Dynamic optimization method and system for digital twin-driven process flow

Abstract

The invention relates to the technical field of data processing, in particular to a method and a system for dynamically optimizing a process flow of digital twin drive. The method comprises the following steps of: constructing a digital twin model by acquiring temperature, pressure and flow data of chip etching production equipment in real time, and identifying abnormal fluctuation of the data and quantifying an etching state by adopting time sequence decomposition analysis based on the model; performing prediction error correction according to the quantization result, determining the monitoring level of each parameter, and calculating an optimized parameter control signal according to the trend slope; performing scheme priority ranking on the optimized signals through data interpolation and periodic pattern recognition, and if configuration redundancy is detected, performing resource allocation balance optimization to obtain simplified control configuration; temperature threshold fine adjustment and pressure linkage synchronization are carried out on the simplified configuration, a precision management instruction is generated, equipment parameters are updated through closed-loop control, and precise control over the production process is achieved. The process parameters can be dynamically adjusted, and the production efficiency and the product quality can be optimized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for dynamic optimization of process flow driven by digital twins. Background Technology

[0002] As the manufacturing industry continues to demand higher production efficiency and product quality, traditional process control methods are no longer sufficient to meet the needs of modern production. In particular, in the chip manufacturing process with micron- and nanometer-level precision requirements, chip etching, as a key process in integrated circuit manufacturing, places extremely high demands on the precise control of various parameters such as temperature, pressure, and flow rate during the production process.

[0003] However, in actual production, changes in the production environment, equipment aging, and subtle differences in operation can all lead to fluctuations in process parameters, thereby affecting product quality and production efficiency. Traditional process control methods often rely on experience and fixed rules for parameter adjustment. This method is not only inefficient but also lacks sufficient flexibility and precision when facing complex production conditions. Therefore, a more precise and dynamic optimization control method is needed to ensure that process parameters can be adjusted in real time in complex and changing production environments, avoiding defects and improving production efficiency.

[0004] To address this issue, this invention proposes a digital twin-driven dynamic optimization method for process flows. This method constructs a digital twin model by real-time acquisition of key parameter data from production equipment, such as temperature, pressure, and flow rate in chip etching equipment. This model is then used for data analysis to identify abnormal fluctuations in the production process and quantify the current production status. Through prediction error correction and trend analysis, the method can dynamically adjust the control signals of various process parameters, further optimizing the production process and ensuring product quality stability and improved production efficiency. Summary of the Invention

[0005] This invention provides a digital twin-driven method and system for dynamic optimization of process flow, which optimizes production processes and improves product quality and production efficiency through real-time data acquisition and dynamic analysis, and is especially suitable for high-precision production processes such as chip etching.

[0006] In a first aspect, the present invention provides a digital twin-driven method for dynamic optimization of process flow, comprising: Step S1: Obtain real-time data of temperature, pressure, and flow parameters from the chip etching production equipment through a sensor network, construct a digital twin model, and use time series decomposition analysis based on the digital twin model to analyze abnormal fluctuations in the real-time data and obtain a quantitative representation of the current etching state. Step S2: Correct the prediction error based on the quantization representation and determine the real-time dependency monitoring level of each parameter; if the real-time dependency monitoring level is higher than the preset threshold, process the real-time data by calculating the trend slope to obtain the optimized parameter control signal; Step S3: Extract data interpolation and filling from the optimized parameter control signal, evaluate the data interpolation and filling using periodic pattern recognition, and generate a priority ranking of the schemes; Step S4: If the priority sorting of the scheme indicates the elimination of redundant configurations, then a simplified control configuration is obtained through iterative optimization of resource allocation balancing. Step S5: Fine-tune the temperature threshold and synchronize the pressure linkage for the simplified control configuration to generate the final precision management command; update the parameter cycle pattern recognition of the production equipment through the final precision management command to obtain the closed-loop control response.

[0007] As a preferred embodiment of the present invention, step S1 includes: Real-time data streams of temperature, pressure, and flow rate are collected by a sensor network distributed across the production equipment. The real-time data streams are preprocessed to obtain standardized time-series data. A digital twin model is constructed based on the standardized time-series data and the structural information of the production equipment. Based on the digital twin model, time-series segments are extracted from the time-series data using a sliding window. Feature extraction is performed on the time-series segments to generate feature vectors containing temperature, pressure, and flow rate parameters. The dynamic correlation between parameters is obtained through the feature vector, and abnormal fluctuations in the real-time data stream are detected through the dynamic correlation. A quantitative representation of the current etching state is generated based on the abnormal fluctuations, wherein the quantitative representation includes the fluctuation amplitude and frequency characteristics of temperature, pressure, and flow rate.

[0008] As a preferred embodiment of the present invention, step S2, which involves correcting the prediction error based on the quantization representation, includes: The time series features of each parameter are extracted based on the quantization representation. Based on the time series features, the expected values ​​of each parameter are predicted using an autoregressive model. The prediction error of each parameter is calculated by comparing the expected values ​​with the real-time data. The time series features are then smoothed using the prediction error to obtain a smoothed parameter sequence. For the smoothed parameter sequence, the real-time dependency monitoring level of each parameter in the etching process is determined, and a dependency weight matrix between parameters is generated based on the real-time dependency monitoring level. The smoothed parameter sequence is then adjusted according to the dependency weight matrix to obtain an optimized parameter sequence for nanoscale defects.

[0009] As a preferred embodiment of the present invention, in step S2, if the real-time dependency monitoring level is higher than a preset threshold, the real-time data is processed by calculating the trend slope, including: Determine whether the real-time dependency monitoring level is higher than a preset threshold; if it is higher than the preset threshold, extract the trend features of the real-time data; calculate the slope of the trend features using the least squares method to obtain a trend slope sequence; The trend slope sequence is used to identify the change patterns in the real-time data; corresponding parameter adjustment vectors are generated for the change patterns; the real-time data is corrected using the parameter adjustment vectors to obtain optimized parameter control signals; and the optimized parameter control signals are used to determine the parameter optimization direction in the etching process.

[0010] As a preferred embodiment of the present invention, step S3, extracting data interpolation and filling from the optimized parameter control signal, includes: Key control points are extracted from the optimized parameter control signal; for the key control points, a continuous parameter curve is generated using spline interpolation; missing data points are filled in according to the continuous parameter curve to obtain a complete parameter sequence; the complete parameter sequence is analyzed using a periodic pattern recognition algorithm to extract periodic features; the stability of the data interpolation filling is evaluated based on the periodic features; the stability generation schemes are prioritized; and a parameter optimization scheme for nanoscale defects is determined based on the priority ranking of the schemes.

[0011] As a preferred embodiment of the present invention, step S4 includes: Determine whether the priority ranking of the scheme indicates configuration redundancy elimination; if it indicates configuration redundancy elimination, extract the redundant parameter configurations from the priority ranking of the scheme; for the redundant parameter configurations, use a resource allocation balancing algorithm to calculate the resource occupancy weights between the parameters; eliminate the redundant parameter configurations according to the resource occupancy weights to generate a preliminary control configuration; The initial control configuration is adjusted through an iterative optimization loop to obtain a simplified control configuration; a resource allocation scheme is generated based on the simplified control configuration; and the stability of the simplified control configuration is verified through the resource allocation scheme.

[0012] As a preferred embodiment of the present invention, step S5, which involves fine-tuning the temperature threshold and synchronizing pressure linkage for the simplified control configuration, includes: The control ranges of temperature and pressure parameters are extracted from the simplified control configuration; for the control range of the temperature parameter, the threshold is fine-tuned using a gradient descent algorithm to obtain the optimized temperature threshold; Based on the optimized temperature threshold, determine the linkage synchronization range of the pressure parameters; generate a coordinated control strategy for temperature and pressure through the linkage synchronization range; adjust the simplified control configuration according to the coordinated control strategy to generate the final precision management command; evaluate the compliance of the final precision management command through verification indicators; and generate a control command sequence for the etching process based on the compliance.

[0013] As a preferred embodiment of the present invention, step S5, updating the parameter cycle pattern recognition of the production equipment through the final precision management instruction, includes: Extract the control parameter sequence from the final precision management instruction; generate corresponding parameter adjustment instructions for the control parameter sequence; update the parameter configuration of the production equipment using the parameter adjustment instructions; analyze the operating status of the production equipment using a periodic pattern recognition algorithm based on the updated parameter configuration; generate a closed-loop control response based on the operating status; adjust the execution frequency of the parameter adjustment instructions based on the closed-loop control response; and optimize the parameter periodic pattern recognition of the production equipment using the execution frequency.

[0014] As a preferred embodiment of the present invention, step S5, obtaining the closed-loop control response, includes: Extract the collaborative application fusion loop from the final precision management instruction; generate parameter update instructions for the production equipment based on the collaborative application fusion loop; adjust the operating parameters of the production equipment using the parameter update instructions; collect real-time operating data of the production equipment based on the adjusted operating parameters; match the real-time operating data with the collaborative application fusion loop to generate a closed-loop control response; update the parameter cycle mode of the production equipment based on the closed-loop control response; verify the stability of the closed-loop control response using the parameter cycle mode.

[0015] Secondly, the present invention also provides a digital twin-driven process flow dynamic optimization system for implementing the above-mentioned method, the system comprising: The modeling unit is used to acquire real-time data of temperature, pressure, and flow parameters from the chip etching production equipment through a sensor network, construct a digital twin model, and use time series decomposition analysis based on the digital twin model to analyze abnormal fluctuations in the real-time data and obtain a quantitative representation of the current etching state. The correction analysis unit is used to correct prediction errors based on the quantization representation and determine the real-time dependency monitoring level of each parameter; if the real-time dependency monitoring level is higher than a preset threshold, the real-time data is processed by trend slope calculation to obtain an optimized parameter control signal. The priority sorting unit is used to extract data interpolation filling from the optimized parameter control signal, evaluate the data interpolation filling using periodic pattern recognition, and generate a priority sorting of schemes. The redundancy optimization unit is used to obtain a simplified control configuration by iterative optimization through resource allocation balancing if the priority ranking of the scheme indicates that redundancy should be eliminated. The equipment update unit is used to fine-tune the temperature threshold and synchronize the pressure linkage for the simplified control configuration, and generate the final precision management command; the parameter cycle pattern recognition of the production equipment is updated through the final precision management command to obtain the closed-loop control response.

[0016] The beneficial effects of the present invention include at least the following: This invention collects key parameter data such as temperature, pressure, and flow rate of production equipment, such as chip etching machines, in real time through a sensor network. Based on this data, a digital twin model is constructed. Using this model, time series decomposition analysis is employed to identify abnormal fluctuations in the data, thereby quantifying the etching state and providing a reliable basis for subsequent optimization. Furthermore, through prediction error correction and trend analysis, the control signals of each parameter are dynamically adjusted. When the monitoring level of certain parameters exceeds a preset threshold, a trend slope method is used to further optimize the parameter control signals. This not only improves the accuracy of prediction but also ensures real-time optimization of process parameters, thus avoiding production defects caused by parameter fluctuations. The optimized control signals are further refined through data interpolation and periodic pattern recognition. The process involves several steps, including prioritizing control schemes, identifying redundant configurations based on analysis results, optimizing resource allocation, simplifying control configurations, and improving system efficiency. After fine-tuning temperature thresholds and synchronously adjusting pressure linkages, final precision management commands are generated. These commands update the parameters of the production equipment and ensure stable process parameters throughout the production process through a closed-loop control mechanism. This refined control scheme significantly improves production efficiency and effectively reduces the occurrence of nanoscale defects. Through the synergy of these technical solutions, real-time monitoring, dynamic analysis, and precise control greatly enhance the stability and precision of the production process, reduce defect rates, improve product quality, and meet the demands of modern high-precision manufacturing. Attached Figure Description

[0017] Figure 1 This is a flowchart of a digital twin-driven process dynamic optimization method in an embodiment. Figure 2 This is a flowchart of the optimized parameter control signal acquisition method in the embodiment; Figure 3 This is a structural diagram of a digital twin-driven process flow dynamic optimization system in an embodiment. Detailed Implementation

[0018] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0019] Chip etching, a critical process in integrated circuit manufacturing, demands extremely precise control over multiple parameters such as temperature, pressure, and flow rate. However, in actual production, factors like equipment aging, environmental changes, and operational variations can cause fluctuations in process parameters, impacting product quality and production efficiency. Traditional process control methods typically rely on experience and fixed rules for adjustments, resulting in low efficiency and a lack of flexibility and precision. Therefore, if... Figure 1 This application proposes a digital twin-driven dynamic optimization method for process flow, which may specifically include: Step S1: Acquire real-time data of temperature, pressure, and flow parameters from the chip etching production equipment via a sensor network, construct a digital twin model, and analyze the abnormal fluctuations of the real-time data using time series decomposition based on the digital twin model to obtain a quantitative representation of the current etching state; specifically including: Real-time data streams of temperature, pressure, and flow rate are collected by a sensor network distributed across the production equipment. The real-time data streams are preprocessed to obtain standardized time-series data. A digital twin model is constructed based on the standardized time-series data and the structural information of the production equipment. Based on the digital twin model, time-series segments are extracted from the time-series data using a sliding window. Feature extraction is performed on the time-series segments to generate feature vectors containing temperature, pressure, and flow rate parameters. The dynamic correlation between parameters is obtained through the feature vector, and abnormal fluctuations in the real-time data stream are detected through the dynamic correlation. A quantitative representation of the current etching state is generated based on the abnormal fluctuations, wherein the quantitative representation includes the fluctuation amplitude and frequency characteristics of temperature, pressure, and flow rate.

[0020] Specifically, in one embodiment, a sensor network acquires real-time data on parameters such as temperature, pressure, and flow rate from the chip etching production equipment. These temperature, pressure, and flow rate are essential parameters in the chip etching process, collectively determining the stability and accuracy of the etching process. The temperature refers to the temperature within the etching reaction chamber. Temperature primarily affects the etching rate and uniformity. Excessively high or low temperatures can affect the rate of the chemical reaction, leading to uneven etching depth and even unstable reactions. Such unevenness can result in nanoscale defects, impacting the final chip performance. The pressure refers to the gas pressure within the etching reaction chamber. During etching, the gas pressure within the reaction chamber affects the interaction between gas molecules and the surface material. The main impacts on etching are on the formation of plasma and the control of gas reactions. Under low pressure conditions, the plasma density is higher, which helps to improve the etching rate and accuracy. However, excessively high or low pressure can lead to plasma instability, thereby affecting the uniformity of etching and potentially causing surface defects or incomplete etching. Flow rate refers to the flow rate of the etching reaction gas or reaction mixture. In etching equipment, the gas flow rate is usually controlled by the gas source and transported to the reaction chamber through pipelines. The gas flow rate directly affects the concentration and rate of the chemical reaction during etching. If the gas flow rate is too low, the concentration of the reaction gas will be insufficient, which may lead to low etching efficiency or uneven etching. On the other hand, excessively high flow rate may lead to excessive accumulation of reactants, producing gas deposits, affecting etching accuracy, and even damaging the equipment.

[0021] The aforementioned real-time data undergoes preprocessing, including noise removal and normalization, to obtain standardized time-series data. Preprocessed data exhibits better stability, eliminating the impact of equipment or environmental noise on data quality and ensuring the accuracy of subsequent analysis. Based on this standardized time-series data and the structural information of the production equipment, a digital twin model is constructed. This structural information includes structural images of the equipment and sensor deployment information. The digital twin model can reflect changes in the equipment's monitoring parameters in real time. Furthermore, based on the digital twin model, time-series segments are extracted from the standardized time-series data using a sliding window method. This sliding window method involves setting a fixed-duration window, such as 10 seconds, and progressively sliding to cover the entire time series. The time series data is segmented into continuous time series segments based on the sliding results. These time series segments are used for subsequent feature extraction, thereby capturing local changes in the data and improving the response speed to short-term fluctuations. In practical applications, by adjusting the window size, such as a 5-second window, the accuracy of identifying instantaneous fluctuations can be further improved, enhancing the ability to avoid defects. After extracting the time series segments, feature extraction is performed on these segments to generate feature vectors containing parameters such as temperature, pressure, and flow rate. The feature extraction process calculates the mean and variance of each segment, converting the features of each segment into a digital vector representation. This vector representation reflects the statistical characteristics of each parameter, providing basic data for subsequent model construction.

[0022] By capturing the relationships between temperature, pressure, and flow rate through the aforementioned feature vectors, their dynamic changes over time are described, thereby obtaining the dynamic correlation between parameters. The dynamic correlation is processed by the vector autoregressive (VAR) model integrated into the digital twin model to handle the time series data. The VAR model estimates the model coefficients of historical data using the least squares method, which are then used to predict future parameter values. Due to the complex dependencies between temperature, pressure, and flow rate, the parameters are iteratively updated to adapt to the real-time data, ensuring that the model can effectively capture changing trends and dependencies. This is especially beneficial when dealing with fine processes such as nanoscale defects, as it improves the accuracy of predictions.

[0023] By utilizing dynamic correlations, abnormal fluctuations in real-time data streams are detected. Specifically, the real-time collected data is first input into a digital twin model to calculate the predicted value for each data point, which is then compared with the actual value to obtain the residual. If the residual exceeds a certain standard, such as twice the standard deviation, it is considered an abnormal fluctuation. This detection process can identify abnormal situations in real time, such as sudden fluctuations or gradual drifts, providing a basis for timely adjustment of process parameters and preventing nanoscale defects caused by these abnormal fluctuations. In practical applications, when an abnormal pressure is detected, an alarm is triggered and process parameters are adjusted in a timely manner, thereby reducing the defect rate and improving production quality.

[0024] Based on detected abnormal fluctuations, a quantitative representation of the current etching state is generated. This quantitative representation includes not only the fluctuation amplitude of temperature, pressure, and flow rate, but also their fluctuation frequency. The fluctuation amplitude refers to the difference between the peak value and the mean value of the data, while the frequency represents the number of fluctuations per unit time. Using these features, a quantitative vector containing multi-dimensional information can be constructed, which comprehensively reflects the stability and defect risk of the current etching state. Based on this quantitative vector, an etching state score can be evaluated, and a comprehensive score is obtained through weighted summation to help determine the stability of the current process. A larger weighted sum indicates a larger fluctuation amplitude or higher fluctuation frequency for each parameter, indicating poorer process stability and a higher defect risk. Conversely, a smaller weighted sum indicates a smaller fluctuation amplitude or lower fluctuation frequency for each parameter, indicating higher process stability and a lower defect risk. This technical solution, using real-time sensor data to obtain a quantitative etching state representation, provides a reliable basis for subsequent process adjustments and optimizations. Through dynamic monitoring, real-time analysis, and precise adjustments, the stability and accuracy of the chip etching process are effectively improved, the occurrence of nanometer-level defects is reduced, and production efficiency and product quality are enhanced.

[0025] Step S2: Correct the prediction error based on the quantization representation and determine the real-time dependency monitoring level of each parameter; if the real-time dependency monitoring level is higher than the preset threshold, process the real-time data by calculating the trend slope to obtain the optimized parameter control signal; In step S2, prediction error correction is performed based on the quantization representation, such as... Figure 2 As shown, it includes: The time series features of each parameter are extracted based on the quantization representation. Based on the time series features, the expected values ​​of each parameter are predicted using an autoregressive model. The prediction error of each parameter is calculated by comparing the expected values ​​with the real-time data. The time series features are then smoothed using the prediction error to obtain a smoothed parameter sequence. For the smoothed parameter sequence, the real-time dependency monitoring level of each parameter in the etching process is determined, and a dependency weight matrix between parameters is generated based on the real-time dependency monitoring level. The smoothed parameter sequence is then adjusted according to the dependency weight matrix to obtain an optimized parameter sequence for nanoscale defects.

[0026] Specifically, time-series features of each parameter—temperature, pressure, and flow rate—are extracted based on the aforementioned quantitative representation. These time-series features include information such as the mean, variance, and trend changes of the series. These features serve as input to an autoregressive model. For the temperature parameter, its fluctuation amplitude during the etching process is calculated based on the time-series data and included as part of the features. For the pressure parameter, its periodic patterns are identified to capture abnormal fluctuations that may trigger nanoscale defects. Based on the extracted time-series features, an autoregressive model is used to predict the expected values ​​of each parameter. This model predicts future parameter states based on past observations. Here, the autoregressive model is used to fit historical data and predict future parameter values. For example, for the temperature parameter, the model can estimate the coefficients using the least squares method to predict the expected temperature value at the next time point. Specifically, a suitable lag order p is selected by calculating the partial correlation function, enabling accurate prediction of the changing trends of each parameter. Especially in complex production environments such as high-temperature etching, capturing the temporal dependencies between parameters reduces prediction bias and improves the accuracy of defect detection.

[0027] After comparing the expected value with the real-time data, the prediction error of each parameter is calculated. Specifically, the prediction error is the difference between the real-time data and the expected value predicted by the model. For example, if the real-time temperature is n degrees Celsius and the predicted value is m degrees Celsius, the error is the difference between the two. This error quantifies the degree of parameter fluctuation and reflects the impact of parameter fluctuation on the etching process. The prediction error is also smoothed by applying a window average to the time series features using a moving average method to reduce noise and smooth the data, ensuring data stability. For example, a moving average with a set window size is used to smooth the data. For the sequence portion with a larger error, the filtering intensity is increased, thus obtaining... More stable sequence data; based on the smoothed parameter sequence, the real-time dependency monitoring level of each parameter in the etching process is determined. The real-time dependency monitoring level is quantified by calculating the correlation coefficient between parameters. For example, the Pearson correlation coefficient is used to measure the dependence strength between temperature and pressure, and flow rate parameters. For parameters with high correlation, such as temperature and pressure, the correlation coefficient may reach 0.8, indicating that there is a strong dependency between them and high-frequency monitoring is required. For the above highly correlated parameters, a low threshold needs to be set, such as above 0.7, which is considered a high monitoring level, to ensure that abnormal fluctuations can be identified and handled in a timely manner, and to avoid nanoscale defects caused by these fluctuations.

[0028] Based on the aforementioned real-time dependency monitoring level, a dependency weight matrix between parameters is generated. The elements in the matrix are filled based on the monitoring level, forming a symmetric matrix that reflects the correlation strength between each parameter. For example, if the correlation coefficient between temperature and pressure is 0.8, then the weight of temperature to pressure in the matrix is ​​0.8, and the weight of pressure to temperature is also 0.8. This matrix can accurately quantify the interaction between different parameters. Based on the generated dependency weight matrix, the smoothed parameter sequence is adjusted to obtain an optimized parameter sequence for nanoscale defects. Specifically, the smoothed sequence is weighted and summed using matrix multiplication to obtain the optimized sequence, making the control values ​​of each parameter more precise and helping to reduce the defect risk caused by parameter fluctuations. For example, in the temperature optimization process, the adjusted temperature value may be the first value, thereby reducing the defect risk and optimizing production efficiency. In the flow rate optimization, after matrix adjustment, the flow rate sequence is stabilized at the second value. The above adjustments support the overall defect avoidance strategy, ensuring the stability and accuracy of the production process. The above technical solution can achieve real-time monitoring, accurate prediction, and optimized control of multiple parameters in the chip etching process, effectively avoiding the occurrence of nanoscale defects and improving the quality and efficiency of the etching process.

[0029] Further, in step S2, if the real-time dependency monitoring level is higher than a preset threshold, the real-time data is processed by calculating the trend slope, such as... Figure 2 As shown, it includes: Determine whether the real-time dependency monitoring level is higher than a preset threshold; if it is higher than the preset threshold, extract the trend features of the real-time data; calculate the slope of the trend features using the least squares method to obtain a trend slope sequence; The trend slope sequence is used to identify the change patterns in the real-time data; corresponding parameter adjustment vectors are generated for the change patterns; the real-time data is corrected using the parameter adjustment vectors to obtain optimized parameter control signals; and the optimized parameter control signals are used to determine the parameter optimization direction in the etching process.

[0030] Specifically, the real-time dependency monitoring level is compared with a preset threshold to determine whether the real-time dependency monitoring level is higher than the preset threshold. If the real-time dependency monitoring level is higher than the preset threshold, the trend features of the real-time data are extracted. Specifically, the time series of real-time collected temperature, pressure, and flow are decomposed to identify the long-term trend components as trend features. These trend features reflect the direction of parameter change over time, laying the foundation for avoiding the accumulation of nanoscale defects. The slope of the trend features is calculated using the least squares method to obtain a trend slope sequence. The slope is calculated by minimizing the sum of squares of the differences between the observed values ​​(i.e., the real data points and the fitted values), thereby accurately capturing the fluctuation trend in the data. The slope is calculated for each local time period using a sliding window to form a trend slope sequence. This trend slope sequence reflects the rate and direction of parameter change, providing a basis for subsequent change pattern recognition.

[0031] Based on the aforementioned trend slope sequence, change patterns in real-time data are identified. Specifically, by applying threshold segmentation, the slope values ​​can be categorized into three modes: rising, falling, and stable. The duration and frequency of these patterns are statistically analyzed to identify patterns such as sudden increases or periodic oscillations, which may indicate potential nanoscale defects. For example, during temperature parameter changes, a sudden temperature increase pattern may lead to uneven etching of the material, resulting in defects. Therefore, identifying these change patterns is crucial for timely adjustment of etching parameters and preventing defects. For each identified change pattern, a corresponding parameter adjustment vector is generated. Each component of the adjustment vector corresponds to a different parameter adjustment magnitude. For instance, when a temperature increase pattern is identified, the adjustment vector component may be negative to reduce the temperature; while in a pressure decrease pattern, the adjustment vector includes a positive component to increase the pressure and maintain the stability of the nanoscale structure. By generating these adjustment vectors, parameter fluctuations in real-time data can be corrected, ensuring that parameters remain within the optimal operating range, thereby reducing the probability of defects and improving chip yield.

[0032] During the calibration process, point-to-point calibration is performed by multiplying the adjustment vector with real-time data to obtain the optimized parameter control signal. Low-pass filtering is applied to remove noise after calibration, further smoothing the signal and ensuring its stability. If the fluctuation of the calibrated signal is less than a preset range, the signal is confirmed as optimized and used for subsequent parameter optimization. Based on this optimized signal, its overall trend is analyzed. For example, if the signal indicates a decrease in temperature, the optimization direction is determined to be a cooling strategy, guiding the control configuration and effectively supporting the closed-loop response mechanism. By real-time monitoring and optimization of multiple parameters such as temperature, pressure, and flow, more precise process control can be achieved, reducing unnecessary adjustments and resource waste, and improving the accuracy and efficiency of the etching process. Specifically, in the temperature parameter scenario, the process of determining the optimization direction involves checking the average value of the control signal. If it is lower than the initial value, the threshold is further fine-tuned, and the pressure parameter is adjusted simultaneously. The above coordinated adjustment ensures good coordination between various parameters, effectively reducing the probability of nanoscale defects. Through the above technical solution, more efficient process optimization can be achieved, improving production stability and increasing chip production yield.

[0033] Step S3: Extract data interpolation and filling from the optimized parameter control signal, evaluate the data interpolation and filling using periodic pattern recognition, and generate a priority ranking of the schemes; specifically including: Key control points are extracted from the optimized parameter control signal; for the key control points, spline interpolation is used to generate continuous parameter curves; missing data points are filled according to the continuous parameter curves to obtain a complete parameter sequence; the complete parameter sequence is analyzed by a periodic pattern recognition algorithm to extract periodic features; the stability of the data interpolation filling is evaluated according to the periodic features, and the stability generation schemes are prioritized; the parameter optimization scheme for nanoscale defects is determined according to the priority ranking of the schemes.

[0034] Specifically, key control points are first extracted from the optimized parameter control signal to identify the peak and valley points of temperature, pressure, and flow parameters in the signal. These peak and valley points are then used as key control points. After identifying the key control points, continuous parameter curves are generated for these key control points using spline interpolation. Spline interpolation connects the key control points using a polynomial function to ensure that the interpolation curve is continuous and its derivative is continuous between the key control points. Specifically, for temperature, pressure, and flow parameters, cubic spline interpolation is applied to generate smooth curves to ensure a smooth transition between points, thereby avoiding abrupt changes in parameters and preventing nanoscale defects caused by excessive parameter fluctuations during the etching process. The above interpolation method can effectively avoid defects caused by excessive fluctuations in different etching scenarios. For example, in a high-temperature etching environment, the interpolation of the temperature parameter ensures that the curve slope does not exceed the set value, thus maintaining temperature stability and preventing the generation of defects.

[0035] After generating continuous parameter curves, the missing data points are filled in using these curves to obtain a complete parameter sequence. Specifically, the values ​​of the missing points are calculated using the curve equations generated by interpolation and then filled into the sequence to form a continuous data stream. This ensures the integrity of the entire data sequence and provides a stable data foundation for subsequent analysis. A preset periodic pattern recognition algorithm is then used to analyze this complete parameter sequence, extracting its periodic features. Specifically, Fourier transform is applied to decompose the sequence, converting the time-domain signal into the frequency domain. By calculating the amplitude and phase of each frequency component, periodic fluctuations in temperature, pressure, and flow parameters are identified, such as temperature fluctuations every minute or pressure fluctuations every 10 seconds. These periodic features help optimize defect avoidance strategies and further enhance the accuracy of the production process. The system assesses data stability and calculates the variance of periodic features. If the variance is below a preset stability threshold, the data sequence is considered stable; otherwise, it is unstable. By evaluating the stability of the data, a priority ranking of solutions is generated, placing the most stable solution in the priority position to ensure the optimization direction in subsequent processing. Based on the generated priority ranking of solutions, the parameter optimization solution for nanoscale defects is finally determined. In practical applications, the above technical solution ensures that various process parameters can be precisely adjusted in the complex microchip etching process, thereby effectively avoiding the occurrence of defects and effectively improving the accuracy and efficiency of the production process. In the microchip etching process, especially in the management of fluctuations in parameters such as pressure, flow rate, and temperature, the system can monitor and adjust process parameters in real time, reduce nanoscale defects, and improve product quality and production yield.

[0036] Step S4: If the priority ranking of the scheme indicates the elimination of redundant configurations, then a simplified control configuration is obtained through iterative optimization of resource allocation balancing; specifically including: Determine whether the priority ranking of the scheme indicates configuration redundancy elimination; if it indicates configuration redundancy elimination, extract the redundant parameter configurations from the priority ranking of the scheme; for the redundant parameter configurations, use a resource allocation balancing algorithm to calculate the resource occupancy weights between the parameters; eliminate the redundant parameter configurations according to the resource occupancy weights to generate a preliminary control configuration; The initial control configuration is adjusted through an iterative optimization loop to obtain a simplified control configuration; a resource allocation scheme is generated based on the simplified control configuration; and the stability of the simplified control configuration is verified through the resource allocation scheme.

[0037] Specifically, in this embodiment, if the scheme priority ranking indicates redundant configuration elimination, a simplified control configuration is obtained through resource allocation balancing and iterative optimization methods. Specifically, the scheme priority ranking data structure is obtained, which contains priority values ​​and indicator flags for multiple parameter configurations. The value of the indicator flag is parsed to determine whether a redundant elimination operation is needed. If redundant configuration elimination is indicated, redundant parameter configurations in the above scheme priority ranking are extracted, and parameter configurations with priorities lower than a preset threshold are selected as redundant parts. Furthermore, detailed information of these redundant configurations, such as temperature thresholds and pressure linkage values, is extracted to form a redundant configuration list. Regarding the aforementioned redundancy... The remaining parameters are configured by calculating the resource occupancy weights of each parameter using a resource allocation balancing algorithm. This algorithm calculates a weighted average based on the parameter's impact on the defect rate during etching. The resources mentioned include at least system control and computational resources. The weight is calculated by multiplying the ratio of the parameter's fluctuation amplitude to the total fluctuation amplitude by an influence factor. This influence factor is preset based on experimental data; for example, the influence factor for temperature is typically set between a first and a second value. This quantifies the resource occupancy weight of each parameter, such as a weight of 0.5 for temperature, 0.4 for pressure, and 0.1 for flow rate. These weights reflect the corresponding parameter's impact on etching. The impact of defects is assessed to generate an initial control configuration. If the resource weight of a parameter is greater than a set threshold, the parameter is considered to have a significant impact on defect rate control and is therefore considered an important parameter, which should be retained in the control configuration. If the resource weight of a parameter is less than the threshold, the parameter is considered to contribute less to defect rate control and is therefore considered redundant and can be removed. After the initial control configuration is generated, it is further adjusted through iterative optimization loops. In each loop, parameter values ​​are adjusted using the gradient descent method, and the parameters are updated based on the error gradient to continuously optimize the control configuration until the error drops below the preset threshold. For example, temperature... The threshold value is adjusted from the initial value to the preset value until the error is lower than the set error. During the iteration process, for example, linkage synchronization is introduced for the adjustment of pressure parameters to ensure that the relationship between temperature and pressure is stably controlled, thereby optimizing the entire control configuration and avoiding the generation of nanoscale defects. The simplified control configuration after iterative optimization will be used to generate a resource allocation scheme, which will map the simplified parameter configuration to the production equipment. The stability of the above simplified control configuration will be verified by simulating the operation of the resource allocation scheme in the etching process to check whether the fluctuation of the configuration parameters is within the preset threshold range. If the stability is verified, the simplified configuration is confirmed to be effective.

[0038] The above-mentioned technical solution, through refined resource allocation, redundant parameter elimination, and multiple iterative optimizations, can significantly improve the precision management in the microchip etching process, reduce the generation of nanoscale defects, thereby improving production efficiency and quality, and ensuring the stability and efficiency of the process.

[0039] Step S5: Fine-tune the temperature threshold and synchronize the pressure linkage for the simplified control configuration to generate the final precision management command; update the parameter cycle pattern recognition of the production equipment through the final precision management command to obtain the closed-loop control response.

[0040] In step S5, the temperature threshold is fine-tuned and pressure linkage is synchronized for the simplified control configuration, including: The control ranges of temperature and pressure parameters are extracted from the simplified control configuration; for the control range of the temperature parameter, the threshold is fine-tuned using a gradient descent algorithm to obtain the optimized temperature threshold; Based on the optimized temperature threshold, determine the linkage synchronization range of the pressure parameters; generate a coordinated control strategy for temperature and pressure through the linkage synchronization range; adjust the simplified control configuration according to the coordinated control strategy to generate the final precision management command; evaluate the compliance of the final precision management command through verification indicators; and generate a control command sequence for the etching process based on the compliance.

[0041] Specifically, for the simplified control configuration, temperature threshold fine-tuning and pressure linkage synchronization are performed. This includes extracting the control range of temperature and pressure parameters from the simplified control configuration, specifically obtaining the upper and lower limits of temperature and pressure by decomposing time-series data previously collected through a sensor network. For the control range of the aforementioned temperature parameters, a gradient descent algorithm is used for threshold fine-tuning to obtain an optimized temperature threshold. This involves initializing the initial threshold of the temperature control range and defining the loss function as the mean square error between the threshold and the nanometer-level defect rate. By calculating the gradient of the loss function, the direction of the threshold change's influence on the defect rate is determined. Then, the threshold is iteratively updated, moving the step size in the opposite direction of the gradient until the loss function converges, ultimately obtaining the optimized temperature threshold. Based on the optimized temperature threshold, the linkage synchronization range of the pressure parameters is determined. The system maps the optimized temperature threshold to the pressure range using a linear function, ensuring that the pressure adjusts synchronously with temperature changes. To ensure the stability of the linkage synchronization, simulated etching data is used to verify that the pressure increases or decreases with each increase or decrease in temperature, thus maintaining a balance between temperature and pressure. Based on this, a collaborative control strategy for temperature and pressure is generated. This collaborative control strategy integrates the linkage range data through a temperature-pressure mapping table and applies the mapping table to generate strategy rules. Specifically, when the temperature approaches the upper limit of the threshold, the pressure is preferentially reduced synchronously to avoid the occurrence of nanoscale defects. Furthermore, in high-density chip etching scenarios, the collaborative control strategy is further extended to include flow rate parameter linkage. For example, after the temperature threshold is optimized, the pressure range adjustment synchronously affects the flow rate to the corresponding value to cover the etching requirements of complex nanostructures.

[0042] Based on the aforementioned collaborative control strategy, the simplified control configuration is adjusted, and the final precision management instruction is generated. After the final instruction is generated, its compliance is evaluated through verification indicators, specifically including checking whether the instruction reduces the defect rate below the preset target value. If the compliance evaluation is passed, a control instruction sequence for the etching process is generated for closed-loop control of the production equipment. The above technical solution achieves an effective combination of precision management and nanoscale defect control through the collaborative optimization of temperature and pressure, thereby significantly improving the stability and efficiency of the production process and ensuring a high-precision and low-defect-rate etching process.

[0043] Further, in step S5, the parameter cycle pattern recognition of the production equipment is updated through the final precision management instruction, including: Extract the control parameter sequence from the final precision management instruction; generate corresponding parameter adjustment instructions for the control parameter sequence; update the parameter configuration of the production equipment using the parameter adjustment instructions; analyze the operating status of the production equipment using a periodic pattern recognition algorithm based on the updated parameter configuration; generate a closed-loop control response based on the operating status; adjust the execution frequency of the parameter adjustment instructions based on the closed-loop control response; and optimize the parameter periodic pattern recognition of the production equipment using the execution frequency.

[0044] Specifically, numerical sequences of key parameters such as temperature, pressure, and flow rate are extracted from the precision management instructions. These parameter sequences are then quantitatively represented through time series decomposition and trend slope calculation, ensuring a high correlation with the real-time status of the production equipment. After obtaining the control parameter sequences, corresponding adjustment instructions are generated based on the deviations in these sequences. For example, when the temperature or pressure value deviates from the target range, a corresponding adjustment instruction is automatically generated, and the parameter configuration of the production equipment is updated in real time through the equipment control interface. The updated equipment configuration is applied in real time, and the equipment operating status is analyzed using a periodic pattern recognition algorithm to identify and extract patterns of periodic fluctuations during the production process, thereby evaluating... The stability of the equipment under different process conditions is assessed using Fourier transform technology to convert time-domain data into frequency-domain data, calculate the amplitude and phase of each frequency component, and generate corresponding stability indicators by identifying the periodicity and stability of fluctuations. Based on the equipment's operating status and stability indicators, a closed-loop control response is further generated. If the operating status is found to be unstable, the response loop will adjust the corresponding control strategy according to the feedback signal, such as by adjusting the temperature or pressure, to ensure that the equipment operates under precise control. In addition, the execution frequency of the closed-loop control response will be dynamically adjusted according to the magnitude of parameter adjustment. For example, when the adjustment magnitude is large, the adjustment frequency will be increased to ensure timely response to changes.

[0045] The optimized frequency will further influence the parameter periodic pattern recognition algorithm of the production equipment, improving its accuracy and real-time performance. In particular, the frequency adjustment enables the periodic pattern recognition algorithm to analyze each parameter sequence more precisely, reducing redundant calculations, thereby optimizing the recognition process and ensuring that the equipment operating status continuously meets the set process requirements. Through the above technical solution, a closed-loop optimization control from equipment operating status to accuracy management commands is realized, ensuring that equipment parameters can be adjusted in real time and accurately in a dynamically changing production environment, avoiding the occurrence of nanoscale defects, and improving overall production efficiency and product quality.

[0046] Further, in step S5, the closed-loop control response is obtained, including: Extract the collaborative application fusion loop from the final precision management instruction; generate parameter update instructions for the production equipment based on the collaborative application fusion loop; adjust the operating parameters of the production equipment using the parameter update instructions; collect real-time operating data of the production equipment based on the adjusted operating parameters; match the real-time operating data with the collaborative application fusion loop to generate a closed-loop control response; update the parameter cycle mode of the production equipment based on the closed-loop control response; verify the stability of the closed-loop control response using the parameter cycle mode.

[0047] Specifically, this invention obtains the collaborative application fusion loop from the final precision management instruction, and then generates parameter update instructions for the production equipment to achieve precise control and optimization of the production equipment. Specifically, the collaborative application fusion loop is extracted from the precision management instruction. This loop, based on the fusion representation of temperature, pressure, and flow parameters, ensures the coordination of various control parameters and provides a quantitative basis for subsequent equipment parameter updates. Based on this loop, a corresponding parameter update vector is generated by calculating temperature threshold fine-tuning and pressure linkage synchronization data. The parameters in the update instruction are weighted and summed according to the quantified data in the loop and the current equipment state, thus obtaining the core instruction for adjusting the equipment parameters. The generated parameter update instruction is used to directly adjust the operating parameters of the production equipment, for example, by modifying the temperature setting and pressure level in the equipment controller to ensure that the equipment parameters are consistent with the optimization target. The adjusted operating parameters are collected in real time through a sensor network to generate new time-series data, including the real-time changing parameter values ​​of temperature, pressure, and flow. By matching the aforementioned real-time data with the collaborative application fusion loop, the Dynamic Time Warping (DTW) algorithm is used to measure and align the time series, ensuring the consistency of equipment parameters in time. DTW effectively eliminates errors caused by time axis asynchrony, accurately aligning real-time data with expected loop values, thereby generating a closed-loop control response. When the data similarity is higher than a set threshold, a positive feedback signal is generated; if the similarity is low, adjustments are made using a correction signal. Next, the closed-loop control response updates the production equipment's parameter cycle mode. Specifically, this involves processing the response data using trend slope calculations and updating the equipment's mode values ​​to better match current production needs. The updated cycle mode is used to verify the stability of the closed-loop control response, i.e., by calculating the variance value to assess data volatility. If the volatility value is less than a preset threshold, it indicates that the system remains stable during control, verifying the response's effectiveness. If the verification result shows insufficient stability, the system will trigger a redundancy elimination iteration to further optimize the control strategy.

[0048] The above technical solution, for micro-processes such as chip etching, uses a dynamic time warping algorithm to ensure the precise alignment of real-time parameters such as temperature and pressure with loop commands. This helps to eliminate nanoscale defects caused by parameter fluctuations, thereby optimizing the overall accuracy of the process. Through dynamic and coordinated control of multiple parameters such as flow rate and temperature, various parameters can be corrected in real time to ensure process stability, avoid defects, and improve overall production efficiency and product quality.

[0049] This invention also provides a digital twin-driven dynamic optimization system for process flow, used to implement the above-mentioned method, such as... Figure 3 As shown, the system includes: The modeling unit is used to acquire real-time data of temperature, pressure, and flow parameters from the chip etching production equipment through a sensor network, construct a digital twin model, and use time series decomposition analysis based on the digital twin model to analyze abnormal fluctuations in the real-time data and obtain a quantitative representation of the current etching state. The correction analysis unit is used to correct prediction errors based on the quantization representation and determine the real-time dependency monitoring level of each parameter; if the real-time dependency monitoring level is higher than a preset threshold, the real-time data is processed by trend slope calculation to obtain an optimized parameter control signal. The priority sorting unit is used to extract data interpolation filling from the optimized parameter control signal, evaluate the data interpolation filling using periodic pattern recognition, and generate a priority sorting of schemes. The redundancy optimization unit is used to obtain a simplified control configuration by iterative optimization through resource allocation balancing if the priority ranking of the scheme indicates that redundancy should be eliminated. The equipment update unit is used to fine-tune the temperature threshold and synchronize the pressure linkage for the simplified control configuration, and generate the final precision management command; the parameter cycle pattern recognition of the production equipment is updated through the final precision management command to obtain the closed-loop control response.

[0050] In summary, this invention uses a sensor network to collect real-time data on key parameters of production equipment, such as temperature, pressure, and flow rate, from chip etching machines. Based on this data, a digital twin model is constructed. Using this model, time series decomposition analysis is employed to identify abnormal fluctuations in the data, thereby quantifying the etching status and providing a reliable basis for subsequent optimization. Furthermore, through prediction error correction and trend analysis, the control signals for each parameter are dynamically adjusted. When the monitoring level of certain parameters exceeds a preset threshold, a trend slope method is used to further optimize the parameter control signals. This not only improves the accuracy of predictions but also ensures real-time optimization of process parameters, thus avoiding production defects caused by parameter fluctuations. The optimized control signals are further processed through data interpolation and periodic pattern recognition. The process involves a single step to prioritize control schemes. Redundant configurations are identified based on analysis results, and resource allocation is optimized for balance, simplifying control configurations and improving system efficiency. After fine-tuning temperature thresholds and synchronously adjusting pressure linkages, final precision management commands are generated. These commands update the parameters of the production equipment, and a closed-loop control mechanism ensures that process parameters remain stable throughout the production process. This refined control scheme significantly improves production efficiency and effectively reduces the occurrence of nanoscale defects. Through the synergy of these technical solutions, real-time monitoring, dynamic analysis, and precise control greatly enhance the stability and precision of the production process, reduce defect rates, improve product quality, and meet the demands of modern high-precision manufacturing.

[0051] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. The present invention has been described in detail with reference to preferred embodiments. Those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications and substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A digital twin-driven method for dynamic optimization of a process flow, characterized in that, include: Step S1: Obtain real-time data of temperature, pressure, and flow parameters from the chip etching production equipment through a sensor network, construct a digital twin model, and use time series decomposition analysis based on the digital twin model to analyze abnormal fluctuations in the real-time data and obtain a quantitative representation of the current etching state. Step S2: Correct the prediction error based on the quantization representation and determine the real-time dependency monitoring level of each parameter; if the real-time dependency monitoring level is higher than the preset threshold, process the real-time data by calculating the trend slope to obtain the optimized parameter control signal; Step S3: Extract data interpolation and filling from the optimized parameter control signal, evaluate the data interpolation and filling using periodic pattern recognition, and generate a priority ranking of the schemes; Step S4: If the priority sorting of the scheme indicates the elimination of redundant configurations, then a simplified control configuration is obtained through iterative optimization of resource allocation balancing. Step S5: Fine-tune the temperature threshold and synchronize the pressure linkage for the simplified control configuration to generate the final precision management command; update the parameter cycle pattern recognition of the production equipment through the final precision management command to obtain the closed-loop control response.

2. The method as described in claim 1, characterized in that, Step S1 includes: Real-time data streams of temperature, pressure, and flow rate are collected by a sensor network distributed across the production equipment. The real-time data streams are preprocessed to obtain standardized time-series data. A digital twin model is constructed based on the standardized time-series data and the structural information of the production equipment. Based on the digital twin model, time-series segments are extracted from the time-series data using a sliding window. Feature extraction is performed on the time-series segments to generate feature vectors containing temperature, pressure, and flow rate parameters. The dynamic correlation between parameters is obtained through the feature vector, and abnormal fluctuations in the real-time data stream are detected through the dynamic correlation. A quantitative representation of the current etching state is generated based on the abnormal fluctuations, wherein the quantitative representation includes the fluctuation amplitude and frequency characteristics of temperature, pressure, and flow rate.

3. The method as described in claim 1, characterized in that, In step S2, prediction error correction is performed based on the quantization representation, including: The time series features of each parameter are extracted based on the quantization representation. Based on the time series features, the expected values ​​of each parameter are predicted using an autoregressive model. The prediction error of each parameter is calculated by comparing the expected values ​​with the real-time data. The time series features are then smoothed using the prediction error to obtain a smoothed parameter sequence. For the smoothed parameter sequence, the real-time dependency monitoring level of each parameter in the etching process is determined, and a dependency weight matrix between parameters is generated based on the real-time dependency monitoring level. The smoothed parameter sequence is then adjusted according to the dependency weight matrix to obtain an optimized parameter sequence for nanoscale defects.

4. The method as described in claim 3, characterized in that, In step S2, if the real-time dependency monitoring level is higher than a preset threshold, the real-time data is processed by calculating the trend slope, including: Determine whether the real-time dependency monitoring level is higher than a preset threshold; if it is higher than the preset threshold, extract the trend features of the real-time data; calculate the slope of the trend features using the least squares method to obtain a trend slope sequence; The trend slope sequence is used to identify the change patterns in the real-time data; corresponding parameter adjustment vectors are generated for the change patterns; the real-time data is corrected using the parameter adjustment vectors to obtain optimized parameter control signals; and the optimized parameter control signals are used to determine the parameter optimization direction in the etching process.

5. The method as described in claim 1, characterized in that, In step S3, data interpolation and filling are extracted from the optimized parameter control signal, including: Key control points are extracted from the optimized parameter control signal; for the key control points, a continuous parameter curve is generated using spline interpolation; missing data points are filled in according to the continuous parameter curve to obtain a complete parameter sequence; the complete parameter sequence is analyzed using a periodic pattern recognition algorithm to extract periodic features; the stability of the data interpolation filling is evaluated based on the periodic features; the stability generation schemes are prioritized; and a parameter optimization scheme for nanoscale defects is determined based on the priority ranking of the schemes.

6. The method as described in claim 1, characterized in that, Step S4 includes: Determine whether the priority ranking of the scheme indicates configuration redundancy elimination; if it indicates configuration redundancy elimination, extract the redundant parameter configurations from the priority ranking of the scheme; for the redundant parameter configurations, use a resource allocation balancing algorithm to calculate the resource occupancy weights between the parameters; eliminate the redundant parameter configurations according to the resource occupancy weights to generate a preliminary control configuration; The initial control configuration is adjusted through an iterative optimization loop to obtain a simplified control configuration; a resource allocation scheme is generated based on the simplified control configuration; and the stability of the simplified control configuration is verified through the resource allocation scheme.

7. The method as described in claim 1, characterized in that, In step S5, temperature threshold fine-tuning and pressure linkage synchronization are performed for the simplified control configuration, including: The control ranges of temperature and pressure parameters are extracted from the simplified control configuration; for the control range of the temperature parameter, the threshold is fine-tuned using a gradient descent algorithm to obtain the optimized temperature threshold; Based on the optimized temperature threshold, determine the linkage synchronization range of the pressure parameters; generate a coordinated control strategy for temperature and pressure through the linkage synchronization range; adjust the simplified control configuration according to the coordinated control strategy to generate the final precision management command; evaluate the compliance of the final precision management command through verification indicators; and generate a control command sequence for the etching process based on the compliance.

8. The method as described in claim 7, characterized in that, In step S5, the parameter cycle pattern recognition of the production equipment is updated through the final precision management instruction, including: Extract the control parameter sequence from the final precision management instruction; generate corresponding parameter adjustment instructions for the control parameter sequence; update the parameter configuration of the production equipment using the parameter adjustment instructions; analyze the operating status of the production equipment using a periodic pattern recognition algorithm based on the updated parameter configuration; generate a closed-loop control response based on the operating status; adjust the execution frequency of the parameter adjustment instructions based on the closed-loop control response; and optimize the parameter periodic pattern recognition of the production equipment using the execution frequency.

9. The method as described in claim 8, characterized in that, In step S5, the closed-loop control response is obtained, including: Extract the collaborative application fusion loop from the final precision management instruction; generate parameter update instructions for the production equipment based on the collaborative application fusion loop; adjust the operating parameters of the production equipment using the parameter update instructions; collect real-time operating data of the production equipment based on the adjusted operating parameters; match the real-time operating data with the collaborative application fusion loop to generate a closed-loop control response; update the parameter cycle mode of the production equipment based on the closed-loop control response; verify the stability of the closed-loop control response using the parameter cycle mode.

10. A digital twin-driven dynamic optimization system for process flow, used to implement the method as described in any one of claims 1-9, characterized in that, The system includes: The modeling unit is used to acquire real-time data of temperature, pressure, and flow parameters from the chip etching production equipment through a sensor network, construct a digital twin model, and use time series decomposition analysis based on the digital twin model to analyze abnormal fluctuations in the real-time data and obtain a quantitative representation of the current etching state. The correction analysis unit is used to correct prediction errors based on the quantization representation and determine the real-time dependency monitoring level of each parameter; if the real-time dependency monitoring level is higher than a preset threshold, the real-time data is processed by trend slope calculation to obtain an optimized parameter control signal. The priority sorting unit is used to extract data interpolation filling from the optimized parameter control signal, evaluate the data interpolation filling using periodic pattern recognition, and generate a priority sorting of schemes. The redundancy optimization unit is used to obtain a simplified control configuration by iterative optimization through resource allocation balancing if the priority ranking of the scheme indicates that redundancy should be eliminated. The equipment update unit is used to fine-tune the temperature threshold and synchronize the pressure linkage for the simplified control configuration, and generate the final precision management command; the parameter cycle pattern recognition of the production equipment is updated through the final precision management command to obtain the closed-loop control response.

Images