A circuit data processing method and system based on an adaptive big data model

Abstract

This invention relates to the field of circuit data processing technology, and discloses a circuit data processing method and system based on an adaptive big data model. By acquiring the energy input parameters and periodic characteristics of the device's operation, and processing sensor data affected by external interference to obtain frequency components, the method can identify and quantify the interference patterns corresponding to the periodic characteristics from these frequency components. Based on this, the interference patterns are separated from the sensor data according to the quantization results, thereby obtaining a data stream reflecting the true state of the device. This data stream is then provided to a self-adjusting system to adjust its parameters. This method effectively solves the problem of model misjudgment and system failure in existing adaptive systems due to sensor data contamination by external interference.

Description

Technical Field

[0001] This invention relates to the field of circuit data processing technology, and in particular to a circuit data processing method and system based on an adaptive big data model. Background Technology

[0002] In modern industrial production, analyzing real-time operating data of complex equipment is crucial for ensuring stable production and predicting potential failures. Traditional circuit data processing methods often rely on pre-set fixed parameters, which struggle to maintain accurate predictive capabilities when facing the constantly changing operating environment of industrial sites. To address this challenge, adaptive data processing mechanisms have been introduced, aiming to enable the system to dynamically adjust internal parameters based on real-time feedback data, thereby better adapting to environmental changes.

[0003] However, in practical applications, when equipment operating conditions change significantly, such as an increase in power, it may be accompanied by some physical phenomena that are not directly monitored, such as increased electromagnetic interference. This interference can subtly contaminate the data collected by sensors, introducing "false signals" with specific patterns into the data stream. Existing adaptive mechanisms often mistake these "false signals" for changes in the actual state of the equipment, causing the model to be misled during the learning process, establishing incorrect associations, and ultimately resulting in significant deviations in predictions, which in turn disrupts the normal order of production.

[0004] Specifically, in a large semiconductor manufacturing plant, the overall distribution of operational status data from a high-precision plasma etching machine under a new high-power process has undergone a fundamental shift. The system continuously collects operational data through multiple sensors installed inside the etching machine, but this data is severely contaminated by electromagnetic interference. While the sensors themselves are functioning normally and the etching machine's vacuum system is stable and reliable, the output electrical signals are superimposed with interference noise synchronized with the radio frequency power cycle before reaching the data acquisition card. This noise is not completely random white noise, but rather structured noise with a certain regularity.

[0005] The adaptive training mechanism was designed with the assumption that all input data reflects the true state of the physical world. Therefore, when it receives sensor data contaminated by electromagnetic interference, it cannot distinguish between genuine pressure fluctuations and external electromagnetic noise. Because this noise is structural, the model tries to learn and fit this "pseudo-pattern," assuming a deep, but not actually existing, correlation between these regular signal fluctuations and the equipment's state. The model continuously adjusts its internal parameters, attempting to explain and predict the noise, severely weakening its ability to judge the true physical process. Ultimately, this supposedly more intelligent adaptive system becomes unreliable due to this "overlearning." It might mistakenly determine that the vacuum system is about to fail based on a specific electromagnetic interference waveform, triggering a high-level alarm. Maintenance personnel, upon receiving the alarm, repeatedly check the vacuum pump and piping, only to find everything normal, wasting resources and severely impacting production schedules. In this scenario of unidentified data contamination, the system's adaptive capability, once an advantage, becomes a source of chaos.

[0006] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0007] This invention provides a circuit data processing method and system based on an adaptive big data model, aiming to solve the technical problem that existing adaptive data processing mechanisms, when faced with complex industrial sites, suffer from sensor data contamination due to external interference, which in turn leads to model misjudgment and system failure.

[0008] The technical solution of this application is as follows: In a first aspect, this application discloses a circuit data processing method based on an adaptive big data model, comprising the following steps: Acquire the energy input parameters of the equipment and the periodic characteristics of the energy input parameters; Process sensor data affected by external interference to obtain frequency components from the sensor data; Identify and quantify interference patterns corresponding to periodic characteristics from the frequency components of sensor data; Based on the quantification results of the interference patterns, the interference patterns are separated from the sensor data to reflect the data stream of the device's true state. The data stream reflecting the actual state of the equipment is provided to the self-adjusting system to adjust the parameters of the self-adjusting system.

[0009] Through this technical solution, this application can effectively identify and separate "false signals" caused by external interference in sensor data, thereby providing a data stream that truly reflects the device status to the self-adjustment system. This avoids misjudgment and failure caused by data pollution in traditional adaptive systems, and significantly improves the accuracy and reliability of the system under complex working conditions.

[0010] Furthermore, based on the above, this application proposes a more specific method for identifying and quantifying interference patterns corresponding to periodic features from frequency components in sensor data, including: Continuously monitor the instantaneous waveform of the energy input parameters and perform real-time spectrum analysis on the instantaneous waveform to obtain the instantaneous frequency drift, amplitude fluctuation rate, and non-harmonic intermodulation products of the energy input parameters; Short-time Fourier transform is performed on sensor data affected by external interference to obtain the frequency components and intensities that change over time in the sensor data; Based on the instantaneous frequency drift, amplitude fluctuation rate, and non-harmonic intermodulation products, the identification window of the interference frequency is dynamically adjusted, and the frequency peaks in the sensor data related to the fundamental frequency, harmonic frequency, and non-harmonic intermodulation products of the energy input parameters are identified. Based on the instantaneous intensity and frequency changes of the identified frequency peaks, the interference mode reference is updated in real time, and the complexity index of the interference mode is calculated to quantify the interference mode.

[0011] Through this technical solution, this application can dynamically adjust the interference identification window by finely monitoring the energy input parameters and performing real-time spectrum analysis, combined with the short-time Fourier transform of sensor data. This allows for more accurate identification of interference frequency peaks related to the periodic characteristics of the energy input parameters, and real-time updates of the interference mode reference, effectively improving the accuracy and quantification capability of interference mode identification.

[0012] More specifically, in some implementations, the reference for the interference mode is updated in real time based on the instantaneous intensity and frequency changes of the identified frequency peaks, and a complexity index of the interference mode is calculated to quantify the interference mode, including: Acquire information on changes in the physical parameters of components in the radio frequency power transmission system; The correlation analysis between the physical parameter change information and the evolution trend of interference mode in historical operation data is carried out to identify the correspondence between the physical parameter change information and the evolution trend of interference mode. Based on the correspondence, the baseline update mechanism for the interference mode reference is triggered; Based on the baseline update mechanism, the frequency component parameters in the interference mode reference are adjusted, and the interference mode reference is updated in real time according to the adjusted interference mode reference. The complexity index of the interference mode is calculated to quantify the interference mode.

[0013] This application introduces information on the physical parameter changes of components in the radio frequency power transmission system and correlates them with historical interference mode evolution trends. This allows for a deeper understanding of the inherent variation patterns of interference modes and triggers a baseline update mechanism, making the update of interference mode references more accurate and adaptive, effectively improving the accuracy and robustness of interference mode quantification.

[0014] Preferably, acquiring information on changes in the physical parameters of components in the radio frequency power transmission system includes: Multiple sensors are deployed on key components of the radio frequency power transmission system to continuously collect physical parameter information of the key components; Time series analysis was performed on the collected physical parameter information to extract the long-term trend of the physical parameter information; Acquire instantaneous operating parameters of the equipment and perform correlation analysis between instantaneous operating parameters and physical parameter information to identify transient changes caused by operating condition fluctuations; By subtracting transient changes caused by fluctuations in operating conditions from the long-term trend of physical parameter information, we can obtain physical parameter change information that reflects the long-term changes caused by component aging.

[0015] Through this technical solution, this application can effectively distinguish between long-term physical parameter changes caused by component aging and transient changes caused by fluctuations in operating conditions by deploying multiple sensors, conducting time series analysis, and performing instantaneous operating condition parameter correlation analysis. This allows for the acquisition of purer and more accurate information on physical parameter changes that reflect the true aging state of components, providing a reliable input for the precise quantification of interference modes.

[0016] Based on the above, this application further proposes a method for identifying the correspondence between physical parameter variation information and interference mode evolution trends in historical operational data through correlation analysis, including: Continuously acquire the evolution trend of interference patterns in historical operational data; Identify the nonlinear or dynamic correlation between changes in physical parameters and the evolution trend of interference modes; Initiate the adaptive association learning mechanism; Adjust the parameters of the correlation function based on real-time data of current physical parameter changes and interference mode evolution trends; Regularly assess the predictive accuracy of the correlation function; When the prediction accuracy drops to a preset threshold, the correlation function is reconstructed or its parameters are optimized.

[0017] Through this technical solution, this application can dynamically establish and optimize the correspondence between changes in physical parameters and the evolution trend of disturbance modes by continuously acquiring historical data, identifying nonlinear correlations, initiating an adaptive correlation learning mechanism, and periodically evaluating the prediction accuracy. This ensures the real-time performance and accuracy of the correlation model, thereby effectively coping with complex and ever-changing industrial environments.

[0018] In one implementation, when the prediction accuracy falls below a preset threshold, a reconstruction or parameter optimization of the correlation function is triggered, including: When the correlation function is refactored or the parameters are optimized, the quality of the data used for learning is assessed, and missing data points and outliers in the data are identified and marked. For the identified missing data points, data is filled in based on the trend of data changes before and after the missing data points and the correlation with relevant physical parameters; For identified outliers, data correction or removal is performed based on the degree and duration of the outlier's deviation from the normal range. When the amount of data after quality assessment and processing is insufficient to support effective learning, synthetic data that matches the characteristics of real data is generated based on the existing evolution patterns of component aging and interference patterns to expand the learning dataset. Based on the expanded dataset, the correlation function is reconstructed or its parameters are optimized.

[0019] Through this technical solution, when the prediction accuracy of the correlation function decreases, this application effectively solves the problem of limited model learning caused by insufficient data or poor data quality by means of data quality assessment, missing data completion, outlier correction or removal, and synthetic data expansion. This ensures the effectiveness of correlation function reconstruction or parameter optimization, thereby maintaining the system's ability to accurately identify the evolution trend of interference patterns.

[0020] Based on the above, this application also proposes a method to expand the learning dataset when the amount of data after quality assessment and processing is insufficient to support effective learning. This method generates synthetic data that matches the characteristics of real data based on existing component aging patterns and interference pattern evolution rules. The method includes: Uncertainty quantification is performed on the existing component aging modes and interference mode evolution patterns to obtain the fluctuation range and distribution characteristics of the interference mode evolution patterns. Based on the fluctuation range and distribution characteristics, random perturbations are introduced when generating synthetic data to simulate uncertainties and data noise in the real world; Real-time monitoring of the statistical characteristics of synthetic data, and comparison of the statistical characteristics of synthetic data with those of real data; If discrepancies exist, the strength and distribution of random perturbations are adjusted to ensure that the synthetic data accurately reflects the complexity and randomness of the real world.

[0021] This application utilizes a technical solution to quantify the uncertainty of existing models and introduce random perturbations into synthetic data. This effectively simulates the complexity and randomness of the real world. By monitoring and comparing the statistical characteristics of synthetic data with those of real data in real time, the random perturbations are dynamically adjusted to ensure the high fidelity of the synthetic data. This provides sufficient and high-quality data support for the effective learning of correlation functions.

[0022] In one implementation, the statistical characteristics of the synthetic data are monitored in real time, and the statistical characteristics of the synthetic data are compared with the statistical characteristics of the real data, including: Real-time acquisition of feature information of synthetic and real data across multiple statistical dimensions, including mean, variance, skewness, kurtosis, and time series autocorrelation; The differences in feature information across multiple statistical dimensions are evaluated to identify statistical feature differences between synthetic and real data, and to quantify the degree of these differences. Based on the degree of difference in statistical characteristics, determine whether the difference is a single-dimensional difference, a multi-dimensional synchronous difference, or a multi-dimensional asynchronous difference. When the difference is determined to be a multidimensional asynchronous difference, the dominant statistical dimension in the multidimensional asynchronous difference is identified, and the contribution weight of the dominant statistical dimension to the difference is calculated. Adjustment suggestions are generated based on the contribution weight of the dominant statistical dimension to the differences, which can be used to adjust the strength and distribution of random disturbances.

[0023] Through this technical solution, this application can accurately identify and quantify statistical feature differences by conducting refined comparison and difference assessment of synthetic and real data across multiple statistical dimensions. In particular, when there are asynchronous differences across multiple dimensions, it can identify the dominant statistical dimension and calculate its contribution weight, thereby generating targeted adjustment suggestions and ensuring a high degree of consistency between synthetic and real data in terms of statistics.

[0024] Building upon the above, this application further proposes a method for adjusting the intensity and distribution of random perturbations if discrepancies exist, to ensure that synthetic data accurately reflects the complexity and randomness of the real world, including: The weights of the contributions of multiple dominant statistical dimensions to the overall difference are similar; Initiate a conflict assessment mechanism to analyze the direction and extent of the impact of the adjustment objectives of each dominant statistical dimension on the strength and distribution of random disturbances; A priority allocation strategy is introduced to assign and adjust the priority of each dominant statistical dimension based on preset weights or the current system’s sensitivity to specific statistical features. Based on the adjustment priority, the intensity and distribution of random disturbances are adjusted in stages or iteratively to meet the adjustment requirements of high-priority dimensions. Real-time monitoring of changes in synthetic data across all relevant statistical dimensions during the adjustment process; The adjustment strategy is fine-tuned based on the monitoring results, so as to gradually optimize the adjustment of other dimensions without significantly worsening the already optimized dimensions.

[0025] Through this technical solution, when the contribution weights of multiple dominant statistical dimensions are close, this application can effectively resolve potential conflicts between multi-dimensional adjustment targets by using conflict assessment, priority allocation, and phased / iterative adjustment strategies. It can also monitor the adjustment effect in real time and make fine adjustments, thereby ensuring the adjustment needs of high-priority dimensions while gradually optimizing other dimensions, and realizing the accurate simulation of the complexity and randomness of the real world by synthetic data.

[0026] Secondly, this application also discloses a circuit data processing system based on an adaptive big data model, comprising: The input terminal is used to acquire the energy input parameters of the device operation and the periodic characteristics of the energy input parameters; it also processes sensor data affected by external interference to obtain the frequency components in the sensor data. The identification end is used to identify and quantify the interference patterns corresponding to periodic features from the frequency components in the sensor data; The adjustment end is used to separate the interference mode from the sensor data based on the quantification result of the interference mode, so as to reflect the data stream of the actual state of the device; and to provide the data stream reflecting the actual state of the device to the self-adjustment system so as to adjust the parameters of the self-adjustment system.

[0027] This application provides a system that can effectively process circuit data affected by external interference and provide a real data stream for the self-adjustment system. Through the collaborative work of the input end, the identification end, and the adjustment end, the system achieves accurate identification, quantification, and separation of interference modes, thereby significantly improving the accuracy and reliability of equipment operation status monitoring and self-adjustment. Beneficial effects

[0028] This application discloses a circuit data processing method based on an adaptive big data model. By acquiring the energy input parameters and periodic characteristics of the device's operation, and processing sensor data affected by external interference to obtain frequency components, the method can identify and quantify the interference patterns corresponding to the periodic characteristics from these frequency components. Based on this, the interference patterns are separated from the sensor data according to the quantization results, thereby obtaining a data stream reflecting the true state of the device. This data stream is then provided to a self-adjusting system to adjust its parameters.

[0029] This method effectively solves the problem of model misjudgment and system failure in existing adaptive systems due to sensor data contamination by external interference. Specifically, in high-precision equipment operation scenarios such as semiconductor manufacturing plants, when changes in equipment operating conditions (such as high-power processes) lead to increased electromagnetic interference, sensor data is superimposed with regular structured noise. Traditional adaptive mechanisms cannot distinguish these "false signals" from real physical state changes, causing the model to "overlearn" the pseudo-patterns of noise, thereby establishing incorrect associations, leading to false alarms and wasted resources.

[0030] This application, by actively identifying and quantifying interference patterns related to the periodic characteristics of energy input parameters, can accurately remove these structured noises from the raw sensor data, ensuring that the data stream provided to the self-adjustment system is pure and accurately reflects the equipment status. This enables the self-adjustment system to adjust parameters based on accurate information, avoiding erroneous judgments and ineffective operations caused by data contamination, significantly improving the system's robustness and predictive accuracy in complex and variable industrial environments, thereby effectively ensuring production stability and efficiency. Attached Figure Description

[0031] Figure 1 This is a flowchart of a circuit data processing method based on an adaptive big data model provided in an embodiment of the present invention; Figure 2 This is a flowchart of a method for identifying and quantifying interference patterns corresponding to periodic features from frequency components in sensor data, provided by an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of a circuit data processing system based on an adaptive big data model provided in an embodiment of the present invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0033] Reference Figure 1 , Figure 1 This is a flowchart of a circuit data processing method based on an adaptive big data model provided by an embodiment of the present invention, including the following steps: S11, acquire the energy input parameters of the device operation and the periodic characteristics of the energy input parameters; S12, Process the sensor data affected by external interference to obtain the frequency components in the sensor data; S13, Identify and quantify the interference pattern corresponding to the periodic feature from the frequency components in the sensor data; S14, Based on the quantization result of the interference mode, separate the interference mode from the sensor data to reflect the data stream of the actual state of the device; S15, the data stream reflecting the actual state of the device is provided to the self-adjustment system to adjust the parameters of the self-adjustment system.

[0034] This application introduces a mechanism for identifying, quantifying, and separating interference patterns, which can effectively extract data streams reflecting the true state of the device from contaminated sensor data. This provides accurate input for the self-adjusting system, avoids misjudgments and incorrect adjustments of system parameters caused by data contamination in traditional methods, and significantly improves the accuracy of data processing and the system's adaptive capabilities.

[0035] The "energy input parameters" mentioned in this application refer to the electrical energy, thermal energy, mechanical energy, etc., consumed or received by the equipment during operation. Their "periodic characteristics" may manifest as periodic fluctuations in voltage, current, and frequency, or periodic patterns of mechanical vibration. These periodic characteristics are often closely related to the equipment's working principle or external power supply method. "Sensor data" refers to the raw data collected in real time by various sensors deployed on the equipment (such as voltage sensors, current sensors, temperature sensors, vibration sensors, etc.). This data may be affected by external factors such as electromagnetic interference and mechanical noise. "Interference modes" refer to non-equipment-specific signals superimposed on the sensor data, possessing specific frequency components and periodic characteristics. These signals typically originate from the external environment or periodic activities of non-core working parts within the equipment. "Self-adjusting system" refers to an intelligent system capable of dynamically adjusting its own operating parameters or model based on input data to optimize equipment performance or predict equipment status, such as predictive maintenance systems and process control systems.

[0036] In one implementation, the circuit data processing method first needs to acquire the energy input parameters of the device operation and the periodic characteristics of the energy input parameters. For example, a high-precision energy meter or oscilloscope can be deployed at the power input terminal of the device to monitor the voltage and current waveforms of the device in real time, and these waveforms can be analyzed using signal processing techniques such as Fourier transform to extract the fundamental frequency, harmonic frequencies, and other periodic components. These periodic characteristics can be fixed or can vary with the device's operating conditions.

[0037] Subsequently, the sensor data affected by external interference is processed to obtain the frequency components within the sensor data. For example, sensor data may be collected by accelerometers, pressure sensors, or temperature sensors installed at critical parts of the equipment. This raw data may contain a significant amount of noise and interference. To extract frequency components, frequency domain analysis methods such as Fast Fourier Transform (FFT) or wavelet transform can be used to convert the time-domain data to the frequency domain, thereby revealing the various frequency components contained in the data and their intensities.

[0038] Next, interference patterns corresponding to the periodic characteristics are identified and quantified from the frequency components of the sensor data. For example, by comparing the frequency components in the sensor data with the periodic characteristics of the energy input parameters, it can be found that certain frequency peaks are highly correlated with the fundamental frequency or harmonic frequency of the energy input parameters. These correlated frequency peaks are likely interference caused by the energy input parameters. Quantifying interference patterns may include calculating parameters such as the amplitude, phase, and bandwidth of these frequency peaks to form a mathematical description of the interference patterns.

[0039] Then, based on the quantization results of the interference pattern, the interference pattern is separated from the sensor data to reflect the data stream that reflects the true state of the device. For example, adaptive filters, notch filters, or machine learning-based signal separation techniques can be used. By constructing a filter that matches the quantized interference pattern, the interference signal is filtered out from the original sensor data. The separated data stream will more accurately reflect the true physical state of the device because it eliminates the influence of external interference.

[0040] Finally, the data stream reflecting the true state of the equipment is provided to the self-adjusting system to adjust its parameters. For example, the self-adjusting system could be a machine learning model for predicting the remaining lifespan of the equipment, or a PID controller for optimizing the production process. Upon receiving the purified data stream, the self-adjusting system can learn and make decisions based on more accurate information, thereby avoiding erroneous adjustments caused by data contamination and improving the system's robustness and accuracy.

[0041] The circuit data processing method of this application, by introducing a mechanism for identifying, quantifying, and separating interference modes, can effectively extract a data stream reflecting the true state of the device from contaminated sensor data. This method first acquires the energy input parameters of the device operation and their periodic characteristics, providing a basis for identifying potential interference sources. Next, frequency domain analysis is performed on the sensor data affected by external interference to reveal the frequency components contained in the data. Crucially, this application can identify and quantify interference modes corresponding to the periodic characteristics of the energy input parameters from these frequency components. For example, in a radio frequency power transmission system, when the power is increased, electromagnetic interference synchronized with the radio frequency may be generated. This interference will exhibit specific frequency peaks in the sensor data. This application can accurately capture and quantify the intensity, frequency, and phase characteristics of these interference modes.

[0042] Once the interference patterns are quantized, this application can accurately separate these interference patterns from the raw sensor data based on the quantization results. For example, by constructing an adaptive filter, frequency components that match the interference patterns are dynamically filtered out, resulting in a "clean" data stream that reflects the true state of the device. This data stream eliminates the influence of "false signals" such as external electromagnetic interference, making the data more accurate and reliable.

[0043] Ultimately, this data stream reflecting the true state of the equipment is provided to the self-adjusting system. Unlike existing technologies that directly input contaminated sensor data into the self-adjusting system, this application provides purified, high-fidelity data. Based on this accurate data, the self-adjusting system can more precisely adjust its internal parameters. For example, in predictive maintenance systems, the model can more accurately learn the aging trends of the equipment and avoid misjudgments caused by interference signals; in process control systems, the controller can adjust based on the actual physical state, thereby optimizing production efficiency and product quality.

[0044] Compared with existing technologies, the core innovation of this application lies in its deep processing capability of "interference patterns." Traditional methods often treat all input data as real signals, causing adaptive systems to incorrectly learn and fit these "pseudo-patterns" when facing structured interference, resulting in misjudgments. This application, by introducing the periodic characteristics of energy input parameters as a reference, can actively identify and quantify interference patterns related to these periodic characteristics and separate them from the original data. This "suppression of false signals" processing method enables the self-adjusting system to receive cleaner and more accurate data, thereby avoiding the drawbacks of "over-learning" interference signals and significantly improving the robustness and prediction accuracy of the system. For example, in the case of plasma etching machines in semiconductor manufacturing plants, this application can identify and separate electromagnetic interference noise synchronized with the RF power cycle, enabling the self-adjusting system to accurately judge the true state of the vacuum system and avoid resource waste and production interruptions caused by false alarms. Therefore, this application not only solves the problem of data pollution causing adaptive system failure in existing technologies, but also provides a more reliable and efficient solution for the intelligent operation and maintenance of complex equipment.

[0045] In some embodiments described above, interference patterns corresponding to periodic characteristics are identified and quantified from frequency components in sensor data. However, during implementation, if the interference patterns are complex and variable or dynamically affected by energy input parameters, it may be difficult to accurately and in real-time identify and quantify these interferences, thus affecting the accuracy of the device's true state data stream. To address this, this application further proposes a more refined and adaptive interference pattern identification and quantification method. By continuously monitoring the dynamic characteristics of energy input parameters and combining real-time spectrum analysis of sensor data, dynamic adjustment and precise quantification of interference patterns are achieved.

[0046] For details, please refer to Figure 2 , Figure 2 This is a flowchart of a method for identifying and quantifying interference patterns corresponding to periodic features from frequency components in sensor data, provided by an embodiment of the present invention, including: S131 continuously monitors the instantaneous waveform of the energy input parameters and performs real-time spectrum analysis on the instantaneous waveform to obtain the instantaneous frequency drift, amplitude fluctuation rate, and non-harmonic intermodulation products of the energy input parameters; performs short-time Fourier transform on sensor data affected by external interference to obtain the frequency components and intensities that change with time in the sensor data. S132, based on the instantaneous frequency drift, amplitude fluctuation rate and non-harmonic intermodulation products, dynamically adjusts the identification window of the interference frequency, and identifies the frequency peaks in the sensor data related to the fundamental frequency, harmonic frequency and non-harmonic intermodulation products of the energy input parameters. S133 updates the interference mode reference in real time based on the instantaneous intensity and frequency change of the identified frequency peaks, and calculates the complexity index of the interference mode to quantify the interference mode.

[0047] The continuous monitoring of instantaneous waveforms of energy input parameters and the real-time spectral analysis of these waveforms aim to capture dynamic changes in these parameters, such as instantaneous frequency fluctuations, amplitude drops, or distortions in grid voltage, as well as non-harmonic intermodulation products caused by nonlinear loads. These parameters are key indicators for evaluating the characteristics of external interference sources. Instantaneous frequency drift reflects the real-time shift of the energy input frequency, amplitude fluctuation indicates the degree of instantaneous change in the energy input amplitude, and non-harmonic intermodulation products reveal the nonlinear distortion components present in the energy input.

[0048] Furthermore, performing a short-time Fourier transform on sensor data affected by external interference aims to obtain joint information about the sensor data in the time and frequency domains. The short-time Fourier transform can decompose non-stationary sensor signals into time-varying frequency components and their corresponding intensities, thereby revealing the spectral characteristics of the interference signal at different time points and providing a refined time-frequency spectrum for subsequent interference identification.

[0049] Based on this, the identification window for interference frequencies is dynamically adjusted according to the instantaneous frequency drift, amplitude fluctuation rate, and non-harmonic intermodulation products, and the frequency peaks in the sensor data related to the fundamental frequency, harmonic frequencies, and non-harmonic intermodulation products of the energy input parameters are identified. Specifically, when the energy input parameters experience instantaneous frequency drift, the identification window for interference frequencies is adjusted accordingly to ensure accurate capture of interference peaks related to the drifted fundamental and harmonic frequencies. When the amplitude fluctuation rate is high, the width or sensitivity of the identification window can be adjusted to better detect broadband or narrowband interference caused by transient events. The presence of non-harmonic intermodulation products guides the identification system to focus on specific non-harmonic frequency ranges, thereby identifying interference peaks caused by these intermodulation products. This dynamic adjustment mechanism enables the interference identification process to adaptively match the real-time changes in energy input parameters, improving the accuracy and robustness of the identification.

[0050] Furthermore, based on the instantaneous intensity and frequency changes of the identified frequency peaks, the interference mode reference is updated in real time, and the complexity index of the interference mode is calculated to quantify the interference mode. The purpose of real-time updating of the interference mode reference is to enable the system to adapt to the evolution of interference modes, such as changes in the intensity of the interference source, minor frequency drifts, or the emergence of new interference sources. The complexity index of the interference mode can be understood as a comprehensive quantification of the characteristics of the current interference mode, calculated based on factors such as the number of identified frequency peaks, intensity distribution, bandwidth, and stability over time. This index provides a unified metric for assessing the severity and characteristics of interference, thus providing a more accurate basis for subsequent interference separation.

[0051] The proposed solution continuously monitors the instantaneous dynamic characteristics of the energy input parameters and combines this with real-time spectral analysis of sensor data to dynamically adjust the interference identification strategy. This adaptive identification window adjustment mechanism enables the system to accurately capture interference frequency peaks closely related to changes in the energy input parameters, maintaining high-precision identification even when the energy input parameters drift or fluctuate. Furthermore, by updating the interference mode reference in real time and calculating the complexity index, this solution allows for refined quantification of interference modes, providing a more accurate and reliable foundation for subsequent separation of interference modes from sensor data.

[0052] Through the above technical solution, this application can significantly improve the accuracy of identifying and quantifying interference modes in circuit data under complex dynamic environments. Compared with traditional static or threshold-based interference identification methods, this solution can better adapt to instantaneous changes in energy input parameters and effectively cope with various forms of interference such as frequency drift, amplitude fluctuations, and non-harmonic intermodulation products. Therefore, it can more accurately separate the data stream reflecting the true state of the device from the interfered sensor data, thereby providing high-quality input for the self-adjusting system and improving the parameter adjustment effect and overall performance of the self-adjusting system.

[0053] In some preferred embodiments, it is assumed that during the operation of an RF power transmission system, its power supply may be affected by factors such as grid fluctuations and load changes, resulting in instantaneous frequency drift and amplitude fluctuations in energy input parameters (such as voltage and current). Simultaneously, nonlinear components within the system may also generate non-harmonic intermodulation products. Sensors deployed on key components of the system collect circuit data containing these interferences. The method of this application first continuously monitors the instantaneous waveform of the power supply, obtaining its instantaneous frequency drift, amplitude fluctuation rate, and non-harmonic intermodulation products through real-time spectrum analysis. For example, if the grid frequency drifts from 50Hz to 50.1Hz, or if a momentary voltage drop occurs, this information is accurately captured. Simultaneously, a short-time Fourier transform is performed on the sensor data to obtain its frequency components and intensities changing over time. Then, the system dynamically adjusts the identification window for interference frequencies based on the monitored instantaneous frequency drift, amplitude fluctuation rate, and non-harmonic intermodulation products. For example, if the power supply fundamental frequency drifts, the identification window will shift accordingly from around 50Hz to around 50.1Hz to identify the associated fundamental frequency and harmonic interference peaks. If specific non-harmonic intermodulation products are detected, the system will focus on peaks within these frequency ranges. Finally, based on the instantaneous intensity and frequency variation of the identified frequency peaks, the system updates the interference mode reference in real time and calculates a complexity index (e.g., based on the number, bandwidth, and energy of interference peaks) to accurately quantify the current interference mode, providing an accurate basis for subsequent interference separation.

[0054] In some embodiments described above, this application proposes updating the interference mode reference in real time based on the instantaneous intensity and frequency changes of identified frequency peaks, and calculating the complexity index of the interference mode to quantify it. However, in its implementation, relying solely on changes in instantaneous spectral characteristics may not adequately capture the gradual drift or structural changes in interference modes caused by the long-term evolution of the physical parameters of RF power transmission system components. This limitation may result in insufficiently comprehensive or forward-looking updates to the interference mode reference, thereby affecting the accurate separation of the device's real-state data stream. To address this, this application further proposes a more robust and adaptive interference mode reference update mechanism. This mechanism integrates information on changes in the physical parameters of RF power transmission system components with the evolution trends of interference modes in historical operating data to more accurately identify and quantify interference modes.

[0055] In response, this application further proposes the following steps: updating the reference of the interference mode in real time based on the instantaneous intensity and frequency change of the identified frequency peak, and calculating the complexity index of the interference mode to quantify the interference mode, including: Acquire information on changes in the physical parameters of components in the radio frequency power transmission system; The physical parameter change information is correlated with the interference mode evolution trend in historical operation data to identify the correspondence between the physical parameter change information and the interference mode evolution trend. Based on the aforementioned correspondence, the baseline update mechanism for the interference mode reference is triggered; According to the baseline update mechanism, the frequency component parameters in the interference mode reference are adjusted, and the interference mode reference is updated in real time according to the adjusted interference mode reference. The complexity index of the interference mode is calculated to quantify the interference mode.

[0056] Specifically, acquiring information on changes in the physical parameters of components in an RF power transmission system refers to continuously collecting physical attribute data of key components (such as power amplifier modules, filters, and antennas) in the RF power transmission system through various sensors or monitoring methods, and extracting trends or anomalies that change over time. These physical parameters may include, but are not limited to, temperature, voltage, current, impedance, and VSWR. By analyzing these parameters, changes in the physical state of components, such as aging, performance degradation, or potential faults, can be identified.

[0057] This involves correlating changes in physical parameters with the evolution trends of interference modes in historical operational data to identify the correspondence between these changes. This can be understood as establishing a model or algorithm to reveal the intrinsic relationship between changes in the physical state of components and the characteristics of the system's output interference modes (such as frequency, amplitude, bandwidth, and modulation scheme). This correlation analysis can employ techniques such as machine learning, statistical regression, or pattern recognition to discover how specific patterns of physical parameter changes lead to or predict specific directions and degrees of interference mode evolution.

[0058] In practical applications, based on the correspondence, the baseline update mechanism for the interference mode reference refers to the system initiating a pre-defined update process once a significant correspondence is identified between changes in physical parameters and the evolution of the interference mode, and this correspondence indicates that the baseline of the interference mode (i.e., its long-term stable characteristics) may drift. This baseline update mechanism aims to ensure that the interference mode reference can adapt to long-term changes in the system's physical state, thereby maintaining its accuracy and effectiveness.

[0059] Furthermore, based on the baseline update mechanism, the frequency component parameters in the interference mode reference are adjusted, and the interference mode reference is updated in real time according to the adjusted reference. The complexity index of the interference mode is then calculated to quantify the interference mode. This means that after the baseline update mechanism is triggered, the system will correct the frequency component parameters (e.g., fundamental frequency, harmonic frequencies, center frequency, bandwidth, relative intensity, etc. of non-harmonic intermodulation products) stored in the interference mode reference according to the identified correspondences. The adjusted interference mode reference will more accurately reflect the interference characteristics under the current physical state of the system. Subsequently, based on the updated interference mode reference, the complexity index of the interference mode can be recalculated. This index can be used to quantify the structural complexity, dynamic change degree, or potential impact on system performance of the interference mode, thus providing a more refined basis for subsequent interference separation and system adjustment.

[0060] This application addresses the limitations of relying solely on instantaneous spectral characteristics to update the interference mode reference by incorporating information on the physical parameter changes of RF power transmission system components and correlating this information with the evolution trends of interference modes in historical operational data. Through this technical solution, the application overcomes the shortcomings of traditional methods in handling interference mode drift caused by long-term evolution of system component physical parameters. By correlating physical parameter change information with historical interference mode evolution trends, the application establishes a deep connection between interference modes and the system's physical state, enabling more forward-looking and robust baseline updates to the interference mode reference. This update mechanism allows interference mode identification and quantification to consider changes at the system's physical level, rather than solely relying on instantaneous spectral characteristics, significantly improving the accuracy and stability of the interference mode reference. Consequently, the interference modes separated from sensor data become more accurate, and the data stream reflecting the true state of the equipment becomes purer and more reliable, thereby enhancing the effectiveness of self-adjusting system parameter adjustments and the overall system performance.

[0061] In some preferred embodiments, a specific example is given below. Suppose a power amplifier module in an RF power delivery system has internal transistors that gradually age due to long-term operation, leading to increased nonlinearity, accompanied by rising operating temperature and a slight drift in output impedance. Conventional methods might only update the interference mode reference when the instantaneous intensity and frequency of the interference mode's peak value change significantly. However, this update may lag behind gradual changes in physical parameters, resulting in a cumulative deviation between the interference mode reference and the actual interference mode.

[0062] According to the scheme of this application, the system will continuously acquire information on the changes in the physical parameters of the power amplifier module, such as monitoring its operating temperature through a temperature sensor and monitoring its output impedance through an impedance analyzer. At the same time, the system will continuously acquire the evolution trend of interference modes related to the aging of the power amplifier module from historical operating data. For example, historical data shows that for every 5 degrees Celsius increase in the temperature of the power amplifier module, the center frequency of a certain harmonic interference will drift 100 Hz towards the higher frequency, and its amplitude will increase by 0.5 dB.

[0063] When the system detects a continuous rise in the power amplifier module's temperature and a slight drift in its output impedance, this information on physical parameter changes is input into the correlation analysis module. This module performs correlation analysis between these real-time physical parameter changes and the evolution trends of interference patterns in historical operating data. For example, using a pre-trained machine learning model, the system identifies a high degree of correspondence between the current physical parameter changes and the evolution trend of a specific interference pattern in historical data (e.g., harmonic frequency drift and amplitude increase).

[0064] Once this correspondence is identified, the system triggers the baseline update mechanism for the interference mode reference. This mechanism adjusts the frequency component parameters related to the power amplifier module in the interference mode reference based on the identified correspondence. For example, the center frequency parameter of a specific harmonic interference might be increased by 100Hz, and its amplitude parameter by 0.5dB. Subsequently, the system updates the interference mode reference in real time based on the adjusted reference and recalculates the complexity index of the interference mode. In this way, even if the instantaneous spectral changes of the interference mode are not significant, the interference mode reference can still be updated promptly and accurately because changes in its physical origins are captured and correlated. This ensures that the subsequent interference separation process can continue effectively and accurately reflect the actual operating status of the equipment.

[0065] In some of the embodiments described above in this application, during the process of updating the interference mode reference in real time and calculating the complexity index of the interference mode to quantify the interference mode, it is necessary to obtain information on the changes in physical parameters of the radio frequency power transmission system components.

[0066] Specifically, obtaining information on changes in the physical parameters of components in a radio frequency power transmission system may include the following steps: Multiple sensors are deployed on key components of the radio frequency power transmission system to continuously collect physical parameter information of the key components; Time series analysis is performed on the collected physical parameter information to extract the long-term variation trend of the physical parameter information; The instantaneous operating condition parameters of the equipment are acquired, and the transient fluctuations of the instantaneous operating condition parameters and the physical parameter information are correlated to identify transient changes caused by fluctuations in operating conditions. Subtracting the transient changes caused by fluctuations in operating conditions from the long-term trend of the physical parameter information yields physical parameter change information that reflects long-term changes caused by component aging.

[0067] Specifically, to accurately obtain information on changes in the physical parameters of components in an RF power transmission system, multiple sensors need to be deployed on key components of the system. These sensors are configured to continuously collect physical parameter information of the key components, such as temperature, vibration, voltage, current, and impedance. By deploying multiple sensors, comprehensive monitoring of the status of key components can be achieved, ensuring the integrity and reliability of the data.

[0068] Furthermore, time series analysis is performed on the collected physical parameter information. Time series analysis is a statistical technique used to analyze a sequence of data points that change over time, with the aim of extracting long-term trends in physical parameter information. For example, moving averages, exponential smoothing, or more complex trend decomposition models can be used to identify slow, gradual changes in component performance over time, which are often related to the aging process of the component.

[0069] In addition, it is necessary to obtain the instantaneous operating parameters of the equipment. These instantaneous parameters may include the equipment's load, operating frequency, ambient temperature, etc., and these parameters cause transient fluctuations in physical parameter information. By performing correlation analysis between instantaneous operating parameters and the transient fluctuations in physical parameter information, transient changes caused by fluctuations in operating conditions can be identified. For example, when the equipment load suddenly increases, the temperature of some components may rise instantaneously. This rise is not caused by component aging, but by changes in operating conditions. Correlation analysis can distinguish between these two types of changes.

[0070] Finally, to obtain information on physical parameter changes that truly reflect the long-term changes caused by component aging, it is necessary to subtract transient changes caused by fluctuations in operating conditions from the long-term trend of physical parameter information. This subtraction operation aims to isolate the influence of transient operating condition changes on physical parameters, thereby revealing the aging state of the component itself more accurately. For example, if the long-term trend shows a slow temperature rise, but includes short-term peaks caused by load fluctuations, subtracting these transient fluctuations can yield a purer temperature rise trend caused by aging.

[0071] This application's solution ensures the accuracy and effectiveness of the acquired information by refining the process of obtaining physical parameter change information of RF power transmission system components step by step. Through the above technical solution, this application provides a more accurate and reliable method for obtaining physical parameter change information of RF power transmission system components. Traditional physical parameter monitoring may struggle to distinguish between long-term changes caused by component aging and transient changes caused by fluctuations in operating conditions, leading to misjudgments of component status or inaccurate updates to interference mode references. This solution effectively eliminates the influence of external interference factors on aging trend judgment by introducing correlation analysis of instantaneous operating parameters and subtracting transient fluctuations from long-term trends. Therefore, the obtained physical parameter change information can more accurately reflect the true aging state of components, providing high-quality input data for the subsequent baseline update mechanism of interference mode references, thereby improving the accuracy of interference mode identification and quantification, and ultimately contributing to enhancing the adaptability and robustness of the entire circuit data processing method.

[0072] In some embodiments of this application, an interference mode reference is updated in real time based on the instantaneous intensity and frequency change of the identified frequency peak, and the complexity index of the interference mode is calculated to quantify the interference mode. This process involves acquiring physical parameter change information of the radio frequency power transmission system components and performing correlation analysis between this physical parameter change information and the interference mode evolution trend in historical operating data to identify the correspondence between the two. Specifically, the aforementioned correlation analysis between physical parameter change information and the interference mode evolution trend in historical operating data to identify the correspondence between physical parameter change information and the interference mode evolution trend may include the following steps: The step of performing correlation analysis between the physical parameter change information and the interference mode evolution trend in historical operational data to identify the correspondence between the physical parameter change information and the interference mode evolution trend includes: Continuously acquire the evolution trend of interference patterns in the historical operational data; Identify the nonlinear or dynamic correlation between the changes in the physical parameters and the evolution trend of the interference mode; Initiate the adaptive association learning mechanism; Based on the real-time data of the current changes in physical parameters and the evolution trend of the interference mode, adjust the parameters of the correlation function; Periodically evaluate the predictive accuracy of the correlation function; When the prediction accuracy drops to a preset threshold, the reconstruction or parameter optimization of the correlation function is triggered.

[0073] Specifically, continuously acquiring the evolution trend of interference patterns in historical operational data refers to the system constantly extracting and updating information about the patterns and characteristics of interference patterns changing over time from the historical database. This can include long-term analysis of historical sensor data to identify periodic, trend-based, or abrupt changes in interference patterns. The purpose is to provide comprehensive background information for subsequent correlation analysis.

[0074] Identifying the nonlinear or dynamic correlation between changes in physical parameters and the evolution trend of interference modes can be understood as the system not merely seeking simple linear relationships, but rather detecting and understanding more complex, time-varying interactions. For example, component aging may have a small impact on interference modes initially, but its impact increases dramatically after reaching a certain critical point; this nonlinear relationship needs to be accurately captured. The aim is to ensure the comprehensiveness and accuracy of the correlation analysis.

[0075] In practical applications, initiating an adaptive association learning mechanism means that the system can automatically select or adjust the method used to build the association model based on real-time changes in the data and the complexity of the associations. For example, machine learning algorithms, such as neural networks, support vector machines, or decision trees, can be used to learn the complex mapping relationship between changes in physical parameters and the evolution of disturbance patterns. The aim is to improve the adaptability and robustness of the association model.

[0076] Furthermore, adjusting the parameters of the correlation function based on real-time data of current physical parameter changes and the evolution trend of interference patterns refers to the system continuously optimizing the established correlation model using the latest data after the learning mechanism is activated. This may include updating the model's weights, bias terms, or other internal parameters to better fit the current data characteristics. The aim is to maintain the timeliness and accuracy of the correlation model.

[0077] Furthermore, periodically evaluating the predictive accuracy of the correlation function refers to the system periodically verifying the performance of the current correlation model in predicting the evolution of disturbance patterns. This can be achieved by comparing the model's predictions with actual observation data and calculating metrics such as mean squared error and R-squared value. The purpose is to monitor the effectiveness of the correlation model.

[0078] When the prediction accuracy drops to a preset threshold, the system triggers a reconstruction or parameter optimization of the association function. This means that once the model's performance falls below an acceptable level, the system will automatically initiate a correction process. Reconstruction may mean building a new association model from scratch, while parameter optimization involves making deeper adjustments to the parameters within the existing model structure. The goal is to ensure that the association model consistently provides reliable association analysis results.

[0079] This application's solution provides a rich data foundation for correlation analysis by continuously acquiring the evolution trend of interference patterns from historical operational data. Based on this, by identifying nonlinear or dynamically changing correlations between physical parameter variations and interference pattern evolution trends, the system can capture more complex and realistic interactions, avoiding over-reliance on simple linear relationships. Furthermore, by initiating an adaptive correlation learning mechanism and adjusting the parameters of the correlation function based on real-time data, the system ensures that the correlation model can dynamically adapt to changes in equipment operating status and environmental conditions. Finally, by periodically evaluating the predictive accuracy of the correlation function and triggering reconstruction or parameter optimization when accuracy declines, a closed-loop adaptive learning process is formed, thereby ensuring that the identification of the correspondence between physical parameter changes and interference pattern evolution remains highly accurate and reliable.

[0080] Through the above technical solution, this application can achieve accurate and dynamic identification of the correspondence between changes in the physical parameters of components in an RF power transmission system and the evolution trend of interference modes. Compared with traditional methods that rely solely on static or simple linear models, this application significantly improves the accuracy and robustness of correlation analysis by introducing nonlinear, dynamic correlation identification and adaptive learning mechanisms. This enables the system to predict changes in interference modes caused by changes in physical parameters such as component aging earlier and more accurately, thus providing a more reliable basis for the baseline update mechanism of interference mode reference. This, in turn, enhances the adaptability and predictive ability of the entire circuit data processing method, effectively ensuring the stability and reliability of equipment operation.

[0081] In some embodiments described above, an adaptive correlation learning mechanism is proposed. This mechanism adjusts the parameters of the correlation function based on real-time data of current physical parameter changes and the evolution trend of interference patterns, and periodically evaluates the prediction accuracy of the correlation function. When the prediction accuracy drops below a preset threshold, the reconstruction or parameter optimization of the correlation function is triggered. However, in practical applications, a decrease in prediction accuracy is often accompanied by a decrease in data quality or insufficient data volume. For example, sensor failure may lead to data loss, or abnormal operating conditions may generate a large number of outliers. If these defective data are used directly for correlation function reconstruction or parameter optimization, the learning effect may be poor, or even new errors may be introduced, thus failing to effectively restore the prediction accuracy of the correlation function and affecting the stability and reliability of the system. To address this, this application further proposes an optimization scheme to ensure that effective learning is based on high-quality and sufficient data when correlation function reconstruction or parameter optimization is triggered, thereby improving the adaptability and robustness of the correlation function in complex and changing environments.

[0082] In response, this application further proposes the aforementioned method of triggering the reconstruction or parameter optimization of the correlation function when the prediction accuracy falls below a preset threshold, including: When the correlation function is reconstructed or the parameters are optimized, the quality of the data used for learning is assessed, and missing data points and outliers in the data are identified and marked. For the identified missing data points, data is filled in based on the trend of data changes before and after the missing data points and the correlation with relevant physical parameters; For the identified outliers, data correction or removal is performed based on the degree and duration of the outlier's deviation from the normal range. When the amount of data after quality assessment and processing is insufficient to support effective learning, synthetic data that matches the characteristics of real data is generated based on the existing evolution patterns of component aging and interference patterns to expand the learning dataset. Based on the expanded dataset, the correlation function is reconstructed or its parameters are optimized.

[0083] Specifically, the data quality assessment aims to systematically examine the dataset used for correlation function learning to identify potential data problems. Missing data points refer to data values ​​that could not be acquired during data acquisition, potentially caused by sensor malfunctions, communication interruptions, or data storage errors. Outliers are data points that significantly deviate from the overall distribution of the dataset, possibly due to transient interference, measurement errors, or abnormal equipment behavior. Identifying and labeling these problematic data points is fundamental to subsequent processing.

[0084] For identified missing data points, data imputation refers to estimating and filling in these missing values ​​using algorithms. This can be achieved through various methods, such as time-series-based interpolation (e.g., linear interpolation, spline interpolation), regression analysis based on relevant physical parameters, or using machine learning models to predict missing values. The goal is to restore data integrity and avoid bias or undertraining of the learning model due to missing data.

[0085] In practical applications, data correction or removal refers to processing outliers based on their characteristics. Correction can involve adjusting outliers to a reasonable range (e.g., by truncation, smoothing, or replacement with nearby normal values), while removal involves removing outliers from the dataset. The choice between correction and removal depends on the degree of impact of the outlier on the overall data distribution and the reason for its occurrence. The goal is to eliminate the negative impact of outliers on model training and improve the model's accuracy and generalization ability.

[0086] Furthermore, when the amount of data after quality assessment and processing is still insufficient to support effective learning—for example, when data accumulation is insufficient in the early stages of system operation, or when data is scarce under specific extreme conditions—this application proposes generating synthetic data that matches the characteristics of real data. This can be achieved through various techniques, such as learning the distribution characteristics of existing data based on generative adversarial networks, variational autoencoders, or statistical models (e.g., Gaussian mixture models) and generating new data with similar statistical properties. The aim is to expand the learning dataset, providing sufficient data support for the reconstruction or parameter optimization of the correlation function, especially when real data is difficult to obtain or too costly.

[0087] Therefore, after obtaining a high-quality dataset that has undergone quality assessment, supplementation, correction, and potential expansion, the reconstruction or parameter optimization of the correlation function can be carried out. This means that a completely new correlation function model can be retrained, or the parameters of an existing model can be finely tuned to better capture the correspondence between changes in physical parameters and the evolution trend of interference patterns.

[0088] This application's solution effectively addresses the limitation of the aforementioned correlation function, where declining prediction accuracy may lead to poor reconstruction or optimization results due to data quality issues or insufficient data volume. Specifically, data quality assessment promptly identifies and locates potential problems in the data, such as missing data points and outliers, laying the foundation for subsequent data processing. For missing data points, data completion is performed based on trends and correlations with relevant physical parameters, maximizing the restoration of data integrity and continuity and avoiding model learning bias due to incomplete information. For outliers, correction or removal effectively eliminates their interference with model training, ensuring the model learns realistic and stable data patterns. Furthermore, if the data volume is still insufficient to support effective learning after the above processing, synthetic data is generated based on existing component aging patterns and interference pattern evolution rules. This intelligently expands the dataset, providing sufficient and representative training samples for correlation function reconstruction or parameter optimization. Therefore, by ensuring that the data used for learning is of high quality and sufficient, the solution of this application enables the correlation function to learn the true correspondence between changes in physical parameters and the evolution of disturbance modes more accurately and effectively during reconstruction or parameter optimization, thereby significantly improving the adaptability and prediction accuracy of the correlation function in the face of complex and ever-changing environments.

[0089] Through the above technical solutions, this application can significantly improve the robustness and reliability of the circuit data processing method based on the adaptive big data model. Especially when the prediction accuracy of the correlation function decreases, this solution can proactively identify and resolve data quality issues, avoiding further performance degradation due to the use of low-quality or insufficient data for model reconstruction or optimization. Data imputation and outlier handling ensure the integrity and accuracy of the training data; synthetic data generation effectively solves the data scarcity problem, providing sufficient learning samples for the model. These measures work together to enable the correlation function to recover its predictive ability more quickly and accurately after reconstruction or parameter optimization, thereby ensuring that the RF power transmission system can continuously and effectively identify and quantify interference patterns during long-term operation, accurately reflecting the true state of the equipment, and thus improving the stability and operating efficiency of the entire system.

[0090] In some preferred embodiments, it is assumed that during long-term operation of the RF power transmission system, due to environmental temperature fluctuations or slight aging of components, the output data of a key sensor may occasionally experience momentary interruptions, resulting in missing data points. Simultaneously, due to a momentary increase in external electromagnetic interference, another sensor may exhibit a short-term abnormally high data value. In this case, based on the prediction accuracy evaluation results of the aforementioned correlation function, it is found that its prediction accuracy has dropped below a preset threshold, triggering the correlation function reconstruction mechanism.

[0091] Specifically, the system first performs a quality assessment on the dataset used for reconstruction. During the assessment, it identifies missing data at multiple consecutive time points caused by sensor interruptions and marks abnormally high values ​​caused by external interference. For missing data points, the system analyzes the data change trend of the sensor before and after the missing time period, and combines this with synchronous changes in other relevant physical parameters (e.g., system power output, ambient temperature, etc.), using time-series-based interpolation algorithms (such as cubic spline interpolation) to complete the data and restore its continuity. For identified abnormally high values, the system determines them to be transient interference based on their degree of deviation from the normal range and their duration, and corrects them by using median filtering or replacing them with the average value of normal data within the same time period to eliminate their negative impact on model training.

[0092] Furthermore, after completing data supplementation and correction, the system evaluation revealed that although data quality had improved, the amount of data available for reconstruction was still insufficient due to inadequate historical data accumulation under this specific operating condition. At this point, the system expands the learning dataset by using a generative adversarial network (GAN) to generate synthetic data that matches the statistical characteristics of real data, based on existing component aging patterns (e.g., the decay of capacitor capacity over time) and interference pattern evolution patterns (e.g., the statistical distribution of interference intensity at a specific frequency with environmental changes). For example, the generator in the GAN learns the distribution of real data, while the discriminator distinguishes between real and synthetic data. Through adversarial training, the generated synthetic data becomes highly similar to real data in multiple statistical dimensions such as mean, variance, and frequency components. Finally, using this improved and expanded dataset, the system reconstructs the correlation function, thereby obtaining a new correlation function model with higher prediction accuracy and robustness under current operating conditions.

[0093] In some embodiments described above, when the amount of data after quality assessment and processing is insufficient to support effective learning, synthetic data is generated based on existing component aging patterns and interference pattern evolution rules to expand the learning dataset. However, in practical applications, simply generating synthetic data based on existing patterns may not adequately simulate the inherent uncertainties, random fluctuations, and complex statistical characteristics of the real world, leading to potential statistical differences between the generated synthetic data and real data. This, in turn, affects the accuracy and robustness of the correlation function learned based on the synthetic data. To address this, this application further proposes a more refined synthetic data generation method. This method quantifies the uncertainty of component aging patterns and interference pattern evolution rules, introduces random perturbations during synthetic data generation, and monitors and adjusts the statistical characteristics of the synthetic data in real time to ensure that the synthetic data more accurately reflects the complexity and randomness of the real world.

[0094] When the amount of data after quality assessment and processing is insufficient to support effective learning, synthetic data consistent with the characteristics of real data is generated based on existing component aging patterns and interference pattern evolution rules to expand the learning dataset, including: Uncertainty quantification is performed on the existing component aging mode and the evolution law of the interference mode to obtain the fluctuation range and distribution characteristics of the evolution law of the interference mode. Based on the fluctuation range and distribution characteristics, random perturbations are introduced when generating the synthetic data to simulate uncertainties and data noise in the real world; The statistical characteristics of the synthesized data are monitored in real time, and the statistical characteristics of the synthesized data are compared with the statistical characteristics of the real data. If discrepancies exist, the intensity and distribution of the random perturbation are adjusted to ensure that the synthetic data accurately reflects the complexity and randomness of the real world.

[0095] Specifically, quantifying the uncertainty of existing component aging and interference patterns involves using methods such as statistical analysis, probabilistic modeling, or expert systems to assess the potential range, frequency, and probability distribution characteristics of these patterns and patterns in actual operation. The aim is to gain a deeper understanding of the inherent randomness and volatility of these patterns, providing a quantitative basis for generating more realistic synthetic data. Based on the quantified range and distribution characteristics of volatility, introducing random perturbations when generating synthetic data can be understood as superimposing random noise or fluctuations conforming to a specific probability distribution onto the base data generated from existing patterns. For example, Gaussian noise, Poisson noise, or other stochastic process models that conform to actual physical processes can be used. The goal is to simulate unpredictable minute changes and environmental noise in the real world, making the synthetic data more realistic and diverse.

[0096] In practical applications, real-time monitoring of the statistical characteristics of synthetic data and comparison with those of real data involves continuously calculating the characteristic values ​​of synthetic data across multiple statistical dimensions, such as mean, variance, skewness, kurtosis, and autocorrelation, and comparing them with the corresponding statistical characteristics of real data from the same period or historical periods. The purpose is to quantify the statistical differences between synthetic and real data, providing an objective basis for subsequent adjustments.

[0097] Furthermore, if the comparison results show differences, the strength and distribution of the random perturbation are adjusted. This refers to dynamically modifying the parameters of the random perturbation introduced when generating the synthetic data based on the degree and direction of the differences in statistical characteristics. For example, if the variance of the synthetic data is less than that of the real data, the strength of the random perturbation can be increased; if the skewness is mismatched, the distribution type of the random perturbation can be adjusted. The aim is to use iterative optimization to make the statistical characteristics of the synthetic data approximate those of the real data as closely as possible, thereby ensuring that it accurately reflects the complexity and randomness of the real world.

[0098] This application's solution effectively addresses the limitations of existing synthetic data generation methods in simulating the complexity and randomness of the real world by introducing uncertainty quantification, random perturbation, and real-time statistical monitoring and adjustment mechanisms. Through these technical solutions, this application can generate synthetic data that is highly consistent with real data in terms of statistical characteristics, significantly improving the quality and reliability of the synthetic data. Compared to methods that generate synthetic data solely based on existing models, this application, by introducing uncertainty quantification and random perturbation, enables the synthetic data to more comprehensively reflect the complexity and randomness of the real world, avoiding the problem of insufficient model generalization ability caused by overly idealized data. Furthermore, the real-time monitoring and adjustment mechanism ensures that the synthetic data can dynamically adapt to changes in the characteristics of real data, further enhancing the robustness and adaptability of the synthetic data. Therefore, the expanded learning dataset can more effectively support the learning and optimization of correlation functions, thereby improving the accuracy and stability of self-adjusting system parameter adjustments, ultimately enhancing the overall performance of circuit data processing.

[0099] In some preferred embodiments, it is assumed that synthetic sensor data needs to be generated for the radio frequency power transmission system to train a model predicting component aging. First, the system analyzes historical component aging data and the evolution of interference patterns. For example, by fitting a probability density function to the historical data, the system quantifies the aging rate fluctuation range of a specific component at different operating temperatures and the distribution characteristics of interference signal frequency drift. Based on these quantification results, when generating synthetic sensor data, the system superimposes random noise conforming to these distribution characteristics onto the simulated aging trend and interference frequency.

[0100] For example, when simulating the aging curve of a component, its aging rate is no longer fixed, but rather a random value with a specific standard deviation is superimposed on the average value. Simultaneously, the system calculates the characteristics of the generated synthetic data in real time across multiple statistical dimensions, such as mean, variance, skewness, and kurtosis, and compares them with the statistical characteristics of real sensor data collected during actual operation. If the variance of the synthetic data is found to be significantly smaller than that of the real data, indicating a lack of sufficient volatility, the system will correspondingly increase the intensity of the generated random perturbation. Conversely, if the skewness of the synthetic data deviates significantly from that of the real data, the system may adjust the distribution type of the random perturbation to better reflect the asymmetry of the real data. Through this continuous monitoring and adjustment, it is ensured that the generated synthetic data not only matches the real data in macroscopic trends but also closely matches its microscopic statistical characteristics and randomness, thus providing high-quality data support for training more accurate and robust prediction models.

[0101] Specifically, the process of monitoring the statistical characteristics of synthetic data in real time and comparing them with the statistical characteristics of real data can be further broken down into the following steps: The synthetic data and the real data are acquired in real time with feature information on multiple statistical dimensions, including mean, variance, skewness, kurtosis and time series autocorrelation. The feature information on the multiple statistical dimensions is evaluated for difference to identify the statistical feature differences between the synthetic data and the real data, and to quantify the degree of the statistical feature differences. Based on the degree of difference in the statistical characteristics, determine whether the difference is a single-dimensional difference, a multi-dimensional synchronous difference, or a multi-dimensional asynchronous difference. When it is determined that the difference is the multidimensional asynchronous difference, the dominant statistical dimension in the multidimensional asynchronous difference is identified, and the contribution weight of the dominant statistical dimension to the difference is calculated. Based on the contribution weight of the dominant statistical dimension to the difference, adjustment suggestions are generated to adjust the strength and distribution of the random disturbance.

[0102] Specifically, real-time acquisition of the characteristics of synthetic and real data across multiple statistical dimensions refers to the system continuously collecting and calculating key statistical indicators of synthetic and real data. These characteristics, such as mean, variance, skewness, kurtosis, and time series autocorrelation, comprehensively reflect the central tendency, dispersion, distribution shape, and temporal correlation of the data. Specifically, the mean represents the average level of the data; variance measures the range of data fluctuation; skewness describes the symmetry of the data distribution; kurtosis reflects the sharpness of the data distribution; and time series autocorrelation reveals the interdependence of the data at different points in time. By acquiring these multi-dimensional statistical characteristics, a comprehensive data foundation can be provided for subsequent discrepancy assessments.

[0103] Furthermore, a difference assessment is performed on the feature information across the multiple statistical dimensions. This aims to systematically compare the performance of synthetic data and real data across each statistical dimension, thereby identifying whether significant differences in statistical characteristics exist between the two and quantifying the degree of these differences. For example, statistical hypothesis testing (such as t-tests, chi-square tests, or Kolmogorov-Smirnov tests) can be used to determine whether there are significant differences in statistical characteristics across different dimensions, or distance metrics (such as Euclidean distance or Mahalanobis distance) can be used to quantify the magnitude of these differences. In this way, it becomes clear in which aspects the synthetic data deviates from the real data, and the severity of these deviations.

[0104] Based on this, and according to the degree of difference in the statistical characteristics, the difference can be determined as a single-dimensional difference, a multi-dimensional synchronous difference, or a multi-dimensional asynchronous difference. A single-dimensional difference indicates a significant deviation only in a specific statistical characteristic, while other characteristics perform well. A multi-dimensional synchronous difference refers to multiple statistical characteristics showing deviations simultaneously, and these deviations exhibit consistent trends or directions. Multi-dimensional asynchronous differences are more complex, manifesting as multiple statistical characteristics showing deviations simultaneously, but these deviations exhibit inconsistent trends or directions, potentially influencing each other or evolving independently. This classification helps to more accurately understand the nature of the difference and provides direction for subsequent adjustments.

[0105] When the discrepancy is determined to be a multi-dimensional asynchronous discrepancy, effective adjustments require identifying the dominant statistical dimension within this discrepancy and calculating its contribution weight. This means the system needs to further analyze the contribution of each statistical dimension to the overall discrepancy, for example, through sensitivity analysis, principal component analysis, or feature importance assessment, to identify the statistical dimensions that have the greatest impact on the difference between synthetic and real data. Calculating the contribution weight quantifies the importance of each dominant dimension in causing the overall discrepancy, thereby guiding subsequent adjustment strategies.

[0106] Finally, based on the contribution weight of the dominant statistical dimension to the differences, adjustment suggestions are generated to adjust the strength and distribution of the random perturbation. These adjustment suggestions are optimization schemes for the random perturbation parameters introduced during the synthesis data generation process. For example, if the mean difference is large, it may be necessary to adjust the central parameter of the random perturbation; if the variance difference is large, it may be necessary to adjust the strength or range of the random perturbation; if the skewness or kurtosis difference is large, it may be necessary to adjust the distribution type or parameter of the random perturbation. Through these targeted adjustment suggestions, it can be ensured that the synthetic data more accurately simulates the complexity and randomness of real data statistically.

[0107] This application's solution, through multi-dimensional and refined statistical feature comparison of synthetic and real data, and by classifying differences and identifying dominant dimensions, enables a deep understanding of the sources of deviation between synthetic and real data. It is precisely this meticulous analysis that allows the system to generate targeted adjustment suggestions, thereby precisely optimizing the strength and distribution of random perturbations during the synthetic data generation process. As a result, synthetic data can more accurately reflect the complexity and randomness of the real world, effectively compensating for the deficiencies of real data.

[0108] Through the aforementioned technical solution, the system can achieve precise control and continuous optimization of the quality of synthetic data. This solution not only identifies statistical differences between synthetic and real data, but also further categorizes these differences in detail and identifies the dominant statistical dimensions causing the differences and their contribution weights. This in-depth analysis makes the generation of adjustment suggestions more targeted and effective, avoiding blind adjustments. Therefore, synthetic data can more accurately simulate the statistical characteristics and randomness of real data, significantly improving the authenticity and usability of synthetic data, providing higher-quality input data for the subsequent self-adjusting system, thereby improving the robustness and accuracy of the entire circuit data processing method.

[0109] In some embodiments described above in this application, when the statistical characteristics of synthetic data differ from those of real data, the intensity and distribution of random perturbations are adjusted to ensure that the synthetic data accurately reflects the complexity and randomness of the real world. However, in practical applications, when the contribution weights of multiple statistical dimensions to the overall difference are similar, simple adjustments may lead to conflicting adjustment objectives, or optimize one dimension while worsening the performance of other dimensions, making it difficult to achieve a comprehensive and accurate reflection of the complexity and randomness of the real world by the synthetic data.

[0110] In response, this application further proposes adjusting the intensity and distribution of the random perturbation if discrepancies exist to ensure that the synthetic data accurately reflects the complexity and randomness of the real world, including: The weights of the contributions of multiple dominant statistical dimensions to the overall difference are similar; Initiate a conflict assessment mechanism to analyze the direction and extent of the influence of the adjustment targets of each of the dominant statistical dimensions on the strength and distribution of the random disturbance; A priority allocation strategy is introduced to assign and adjust the priority of each dominant statistical dimension based on preset weights or the sensitivity of the current system to specific statistical features. Based on the adjustment priority, the intensity and distribution of the random disturbance are adjusted in stages or iteratively to meet the adjustment requirements of high-priority dimensions. Real-time monitoring of changes in the synthetic data across all relevant statistical dimensions during the adjustment process; The adjustment strategy is fine-tuned based on the monitoring results, so as to gradually optimize the adjustment of other dimensions without significantly worsening the already optimized dimensions.

[0111] Specifically, identifying that the contribution weights of multiple dominant statistical dimensions to the overall difference are similar means that, after comparing the statistical characteristics of synthetic and real data, it is found that the differences in multiple statistical dimensions (such as mean, variance, skewness, kurtosis, or time series autocorrelation) are all significant, and their contribution weights to the overall difference are close to each other within a preset threshold range. This indicates that these dimensions need to be given special attention and adjustment.

[0112] The mechanism involves initiating a conflict assessment to analyze the direction and extent of the impact of the adjustment targets of each dominant statistical dimension on the strength and distribution of the random disturbance. This can be understood as simulating adjustments to each dominant statistical dimension to predict how changes to random disturbance parameters (such as mean, standard deviation, and distribution type) will affect other statistical dimensions. For example, increasing the strength of the random disturbance might improve the fit of variance, but it might simultaneously worsen the fit of skewness. This mechanism aims to quantify these potential interactions and conflicts.

[0113] In practical applications, a priority allocation strategy is introduced. Based on preset weights or the current system's sensitivity to specific statistical features, an adjustment priority is assigned to each dominant statistical dimension. For example, based on business needs or data sensitivity, the matching degree of mean and variance can be set as high priority, while skewness and kurtosis can be set as secondary priority. Preset weights can be determined based on expert experience or historical data analysis, while system sensitivity reflects the current system's tolerance to changes in specific statistical features.

[0114] Furthermore, based on the adjustment priority, the intensity and distribution of the random disturbance are adjusted in stages or iteratively to meet the adjustment requirements of high-priority dimensions. This means first adjusting the highest-priority statistical dimension until its difference meets the requirements or reaches the optimal state. Then, while maintaining the optimization results of high-priority dimensions, the next lower-priority dimensions are gradually adjusted. This adjustment can be staged, i.e., completing the adjustment of one dimension before moving to the next; or it can be iterative, i.e., considering all dimensions simultaneously in each iteration, but prioritizing the adjustment goals of high-priority dimensions. In addition, real-time monitoring of the changes of the synthetic data across all relevant statistical dimensions during the adjustment process ensures the transparency and controllability of the adjustment process. Through continuous monitoring, unexpected impacts of the adjustment on other dimensions can be detected in a timely manner, providing a basis for subsequent fine-tuning.

[0115] Therefore, fine-tuning the adjustment strategy based on monitoring results, in order to gradually optimize the adjustment of other dimensions without significantly worsening the already optimized dimensions, means that if it is found that the optimization of a certain dimension has led to a significant deterioration of other already optimized dimensions, then the adjustment strategy needs to be re-evaluated and refined, such as reducing the adjustment step size, changing the adjustment direction, or reallocating priorities, in order to seek the overall optimal solution.

[0116] This application's solution effectively addresses the conflicts and suboptimal problems that may arise from simple adjustments when multiple statistical dimensions have similar contribution weights to the differences between synthetic and real data, by introducing a refined adjustment strategy. Through this technical solution, the application can more intelligently and precisely adjust the random perturbation parameters during the synthetic data generation process. Especially when facing multi-dimensional and complex difference scenarios, this solution effectively avoids adjustment conflicts, ensuring that while optimizing one or more key statistical features, it maximizes the matching degree of other statistical features, thereby significantly improving the accuracy and robustness of synthetic data in simulating the complexity and randomness of the real world. This allows the generated synthetic data to more accurately reflect the inherent patterns and fluctuation characteristics of real data when expanding the learning dataset, providing high-quality training data for subsequent correlation function reconstruction or parameter optimization, thus improving the accuracy and adaptability of the entire circuit data processing method.

[0117] refer to Figure 3 , Figure 3 This is a schematic diagram of the structure of a circuit data processing system based on an adaptive big data model provided in an embodiment of the present invention, including: The input terminal is used to acquire the energy input parameters of the device operation and the periodic characteristics of the energy input parameters; and to process sensor data affected by external interference to obtain the frequency components in the sensor data. The identification end is used to identify and quantify the interference pattern corresponding to the periodic feature from the frequency components in the sensor data; The adjustment end is used to separate the interference mode from the sensor data based on the quantization result of the interference mode, so as to reflect the data stream of the actual state of the device; and to provide the data stream reflecting the actual state of the device to the self-adjustment system to adjust the parameters of the self-adjustment system.

[0118] This system employs a modular design, assigning key steps in the data processing flow to specialized functional modules to achieve efficient and accurate processing of circuit data. The input end handles data acquisition and preliminary processing, the identification end focuses on the identification and quantification of interference patterns, and the adjustment end is responsible for interference separation and data stream output. This structured processing approach effectively addresses the misjudgments and incorrect system parameter adjustments caused by data contamination in traditional methods, significantly improving the accuracy of data processing and the system's adaptability, providing a more reliable and efficient solution for the intelligent operation and maintenance of complex equipment.

[0119] In some embodiments of this application, the various functional modules in the above-described circuit data processing system are designed to work together to achieve accurate processing of circuit data.

[0120] Specifically, the input terminal is configured to acquire the energy input parameters of the device operation and the periodic characteristics of the energy input parameters, and to process sensor data affected by external interference to obtain the frequency components in the sensor data. This input terminal may include a series of sensor interfaces, a data acquisition unit, and a preprocessing module. For example, the sensor interface can be used to connect to a high-precision energy meter or oscilloscope to monitor the voltage and current waveforms of the device in real time; the data acquisition unit is responsible for converting analog signals into digital signals; the preprocessing module can perform signal processing operations such as Fourier transform or wavelet transform to convert time-domain data to the frequency domain, thereby revealing the various frequency components and their intensities contained in the data. The specific acquisition and processing methods of energy input parameters, periodic characteristics, and sensor data have been described in detail in other embodiments and will not be repeated here. It is important to emphasize that the design of the input terminal aims to ensure that the raw data can be accurately and completely acquired and initially converted into a frequency domain form suitable for subsequent analysis.

[0121] The identification unit is configured to identify and quantify interference patterns corresponding to the periodic features from the frequency components of the sensor data. This identification unit can be a standalone signal analysis module or an embedded processor running specific algorithms. For example, the module can perform spectrum comparison, pattern matching, or machine learning-based classification algorithms to detect frequency peaks in the sensor data that are highly correlated with the fundamental frequency, harmonic frequencies, or other periodic features of the energy input parameters. Quantifying the interference patterns may include calculating parameters such as the amplitude, phase, and bandwidth of these frequency peaks to form a mathematical description of the interference patterns. Details regarding the identification and quantification of interference patterns have been described in detail in other embodiments and will not be repeated here. The design goal of the identification unit is to accurately identify and quantify structured interference caused by external factors in the data, providing an accurate basis for subsequent interference separation.

[0122] The adjustment terminal is configured to separate the interference pattern from the sensor data based on the quantization result of the interference pattern, to obtain a data stream reflecting the true state of the device; and to provide the data stream reflecting the true state of the device to the self-adjustment system to adjust the parameters of the self-adjustment system. This adjustment terminal may include an adaptive filter module, a notch filter module, or a data output interface. For example, the adaptive filter module can dynamically adjust the characteristics of the filter based on the interference pattern parameters quantized by the identification terminal, thereby filtering out interference signals from the original sensor data. The data output interface is responsible for formatting the purified data stream and transmitting it to the self-adjustment system. Details regarding the separation of interference patterns and the provision and adjustment of the data stream have been described in detail in other embodiments and will not be repeated here. The design of the adjustment terminal aims to ensure that the output data stream can truly and accurately reflect the physical state of the device, avoiding misjudgments caused by interference signals, thereby providing reliable input to the self-adjustment system, enabling it to learn and make decisions based on more accurate information.

[0123] The circuit data processing system of this application, through its modular input, identification, and adjustment terminals, achieves in-depth processing of interference patterns in circuit data. Compared with the shortcomings of existing technologies that treat all input data as real signals, leading to the adaptive system's erroneous learning and fitting of "pseudo-patterns," the system of this application can actively identify and quantify interference patterns related to the periodic characteristics of energy input parameters and separate them from the original data. This "spurious-to-true" processing method enables the self-adjusting system to receive cleaner and more accurate data, thereby avoiding the problem of "over-learning" interference signals and significantly improving the system's robustness and prediction accuracy. For example, in the case of plasma etching machines in semiconductor manufacturing plants, the system of this application can accurately capture and quantify electromagnetic interference noise synchronized with the RF power cycle through the identification terminal and effectively separate it through the adjustment terminal. This allows the self-adjusting system to accurately determine the true state of the vacuum system, avoiding resource waste and production interruptions caused by false alarms. Therefore, the system of this application not only solves the problem of data pollution causing adaptive system failure in existing technologies but also provides a more reliable and efficient solution for the intelligent operation and maintenance of complex equipment.

[0124] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A circuit data processing method based on an adaptive big data model, characterized in that, include: Obtain the energy input parameters for device operation and the periodicity characteristics of the energy input parameters; The sensor data affected by external interference is processed to obtain the frequency components in the sensor data; Identify and quantify the interference patterns corresponding to the periodic features from the frequency components of the sensor data; Based on the quantification results of the interference pattern, the interference pattern is separated from the sensor data to reflect the data stream of the device's true state. The data stream reflecting the actual state of the device is provided to the self-adjusting system to adjust the parameters of the self-adjusting system.

2. The circuit data processing method based on an adaptive big data model according to claim 1, characterized in that, The step of identifying and quantifying the interference pattern corresponding to the periodic feature from the frequency components of the sensor data includes: The instantaneous waveform of the energy input parameter is continuously monitored, and real-time spectrum analysis is performed on the instantaneous waveform to obtain the instantaneous frequency drift, amplitude fluctuation rate, and non-harmonic intermodulation products of the energy input parameter. A short-time Fourier transform is performed on the sensor data affected by external interference to obtain the time-varying frequency components and intensities in the sensor data; Based on the instantaneous frequency drift, the amplitude fluctuation rate, and the non-harmonic intermodulation products, the identification window for the interference frequency is dynamically adjusted, and the frequency peaks in the sensor data related to the fundamental frequency, harmonic frequency, and non-harmonic intermodulation products of the energy input parameters are identified. Based on the instantaneous intensity and frequency change of the identified frequency peak, the interference mode reference is updated in real time, and the complexity index of the interference mode is calculated to quantify the interference mode.

3. The circuit data processing method based on an adaptive big data model according to claim 2, characterized in that, The step of updating the reference of the interference mode in real time based on the instantaneous intensity and frequency change of the identified frequency peak, and calculating the complexity index of the interference mode to quantify the interference mode, includes: Acquire information on changes in the physical parameters of components in the radio frequency power transmission system; The physical parameter change information is correlated with the interference mode evolution trend in historical operation data to identify the correspondence between the physical parameter change information and the interference mode evolution trend. Based on the aforementioned correspondence, the baseline update mechanism for the interference mode reference is triggered; According to the baseline update mechanism, the frequency component parameters in the interference mode reference are adjusted, and the interference mode reference is updated in real time according to the adjusted interference mode reference. The complexity index of the interference mode is calculated to quantify the interference mode.

4. The circuit data processing method based on an adaptive big data model according to claim 3, characterized in that, The acquisition of physical parameter change information of radio frequency power transmission system components includes: Multiple sensors are deployed on key components of the radio frequency power transmission system to continuously collect physical parameter information of the key components; Time series analysis is performed on the collected physical parameter information to extract the long-term variation trend of the physical parameter information; The instantaneous operating condition parameters of the equipment are acquired, and the transient fluctuations of the instantaneous operating condition parameters and the physical parameter information are correlated to identify transient changes caused by fluctuations in operating conditions. Subtracting the transient changes caused by fluctuations in operating conditions from the long-term trend of the physical parameter information yields physical parameter change information that reflects long-term changes caused by component aging.

5. The circuit data processing method based on an adaptive big data model according to claim 3, characterized in that, The step of performing correlation analysis between the physical parameter change information and the interference mode evolution trend in historical operational data to identify the correspondence between the physical parameter change information and the interference mode evolution trend includes: Continuously acquire the evolution trend of interference patterns in the historical operational data; Identify the nonlinear or dynamic correlation between the changes in the physical parameters and the evolution trend of the interference mode; Initiate the adaptive association learning mechanism; Based on the real-time data of the current changes in physical parameters and the evolution trend of the interference mode, adjust the parameters of the correlation function; Periodically evaluate the predictive accuracy of the correlation function; When the prediction accuracy drops to a preset threshold, the reconstruction or parameter optimization of the correlation function is triggered.

6. The circuit data processing method based on an adaptive big data model according to claim 5, characterized in that, When the prediction accuracy drops below a preset threshold, triggering the reconstruction or parameter optimization of the correlation function includes: When the correlation function is reconstructed or the parameters are optimized, the quality of the data used for learning is assessed, and missing data points and outliers in the data are identified and marked. For the identified missing data points, data is filled in based on the trend of data changes before and after the missing data points and the correlation with relevant physical parameters; For the identified outliers, data correction or removal is performed based on the degree and duration of the outlier's deviation from the normal range. When the amount of data after quality assessment and processing is insufficient to support effective learning, synthetic data that matches the characteristics of real data is generated based on the existing evolution patterns of component aging and interference patterns to expand the learning dataset. Based on the expanded dataset, the correlation function is reconstructed or its parameters are optimized.

7. The circuit data processing method based on an adaptive big data model according to claim 6, characterized in that, When the amount of data after quality assessment and processing is insufficient to support effective learning, synthetic data consistent with the characteristics of real data is generated based on existing component aging patterns and interference pattern evolution rules to expand the learning dataset, including: Uncertainty quantification is performed on the existing component aging mode and the evolution law of the interference mode to obtain the fluctuation range and distribution characteristics of the evolution law of the interference mode. Based on the fluctuation range and distribution characteristics, random perturbations are introduced when generating the synthetic data to simulate uncertainties and data noise in the real world; The statistical characteristics of the synthesized data are monitored in real time, and the statistical characteristics of the synthesized data are compared with the statistical characteristics of the real data. If discrepancies exist, the intensity and distribution of the random perturbation are adjusted to ensure that the synthetic data accurately reflects the complexity and randomness of the real world.

8. The circuit data processing method based on an adaptive big data model according to claim 7, characterized in that, The real-time monitoring of the statistical characteristics of the synthetic data, and the comparison of the statistical characteristics of the synthetic data with the statistical characteristics of the real data, includes: The synthetic data and the real data are acquired in real time with feature information on multiple statistical dimensions, including mean, variance, skewness, kurtosis and time series autocorrelation. The feature information on the multiple statistical dimensions is evaluated for difference to identify the statistical feature differences between the synthetic data and the real data, and to quantify the degree of the statistical feature differences. Based on the degree of difference in the statistical characteristics, determine whether the difference is a single-dimensional difference, a multi-dimensional synchronous difference, or a multi-dimensional asynchronous difference. When it is determined that the difference is the multidimensional asynchronous difference, the dominant statistical dimension in the multidimensional asynchronous difference is identified, and the contribution weight of the dominant statistical dimension to the difference is calculated. Based on the contribution weight of the dominant statistical dimension to the difference, adjustment suggestions are generated to adjust the strength and distribution of the random disturbance.

9. A circuit data processing method based on an adaptive big data model according to claim 7, characterized in that, If discrepancies exist, the intensity and distribution of the random perturbation are adjusted to ensure that the synthetic data accurately reflects the complexity and randomness of the real world, including: The weights of the contributions of multiple dominant statistical dimensions to the overall difference are similar; Initiate a conflict assessment mechanism to analyze the direction and extent of the influence of the adjustment targets of each of the dominant statistical dimensions on the strength and distribution of the random disturbance; A priority allocation strategy is introduced to assign and adjust the priority of each dominant statistical dimension based on preset weights or the sensitivity of the current system to specific statistical features. Based on the adjustment priority, the intensity and distribution of the random disturbance are adjusted in stages or iteratively to meet the adjustment requirements of high-priority dimensions. Real-time monitoring of changes in the synthetic data across all relevant statistical dimensions during the adjustment process; The adjustment strategy is fine-tuned based on the monitoring results, so as to gradually optimize the adjustment of other dimensions without significantly worsening the already optimized dimensions.

10. A circuit data processing system based on an adaptive big data model, characterized in that, include: The input terminal is used to acquire the energy input parameters of the device operation and the periodic characteristics of the energy input parameters; The sensor data affected by external interference is processed to obtain the frequency components in the sensor data; The identification end is used to identify and quantify the interference pattern corresponding to the periodic feature from the frequency components in the sensor data; The adjustment end is used to separate the interference mode from the sensor data based on the quantization result of the interference mode, so as to reflect the data stream of the actual state of the device; The data stream reflecting the actual state of the device is provided to the self-adjusting system to adjust the parameters of the self-adjusting system.

Images