Wafer baking control method, apparatus and medium

Abstract

The application discloses a wafer drying control method, equipment and medium, relates to the field of semiconductor manufacturing, and comprises the following steps: collecting wafer surface state data, temperature distribution data and airflow state data to construct multi-modal state characteristics; analyzing wafer surface drying uniformity variation trend, temperature field stability index and airflow distribution uniformity to generate process health degree evaluation results; iteratively solving the optimal control parameter combination of the wafer drying control system, dynamically generating control action instructions and converting them into driving signals executable by the wafer drying control system, controlling the partition heating unit of the heating device, the speed regulation mechanism of the variable frequency fan and the opening actuator of the multi-way valve through a distributed driving circuit; monitoring the actual operation state of each actuator in the wafer drying control system during the drying process, comparing the monitoring data with the expected state, and triggering control parameter updating when a state deviation is detected.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor manufacturing, and in particular to a wafer drying control method, device and medium. BACKGROUND

[0002] As a key link in the semiconductor manufacturing process, the control precision of wafer drying process directly determines the cleanliness of wafer surface and product quality. The traditional wafer drying equipment mainly adopts a control strategy based on a fixed process curve to realize programmed control of the drying process through preset temperature, airflow and time parameters. This simple feedback control method cannot adapt to the dynamic changes of wafer surface state, environmental conditions and equipment running state, resulting in serious challenges to drying uniformity and process stability.

[0003] In the prior art, the control of wafer drying process often relies on single or limited dimensional sensor data, lacking comprehensive perception and comprehensive evaluation of process state. For example, only local parameter adjustment through temperature sensor or pressure sensor cannot accurately reflect the actual drying state of wafer surface and the environmental distribution characteristics in the cavity. This one-sided monitoring method is easy to cause local over-drying or residual water stains and other defects, affecting the final quality of wafer products. SUMMARY

[0004] The embodiments of the present application provide a wafer drying control method, device and medium to solve the above technical problems.

[0005] In one aspect, the embodiments of the present application provide a wafer drying control method, comprising:

[0006] Through the multi-sensor pre-deployed in the wafer drying equipment, wafer surface state data, temperature distribution data and airflow state data in the wafer drying process cavity are collected in real time to construct multi-modal state features containing time sequence characteristics; the process cavity includes an air inlet device, a wafer supporting device and an air outlet device;

[0007] Based on the multi-modal state features, the drying uniformity variation trend of wafer surface, the temperature field stability index and the airflow distribution uniformity are analyzed to generate a process health degree evaluation result of the current drying stage;

[0008] Based on the process health degree evaluation result, the optimal control parameter combination of the wafer drying control system is iteratively solved, and the control action instruction containing the heating device partition power set value, the variable frequency fan speed gradient set value and the multi-way valve opening degree coordination set value is dynamically generated according to the optimal control parameter combination;

[0009] The control action instruction is converted into a driving signal executable by the wafer baking control system, and the driving signal is used to control the partition heating unit of the heating device, the speed regulation mechanism of the variable frequency fan, and the opening actuator of the multi-way valve through a distributed driving circuit.

[0010] During the baking process, the actual operating state of each actuator in the wafer baking control system is continuously monitored, the monitoring data is compared with the expected state in real time, and the control parameter is updated when a state deviation is detected to form a closed-loop control.

[0011] In an implementation manner of the present application, the drying uniformity change trend of the wafer surface is analyzed, specifically including:

[0012] The drying progress data of different regions of the wafer surface is extracted from the multi-modal state features, and the drying rate change curve of each region in the current time window is calculated through a sliding window algorithm.

[0013] Based on the drying rate change curve, the drying rate trend is identified, and the synchronization index of the drying rate between regions is calculated; the drying rate trend includes an acceleration trend and a deceleration trend.

[0014] The drying uniformity index is calculated by statistically analyzing the dispersion degree of the drying progress data of each region, and the drying uniformity change trend evaluation report is generated in combination with the drying rate trend and the synchronization index.

[0015] The drying uniformity change trend evaluation report is compared with the preset process benchmark, and a quantitative drying uniformity health score is output as a component of the process health evaluation result.

[0016] In an implementation manner of the present application, based on the multi-modal state features, the temperature field stability index is analyzed, specifically including:

[0017] The real-time readings of the temperature sensor array are extracted from the multi-modal state features, and a three-dimensional temperature field distribution model in the process cavity is reconstructed through a spatial interpolation algorithm.

[0018] Based on the three-dimensional temperature field distribution model, the maximum and average values of the temperature gradient are calculated, and the distribution characteristics of the high-temperature region and the low-temperature region are analyzed to generate a temperature field uniformity index.

[0019] The historical temperature data is processed through a time series analysis method to detect the fluctuation frequency and amplitude of the temperature field, and a temperature stability coefficient is calculated.

[0020] The temperature field uniformity index and the temperature stability coefficient are fused to generate a comprehensive temperature field stability evaluation result, and the comprehensive temperature field stability evaluation result is mapped as a temperature field health degree score as a component of the process health degree evaluation result.

[0021] In an implementation manner of the present application, the gas flow distribution uniformity is analyzed, specifically including:

[0022] The wind speed measurement data of the multiple wind speed sensors in the process cavity are extracted from the multi-modal state features, and the variation coefficient and the range of the gas flow speed of each measurement point are calculated to generate a gas flow distribution uniformity primary index;

[0023] The autocorrelation features of the gas flow speed time series data are analyzed to detect periodic fluctuation patterns of the gas flow and identify abnormal fluctuation events to generate gas flow stability evaluation data;

[0024] The gas flow distribution uniformity primary index and the gas flow stability evaluation data are combined to calculate a gas flow distribution uniformity level by a fuzzy logic inference system;

[0025] The gas flow distribution uniformity level is matched with a preset process requirement to output a gas flow health degree score, and the gas flow health degree score is integrated into the process health degree evaluation result.

[0026] In an implementation manner of the present application, based on the process health degree evaluation result, the optimal control parameter combination of the wafer drying control system is iteratively solved, specifically including:

[0027] The control parameter candidate solutions in the wafer drying control system are initialized; the control parameters include heating device partition power values, variable frequency fan speed values, and multi-way valve opening degree values;

[0028] The population is iteratively evolved by selection, crossover and mutation operations of the multi-objective genetic algorithm, the fitness value of each control parameter candidate solution is calculated in each generation evolution process, and a new generation of candidate solutions is generated based on the fitness value;

[0029] The iterative evaluation process is repeated until the convergence condition is met, and the optimal control parameter combination of the wafer drying control system is output.

[0030] In an implementation manner of the present application, according to the optimal control parameter combination, a control action instruction containing heating device partition power setting values, variable frequency fan speed gradient setting values, and multi-way valve opening degree coordination setting values is dynamically generated, specifically including:

[0031] The heating device partition power values in the optimal control parameter combination are mapped as the power setting values corresponding to each partition heating unit, and a coupling relationship model between each partition heating unit is established;

[0032] Convert the variable frequency fan speed value in the optimal control parameter combination into a speed gradient setting value, define the speed ratio relationship of the fans in different regions and the acceleration and deceleration curve;

[0033] Resolve the multi-way valve opening value in the optimal control parameter combination into an opening coordination setting value, ensure that the valve opening changes are synchronized and the air flow distribution is balanced;

[0034] Pack the power setting value, speed gradient setting value and opening coordination setting value according to the device communication protocol format, and generate a control action instruction with a time stamp and priority identifier.

[0035] In an implementation manner of the present application, the actual running state of each actuator in the wafer drying control system is continuously monitored, the monitoring data is compared with the expected state in real time, and the control parameter update is triggered when the state deviation is detected, so as to form a closed loop control, specifically including:

[0036] The actual power data of each partition heating unit of the heating device, the actual speed data of the variable frequency fan and the actual opening data of the opening actuator are collected in real time through the sensor network to form an actual running state data set of the actuator;

[0037] The actual running state data set is compared with the expected state data in the control action instruction item by item, and the absolute values and change trends of the power deviation, speed deviation and opening deviation are calculated;

[0038] When the absolute value of any deviation exceeds the preset threshold or the change trend of the deviation is abnormal, the current actual running state is taken as an initial condition, and the iterative solving execution process is retriggered;

[0039] The updated optimal control parameter combination is generated by re-executing the iterative solving, and a new control action instruction is generated based on the updated optimal control parameter combination, so as to realize dynamic adjustment and closed loop optimization of the control parameter.

[0040] In an implementation manner of the present application, through the multi-sensor pre-deployed in the wafer drying device, wafer surface state data, temperature distribution data and air flow state data in the wafer drying process cavity are collected in real time, and multi-modal state features containing time sequence characteristics are constructed, specifically including:

[0041] The wafer surface image data is collected through a high-resolution optical sensor, and the wafer surface image data is subjected to grayscale and region segmentation processing, the average grayscale value of each segmented region is calculated, the suspicious wetting area with an average grayscale value lower than a preset grayscale threshold is identified, and a suspicious wetting area coordinate map is generated;

[0042] According to the suspicious wet area coordinate map, a near-infrared spectrum sensor is used to perform a point spectrum measurement on each suspicious wet area, corresponding near-infrared reflection spectrum data is collected, and spectrum absorption intensity in a water feature absorption band is analyzed; and according to a pre-calibrated absorption intensity and water content relationship model, water content data of each suspicious wet area is calculated;

[0043] The water content data is fused with the average gray value of the corresponding suspicious wet area, a local linear mapping model between the gray value and the water content is established, the gray value of the wafer surface is converted into the water content data, a wafer full-field water content distribution map is generated, the maximum water content, the average water content and the water content distribution uniformity index are extracted from the full-field water content distribution map, and a wafer surface state sub-feature is constructed;

[0044] Temperature readings of multiple spatial positions in a process cavity are collected by a distributed temperature sensor array, an average value and a standard deviation of the temperature readings are calculated to represent temperature distribution uniformity, and a temperature change rate feature is extracted based on time series analysis to construct a temperature distribution data sub-feature;

[0045] Airflow velocity data of key positions in the cavity are collected by an airflow speed sensor group, a coefficient of variation and a fluctuation amplitude of the airflow velocity data are calculated, and an airflow distribution pattern is evaluated in combination with airflow direction sensor data to construct an airflow state data sub-feature;

[0046] The wafer surface state sub-feature, the temperature distribution data sub-feature and the airflow state data sub-feature are aligned and spliced in time series to form a multi-modal state feature; the multi-modal state feature includes historical state and real-time state.

[0047] On the other hand, the embodiments of the present application also provide a wafer drying control device, the device comprising:

[0048] At least one processor;

[0049] and a memory in communication connection with the at least one processor;

[0050] Among them, the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the wafer drying control method as described above.

[0051] On the other hand, the embodiments of the present application also provide a non-volatile computer storage medium, which stores computer executable instructions, and the computer executable instructions are executed to implement the wafer drying control method as described above.

[0052] The embodiments of the present application provide a wafer drying control method, device and medium, which at least have the following beneficial effects:

[0053] By constructing a multi-modal state feature containing time series features, a comprehensive perception ability of the wafer drying process state is established, wafer surface state data, temperature distribution data and airflow state data are comprehensively collected, the dynamic characteristics of the drying process are described from multiple dimensions, and the limitations of traditional single sensor monitoring are overcome; by analyzing the wafer surface drying uniformity trend, temperature field stability index and airflow distribution uniformity, not only the static parameters are concerned, but also the change trend and mutual relationship of each parameter are focused, and the quantitative evaluation of the process health degree is realized; by finding the best matching scheme of each actuator, the problem that multiple performance indicators are difficult to be considered in the traditional control method is solved; through the closed-loop control mechanism of continuously monitoring the actuator state and triggering parameter updating, the stable operation of the system under various working conditions is ensured. BRIEF DESCRIPTION OF DRAWINGS

[0054] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and serve to explain the principles of the application, and do not limit the application. In the drawings:

[0055] Figure 1 A flowchart of a wafer drying control method provided by an embodiment of the application is shown.

[0056] Figure 2 A schematic diagram of a wafer drying process chamber provided by an embodiment of the application is shown.

[0057] Figure 3 A schematic diagram of communication among components in wafer drying control provided by an embodiment of the application is shown.

[0058] Figure 4 A schematic diagram of the internal structure of a wafer drying control device provided by an embodiment of the application is shown. DETAILED DESCRIPTION

[0059] To make the objectives, technical solutions and advantages of the application clearer, the technical solutions of the application will be described below in detail with reference to the embodiments of the application and the corresponding drawings. Obviously, the described embodiments are only some of the embodiments of the application, but not all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the application.

[0060] The technical solutions provided by the embodiments of the application will be described in detail below with reference to the drawings.

[0061] Figure 1 A flowchart of a wafer drying control method provided by an embodiment of the application is shown.

[0062] The analysis method involved in the embodiments of this application can be implemented by a terminal device or a server, and this application does not impose any special limitations on it. For ease of understanding and description, the following embodiments are all described in detail using a server as an example.

[0063] It should be noted that the server can be a single device or a system composed of multiple devices, i.e., a distributed server. This application does not make any specific limitations on this.

[0064] like Figure 1 As shown in the embodiment of this application, a wafer drying control method includes:

[0065] Step 101: By using multiple sensors pre-deployed in the wafer drying equipment, real-time data on wafer surface state, temperature distribution, and airflow state are collected within the wafer drying process chamber to construct a multimodal state feature containing time-series characteristics.

[0066] It should be noted that the process chamber of the wafer drying equipment in this embodiment includes an air inlet device, a wafer carrier device, and an air outlet device.

[0067] Figure 2 This is a schematic diagram of a wafer drying process cavity provided in an embodiment of this application. Figure 2 As shown, the wafer drying process chamber includes an air inlet device, a wafer carrier, an air outlet device, a heating device, and a variable frequency fan. Within the sealed wafer drying process chamber, the wafer is placed on a movable wafer carrier. It is understood that the process chamber employs a highly airtight design to ensure isolation between the internal environment and the external environment, preventing contaminants from entering. Exemplarily, the movable wafer carrier is designed according to the wafer size and process requirements, typically made of special materials that are resistant to high temperatures and have low gas release. Its structure ensures the wafer remains stable during movement while avoiding mechanical damage to the wafer surface. After the wafer is precisely placed on the positioning mechanism of the wafer carrier, the entire wafer carrier enters the sealed wafer drying process chamber, preparing for subsequent drying processes.

[0068] The air intake device includes a first spray pipe and a second spray pipe, used to uniformly spray heated nitrogen gas into the process chamber. Exemplarily, the first spray pipe is responsible for forming a uniform airflow layer in the horizontal direction, while the second spray pipe supplements the airflow from top to bottom. This effectively eliminates drying dead zones and ensures that all parts of the wafer are in contact with stable-temperature hot nitrogen gas. It is understood that this multi-directional airflow distribution method is suitable for wafer carriers with complex three-dimensional structures, enabling a more thorough drying effect.

[0069] The nitrogen gas flowing through the spray pipeline is continuously heated by the heating device integrated in the spray pipeline, which is the fundamental way to ensure the stability of the process temperature. It should be noted that the heating device usually adopts a high-precision electric heating element, and the power output is adjusted by a precise temperature control system. Specifically, the nitrogen gas is fully heated by the heating device when flowing through the spray pipeline, ensuring that the output nitrogen gas reaches the set temperature. Exemplarily, the heating device adopts a segmented control mode, which can dynamically adjust the heating power of each heating unit according to the real-time monitored nitrogen gas temperature, avoiding the temperature uneven phenomenon caused by pipeline heat dissipation.

[0070] The gas outlet device is usually arranged at the bottom of the wafer drying process chamber to discharge the waste gas containing water vapor in the wafer drying process chamber from the wafer drying process chamber, so as to maintain the pressure balance in the wafer drying process chamber.

[0071] Figure 3 A schematic diagram of communication between components in wafer drying control is provided for the embodiments of the present application. As shown in Figure 3 The computer control system in the present application is connected to the Ethernet switch through the interface, and the Ethernet switch is connected to the heating device controller, the frequency conversion fan controller, and the air inlet device and the air outlet device controller. The computer control system sends control signals to the corresponding heating device controller, frequency conversion fan controller, or air inlet device and air outlet device controller through the Ethernet switch, thereby realizing wafer drying control.

[0072] In the embodiment, a digital representation reflecting the drying process state comprehensively is established through collaborative collection and feature fusion of multi-source sensor data. Specifically, the high-resolution optical sensor refers to a visual detection device equipped with a specific waveband light source and a high-quality imaging element, which continuously collects wafer surface image data at a fixed sampling frequency, ensuring that a wafer surface image with clear contrast is obtained, specifically through an industrial-grade area array camera installed above the process cavity. It can be understood that these image data contain rich surface state information. By performing grayscale processing on the wafer surface image data, the color image is converted into a grayscale image to highlight the water distribution features, which helps to simplify the subsequent computational complexity. Then, a region segmentation algorithm is used to divide the wafer surface into multiple analysis regions, and the size of each region is pre-set according to the wafer diameter and detection accuracy requirements to facilitate subsequent local feature extraction. It should be noted that the average grayscale value of each segmented region is calculated by statistically analyzing the grayscale values of all pixels in the region and taking the arithmetic mean, which can reflect the overall brightness of the region. When the average grayscale value of a certain region is lower than the pre-set grayscale threshold, it indicates that there may be water residue in the region, because the water film will absorb and scatter light, thereby reducing the intensity of reflected light. The system automatically records the center coordinates of these suspiciously wet regions to form a suspiciously wet region coordinate map, which provides target positioning for subsequent accurate measurement.

[0073] Further, the near-infrared spectral sensor is installed on a precisely positionable motion mechanism. The system moves the probe head of the spectral sensor above each suspicious region in sequence according to the position indicated by the suspiciously wet region coordinate map. Exemplarily, when performing point spectral measurement, the near-infrared spectral sensor emits a wide-spectrum light containing a near-infrared waveband to the target region and receives its reflected light to generate near-infrared reflectance spectrum data of the region. It should be noted that analyzing the spectral absorption intensity in the water feature absorption waveband is based on the characteristic absorption peaks of water molecules at specific wavelengths (such as near 1450 nm or 1940 nm), and the spectral response signal of water is quantified by calculating the ratio of the absorption intensity of these characteristic peak positions to the baseline intensity. Exemplarily, the pre-calibrated absorption intensity and water content relationship model is a standard curve established by training sample data, and the system can convert the measured spectral absorption intensity into the corresponding water content value, realizing the quantitative conversion from optical signal to physical quantity, and ensuring the accuracy and reliability of the measurement results.

[0074] It can be understood that the fusion of the water content data and the average gray value corresponding to the suspiciously humid region is realized by establishing a statistical correlation model between the two. Specifically, the local linear mapping model between the gray value and the water content is established by using the least squares method to perform linear regression analysis on the gray value and the water content value obtained by near-infrared measurement in the same region to obtain the slope and intercept parameters. Exemplarily, the conversion of the gray value of the wafer surface into water content data is to substitute the gray value of each pixel in the optical image into the established local linear mapping model to calculate the corresponding water content value, thereby converting the optical intensity image into a quantitative water distribution map. It should be noted that by statistically analyzing the distribution map, the maximum water content, the average water content and the water distribution uniformity index are extracted from the full-field water content distribution map, and these features together constitute the wafer surface state sub-feature reflecting the water state of the wafer surface. Among them, the water distribution uniformity is quantified by calculating the standard deviation of the water content of all pixel points.

[0075] Exemplarily, in each segmented region, the surface roughness variation is quantified by calculating the texture feature value of the image, and the water residue index is extracted by analyzing the distribution characteristics of the pixel gray value, and these indexes together constitute the wafer surface state sub-feature.

[0076] It should be further noted that the construction of the temperature distribution data sub-feature depends on the deployment of the distributed temperature sensor array. In the present embodiment, the distributed temperature sensor array refers to a plurality of temperature detection nodes arranged in a specific spatial pattern in the process chamber, which synchronously collect temperature readings at their respective positions. Specifically, the overall temperature level is obtained by calculating the average value of all temperature readings, and the uniformity degree of the temperature distribution is quantified by calculating the standard deviation, and these two parameters together represent the spatial distribution characteristics of the temperature field. It can be understood that the extraction of the temperature change rate feature based on time series analysis is realized by performing difference operation on the temperature data of a plurality of consecutive sampling periods, which helps to identify the trend and dynamic characteristics of the temperature change. Exemplarily, the system records the temperature values of each sensor position in the last several sampling periods, and calculates the temperature change slope of the position by linear regression analysis, and these change rate features and the spatial distribution characteristics together construct the temperature distribution data sub-feature, providing complete data support for the temperature field stability evaluation.

[0077] Regarding the construction of the airflow state data sub-feature, the embodiment adopts a collaborative measurement scheme of multiple types of sensors. The wind speed sensor group refers to multiple wind speed detection devices arranged at key positions in the cavity. These devices collect airflow speed data at a high frequency. Specifically, the relative difference degree of airflow distribution is evaluated by calculating the coefficient of variation of airflow speed data at all measurement points, and the absolute change range of airflow is quantified by calculating the fluctuation amplitude. These two statistical indicators together reflect the uniformity of airflow distribution. It can be understood that the system combines the airflow direction sensor data to synthesize airflow vectors from speed data and direction data, and analyzes the spatial distribution of these vectors to identify the main mode and abnormal conditions of the airflow, thereby evaluating the airflow distribution pattern. Exemplarily, wind speed sensors and wind direction sensors are deployed in some key areas at the same time, and the correlation characteristics of the two are analyzed to construct a more complete airflow state data sub-feature, providing multi-dimensional data input for subsequent airflow distribution uniformity evaluation.

[0078] Finally, the synthesis of multi-modal state features is a key data fusion process. It should be noted that aligning the wafer surface state sub-feature, temperature distribution data sub-feature and airflow state data sub-feature in time sequence is achieved through a unified timestamp mechanism, ensuring that data from different sources remain synchronized in the time dimension. Specifically, the system will mark the same time identifier for all sensor data within each sampling period, and strictly organize it in chronological order when performing feature splicing. It can be understood that the feature splicing process also needs to consider the differences in the dimensions and numerical ranges of each sub-feature, so before splicing, each sub-feature will be standardized to eliminate the bias that may be caused by different dimensions. Exemplarily, the system will maintain a fixed-dimensional feature vector template, and fill the processed sub-features into the corresponding positions in the predetermined order to form a complete multi-modal state feature containing historical state and real-time state.

[0079] Step 102, based on the multi-modal state feature, analyze the wafer surface drying uniformity change trend, temperature field stability index and airflow distribution uniformity to generate the process health degree evaluation result of the current drying stage.

[0080] In this embodiment, it can be understood that the extraction of drying progress data of different regions of the wafer surface from the multi-modal state features is based on the pre-divided wafer surface regions, and the drying progress data of each region is derived from the moisture residue index obtained by processing the surface image collected by the optical sensor. Illustratively, when calculating the drying rate change curve of each region in the current time window by the sliding window algorithm, the system will maintain a fixed length data window, in which the drying progress data of each region is linearly fitted to obtain the instantaneous drying rate of the region, and the complete drying rate change curve is constructed by continuously moving the time window. It should be noted that based on the drying rate change curve, the acceleration or deceleration trend of the drying rate is realized by calculating the first and second derivatives of the drying rate curve, which can accurately capture the change direction and intensity of the drying rate.

[0081] Further, the synchronization index of the drying rate between regions is calculated by analyzing the correlation coefficient and phase difference of the drying rate curves of different regions, which helps to evaluate the drying coordination of the entire wafer surface. Finally, by statistically analyzing the dispersion degree of the drying progress data of each region, the drying uniformity index is calculated using a combination algorithm of standard deviation and coefficient of variation, and combined with the drying rate trend and synchronization index mentioned above, a drying uniformity change trend evaluation report containing quantitative analysis and qualitative judgment is generated, which is finally converted into a quantitative drying uniformity health score as an important part of the process health evaluation result.

[0082] For temperature field stability index analysis, this embodiment adopts a spatial and temporal evaluation strategy. It can be understood that after extracting the real-time readings of the temperature sensor array from the multi-modal state features, the three-dimensional temperature field distribution model in the process cavity is reconstructed by a spatial interpolation algorithm based on the inverse distance weighted interpolation method or the Kriging interpolation method, which can construct a continuous three-dimensional temperature field from discrete temperature sensor data. Illustratively, based on the three-dimensional temperature field distribution model, the maximum and average values of the temperature gradient are obtained by spatial differentiation of the temperature field, and these gradient parameters can accurately reflect the uniformity characteristics of the temperature field. It should be noted that the analysis of the distribution characteristics of high temperature and low temperature regions is realized by cluster analysis of the temperature field, and the system will identify high temperature clusters and low temperature clusters in the temperature field and calculate the spatial distribution of these regions to generate the temperature field uniformity index.

[0083] In the time dimension, the time series of historical temperature data is analyzed, and methods such as fast Fourier transform or wavelet analysis are used to detect the fluctuation frequency and amplitude of the temperature field, and then the temperature stability coefficient is calculated. Finally, combined with the temperature field uniformity index and the temperature stability coefficient, a multi-index fusion algorithm based on weighted summation or fuzzy reasoning is used to generate a comprehensive temperature field stability evaluation result, and the result is converted into a temperature field health degree score through a pre-set mapping relationship, and is included in the process health degree evaluation system.

[0084] For the analysis of airflow distribution uniformity, a complete evaluation process from basic statistics to advanced pattern recognition is established. It can be understood that after extracting the wind speed measurement data of multiple wind speed sensors in the process cavity from the multi-modal state characteristics, the variation coefficient and the range of the airflow velocity of each measurement point are calculated through basic statistical analysis. The variation coefficient reflects the relative dispersion degree of the airflow velocity, and the range reflects the absolute fluctuation range of the airflow velocity. These two indicators together constitute the primary index of airflow distribution uniformity. For example, the autocorrelation characteristics of the airflow velocity time series data are analyzed by calculating the correlation between the time series and its lag version, which can effectively detect the periodic fluctuation pattern in the airflow. It should be noted that the identification of abnormal fluctuation events is completed by setting a dynamic threshold and combining a mutation detection algorithm. When a sharp change in airflow velocity is detected, the system will record these abnormal events and generate corresponding airflow stability evaluation data.

[0085] Further, combined with the primary index of airflow distribution uniformity and the airflow stability evaluation data, the process of calculating the airflow distribution uniformity level by the fuzzy logic reasoning system includes: first, define the fuzzy set and membership function for each input index, then establish a fuzzy rule base based on expert knowledge, and finally get the accurate airflow distribution uniformity level through fuzzy reasoning and defuzzification. Finally, the airflow distribution uniformity level is matched with the pre-set process requirements, the accurate airflow health degree score is output by querying the pre-set level-score corresponding table, and the score is integrated into the process health degree evaluation result, completing the quantitative evaluation of the entire drying process airflow state.

[0086] Step 103, based on the process health degree evaluation result, iteratively solve the optimal control parameter combination of the wafer drying control system, and according to the optimal control parameter combination, dynamically generate control action instructions including heating device partition power setting value, variable frequency fan speed gradient setting value and multi-way valve opening degree coordination setting value.

[0087] In the present embodiment, the control parameter candidate solution refers to a complete parameter set containing heating device partition power values, variable frequency fan rotating speed values and multi-way valve opening degree values. Exemplarily, the heating device can be divided into multiple partition heating units, and the heating device partition power values contain multiple power setting values, each of which corresponds to the output power of a partition heating unit. The variable frequency fan rotating speed value refers to the speed setting value of the fan motor driving the gas circulation in the process chamber. The multi-way valve opening degree value refers to the valve opening degree setting value set controlling the air inlet device and the air outlet device.

[0088] Exemplarily, the initialization process generates an initial population within a pre-set parameter feasible region, which can ensure the uniform distribution and diversity of the initial solution in the parameter space. It should be noted that the initialization of the heating device partition power values needs to consider the power ratio relationship between the partition heating units to avoid local overheating or power conflict; the initialization of the variable frequency fan rotating speed value is based on the safe working range of the fan to ensure that the rotating speed is set within the effective range; and the initialization of the multi-way valve opening degree value needs to meet the basic requirement of airflow balance to prevent airflow turbulence caused by improper valve opening degree setting.

[0089] Further, when the population is iteratively evolved through selection, crossover and mutation operations of the multi-objective genetic algorithm, the selection operation selects high-quality individuals by comparing the performance of the candidate solutions on each objective function; the crossover operation adopts a simulated binary crossover method to generate offspring solutions by exchanging parameter information between parent solutions, which can effectively maintain population diversity and promote the spread of excellent genes; the mutation operation adopts a polynomial mutation strategy to randomly disturb some dimensions of the solution vector with a small probability, which helps to jump out of the local optimum and explore new search space. In each generation evolution process, the fitness value of each control parameter candidate solution is calculated, and based on these fitness values, a new generation of candidate solutions is generated through genetic operations. Finally, the iterative evaluation process is repeated until the convergence condition is met, which can be the maximum number of iterations, the improvement degree of fitness value or the distribution characteristics of the solution set, and finally the optimal control parameter combination of the wafer drying control system is output.

[0090] For the control action instruction generation process, the embodiment establishes a complete conversion mechanism from the optimization result to the equipment instruction. It can be understood that when the heating device partition power value in the optimal control parameter combination is mapped to the power setting value corresponding to each partition heating unit, the thermal coupling effect between partitions needs to be considered. Exemplarily, the coupling relationship model between partitions is established through heat conduction equation and computational fluid dynamics analysis, which can predict the influence of power change in a certain partition on the temperature field of other partitions, thereby avoiding thermal field conflicts. It should be noted that when the variable frequency fan speed value in the optimal control parameter combination is converted into the speed gradient setting value, the system will define the speed ratio relationship of the fans in different areas, which is based on the airflow demand difference of each area, and will also design a smooth acceleration and deceleration curve to avoid airflow sudden change causing impact on the wafer surface.

[0091] Further, when the multi-way valve opening value in the optimal control parameter combination is analyzed into the opening coordination setting value, the system will establish a cooperative control logic between the valves to ensure that the valve opening changes are synchronized in time and evenly distributed in space. Finally, when the power setting value, speed gradient setting value and opening coordination setting value are packaged according to the equipment communication protocol format, the system will organize all parameters according to the predetermined data frame structure, add a time stamp to ensure timing accuracy, set a priority identifier to distinguish between key instructions and ordinary instructions, and finally generate a complete control action instruction.

[0092] Step 104, convert the control action instruction into a driving signal executable by the wafer drying control system, control the partition heating units of the heating device, the speed regulation mechanism of the variable frequency fan and the opening actuator of the multi-way valve through the distributed drive circuit.

[0093] In the embodiment, it should be noted that the driving signal conversion process involves the cooperative work of multiple special driving modules. Specifically, the power driving module converts the heating device partition power setting value into a pulse width modulation signal with a corresponding duty cycle, accurately controls the output power of each partition heating unit by adjusting the pulse width. Exemplarily, the motor driving module converts the variable frequency fan speed gradient setting value into an alternating current signal with a corresponding frequency or an analog voltage signal with a corresponding amplitude, to achieve smooth adjustment of the variable frequency fan speed. It can be understood that the valve driving module converts the multi-way valve opening coordination setting value into an analog current signal with a corresponding intensity or an accurate digital pulse signal to drive each valve actuator to reach the specified opening position. It should be further noted that the distributed drive circuit adopts a modular design, and each driving module exchanges data and synchronously controls through an industrial bus, ensuring that all actuator mechanisms can work cooperatively according to the requirements of the control action instruction.

[0094] Step 105, during the drying process, the actual running state of each actuator in the wafer drying control system is continuously monitored, the monitoring data is compared with the expected state in real time, and the control parameter is updated when the state deviation is detected to form a closed-loop control.

[0095] In this embodiment, the actual power data of each partition heating unit of the heating device, the actual speed data of the frequency conversion fan and the actual opening data of the opening actuator are collected in real time through the sensor network, which is realized through the distributed monitoring system. It can be understood that the sensor network includes a power sensor installed on each partition heating unit, a speed encoder integrated in the frequency conversion fan, and a position sensor arranged on the opening actuator. Exemplarily, the power sensor calculates the actual power output by measuring the current and voltage values of the heating element, the speed encoder obtains the real-time speed of the fan through pulse counting, and the position sensor detects the accurate opening of the valve through a potentiometer or an optical encoder. These sensors synchronously collect data at a fixed sampling period and transmit them to the central processing unit through the industrial bus to form a complete set of actual running state data of the actuator.

[0096] It needs to be further explained that the actual running state data set is compared with the expected state data in the control action instruction item by item through a special deviation analysis algorithm. Specifically, the system extracts the expected power set value, speed set value and opening set value from the currently executed control action instruction, and compares them with the actual data collected by the sensor one by one. It can be understood that the absolute values of the power deviation, speed deviation and opening deviation are calculated by simple arithmetic difference operation, and the deviation trend is analyzed by linear regression analysis of the deviation values of continuous multiple sampling periods. Exemplarily, the system maintains a sliding time window, analyzes the change slope and fluctuation characteristics of each deviation value in the window, and judges whether the running state of the actuator is stable, improved or deteriorated.

[0097] In this embodiment, when any deviation absolute value exceeds the preset threshold or the deviation trend is abnormal, the system will start the automatic optimization trigger mechanism. It needs to be noted that the preset threshold is set in advance according to the technical specifications and control accuracy requirements of each actuator, and different actuators can set different threshold ranges. It can be understood that the abnormal deviation trend includes but is not limited to the continuous increase, periodic oscillation or mutation of the deviation value, and these abnormal trends often indicate the performance degradation of the equipment or the change of the external environment. Specifically, when the abnormal situation is detected, the system automatically records the current process state data, including the multi-modal state characteristics and the actual running state of the actuator, and uses these data as the initial conditions for re-execution of the multi-objective optimization algorithm.

[0098] Finally, the process of generating an updated optimal control parameter combination through the re-executed multi-objective optimization algorithm embodies the self-learning ability of the system. Exemplarily, the system retains effective information in the previous optimization process, such as population diversity, convergence characteristics, etc., to accelerate the optimization solution of the new time. It can be understood that when generating a new control action instruction based on the updated optimal control parameter combination, the system will consider the smooth transition of the control instruction to avoid the impact on the equipment and process due to parameter mutation. This closed-loop optimization mechanism enables the entire control system to continuously adapt to changes in the operating environment, continuously adjust and improve itself, and always keep the drying process running in an optimal state, thereby ensuring the continuous improvement of the quality and production efficiency of the wafer product.

[0099] The above is the method embodiment of the present application. Based on the same inventive concept, the present embodiment also provides a wafer drying control device, the structure of which is shown in Figure 4 .

[0100] Figure 4 The internal structure of a wafer drying control device provided by the present embodiment is shown in Figure 4 . The device includes:

[0101] at least one processor;

[0102] and a memory in communication connection with the at least one processor;

[0103] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0104] Through the multi-sensor pre-deployed in the wafer drying device, real-time collection of wafer surface state data, temperature distribution data and airflow state data in the wafer drying process cavity is performed to construct multi-modal state features containing time sequence characteristics; the process cavity includes an air inlet device, a wafer supporting device and an air outlet device;

[0105] Based on the multi-modal state features, the wafer surface drying uniformity change trend, the temperature field stability index and the airflow distribution uniformity are analyzed to generate a process health degree evaluation result of the current drying stage;

[0106] Based on the process health degree evaluation result, the optimal control parameter combination of the wafer drying control system is iteratively solved, and based on the optimal control parameter combination, a control action instruction containing heating device partition power setting value, variable frequency fan rotating speed gradient setting value and multi-way valve opening degree coordination setting value is dynamically generated;

[0107] The control action instruction is converted into a drive signal executable by the wafer drying control system, and the drive signal is used to control the partition heating unit of the heating device, the speed regulation mechanism of the variable frequency fan, and the opening actuator of the multi-way valve through a distributed drive circuit.

[0108] During the drying process, the actual operating state of each execution mechanism in the wafer drying control system is continuously monitored, the monitoring data is compared with the expected state in real time, and the control parameter is updated when the state deviation is detected, so as to form a closed loop control.

[0109] The embodiments of the present application also provide a non-volatile computer storage medium, which stores computer executable instructions, and the computer executable instructions can be executed to:

[0110] Through the multi-sensor pre-deployed in the wafer drying equipment, wafer surface state data, temperature distribution data and airflow state data in the wafer drying process cavity are collected in real time, and multi-modal state features containing time sequence features are constructed; the process cavity includes an air inlet device, a wafer supporting device and an air outlet device;

[0111] Based on the multi-modal state features, the wafer surface drying uniformity change trend, the temperature field stability index and the airflow distribution uniformity are analyzed to generate a process health degree evaluation result of the current drying stage;

[0112] Based on the process health degree evaluation result, the optimal control parameter combination of the wafer drying control system is iteratively solved, and the control action instruction containing the heating device partition power set value, the variable frequency fan speed gradient set value and the multi-way valve opening coordination set value is dynamically generated according to the optimal control parameter combination;

[0113] The control action instruction is converted into a drive signal executable by the wafer drying control system, and the drive signal is used to control the partition heating unit of the heating device, the speed regulation mechanism of the variable frequency fan, and the opening actuator of the multi-way valve through a distributed drive circuit;

[0114] During the drying process, the actual operating state of each execution mechanism in the wafer drying control system is continuously monitored, the monitoring data is compared with the expected state in real time, and the control parameter is updated when the state deviation is detected, so as to form a closed loop control.

[0115] Each of the embodiments in the present application is described in a progressive manner, and the same or similar parts of each of the embodiments can be referred to each other. Each of the embodiments mainly describes the difference from other embodiments. In particular, for the device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the related parts can be referred to the part of the method embodiment.

[0116] The device and medium provided by the embodiments of the present application are one-to-one corresponding, and therefore, the device and medium also have similar beneficial technical effects to the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the device and medium will not be described here again.

[0117] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

Claims

1. A wafer baking control method, characterized by, The method comprises: Real-time acquisition of wafer surface state data, temperature distribution data and airflow state data in the wafer baking process cavity by a multi-sensor pre-deployed in the wafer baking equipment, to construct a multi-modal state feature containing time series characteristics; the process cavity comprises an air inlet device, a wafer support device and an air outlet device; Based on the multi-modal state feature, analyze the wafer surface drying uniformity change trend, temperature field stability index and airflow distribution uniformity to generate the process health degree evaluation result of the current baking stage; Based on the process health degree evaluation result, iteratively solve the optimal control parameter combination of the wafer baking control system, and dynamically generate the control action instruction containing the heating device partition power setting value, the variable frequency fan speed gradient setting value and the multi-way valve opening degree coordination setting value according to the optimal control parameter combination; Convert the control action instruction into a driving signal executable by the wafer baking control system, and control the partition heating unit of the heating device, the speed regulation mechanism of the variable frequency fan and the opening actuator of the multi-way valve through a distributed driving circuit; During the baking process, continuously monitor the actual running state of each execution mechanism in the wafer baking control system, compare the monitoring data with the expected state in real time, and trigger control parameter update when state deviation is detected to form a closed-loop control.

2. The wafer baking control method of claim 1, wherein The analysis of the drying uniformity change trend of the wafer surface specifically comprises: Extract the drying progress data of different regions of the wafer surface from the multi-modal state feature, and calculate the drying rate change curve of each region in the current time window through a sliding window algorithm; Based on the drying rate change curve, identify the drying rate trend, and calculate the synchronization index of the drying rate between regions; the drying rate trend includes acceleration trend and deceleration trend; Through statistical analysis of the dispersion degree of the drying progress data of each region, calculate the drying uniformity index, and generate a drying uniformity change trend evaluation report in combination with the drying rate trend and the synchronization index; Compare the drying uniformity change trend evaluation report with the preset process benchmark, and output the quantitative drying uniformity health degree score as a part of the process health degree evaluation result.

3. The wafer baking control method of claim 1, wherein The analysis of the temperature field stability index specifically comprises: Extract the real-time readings of the temperature sensor array from the multi-modal state feature, and reconstruct a three-dimensional temperature field distribution model in the process cavity through a spatial interpolation algorithm; Based on the three-dimensional temperature field distribution model, calculate the maximum and average values of the temperature gradient, and analyze the distribution characteristics of the high-temperature and low-temperature regions to generate a temperature field uniformity index; Process historical temperature data through a time series analysis method to detect the fluctuation frequency and amplitude of the temperature field, and calculate a temperature stability coefficient; Fuse the temperature field uniformity index and the temperature stability coefficient to generate a comprehensive temperature field stability evaluation result, and map the comprehensive temperature field stability evaluation result to a temperature field health degree score as a part of the process health degree evaluation result.

4. The wafer baking control method of claim 1, wherein The analysis of the airflow distribution uniformity specifically comprises: extracting wind speed measurement data of multiple wind speed sensors in the process cavity from the multi-modal state features, and calculating the variation coefficient and range of the air flow velocity of each measurement point to generate an air flow distribution uniformity primary index; analyzing the autocorrelation characteristics of the air flow velocity time series data, detecting periodic fluctuation patterns of the air flow, and identifying abnormal fluctuation events to generate air flow stability evaluation data; combining the air flow distribution uniformity primary index and the air flow stability evaluation data, calculating the air flow distribution uniformity level through a fuzzy logic reasoning system; matching the air flow distribution uniformity level with the preset process requirements, outputting an air flow health score, and integrating the air flow health score into the process health evaluation result.

5. The wafer baking control method of claim 1, wherein Based on the process health evaluation result, the optimal control parameter combination of the wafer drying control system is iteratively solved, specifically including: initializing the control parameter candidate solutions in the wafer drying control system; the control parameters include heating device partition power values, variable frequency fan speed values, and multi-way valve opening values; evolving the population through selection, crossover and mutation operations of the multi-objective genetic algorithm, calculating the fitness value of each control parameter candidate solution in each generation evolution process, and generating a new generation of candidate solutions based on the fitness value; repeating the iterative evaluation process until the convergence condition is met, and outputting the optimal control parameter combination of the wafer drying control system.

6. The wafer baking control method of claim 1, wherein According to the optimal control parameter combination, dynamic control action instructions containing heating device partition power setting values, variable frequency fan speed gradient setting values, and multi-way valve opening coordination setting values are generated, specifically including: mapping the heating device partition power values in the optimal control parameter combination to the corresponding power setting values of each partition heating unit, and establishing a coupling relationship model between each partition heating unit; convert the variable frequency fan speed value in the optimal control parameter combination into a speed gradient setting value, define the speed ratio relationship and acceleration-deceleration curve of the fans in different regions; analyzing the multi-way valve opening value in the optimal control parameter combination into an opening coordination setting value, ensuring that the valve opening changes are synchronized and the air flow distribution is balanced; packaging the power setting value, speed gradient setting value and opening coordination setting value according to the device communication protocol format to generate control action instructions with time stamp and priority identifier.

7. The wafer baking control method of claim 1, wherein Continuously monitor the actual running state of each actuator in the wafer drying control system, compare the monitoring data with the expected state in real time, and trigger control parameter update when state deviation is detected to form a closed loop control, specifically including: real-time acquisition of actual power data of each partition heating unit of the heating device, actual speed data of the variable frequency fan, and actual opening data of the opening actuator through the sensor network to form an actual running state data set of the actuator; comparing the actual running state data set with the expected state data in the control action instruction item by item, calculating the absolute value and change trend of the power deviation, speed deviation and opening deviation; when the absolute value of any deviation exceeds the preset threshold or the change trend of the deviation is abnormal, the current actual running state is taken as the initial condition, and the iterative solution execution process is triggered again; An updated optimal control parameter combination is generated by re-executing the iterative solving, and a new control action instruction is generated based on the updated optimal control parameter combination, so as to realize dynamic adjustment and closed-loop optimization of the control parameters.

8. The wafer baking control method of claim 1, wherein Through the multi-sensor pre-deployed in the wafer baking equipment, wafer surface state data, temperature distribution data and airflow state data in the wafer baking process cavity are collected in real time, multi-modal state features containing time sequence characteristics are constructed, specifically including: Wafer surface image data is collected by a high-resolution optical sensor, and the wafer surface image data is subjected to grayscale and region segmentation processing, the average grayscale value of each segmented region is calculated, to identify suspicious wetting areas with an average grayscale value lower than a preset grayscale threshold, and a suspicious wetting area coordinate map is generated; According to the suspicious wetting area coordinate map, a near-infrared spectrum sensor is used to perform point spectrum measurement on each suspicious wetting area, corresponding near-infrared reflection spectrum data is collected, and the spectrum absorption intensity in the water feature absorption band is analyzed, and a pre-calibrated absorption intensity and water content relationship model is used to calculate the water content data of each suspicious wetting area; The water content data and the average grayscale value of the corresponding suspicious wetting area are fused to establish a local linear mapping model between the grayscale value and the water content, to convert the grayscale value of the wafer surface to the water content data, generate a wafer full-field water content distribution map, and extract the maximum water content, average water content and water distribution uniformity index from the full-field water content distribution map to construct wafer surface state sub-features; A distributed temperature sensor array is used to collect temperature readings at multiple spatial positions in the process cavity, the average value and standard deviation of the temperature readings are calculated to represent the temperature distribution uniformity, and a temperature change rate feature is extracted based on time series analysis to construct temperature distribution data sub-features; An airflow speed sensor group is used to collect airflow speed data at key positions in the cavity, the coefficient of variation and fluctuation amplitude of the airflow speed data are calculated, and the airflow distribution pattern is evaluated in combination with airflow direction sensor data to construct airflow state data sub-features; The wafer surface state sub-features, temperature distribution data sub-features and airflow state data sub-features are aligned and spliced in time sequence to form multi-modal state features; the multi-modal state features contain historical state and real-time state.

9. A wafer baking control apparatus characterized by comprising: The device comprises: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the wafer baking control method of any one of claims 1-8.

10. A non-transitory computer storage medium storing computer-executable instructions, the computer-executable instructions comprising instructions for: receiving a request to access a file; determining whether the file is stored in a cache; and in response to determining that the file is stored in the cache, providing access to the file from the cache. The computer executable instructions, when executed, implement the wafer baking control method of any one of claims 1-8.

Images