively controlling the RF power of a remote plasma source, aiming to solve the technical problems in the prior art of insufficient RF power control accuracy and inability to adapt to load changes in real time. By adopting a multi-parameter sensor network to collect discharge state parameters in real time and dynamically controlling them in combination with the target power curve, the technical effect of improving the RF power control accuracy and stability and enhancing the operating efficiency of the plasma source is achieved.

[0012] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0013] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0014] Example 1, as Figure 1 As shown, the present application provides a method for adaptively controlling the radio frequency power of a remote plasma source, the method comprising: Step S100: deploying a multi-parameter sensor network in a remote plasma source system, and acquiring a discharge state parameter flow through the multi-parameter sensor network.

[0015] In the embodiments of the present application, a multi-parameter sensor network is first deployed in a remote plasma source system. This network includes multiple sensor types, such as voltage sensors, current sensors, and gas flow sensors. During the operation of the plasma source, the multi-parameter sensor network continuously acquires and collects discharge state parameter streams in real time, including voltage, current, gas flow, and other data.

[0016] Step S200: Preset a target power curve according to target process requirements, identify and analyze the discharge state parameter flow and the target power curve, and output a load state trend parameter.

[0017] In the embodiment of the present application, a preset target process requirement is first obtained, and then a technical expert sets a target power curve based on the target process requirement.

[0018] Next, the discharge state parameter stream and target power curve are identified and analyzed. This process first ensures data consistency and availability by aligning the discharge state parameter stream with the target power curve in time series. Key features are then extracted from the aligned parameter stream to generate static and dynamic discharge state feature sets. These features are used to train load prediction based on the target power curve and construct a load state predictor. Finally, the load state predictor performs identification analysis on the discharge state feature set and outputs load state trend parameters.

[0019] Furthermore, in the method provided in the embodiment of the application, the output load state trend parameter further includes: The discharge state parameter stream is data-aligned with the target power curve in time series to obtain an available discharge state parameter stream; key features are extracted from the available discharge state parameter stream to obtain a discharge state key feature set, wherein the discharge state key feature set includes a static discharge feature set and a dynamic discharge feature set; load prediction training is performed based on the target power curve to construct a load state predictor; the load state predictor is used to identify and analyze the discharge state key feature set to output the load state trend parameter.

[0020] In the embodiment of the present application, the discharge state parameter stream is first aligned with the target power curve in time series. In this process, the discharge state parameter stream and the target power curve are synchronized in time using an interpolation method to obtain a usable discharge state parameter stream.

[0021] Next, wavelet transform is used to extract key features from the available discharge state parameter stream to obtain the discharge state key feature set. In this step, the discharge state key feature set includes a static discharge feature set and a dynamic discharge feature set. The static discharge feature set includes parameters such as voltage and current, while the dynamic discharge feature set extracts features of instantaneous changes such as power fluctuations and gas flow.

[0022] Load prediction training is then performed based on the target power curve. During this process, a historical RF power operating feature set is first obtained through associated data mining and key feature extraction. This feature set is then trained for state classification and identification according to the target power curve to generate a load state classifier. Load trend identification and prediction supervision training are then performed based on the historical RF power operating feature set to obtain a load trend predictor. Finally, the load state classifier and load trend predictor are fused in parallel to construct a load state predictor.

[0023] Finally, the load state predictor is used to identify and analyze the key feature set of the discharge state. That is, the key feature set of the discharge state is input into the load state predictor to identify and analyze the load state trend, and the load state trend parameter is obtained. The load state trend parameter reflects the future change direction of the load.

[0024] Furthermore, in the method provided in the embodiment of the application, the constructing of the load state predictor further includes: Based on the target power curve, associated data mining and key feature extraction are performed to obtain a historical RF power working feature set; state classification identification training is performed on the historical RF power working feature set according to the target power curve to generate a load state classifier; based on the historical RF power working feature set, load trend identification and prediction supervision training are performed to obtain a load trend predictor; the load state classifier and the load trend predictor are fused in parallel to construct the load state predictor.

[0025] In this embodiment, the historical RF power operating feature set is first analyzed through correlation data mining. Correlation data mining identifies key factors influencing RF power, such as voltage, current, and gas flow, by exploring the potential relationships between historical RF power data and the target power curve. This data is then used to extract valuable features using feature extraction methods (e.g., data clustering, association rules, etc.), thereby forming the historical RF power operating feature set.

[0026] Next, based on the target power curve, the RF power historical operating feature set is trained for state classification and identification to generate a load state classifier. This process begins by labeling the RF power historical operating feature set through the identification step, classifying the data into different load states, such as high load, low load, or normal load. These labels are manually annotated. After labeling, classification algorithms such as support vector machines and decision trees are used to train this labeled data to generate a load state classifier.

[0027] Then, based on the RF power historical operating feature set, load trend identification is performed. At this stage, load trend identification is performed using a time series analysis method. A sliding average is used to perform trend analysis on the time series data in the RF power historical operating feature set. This step identifies the load change pattern over time, including trends such as load increases, decreases, or fluctuations. Through these methods, load trend identification can accurately capture the dynamic changes in the load. Then, based on the load trend identification, load trend prediction training is performed to obtain a load trend predictor. In this process, deep learning methods such as long short-term memory networks are used, combined with the identified load trend data, for supervised training to learn the long-term change pattern of the load, and ultimately build a load trend predictor that can predict future load fluctuations.

[0028] Finally, the load state classifier and the load trend predictor are fused in parallel to construct the final load state predictor. During this process, the two predictors run in parallel, providing the current load state and future load trend, respectively. Through parallel fusion, the final load state predictor outputs load state trend parameters.

[0029] Step S300: performing hierarchical index decomposition and evaluation factor fitting on the radio frequency power control target to determine radio frequency power control evaluation rules.

[0030] In an embodiment of the present application, when performing hierarchical index decomposition and evaluation factor fitting for the RF power control target, a target hierarchical architecture is first constructed, and the hierarchical index decomposition of the RF power control target is performed through the architecture to construct an RF power control evaluation index system. On this basis, a critical analysis is performed on each evaluation indicator in the evaluation index system to determine the indicator evaluation factor set. Then, based on the evaluation index system, a hierarchical objective function is designed to obtain a multi-level evaluation objective function. Finally, according to the indicator evaluation factor set, the multi-level evaluation objective function is hierarchically weighted and fused to finally determine the RF power control evaluation rule.

[0031] Furthermore, in the method provided in the embodiment of the application, determining the RF power control evaluation rule further includes: Build a target hierarchical architecture, decompose the hierarchical indicators of the RF power control target according to the target hierarchical architecture, and construct a RF power control evaluation index system; perform key analysis on each evaluation indicator in the RF power control evaluation index system to determine the indicator evaluation factor set; design a hierarchical objective function based on the RF power control evaluation index system to obtain a multi-level evaluation objective function; perform hierarchical weighted fusion of the multi-level evaluation objective function according to the indicator evaluation factor set to determine the RF power control evaluation rule.

[0032] In an embodiment of the present application, a target hierarchical architecture is first established, that is, the RF power control target is decomposed into multiple sub-targets and a hierarchical structure. A hierarchical structure is constructed by analyzing the various dimensions of the RF power control target. For example, the RF power control target can be decomposed into several main sub-targets, such as stability, energy efficiency, and temperature control. These sub-targets are further decomposed into specific evaluation indicators, such as stability can include current stability and voltage fluctuation, and energy efficiency can include power loss rate and conversion efficiency. By using the analytic hierarchy process (AHP), the RF power control target is decomposed layer by layer to ensure that each sub-target and evaluation indicator have a clear structural relationship, thereby building a target hierarchical architecture.

[0033] Next, based on the established hierarchical target architecture, the RF power control target was broken down into its hierarchical indicators. In this step, specific evaluation indicators were identified based on the specific meaning of each sub-target. For example, under the stability sub-target, current stability and voltage fluctuation were the primary evaluation indicators; under the energy efficiency sub-target, power loss rate and conversion efficiency were the key indicators. By analyzing the implementation requirements of each sub-target, the final RF power control evaluation indicator system was determined.

[0034] Next, a criticality analysis is performed on each evaluation indicator in the RF power control evaluation index system to determine the set of indicator evaluation factors. This process uses methods such as multivariate regression analysis or sensitivity analysis to assess the impact of each indicator on the RF power control objectives. For example, regression analysis is used to analyze the impact of voltage fluctuation and current stability on RF power control and determine their contribution to the ultimate objectives (such as stability and energy efficiency). This step identifies the most critical evaluation factors for RF power control and forms the set of indicator evaluation factors.

[0035] Next, based on the RF power control evaluation index system, a hierarchical objective function is designed to obtain a multi-level evaluation objective function. In this step, the evaluation indicators and weights of related factors for each sub-goal are quantified to form a set of mathematical functions. The objective function of each level is composed of multiple evaluation indicators, and each indicator contributes to the value of the objective function according to its weight. For example, in the objective function of the stability sub-goal, the contributions of current stability and voltage fluctuation can be summed according to their preset weights. In this way, a set of multi-level evaluation objective functions is designed to comprehensively reflect the realization of each level goal and quantify the contribution of different evaluation indicators to the final goal.

[0036] Finally, based on the set of indicator evaluation factors, a hierarchical weighted fusion of the multi-level evaluation objective functions is performed to ultimately determine the RF power control evaluation rules. In this step, a weighted average method is used to weight the evaluation objective functions at each level. For example, current stability and voltage fluctuation may have weights of 0.6 and 0.4, respectively, in the stability sub-objective. In the energy efficiency sub-objective, power loss rate and conversion efficiency may have weights of 0.7 and 0.3, respectively. Through the weighted fusion of these evaluation factors, the RF power control evaluation rules are ultimately derived.

[0037] Step S400: introducing the RF power control evaluation rule to perform power control learning based on the RF power historical working data set of the remote plasma source system, and constructing a RF power adaptive control channel.

[0038] In this embodiment, RF power control evaluation rules are first introduced. The power control effect is evaluated based on the RF power historical operating dataset of the remote plasma source system to obtain an RF power control effect set. Based on the RF power control effect set, the RF power historical operating dataset is then optimized to obtain an available RF power operating dataset.

[0039] Then, based on the available RF power data set and the power accuracy control requirements, a power impact threshold is set. Finally, based on the set power impact threshold, branch power control learning is performed on the available RF power data set, training a model that can adaptively adjust RF power and ultimately generating an RF power adaptive control channel.

[0040] Furthermore, in the method provided in the embodiment of the application, the constructing of the RF power adaptive control channel further includes: The RF power control evaluation rule is introduced to evaluate the control effect of the RF power historical working data set of the remote plasma source system to obtain a RF power control effect set; based on the RF power control effect set, the effect of the RF power historical working data set is optimized to obtain an available RF power working data set; according to the available RF power working data set and the power accuracy control requirement, a power influence threshold is set; based on the power influence threshold, branch power control learning is performed on the available RF power working data set to generate a RF power adaptive control channel.

[0041] In this embodiment, RF power control evaluation rules are first introduced and used to evaluate the control effects of a historical RF power operating dataset for a remote plasma source system. This process uses the RF power control evaluation rules to analyze the control effects under each operating condition in the historical RF power data, identifying key factors influencing the RF power control results. These factors, such as power stability and conversion efficiency, are comprehensively analyzed using the evaluation rules to ultimately produce a RF power control effect set, which contains data on the effects of RF power control under different operating conditions.

[0042] Next, based on the RF power control effect set, the historical RF power operating dataset is optimized to obtain a usable RF power operating dataset. This process uses techniques such as data cleaning and feature selection to filter the historical data and select those with a significant impact on RF power control. Specifically, a feature importance assessment method is used to identify key indicators in RF power control, such as current fluctuation and power loss rate, and data samples with minimal impact on power control are removed. After screening, a usable RF power operating dataset is obtained.

[0043] The power impact threshold is then set based on the available RF power working data set and the power precision control requirements. This process first extracts power impact factors and assesses their impact based on the available RF power working data set. The primary influencing factors related to RF power control are identified and their impact on power output is assessed, resulting in the RF power impact factor set and the power factor impact factor set. Next, the power factor impact factor set is weighted and fitted to the RF power impact factor set to generate the RF power impact function. Finally, based on the power precision control requirements, the RF power impact function is divided into impact thresholds to determine the power impact threshold.

[0044] Finally, branch power control learning is performed on the available RF power working data set based on the power impact threshold. If the power impact threshold is not reached, conventional power control learning is performed on the data set using a PID controller to generate a conventional RF power control channel. Once the set power impact threshold is reached, the RF power impact function is introduced to modify the conventional RF power control channel to generate an aggregated RF power control channel. Finally, the conventional RF power control channel and the aggregated RF power control channel are merged to generate the final RF power adaptive control channel.

[0045] Furthermore, in the method provided in the embodiment of the application, setting the power impact threshold further includes: Based on the available RF power working data set, power influencing factors are extracted and the degree of influence is evaluated to obtain a RF power influencing factor set and a power factor influencing factor set; the RF power influencing factor set is weightedly fitted according to the power factor influencing factor set to generate a RF power influence function; based on the power precision control requirement, the RF power influence function is divided into influence thresholds to determine the power influence degree threshold.

[0046] In an embodiment of the present application, power influencing factors are first extracted and their degree of influence is evaluated based on the available RF power working data set. This process extracts the most important influencing factors in RF power regulation through the principal component analysis (PCA) method. Through this method, the influencing factors most relevant to power regulation, such as current fluctuations, temperature changes, etc., are identified from the RF power historical data set. Next, regression analysis (such as linear regression or multivariate regression) is used to evaluate the degree of influence of these extracted influencing factors, and by fitting the relationship between the RF power data and each factor, the influencing factor of each factor is obtained, that is, the quantified contribution of each factor to the RF power regulation effect. These influencing factors reflect the influence of each factor on RF power regulation under different working conditions, and ultimately form a set of RF power influencing factors and a set of power factor influencing factors.

[0047] Next, the RF power influencing factor set is weightedly fitted according to the power factor influencing factor set to generate an RF power influence function. In this step, the weighted least squares method is used to weight the factor weights in each power factor influencing factor set and the corresponding RF power influencing factor. The weighted least squares method comprehensively considers the importance (weight) of each influencing factor by minimizing the weighted residual, generating a mathematical model, namely the RF power influence function, which quantifies the specific impact of each factor on RF power regulation. For example, assuming that the weight of the influence factor of current fluctuation is 0.6 and the weight of the influence factor of temperature fluctuation is 0.4, a comprehensive function is generated through weighted fitting to reflect the role of different influencing factors in RF power regulation. The RF power influence function is used to evaluate the influence degree and obtain the influence degree of the plasma source system.

[0048] Finally, based on the power precision control requirements, the RF power impact function is divided into impact thresholds and a power impact threshold is determined. This process uses Otsu's method to determine the critical threshold of the influencing factors. By maximizing the inter-class variance, Otsu's method automatically analyzes the output of the RF power impact function to determine which influencing factors have a significant impact on RF power control, thereby setting a power impact threshold. If a factor's change exceeds this threshold, it indicates that the factor has a significant impact on power control and requires adjustment.

[0049] Furthermore, in the method provided in the embodiment of the application, the generating of the RF power adaptive control channel further includes: If the power impact threshold is not reached, a PID controller is used to perform power control learning on the available RF power working data set to obtain a conventional RF power control channel; when the power impact threshold is reached, the RF power impact function is introduced to perform impact correction on the conventional RF power control channel to construct an aggregated RF power control channel; the conventional RF power control channel and the aggregated RF power control channel are branched and merged to generate the RF power adaptive control channel.

[0050] In this embodiment of the present application, if the plasma source system influence calculated using the RF power influence function does not reach the power influence threshold, a PID controller is used to learn power control using the available RF power operating data set and obtain a conventional RF power control channel. During this process, the PID controller (proportional, integral, and differential controller) adjusts the RF power output through a real-time feedback mechanism. This controller calculates the error between the current RF power value and the target value and, using a combination of proportional gain, integral gain, and differential gain, adjusts the power output so that the RF power gradually approaches the target power value.

[0051] When the plasma source system influence, calculated using the RF power influence function, reaches the power influence threshold, the RF power influence function is introduced to modify the influence of the conventional RF power control channel, generating an aggregated RF power control channel. At this point, the RF power influence function dynamically modifies the RF power control channel based on the results previously obtained through regression analysis and weighted fitting methods. This function adjusts the control strategy by weighting the effect of each influencing factor, enabling the RF power to more accurately respond to environmental changes and load fluctuations. For example, if current fluctuations significantly affect RF power, they are corrected using the RF power influence function to ensure that the system maintains a high level of power control accuracy even in this environment. This correction results in an aggregated RF power control channel.

[0052] Finally, the conventional RF power control channel and the aggregated RF power control channel are branched and merged to form an adaptive RF power control channel. This step integrates the outputs of the two control channels, ensuring that the conventional RF power control channel can be used for simple adjustments while combining it with the aggregated RF power control channel for more precise corrections. This merging ensures that traditional PID control can be used to handle simple load fluctuations, while corrective measures based on the RF power influence function are introduced to achieve optimal power regulation when encountering complex operating conditions.

[0053] Step S500: using the RF power adaptive control channel to perform control calculations on the target power curve and the load state trend parameter, determine the RF power correction amount, and perform power adaptive control on the remote plasma source system based on the RF power correction amount.

[0054] In this embodiment, the RF power impact function is first used to evaluate the impact of the remote plasma source system's application conditions, thereby obtaining the plasma source system impact. Based on this impact, the RF power adaptive control channel is then activated. Channel matching and control calculations are performed on the target power curve and load state trend parameters to determine the RF power correction amount.

[0055] Based on the calculated RF power correction, the remote plasma source system is then adaptively controlled for power. Specifically, the RF power correction is used as a control signal and input into the power control mechanism of the remote plasma source system. The RF power output is adjusted based on the correction so that the actual output power is closer to the target power curve, ensuring that the remote plasma source system maintains stable operation under different load conditions. For example, if a large power fluctuation is detected and deviates from the target power curve, the RF power correction will be increased to increase the power output, and vice versa. In this way, the RF power correction ensures that the remote plasma source system can dynamically adjust the RF power output under real-time changes in operating conditions to achieve precise adaptive regulation.

[0056] Furthermore, in the method provided in the embodiment of the application, the determining of the radio frequency power correction amount further includes: The RF power influence function is used to evaluate the impact of the application status of the remote plasma source system to obtain the plasma source system impact; based on the plasma source system impact, the RF power adaptive control channel is activated to perform channel matching and control calculation on the target power curve and the load state trend parameter to determine the RF power correction amount.

[0057] In this embodiment, an RF power influence function is first used to evaluate the impact of the remote plasma source system's application conditions, thereby obtaining the plasma source system impact. This process involves substituting various input parameters (such as current, voltage, and gas flow) under current operating conditions into the RF power influence function to quantify the combined impact of these factors on RF power regulation. This calculation yields the plasma source system impact.

[0058] Next, based on the plasma source system's influence, the RF power adaptive control channel is activated. The plasma source system's influence is compared with the power influence threshold. If the power influence threshold is not reached, the conventional RF power control channel within the RF power adaptive control channel is selected. If the plasma source system's influence reaches the power influence threshold, the aggregated RF power control channel within the RF power adaptive control channel is selected. After channel matching is complete, the target power curve and load state trend parameters are input into the corresponding channel for control calculation to obtain the RF power correction.

[0059] Furthermore, the method provided in the application embodiment also includes: The power control delay time of the remote plasma source system is monitored and acquired; a radio frequency power loss coefficient is determined according to the power control delay time, and feedback compensation is performed on the radio frequency power correction amount based on the radio frequency power loss coefficient.

[0060] In the embodiment of the present application, the issuing time of the power adjustment command and the actual time of the power adjustment response are first recorded by timestamp, and the power adjustment delay time is obtained by calculating the difference between the actual time and the issuing time of the power adjustment response.

[0061] Next, the RF power loss coefficient is determined based on the power control delay time. This process is performed using preset rules. Specifically, a set of rules is set that assigns a fixed RF power loss coefficient based on different delay times. For example, when the control delay time is 100 milliseconds, the preset rules set the RF power loss coefficient to 0.1.

[0062] Finally, feedback compensation is performed on the RF power correction value based on the RF power loss coefficient. In this process, the calculated RF power correction value is compensated using the RF power loss coefficient. Specifically, the RF power loss coefficient is multiplied by the RF power correction value to obtain the compensated correction value. For example, if the initially calculated RF power correction value is 10W and the RF power loss coefficient is 0.1 (indicating a 10% power loss due to delay), the compensated RF power correction value is 10W × (1 + 0.1) = 11W.

[0063] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects: The present application deploys a multi-parameter sensor network in a remote plasma source system, collects and obtains a discharge state parameter stream through the multi-parameter sensor network; presets a target power curve according to target process requirements, identifies and analyzes the discharge state parameter stream and the target power curve, and outputs a load state trend parameter; performs hierarchical indicator decomposition and evaluation factor fitting on the RF power control target to determine an RF power control evaluation rule; introduces the RF power control evaluation rule to perform power control learning based on the RF power historical working data set of the remote plasma source system, and constructs an RF power adaptive control channel; uses the RF power adaptive control channel to perform control calculation on the target power curve and the load state trend parameter, determines an RF power correction amount, and performs power adaptive control on the remote plasma source system based on the RF power correction amount. The present invention solves the technical problems of insufficient RF power control accuracy and inability to adapt to load changes in the prior art. By using a multi-parameter sensor network to collect discharge state parameters in real time and dynamically controlling them in combination with the target power curve, the present invention achieves the technical effect of improving RF power control accuracy and stability and improving plasma source operation efficiency.

[0064] Embodiment 2 is based on the same inventive concept as the adaptive control method of the radio frequency power of a remote plasma source in the above embodiment. Figure 2 As shown, the present application provides an adaptive control system for radio frequency power of a remote plasma source. The system and method embodiments in the present application are based on the same inventive concept. The system includes: A parameter acquisition module 11 is used to deploy a multi-parameter sensor network in the remote plasma source system and acquire a discharge state parameter stream through the multi-parameter sensor network; an identification and analysis module 12 is used to preset a target power curve according to target process requirements, identify and analyze the discharge state parameter stream and the target power curve, and output a load state trend parameter; a rule determination module 13 is used to perform hierarchical indicator decomposition and evaluation factor fitting on the RF power control target to determine the RF power control evaluation rule; a control channel construction module 14 is used to introduce the RF power control evaluation rule to perform power control learning based on the RF power historical working data set of the remote plasma source system, and construct an RF power adaptive control channel; an adaptive control module 15 is used to use the RF power adaptive control channel to perform control calculation on the target power curve and the load state trend parameter, determine the RF power correction amount, and perform power adaptive control on the remote plasma source system based on the RF power correction amount.

[0065] Furthermore, the system is also used to implement the following functions: The discharge state parameter stream is data-aligned with the target power curve in time series to obtain an available discharge state parameter stream; key features are extracted from the available discharge state parameter stream to obtain a discharge state key feature set, wherein the discharge state key feature set includes a static discharge feature set and a dynamic discharge feature set; load prediction training is performed based on the target power curve to construct a load state predictor; the load state predictor is used to identify and analyze the discharge state key feature set to output the load state trend parameter.

[0066] Furthermore, the system is also used to implement the following functions: Based on the target power curve, associated data mining and key feature extraction are performed to obtain a historical RF power working feature set; state classification identification training is performed on the historical RF power working feature set according to the target power curve to generate a load state classifier; based on the historical RF power working feature set, load trend identification and prediction supervision training are performed to obtain a load trend predictor; the load state classifier and the load trend predictor are fused in parallel to construct the load state predictor.

[0067] Furthermore, the system is also used to implement the following functions: Build a target hierarchical architecture, decompose the hierarchical indicators of the RF power control target according to the target hierarchical architecture, and construct a RF power control evaluation index system; perform key analysis on each evaluation indicator in the RF power control evaluation index system to determine the indicator evaluation factor set; design a hierarchical objective function based on the RF power control evaluation index system to obtain a multi-level evaluation objective function; perform hierarchical weighted fusion of the multi-level evaluation objective function according to the indicator evaluation factor set to determine the RF power control evaluation rule.

[0068] Furthermore, the system is also used to implement the following functions: The RF power control evaluation rule is introduced to evaluate the control effect of the RF power historical working data set of the remote plasma source system to obtain a RF power control effect set; based on the RF power control effect set, the effect of the RF power historical working data set is optimized to obtain an available RF power working data set; according to the available RF power working data set and the power accuracy control requirement, a power influence threshold is set; based on the power influence threshold, branch power control learning is performed on the available RF power working data set to generate a RF power adaptive control channel.

[0069] Furthermore, the system is also used to implement the following functions: Based on the available RF power working data set, power influencing factors are extracted and the degree of influence is evaluated to obtain a RF power influencing factor set and a power factor influencing factor set; the RF power influencing factor set is weightedly fitted according to the power factor influencing factor set to generate a RF power influence function; based on the power precision control requirement, the RF power influence function is divided into influence thresholds to determine the power influence degree threshold.

[0070] Furthermore, the system is also used to implement the following functions: If the power impact threshold is not reached, a PID controller is used to perform power control learning on the available RF power working data set to obtain a conventional RF power control channel; when the power impact threshold is reached, the RF power impact function is introduced to perform impact correction on the conventional RF power control channel to construct an aggregated RF power control channel; the conventional RF power control channel and the aggregated RF power control channel are branched and merged to generate the RF power adaptive control channel.

[0071] Furthermore, the system is also used to implement the following functions: The RF power influence function is used to evaluate the impact of the application status of the remote plasma source system to obtain the plasma source system impact; based on the plasma source system impact, the RF power adaptive control channel is activated to perform channel matching and control calculation on the target power curve and the load state trend parameter to determine the RF power correction amount.

[0072] Furthermore, the system is also used to implement the following functions: The power control delay time of the remote plasma source system is monitored and acquired; a radio frequency power loss coefficient is determined according to the power control delay time, and feedback compensation is performed on the radio frequency power correction amount based on the radio frequency power loss coefficient.

[0073] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0074] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

[0075] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A method for adaptively controlling radio frequency power of a remote plasma source, characterized in that: The method comprises: Deploying a multi-parameter sensor network in a remote plasma source system, and acquiring a discharge state parameter flow through the multi-parameter sensor network; According to the target process requirements, a target power curve is preset, the discharge state parameter flow and the target power curve are identified and analyzed, and a load state trend parameter is output; Decompose the hierarchical indicators and fit the evaluation factors for the RF power control targets to determine the RF power control evaluation rules; Introducing the radio frequency power control evaluation rule to perform power control learning based on the radio frequency power historical working data set of the remote plasma source system, and constructing a radio frequency power adaptive control channel; The RF power adaptive control channel is used to perform control calculations on the target power curve and the load state trend parameter to determine a RF power correction amount, and the remote plasma source system is power adaptively controlled based on the RF power correction amount.

2. The method for adaptively controlling radio frequency power of a remote plasma source according to claim 1, wherein: The output load state trend parameters include: Aligning the discharge state parameter stream with the target power curve according to a time series to obtain an available discharge state parameter stream; Extracting key features from the available discharge state parameter stream to obtain a discharge state key feature set, wherein the discharge state key feature set includes a static discharge feature set and a dynamic discharge feature set; Perform load prediction training based on the target power curve to build a load state predictor; The load state predictor is used to identify and analyze the discharge state key feature set, and output the load state trend parameter.

3. The method for adaptively controlling radio frequency power of a remote plasma source according to claim 2, wherein: The constructing of the load state predictor comprises: Performing correlation data mining and key feature extraction based on the target power curve to obtain a radio frequency power historical working feature set; Performing state classification identification training on the radio frequency power historical operating feature set according to the target power curve to generate a load state classifier; Performing load trend identification and prediction supervision training based on the radio frequency power historical operating feature set to obtain a load trend predictor; The load state classifier and the load trend predictor are connected in parallel and fused to construct the load state predictor.

4. The method for adaptively controlling radio frequency power of a remote plasma source according to claim 1, wherein: Determining the radio frequency power control evaluation rule includes: Establishing a target hierarchical architecture, breaking down the radio frequency power control target into hierarchical indicators according to the target hierarchical architecture, and constructing a radio frequency power control evaluation indicator system; Performing a critical analysis on each evaluation indicator in the radio frequency power control evaluation indicator system to determine an indicator evaluation factor set; Designing a hierarchical objective function based on the RF power control evaluation index system to obtain a multi-level evaluation objective function; The multi-level evaluation objective function is hierarchically weighted and fused according to the indicator evaluation factor set to determine the radio frequency power control evaluation rule.

5. The method for adaptively controlling radio frequency power of a remote plasma source according to claim 1, wherein: The step of constructing a radio frequency power adaptive control channel includes: Introducing the radio frequency power control evaluation rule to evaluate the control effect of the radio frequency power historical working data set of the remote plasma source system to obtain a radio frequency power control effect set; Based on the radio frequency power control effect set, the radio frequency power historical working data set is optimized to obtain an available radio frequency power working data set; Setting a power impact threshold according to the available radio frequency power working data set and power accuracy control requirements; Based on the power impact threshold, branch power control learning is performed on the available radio frequency power working data set to generate a radio frequency power adaptive control channel.

6. The method for adaptively controlling radio frequency power of a remote plasma source according to claim 5, wherein: The setting of the power impact threshold includes: Extracting power influencing factors and evaluating their impact based on the available radio frequency power working data set to obtain a radio frequency power influencing factor set and a power factor influencing factor set; Performing weighted fitting on the radio frequency power influencing factor set according to the power factor influencing factor set to generate a radio frequency power influencing function; The radio frequency power impact function is divided into impact thresholds based on the power precision control requirement to determine the power impact threshold.

7. The method for adaptively controlling radio frequency power of a remote plasma source according to claim 6, wherein: Generating a radio frequency power adaptive control channel includes: If the power impact threshold is not reached, a PID controller is used to perform power control learning on the available radio frequency power working data set to obtain a conventional radio frequency power control channel; When the power impact threshold is reached, the radio frequency power impact function is introduced to modify the impact of the conventional radio frequency power control channel to construct an aggregated radio frequency power control channel; The conventional radio frequency power control channel and the aggregated radio frequency power control channel are branched and merged to generate the radio frequency power adaptive control channel.

8. The method for adaptively controlling radio frequency power of a remote plasma source according to claim 7, wherein: Determining the radio frequency power correction amount includes: Using the radio frequency power influence function to evaluate the impact of the application status of the remote plasma source system to obtain the plasma source system impact; The RF power adaptive control channel is activated based on the influence of the plasma source system to perform channel matching and control calculation on the target power curve and the load state trend parameter to determine the RF power correction amount.

9. The method for adaptively controlling radio frequency power of a remote plasma source according to claim 1, wherein: The method further comprises: Monitoring and obtaining a power control delay time of the remote plasma source system; A radio frequency power loss coefficient is determined according to the power control delay time, and feedback compensation is performed on the radio frequency power correction amount based on the radio frequency power loss coefficient.

10. An adaptive control system for radio frequency power of a remote plasma source, characterized in that: The system is used to perform the adaptive control method of radio frequency power of a remote plasma source according to any one of claims 1 to 9, and the system comprises: A parameter acquisition module is used to deploy a multi-parameter sensor network in the remote plasma source system and acquire a discharge state parameter stream through the multi-parameter sensor network; an identification and analysis module, configured to preset a target power curve according to target process requirements, identify and analyze the discharge state parameter flow and the target power curve, and output a load state trend parameter; The rule determination module is used to decompose the hierarchical indicators and fit the evaluation factors of the RF power control target to determine the RF power control evaluation rules; A control channel construction module is used to introduce the radio frequency power control evaluation rule to perform power control learning based on the radio frequency power historical working data set of the remote plasma source system, and to construct a radio frequency power adaptive control channel; The adaptive control module is used to use the RF power adaptive control channel to perform control calculations on the target power curve and the load state trend parameter, determine the RF power correction amount, and perform power adaptive control on the remote plasma source system based on the RF power correction amount.

Images