Wafer cleaning equipment multi-cavity liquid medicine intelligent circulation control method

Abstract

The invention relates to the technical field of information, in particular to a wafer cleaning equipment multi-cavity liquid medicine intelligent circulation control method of a semiconductor wafer cleaning technology, which comprises the following steps of: sending optimized parameters obtained by calculation to a control unit of a corresponding cleaning cavity, and adjusting the actual working flow of a cavity circulating pump and the actual working state of a filtering device; the recycling process of the liquid medicine in the cavity is dynamically optimized, so that the quality of the liquid medicine is recovered to be within the range required by the process; according to cleaning process requirements of different wafer products, a liquid medicine quality parameter threshold range of each cleaning cavity is pre-established, and when the product model of the wafer cleaning equipment is switched, a liquid medicine quality control threshold corresponding to each cavity is automatically matched according to the product model, so that dynamic adjustment of a liquid medicine circulation mode and parameters of the sub-cavities is realized.

Description

Technical Field

[0001] The present invention relates to the field of information technology, in particular to semiconductor wafer cleaning technology, and in particular to a multi-cavity liquid intelligent circulation control method for wafer cleaning equipment. Background Art

[0002] In the semiconductor wafer manufacturing process, the wafer needs to be cleaned. In the wafer cleaning process, there are interrelated technical problems in the recycling of liquid medicine in multiple cleaning chambers. First, the quality parameters of the liquid medicine in each chamber, such as concentration, temperature and pH, will gradually deviate from the process requirements over time, affecting the cleaning effect. Secondly, different wafer products have different requirements for liquid medicine quality, and the liquid medicine circulation parameters of each chamber need to be dynamically adjusted. Furthermore, when multiple chambers are collaboratively cleaned, how to optimize scheduling to improve overall efficiency and liquid medicine utilization is also a major challenge. These problems are interrelated and require a comprehensive solution. By installing sensors in each chamber to monitor the quality of the liquid medicine in real time, and combining intelligent control algorithms to dynamically adjust the flow rate of the circulation pump and the parameters of the filter device, the quality of the liquid medicine can be kept within the process requirements. At the same time, based on historical data, a correlation model between the use time of the liquid medicine and the quality attenuation is established to predict the optimal replacement cycle and reduce waste. In addition, the central controller monitors the operating status of each chamber in real time, automatically selects the appropriate chamber according to the wafer cleaning progress and process requirements, and calls the corresponding liquid medicine circulation control strategy to achieve continuous cleaning of multiple products and optimal resource utilization. Summary of the invention

[0003] The present invention provides a method for intelligent circulation control of multi-chamber chemical liquid in wafer cleaning equipment, the method comprising the following steps:

[0004] Step S1, obtaining real-time data of the circulation process of the medicine solution in multiple cleaning cavities, specifically including the parameters of the concentration, temperature and pH of the medicine solution in each cavity, using the concentration sensor, temperature sensor and pH sensor installed at the key position of the circulation pipeline of each cavity to collect the data, and recording the use time of the medicine solution in each cavity;

[0005] Step S2, inputting the acquired real-time data of the quality parameters of the drug liquid in each cavity into the intelligent control algorithm model, and judging whether the current drug liquid quality of each cavity meets the preset wafer cleaning process requirements through the algorithm model, if it meets the requirements, continuing to acquire real-time data, if it does not meet the requirements, then according to the deviation degree between the drug liquid quality parameters and the preset process requirements, the intelligent control algorithm is used to calculate the optimized flow rate of the drug liquid circulation pump of the cavity and the working state parameters of the filter device;

[0006] Step S3, sending the calculated optimization parameters to the control unit of the corresponding cleaning chamber, adjusting the actual working flow of the circulation pump of the chamber and the actual working state of the filter device, dynamically optimizing the circulation process of the liquid medicine in the chamber, and restoring the quality of the liquid medicine to within the range of process requirements;

[0007] Step S4, obtaining the working parameters of each cleaning chamber and the corresponding liquid use time data during the actual operation of the wafer cleaning equipment, establishing a correlation model between the liquid use time of each chamber and the attenuation of the liquid quality through a machine learning algorithm, dynamically predicting the optimal liquid replacement cycle of each chamber, guiding the regular replacement of the liquid, and reducing liquid waste;

[0008] Step S5, obtaining the liquid circulation process parameters of each cleaning chamber and the corresponding cleaning effect detection results during the actual operation of the wafer cleaning equipment, extracting the intrinsic correlation between the liquid quality, circulation parameters and cleaning effect of each chamber from the historical big data through a data mining algorithm, forming a knowledge base of optimized liquid circulation control strategies for each chamber, and guiding the dynamic optimization of the intelligent control algorithm of each chamber;

[0009] Step S6: Pre-establish the threshold range of the liquid quality parameter of each cleaning chamber according to the cleaning process requirements of different wafer products. When the wafer cleaning equipment switches the product model, the liquid quality control threshold corresponding to each chamber is automatically matched according to the product model, so as to realize the dynamic adjustment of the liquid circulation mode and parameters of each chamber.

[0010] Step S7, obtaining real-time operation status data of the wafer cleaning equipment, specifically including the wafer cleaning progress and liquid medicine circulation parameters of each cleaning chamber, and monitoring the operation status of each chamber in real time through the central controller. When the wafer cleaning of a chamber is completed, it is immediately switched to the next pre-configured idle chamber, and according to the process requirements of the wafer to be cleaned, the corresponding liquid medicine circulation control strategy is retrieved from the knowledge base, and the liquid medicine circulation parameters of the new chamber are preset to achieve continuous multi-chamber cleaning of different wafer products;

[0011] Step S8, based on the optimized intelligent control algorithm and the chamber knowledge base, the liquid circulation process of multiple cleaning chambers is monitored and dynamically adjusted in real time. When the liquid quality parameters of a certain chamber deviate from the preset threshold, the optimization control algorithm of the chamber is triggered, and the new liquid circulation pump flow and filter device parameters are calculated. The adjustment is performed through the control unit of the chamber. At the same time, the liquid usage time and cleaning efficiency of each chamber are comprehensively analyzed, and the scheduling strategy of multi-chamber collaborative cleaning is dynamically optimized to improve the overall efficiency of wafer cleaning and liquid utilization.

[0012] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0013] The present invention discloses a multi-chamber liquid intelligent circulation control method for wafer cleaning equipment. The method collects liquid quality parameters in real time by installing sensors on the circulation pipelines of each cleaning chamber, and inputs them into an intelligent control algorithm to determine whether the process requirements are met. If not, the optimized circulation pump flow and filter device parameters are calculated and adjusted. At the same time, a machine learning algorithm is used to establish a correlation model between the use time of liquid and quality attenuation, and dynamically predict the optimal replacement cycle. Through data mining, the correlation law between liquid quality, circulation parameters and cleaning effect is extracted from historical data to form a knowledge base of sub-chamber optimization control strategies. Liquid quality thresholds are preset for different products to achieve dynamic adjustment. The operating status of each chamber is monitored in real time to achieve continuous multi-chamber cleaning. The use of each chamber is comprehensively analyzed to optimize the multi-chamber collaborative cleaning scheduling strategy. The present invention realizes the intelligent dynamic optimization control of multi-chamber liquid circulation in the wafer cleaning process, improves the cleaning efficiency and liquid utilization rate, and reduces production costs. DETAILED DESCRIPTION

[0014] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the embodiments. The present application is further described in detail below in conjunction with the embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention.

[0015] In this embodiment, a multi-chamber liquid intelligent circulation control method for wafer cleaning equipment may specifically include:

[0016] Step S1, obtaining real-time data of the circulation process of the medicine solution in multiple cleaning chambers, specifically including the parameters of the concentration, temperature and pH of the medicine solution in each chamber, using concentration sensors, temperature sensors and pH sensors installed at key positions of the circulation pipelines of each chamber to collect the data, and record the use time of the medicine solution in each chamber.

[0017] Step S1 includes: obtaining real-time collected data from concentration sensors, temperature sensors and pH sensors installed at key positions of circulation pipelines of each cleaning cavity; judging whether the real-time concentration of the drug solution in each cleaning cavity meets the preset range according to the data collected by the concentration sensor, and if it is lower than the preset range, controlling the drug adding device to add drug solution to the cleaning cavity until the concentration reaches the preset range; judging whether the real-time temperature of the drug solution in each cleaning cavity meets the preset range according to the data collected by the temperature sensor, and if it is lower than the preset range, controlling the heating device to heat the drug solution in the cleaning cavity until the temperature reaches the preset range; judging whether the real-time temperature of the drug solution in each cleaning cavity meets the preset range according to the data collected by the pH sensor whether the real-time pH value of the cleaning cavity meets the preset range; if it exceeds the preset range, the acid-adding or alkali-adding device is controlled to add acid or alkali into the cleaning cavity until the pH value reaches the preset range; the real-time concentration, the real-time temperature and the real-time pH value data of the medicine in each cleaning cavity are summarized to obtain the comprehensive quality parameters of the medicine in each cleaning cavity; the start time and the end time of the use of each cleaning cavity medicine are recorded, and the use time of each cleaning cavity medicine is calculated; the comprehensive quality parameters of each cleaning cavity medicine and the use time are correlated and analyzed, and a medicine quality prediction model is established by using a support vector machine algorithm; according to the medicine quality prediction model, the optimal replacement time of the medicine in each cleaning cavity is predicted.

[0018] Specifically, concentration sensors, temperature sensors and pH sensors are installed at key positions of the circulation pipeline of the cleaning chamber to collect the concentration, temperature and pH data of the liquid medicine in real time. When the concentration sensor detects that the concentration of the liquid medicine is lower than the lower limit of the preset range by 10%, the dosing device is controlled to add liquid medicine to the cleaning chamber at a speed of 20 ml / s until the concentration reaches 12% of the preset range. When the temperature sensor detects that the temperature of the liquid medicine is lower than the lower limit of the preset range by 40°C, the heating device is controlled to heat the liquid medicine in the cleaning chamber at a speed of 1°C / min until the temperature reaches 45°C in the preset range. When the pH sensor detects that the pH value of the liquid medicine exceeds the preset range of 5 to 5, the acid adding device is controlled to add a hydrochloric acid solution with a concentration of 1 mol / L to the cleaning chamber at a speed of 1 ml / s, or the alkali adding device is controlled to add a sodium hydroxide solution with a concentration of 1 mol / L at a speed of 1 ml / s until the pH value reaches 0 in the preset range. The real-time concentration, temperature and pH data of the liquid medicine in each cleaning chamber are weighted averaged to obtain the comprehensive quality parameters of the liquid medicine. Record the start and end time of the use of each cleaning chamber liquid, and calculate the liquid use time. Use the support vector machine algorithm, take the comprehensive quality parameters of the liquid and the use time as input, and take whether the liquid needs to be replaced as output to establish a liquid quality prediction model. By analyzing historical data, it is concluded that when the liquid is used for more than 48 hours and the comprehensive quality parameters of the liquid are lower than 85 points, it is predicted that the liquid needs to be replaced, thereby ensuring the quality stability of the liquid recycling process.

[0019] Step S2, input the real-time data of the quality parameters of the chemical liquid in each cavity into the intelligent control algorithm model, and use the algorithm model to determine whether the current quality of the chemical liquid in each cavity meets the preset wafer cleaning process requirements. If the requirements are met, continue to obtain real-time data; if not, the optimized flow rate of the chemical liquid circulation pump and the working state parameters of the filter device in the cavity are calculated by the intelligent control algorithm according to the degree of deviation between the chemical liquid quality parameters and the preset process requirements.

[0020] Step S2 includes: obtaining real-time data of the quality parameters of the drug solution in each cavity, wherein the drug solution quality parameters include pH value, temperature and concentration; using the drug solution quality parameters as the input of the intelligent control algorithm model; obtaining the pre-established wafer cleaning process quality requirement standards; judging whether the drug solution quality parameters meet the wafer cleaning process quality requirement standards; if the drug solution quality parameters meet the wafer cleaning process quality requirement standards, returning to continue to obtain the real-time data of the drug solution quality parameters; if the drug solution quality parameters do not meet the wafer cleaning process quality requirement standards, determining the degree of deviation between the drug solution quality parameters and the wafer cleaning process quality requirement standards , and determine the target value for optimization and adjustment; use a machine learning algorithm to establish a mapping relationship model between the quality parameters of the liquid medicine and the flow rate of the liquid medicine circulation pump and the working state parameters of the filter device; input the degree of deviation of the liquid medicine quality parameters into the mapping relationship model to obtain the optimized flow rate of the liquid medicine circulation pump and the working state parameters of the filter device; according to the optimized flow rate of the liquid medicine circulation pump, control the liquid medicine circulation pump to adjust the flow rate; according to the optimized working state parameters of the filter device, adjust the backwashing frequency and backwashing time of the filter device; after executing the optimization control, return to continue to obtain the real-time data of the liquid medicine quality parameters, and perform cycle judgment and optimization control.

[0021] Specifically, during the wafer cleaning process, pH meters, temperature sensors, concentration meters and other instruments are installed in each chamber to collect the pH value, temperature and concentration of the liquid in real time. These parameters are used as the input of the intelligent control algorithm model and compared with the pre-established wafer cleaning process quality standards to determine whether the liquid quality meets the standards. When the parameters deviate from the standard value, the support vector machine algorithm is used to establish a mapping relationship model between the liquid quality parameters and the liquid circulation pump flow and the filter device working state parameters according to the degree of deviation. For example, when the pH value is lower than 5, the model calculates that the liquid circulation pump flow needs to be increased by 10%, the filter device backwashing frequency is increased to 2 times per hour, and the backwashing time is extended to 40 seconds. The control system automatically adjusts the liquid circulation pump flow and the working state of the filter device according to the optimized parameters, and continuously monitors the changes in the liquid quality parameters to form a closed-loop control. In this way, real-time monitoring and dynamic optimization of liquid quality can be achieved to ensure the stability and consistency of the wafer cleaning process and improve product yield.

[0022] Step S3, sending the calculated optimization parameters to the control unit of the corresponding cleaning chamber, adjusting the actual working flow of the chamber circulation pump and the actual working state of the filter device, dynamically optimizing the liquid medicine recycling process of the chamber, and restoring the liquid medicine quality to within the process requirements.

[0023] Step S3 includes: obtaining real-time data on the quality parameters of the medicinal liquid, the working flow of the circulation pump and the working status of the filtering device during the circulation of the medicinal liquid, and inputting the real-time data into the machine learning model; if the machine learning model determines that the current quality of the medicinal liquid does not meet the preset process requirements, a reinforcement learning algorithm is used to obtain a control strategy for optimizing the working flow of the circulation pump and the working status of the filtering device; according to the control strategy, the actual working flow of the circulation pump and the actual working status of the filtering device are dynamically adjusted to achieve optimized control of the medicinal liquid circulation process.

[0024] Specifically, according to the optimized parameters, the control strategy of the cleaning chamber control unit can be determined as follows: when the quality parameters of the liquid medicine are lower than the threshold value of 85%, the working flow of the circulation pump is adjusted to 120L / min, and the filtering accuracy of the filter device is adjusted to 22μm; when the quality parameters of the liquid medicine are higher than the threshold value of 85%, the working flow of the circulation pump is maintained at 100L / min, and the filtering accuracy of the filter device is maintained at 45μm. During the circulation of the liquid medicine, the pH value, conductivity, turbidity and other quality parameters of the liquid medicine are obtained every 5 minutes, and the real-time working flow of the circulation pump and the actual working status of the filter device are recorded at the same time. These real-time data are input into the machine learning model based on the long short-term memory network (LSTM) for analysis. Through the training of historical data, the machine learning model establishes a nonlinear relationship between the quality parameters of the liquid medicine and the working flow of the circulation pump and the working status of the filter device. Based on the input real-time data, it can be judged whether the current liquid medicine quality meets the process requirements. If the quality parameters of the liquid medicine are lower than the threshold, the DeepQNetwork (DQN) reinforcement learning algorithm is used to simulate different combinations of circulating pump working flow and filter device working state in the simulation environment, and the reward function (such as the magnitude of the improvement of the liquid medicine quality parameters) is used to learn the strategy of optimizing the circulating pump working flow and the filter device working state. The optimal strategy may be to increase the circulating pump working flow to 140L / min and adjust the filtration accuracy of the filter device to 1μm. According to the optimization strategy output by the machine learning model, the PLC control system dynamically issues instructions to adjust the actual working flow of the circulating pump to 140L / min and the actual filtration accuracy of the filter device to 1μm, optimizing the control of the liquid medicine recycling process. During the dynamic optimization process, if the drug solution quality parameters obtained for three consecutive times are restored to above the threshold of 85%, the dynamic optimization process is judged to be over, and the circulation pump and the filter device are restored to normal working conditions; if the drug solution quality parameters still fail to recover to above the threshold after three rounds of optimization and adjustment, the latest drug solution quality parameters, circulation pump working flow rate, and filter device working status data are input into the machine learning model for the next round of analysis and optimization. Through the exploration mechanism of the DQN algorithm, new optimization strategies are tried, such as further increasing the circulation pump working flow rate to 160L / min and adjusting the filtration accuracy of the filter device to 0.5μm. The above process is repeated continuously. Through continuous learning and optimization of the machine learning model, the circulation pump working flow rate and the filter device working status are dynamically adjusted until the drug solution quality is stabilized within the process requirements, completing the dynamic optimization of the drug solution recycling process.

[0025] Step S4, obtaining the operating parameters of each cleaning chamber and the corresponding liquid usage time data during the actual operation of the wafer cleaning equipment, and establishing a correlation model between the liquid usage time of each chamber and the liquid quality attenuation through a machine learning algorithm, dynamically predicting the optimal liquid replacement cycle of each chamber, guiding the regular replacement of the liquid, and reducing liquid waste.

[0026] Step S4 includes: obtaining the operating parameters of each cleaning chamber during the operation of the wafer cleaning equipment to obtain an operating parameter data set; preprocessing the operating parameter data set to obtain a preprocessed data set; using a support vector machine algorithm to construct a correlation model between the use time of the chamber liquid and the attenuation of the liquid quality to obtain model parameters; if the predicted liquid quality is lower than a preset threshold, it is determined that the chamber needs to replace the liquid; according to the chamber that needs to replace the liquid, combined with historical liquid replacement cycle data, the optimal liquid replacement cycle of each chamber is determined by a clustering algorithm.

[0027] Specifically, during the operation of the wafer cleaning equipment, the temperature, pressure, flow rate and other working parameters of each cleaning chamber are collected in real time through sensors, and a working parameter data set is formed at a frequency of once per minute. At the same time, the time of use and replacement of the corresponding liquid in each cleaning chamber are recorded to form a liquid duration data set. The working parameter data set and the liquid duration data set are cleaned, outliers and missing values ​​are removed, and key features such as temperature mean and pressure peak are extracted. Then, the maximum and minimum normalization method is used to normalize the data to the range of 0 to 1 to obtain the preprocessed data set. Using the preprocessed data set, the support vector machine algorithm of the Gaussian kernel function is selected to construct a correlation model between the use time of the chamber liquid and the quality attenuation of the liquid. The penalty coefficient C and kernel function parameter γ of the model are optimized by grid search, and the model is trained by 5-fold cross validation to obtain the optimal model parameters. During the operation of the wafer cleaning equipment, the real-time working parameters of each cleaning chamber are obtained every 10 minutes, input into the correlation model, and the quality attenuation value of the liquid in each chamber is predicted. If the predicted medicine quality attenuation value is lower than the preset threshold of 8, it is determined that the medicine in the cavity needs to be replaced. For cavities that need to replace medicine, collect their historical medicine replacement cycle data, and use the K-means clustering algorithm to divide the replacement cycle into three categories: long, medium, and short. Take the mean of each category as the optimal medicine replacement cycle for the cavity, for example, a short cycle is 7 days, a medium cycle is 14 days, and a long cycle is 30 days. Based on the optimal replacement cycle, generate a regular medicine replacement schedule for each cavity to guide staff to regularly replace medicine and reduce waste of medicine caused by premature replacement.

[0028] Step S5, obtaining the liquid circulation process parameters of each cleaning chamber and the corresponding cleaning effect detection results during the actual operation of the wafer cleaning equipment, and extracting the inherent correlation between the liquid quality, circulation parameters and cleaning effect of each chamber from the historical big data through data mining algorithms. Form a knowledge base of chamber-by-chamber optimized liquid circulation control strategy to guide the dynamic optimization of each chamber intelligent control algorithm.

[0029] Step S5 includes: obtaining the liquid circulation parameters of each cleaning chamber and the corresponding cleaning effect detection results during the operation of the wafer cleaning equipment, wherein the liquid circulation parameters include liquid concentration, temperature, flow rate and circulation time, and the cleaning effect detection results include the number of residual particles on the wafer surface and the amount of residual organic matter; preprocessing the liquid circulation parameters and the cleaning effect detection results to obtain a historical data set suitable for data mining, wherein the preprocessing includes removing abnormal data, filling missing values ​​and standardizing the data; using an association rule mining algorithm to mine the inherent association rules between the liquid quality parameters, circulation parameters and cleaning effects of each chamber from the historical data set to obtain an association rule set, wherein the association rules in the association rule set are in the form of "if the liquid concentration is within the range of A, the temperature is within the range of B, and the circulation time is C, then the wafer The number of residual particles on the surface is less than D"; the association rule set is evaluated, and the rules whose confidence and support are lower than the preset threshold are eliminated to obtain an optimized association rule set; the optimized association rule set is stored in the knowledge base to form an optimized control strategy for the liquid circulation parameters of the sub-cavities; during the operation of the wafer cleaning equipment, the intelligent control algorithm of each cleaning cavity obtains the current liquid circulation parameters and queries the optimized control strategy in the knowledge base; if a matching optimized control strategy is queried, the intelligent control algorithm dynamically adjusts the liquid circulation parameters according to the strategy; if no matching optimized control strategy is queried, the intelligent control algorithm maintains the current liquid circulation parameters unchanged, and adds the current parameters and cleaning effects to the historical data set for subsequent data mining processes to continuously optimize the control strategy of the knowledge base.

[0030] Specifically, during the operation of the wafer cleaning equipment, the liquid circulation parameters of each cleaning chamber are obtained in real time, such as the liquid concentration in the range of 20-30%, the temperature in the range of 50-60°C, the flow rate of 2-3L / min, the cycle time of 10-15min, etc. At the same time, the corresponding cleaning effect test results are obtained, such as the number of residual particles on the wafer surface is less than 10 / cm 2 , organic matter residue is less than 5ng / cm 2 The acquired historical data was preprocessed, abnormal data was eliminated using the 3σ criterion, missing values ​​were filled using the nearest neighbor interpolation method, and the data was normalized using the Z-score standardization method to form a data set suitable for the April association rule mining algorithm. By setting the minimum support to 10% and the minimum confidence to 80%, a pattern such as "If the concentration of the drug solution is within the range of 25-28%, the temperature is within the range of 55-58℃, and the cycle time is 12-13min, the number of residual particles on the wafer surface is less than 5 / cm 2The high-quality association rules of "are stored in the knowledge base to form an optimized control strategy. The intelligent control algorithm obtains the current parameters of 27% solution concentration, 56°C temperature and 11min cycle time in real time, queries the knowledge base to match the above-mentioned optimized control strategy, and dynamically adjusts the cycle time to 15min to further reduce the number of residual particles on the wafer surface. For the current parameters of 31% solution concentration, 62°C temperature and 16min cycle time, there is no matching optimized control strategy in the knowledge base. The control algorithm maintains the current parameters unchanged, and adds the parameters and the corresponding cleaning effect test results to the historical data set. In the subsequent data mining iteration process, the control strategy of the knowledge base is continuously optimized and enriched.

[0031] Step S6, pre-establish the threshold range of the liquid quality parameters of each cleaning chamber according to the cleaning process requirements of different wafer products. When the wafer cleaning equipment switches the product model, the liquid quality control threshold corresponding to each chamber is automatically matched according to the product model to realize the dynamic adjustment of the liquid circulation mode and parameters of each chamber.

[0032] Step S6 includes: obtaining product model information of wafers to be processed by the equipment, and obtaining cleaning process information corresponding to the current product model according to a pre-established product model and cleaning process database; extracting standard values ​​of quality parameters of cleaning chamber liquid medicine corresponding to the current product model from the cleaning process information, and determining threshold intervals of quality parameters of cleaning chamber liquid medicine in combination with preset quality parameter floating ranges; determining whether the current values ​​of the liquid medicine quality parameters are within the threshold intervals by real-time monitoring of the current values ​​of the liquid medicine quality parameters of each cleaning chamber, and generating a liquid medicine quality parameter adjustment signal if the current values ​​are beyond the corresponding threshold intervals; obtaining a corresponding cleaning chamber liquid medicine circulation mode adjustment instruction according to the liquid medicine quality parameter adjustment signal, and executing the liquid medicine circulation mode adjustment instruction to adjust the liquid medicine circulation mode of the cleaning chamber. The control system generates an adjustment instruction to obtain a new state of the medicine liquid circulation mode of the corresponding cleaning cavity; based on the new state of the medicine liquid circulation mode, obtains the target value of the medicine liquid circulation parameter of the corresponding cleaning cavity, and obtains the adjustment value of the medicine liquid circulation parameter by adjusting the operating state of the medicine liquid circulation system component; determines whether the deviation between the adjustment value of the medicine liquid circulation parameter and the target value of the medicine liquid circulation parameter is within a preset range, and if the deviation exceeds the preset range, generates a medicine liquid circulation parameter fine-tuning instruction, and finely adjusts the operating state of the medicine liquid circulation system component according to the fine-tuning instruction; determines whether the medicine liquid quality parameter meets the preset quality standard by collecting the medicine liquid quality parameter sensor data, and if not, returns to the step of real-time monitoring the current value of the medicine liquid quality parameter of each cleaning cavity, and executes the cycle until the medicine liquid quality parameter meets the standard.

[0033] Specifically, assuming that the wafer product model to be processed by the current equipment is A123, the system has pre-established a product model and cleaning process database, which can be accessed through the query interface. The system queries the database, inputs A123, and the database returns the corresponding cleaning process information, such as the need to use a 5% hydrofluoric acid solution for cleaning chamber 1, a 10% ammonia solution for cleaning chamber 2, and pure water for cleaning chamber 3. Extract the standard value of the quality parameter of each cleaning chamber liquid corresponding to the A123 model from the cleaning process information. For example, set the standard value of the hydrofluoric acid concentration of cleaning chamber 1 to 5%, and the standard value of the ammonia solution concentration of cleaning chamber 2 to 10%. Combined with the preset quality parameter floating range of ±2%, the threshold interval of the hydrofluoric acid concentration of cleaning chamber 1 is 8% to 2%, and the threshold interval of the ammonia solution concentration of cleaning chamber 2 is 8% to 12%. Assume that at a certain moment, the system monitors in real time through the sensor that the current value of the hydrofluoric acid concentration of the cleaning chamber 1 is 7%, and the current value of the ammonia solution concentration of the cleaning chamber 2 is 11%. Because 7% is lower than 8%, the hydrofluoric acid concentration of the cleaning chamber 1 is lower than the threshold interval, and the system generates a liquid medicine quality parameter adjustment signal. According to the liquid medicine quality parameter adjustment signal, the system queries the preset adjustment rule database. For example, the hydrofluoric acid concentration is lower than the lower threshold, corresponding to the concentration increase mode, and obtains the instruction that the cleaning chamber 1 needs to execute the concentration increase mode. By executing the instruction, the cleaning chamber 1 enters the concentration increase mode. After the cleaning chamber 1 enters the concentration increase mode, the system queries the preset liquid medicine circulation parameter database to obtain, for example, in the concentration increase mode, the opening of the hydrofluoric acid supply valve needs to be increased by 20%, and the speed of the circulation pump needs to be increased by 100 rpm, and these values ​​are used as the liquid medicine circulation parameter target values. The hydrofluoric acid supply valve and the circulation pump are adjusted by the control system, and the valve opening is actually increased by 19%, and the speed of the circulation pump is increased by 95 rpm, and the liquid medicine circulation parameter adjustment value is obtained. Calculate the deviation, the valve opening deviation is |20%-19%|=1%, the circulating pump speed deviation is |100-95|=5 rpm, assuming that the preset deviation range is that the valve opening deviation does not exceed 2%, and the circulating pump speed deviation does not exceed 10 rpm, then the deviation is within the preset range and no fine-tuning is required. At this time, the system collects liquid concentration data through the sensor. Assuming that the collected hydrofluoric acid concentration is 9%, which is still lower than the standard value of 5%, it returns to the previous step and continues to cycle. Until a certain moment, the collected hydrofluoric acid concentration is 05%, which meets the preset quality standard, then the current cycle ends and enters the cleaning process of the next product model.

[0034] Step S7, obtain the real-time operation status data of the wafer cleaning equipment, including the wafer cleaning progress and liquid medicine circulation parameters of each cleaning chamber, and monitor the operation status of each chamber in real time through the central controller. When the wafer cleaning of a chamber is completed, immediately switch to the next pre-configured idle chamber, and according to the process requirements of the wafer to be cleaned, call the corresponding liquid medicine circulation control strategy from the knowledge base, preset the liquid medicine circulation parameters of the new chamber, and realize continuous multi-chamber cleaning of different wafer products.

[0035] Step S7 includes: obtaining real-time operation status data of each cavity of the wafer cleaning equipment, the real-time operation status data including wafer cleaning progress and liquid medicine circulation parameters; transmitting the real-time operation status data to the central controller; the central controller determines whether the wafer cleaning of each cavity is completed according to the received real-time operation status data; if a cavity is cleaned, an idle cavity is selected from the pre-configured idle cavity list as the next target cavity to be switched; obtaining process requirement information of the wafer to be cleaned; retrieving a matching liquid medicine circulation control strategy from a pre-built knowledge base according to the process requirement information; if a matching liquid medicine circulation control strategy is found, the liquid medicine circulation parameters in the strategy are used as preset parameters; if no matching liquid medicine circulation control strategy is found, a new set of liquid medicine circulation parameters are generated according to the process requirement information as preset parameters; The liquid medicine circulation parameters of the target cavity are set to the preset parameters, and the liquid medicine circulation system of the cavity is started to pre-treat the internal environment of the target cavity; after the pre-treatment of the target cavity is completed, the conveying system of the wafer cleaning equipment is switched to the target cavity, the wafer to be cleaned is transferred to the target cavity, and the cleaning process is started; the central controller collects the operating status data of the target cavity in real time, and the operating status data includes cleaning progress, liquid medicine temperature, liquid medicine flow rate and liquid medicine concentration; the collected operating status data is compared with the preset parameters; if the operating status data deviates from the preset parameters beyond the allowable range, an alarm is triggered, and the liquid medicine circulation parameters of the target cavity are adjusted according to the degree of deviation; when the wafer cleaning of the target cavity is completed, the cleaned wafer is transferred out of the cavity, and the cavity is marked as idle.

[0036] Specifically, the central controller obtains the cleaning progress and liquid circulation parameters of each cavity, such as temperature, flow rate, concentration, etc., every 1 second, and transmits them through the TCP / IP protocol. When the cleaning progress of a cavity reaches 100%, the controller selects the cavity with the longest idle time from the list of 5 idle cavities as the target cavity. According to the process requirements such as the material and contaminant type of the wafer, a search is conducted in the knowledge base containing 100 strategies. If no complete match is found, the decision tree algorithm is used to generate parameters such as liquid temperature, flow rate, concentration, and time. The controller sends the parameters to the target cavity and starts the liquid circulation system to pre-treat the cavity for 2 minutes, with the temperature controlled at 50±1℃ and the flow rate controlled at 10±5m / s. After the pre-treatment is completed, the conveying system transfers the wafer into the cavity and monitors the original cavity at the same time. If the actual parameters deviate from the preset value by more than 5% during the cleaning process, an alarm is issued and the PID algorithm is used to adjust in real time. After the cleaning is completed, the wafer is transferred out and the cavity status is updated to idle. If there are wafers to be cleaned, the above steps are repeated, otherwise the cleaning process ends.

[0037] Step S8, based on the optimized intelligent control algorithm and the sub-chamber knowledge base, the liquid circulation process of multiple cleaning chambers is monitored in real time and dynamically adjusted. When the liquid quality parameters of a certain chamber deviate from the preset threshold, the optimization control algorithm of the chamber is triggered, the new liquid circulation pump flow and filter device parameters are calculated, and the adjustment is performed through the control unit of the chamber. At the same time, the liquid use time and cleaning efficiency of each chamber are comprehensively analyzed, and the scheduling strategy of multi-chamber collaborative cleaning is dynamically optimized to improve the overall efficiency of wafer cleaning and liquid utilization.

[0038] Step S8 includes: obtaining real-time liquid medicine circulation data of multiple cleaning chambers, the real-time liquid medicine circulation data including liquid medicine quality parameters, circulation pump flow and filter device parameters; transmitting the real-time liquid medicine circulation data to an intelligent control system; the intelligent control system determines whether the liquid medicine quality parameters of each chamber deviate from a preset threshold value according to a preset sub-chamber knowledge base; if the liquid medicine quality parameters deviate from the preset threshold value, triggering an optimization control algorithm for the corresponding chamber; the optimization control algorithm calculates the optimized liquid medicine circulation pump flow and filter device parameters using a machine learning method according to the degree of deviation of the liquid medicine quality parameters and historical data; transmitting the optimized liquid medicine circulation pump flow and filter device parameters to a control unit of the corresponding chamber; after receiving the optimized liquid medicine circulation pump flow and filter device parameters, the control unit dynamically adjusts the liquid medicine circulation pump flow and the working parameters of the filter device, and the intelligent control system dynamically generates an optimal scheduling strategy for multi-chamber collaborative cleaning; the optimal scheduling strategy includes reasonably allocating the cleaning tasks and liquid medicine usage time of each chamber to improve the overall efficiency of wafer cleaning and liquid medicine utilization.

[0039] Specifically, the intelligent control system obtains real-time liquid circulation data of multiple cleaning chambers every 5 seconds, including quality parameters such as liquid pH value, conductivity, temperature, and parameters such as the pressure difference and flow rate of the circulating pump flow and the filter device. After the system transmits the data to the intelligent control system, it compares it with the standard threshold of each parameter in the preset sub-chamber knowledge base. For example, the standard range of pH value is 0-0. If the actual pH value is 5, it is judged as deviation. In the case of pH value deviation, the system triggers the optimization control algorithm of the cavity. According to the deviation degree 5 and the association law between pH value and liquid pump flow and filter device parameters in historical data, the support vector machine algorithm is used to calculate the optimal parameter combination that can make the pH value return to 8 within 5 minutes by adjusting the liquid pump flow from 2L / min to 5L / min and shortening the backwash cycle of the filter device from 60s / time to 45s / time, and transmit the optimized parameters to the control unit of the corresponding cavity. After receiving the optimized parameters, the control unit immediately adjusts the liquid pump flow to 5L / min and the backwash cycle of the filter device to 45s / time. At the same time, the intelligent control system continuously monitors the duration of the liquid medicine usage and cleaning efficiency data of the cavity. Through data mining algorithm analysis, it is found that after the liquid medicine usage time exceeds 2 hours, the cleaning efficiency begins to decline, and the decline rate is 10% / h. According to the analysis results, the intelligent control system dynamically generates a collaborative cleaning scheduling strategy, shortening the cleaning task of the cavity by 5 hours, controlling the liquid medicine usage time within 5 hours, and scheduling adjacent idle cavities to share the remaining cleaning tasks. After analysis, the overall cleaning time can be reduced by 5%, and the liquid medicine usage can be saved by 10%. After the scheduling strategy is sent to each cavity control unit, the working hours and liquid medicine usage of multiple cavities are coordinated to maximize the cleaning efficiency and liquid medicine utilization while ensuring the cleaning effect.

[0040] The above is only a preferred embodiment of the present invention. It should be pointed out that ordinary technicians in this technical field can make several improvements and supplements without departing from the principle of the present invention. These improvements and supplements should also be regarded as the scope of protection of the present invention.

Claims

1. A multi-chamber liquid intelligent circulation control method for wafer cleaning equipment, characterized in that: The method comprises the following steps: Step S1, obtaining real-time data of the circulation process of the medicine solution in multiple cleaning cavities, specifically including the parameters of the concentration, temperature and pH of the medicine solution in each cavity, using the concentration sensor, temperature sensor and pH sensor installed at the key position of the circulation pipeline of each cavity to collect the data, and recording the use time of the medicine solution in each cavity; Step S2, inputting the acquired real-time data of the quality parameters of the drug liquid in each cavity into the intelligent control algorithm model, and judging whether the current drug liquid quality of each cavity meets the preset wafer cleaning process requirements through the algorithm model, if it meets the requirements, continuing to acquire real-time data, if it does not meet the requirements, then according to the deviation degree between the drug liquid quality parameters and the preset process requirements, the intelligent control algorithm is used to calculate the optimized flow rate of the drug liquid circulation pump of the cavity and the working state parameters of the filter device; Step S3, sending the calculated optimization parameters to the control unit of the corresponding cleaning chamber, adjusting the actual working flow of the circulation pump of the chamber and the actual working state of the filter device, dynamically optimizing the circulation process of the liquid medicine in the chamber, and restoring the quality of the liquid medicine to within the range of process requirements; Step S4, obtaining the working parameters of each cleaning chamber and the corresponding liquid use time data during the actual operation of the wafer cleaning equipment, establishing a correlation model between the liquid use time of each chamber and the attenuation of the liquid quality through a machine learning algorithm, dynamically predicting the optimal liquid replacement cycle of each chamber, guiding the regular replacement of the liquid, and reducing liquid waste; Step S5, obtaining the liquid circulation process parameters of each cleaning chamber and the corresponding cleaning effect detection results during the actual operation of the wafer cleaning equipment, extracting the intrinsic correlation between the liquid quality, circulation parameters and cleaning effect of each chamber from the historical big data through a data mining algorithm, forming a knowledge base of optimized liquid circulation control strategies for each chamber, and guiding the dynamic optimization of the intelligent control algorithm of each chamber; Step S6: Pre-establish the threshold range of the liquid quality parameter of each cleaning chamber according to the cleaning process requirements of different wafer products. When the wafer cleaning equipment switches the product model, the liquid quality control threshold corresponding to each chamber is automatically matched according to the product model, so as to realize the dynamic adjustment of the liquid circulation mode and parameters of each chamber. Step S7, obtaining real-time operation status data of the wafer cleaning equipment, specifically including the wafer cleaning progress and liquid medicine circulation parameters of each cleaning chamber, and monitoring the operation status of each chamber in real time through the central controller. When the wafer cleaning of a chamber is completed, it is immediately switched to the next pre-configured idle chamber, and according to the process requirements of the wafer to be cleaned, the corresponding liquid medicine circulation control strategy is retrieved from the knowledge base, and the liquid medicine circulation parameters of the new chamber are preset to achieve continuous multi-chamber cleaning of different wafer products; Step S8, based on the optimized intelligent control algorithm and the chamber knowledge base, the liquid circulation process of multiple cleaning chambers is monitored and dynamically adjusted in real time. When the liquid quality parameters of a certain chamber deviate from the preset threshold, the optimization control algorithm of the chamber is triggered, and the new liquid circulation pump flow and filter device parameters are calculated. The adjustment is performed through the control unit of the chamber. At the same time, the liquid usage time and cleaning efficiency of each chamber are comprehensively analyzed, and the scheduling strategy of multi-chamber collaborative cleaning is dynamically optimized to improve the overall efficiency of wafer cleaning and liquid utilization.

2. According to claim 1, a multi-chamber liquid intelligent circulation control method for wafer cleaning equipment is characterized in that: The step S1 comprises: Obtain real-time data collected from concentration sensors, temperature sensors and pH sensors installed at key locations of the circulation pipelines of each cleaning chamber; According to the data collected by the concentration sensor, it is judged whether the real-time concentration of the drug solution in each cleaning cavity meets the preset range. If it is lower than the preset range, the drug adding device is controlled to add drug solution into the cleaning cavity until the concentration reaches the preset range; According to the data collected by the temperature sensor, it is judged whether the real-time temperature of the liquid medicine in each cleaning cavity meets the preset range. If it is lower than the preset range, the heating device is controlled to heat the liquid medicine in the cleaning cavity until the temperature reaches the preset range; According to the data collected by the pH sensor, it is judged whether the real-time pH value of the liquid in each cleaning cavity meets the preset range. If it exceeds the preset range, the acid or alkali adding device is controlled to add acid or alkali into the cleaning cavity until the pH value reaches the preset range; Summarizing the real-time concentration, real-time temperature and real-time pH data of the liquid medicine in each cleaning cavity to obtain comprehensive quality parameters of the liquid medicine in each cleaning cavity; Record the start time and end time of use of each cleaning cavity liquid, and calculate the use time of each cleaning cavity liquid; Perform correlation analysis on the comprehensive quality parameters of the cleaning cavity liquid and the usage time, and establish a liquid quality prediction model using a support vector machine algorithm; The optimal replacement time of the medicine liquid in each cleaning cavity is predicted according to the medicine liquid quality prediction model.

3. According to claim 1, a multi-chamber liquid intelligent circulation control method for wafer cleaning equipment is characterized in that: The step S2 comprises: Acquire real-time data of the quality parameters of the drug solution in each cavity, wherein the quality parameters of the drug solution include pH value, temperature and concentration; Using the drug solution quality parameters as input to an intelligent control algorithm model; Obtain pre-established wafer cleaning process quality requirements; Determining whether the quality parameters of the chemical solution meet the quality requirements of the wafer cleaning process; If the quality parameters of the chemical solution meet the quality requirements of the wafer cleaning process, returning to continue to obtain real-time data of the quality parameters of the chemical solution; If the quality parameters of the chemical solution do not meet the quality requirements of the wafer cleaning process, then determining the degree of deviation between the quality parameters of the chemical solution and the quality requirements of the wafer cleaning process, and determining a target value for optimization adjustment; A machine learning algorithm is used to establish a mapping relationship model between the quality parameters of the liquid medicine and the flow rate of the liquid medicine circulation pump and the working state parameters of the filtering device; Inputting the degree of deviation of the drug liquid quality parameter into the mapping relationship model to obtain optimized drug liquid circulation pump flow rate and filter device working state parameters; According to the optimized flow rate of the liquid medicine circulation pump, controlling the liquid medicine circulation pump to adjust the flow rate; According to the optimized working state parameters of the filter device, adjusting the backwashing frequency and backwashing time of the filter device; After executing the optimization control, the system returns to continue acquiring the real-time data of the quality parameters of the liquid medicine, and performs cyclic judgment and optimization control.

4. A multi-chamber liquid intelligent circulation control method for wafer cleaning equipment according to any one of claims 1 to 3, characterized in that: The step S5 comprises: Obtaining the liquid circulation parameters of each cleaning chamber and the corresponding cleaning effect detection results during the operation of the wafer cleaning equipment, wherein the liquid circulation parameters include liquid concentration, temperature, flow rate and circulation time, and the cleaning effect detection results include the number of residual particles and organic matter residues on the wafer surface; Preprocessing the liquid circulation parameters and cleaning effect test results to obtain a historical data set suitable for data mining, wherein the preprocessing includes removing abnormal data, filling missing values, and standardizing the data; Using an association rule mining algorithm to mine the inherent association rules between the quality parameters and circulation parameters of each cavity liquid and the cleaning effect from the historical data set, to obtain an association rule set; Evaluating the association rule set, eliminating rules whose confidence and support are lower than preset thresholds, and obtaining an optimized association rule set; The optimized association rule set is stored in a knowledge base to form an optimized control strategy for the circulation parameters of the liquid medicine in the sub-cavities; During the operation of the wafer cleaning equipment, the intelligent control algorithm of each cleaning chamber obtains the current liquid circulation parameters and queries the optimization control strategy in the knowledge base; If a matching optimization control strategy is found, the intelligent control algorithm dynamically adjusts the liquid medicine circulation parameters according to the strategy; If no matching optimization control strategy is found, the intelligent control algorithm maintains the current liquid circulation parameters unchanged, and adds the current parameters and cleaning effects to the historical data set for subsequent data mining processes to continuously optimize the control strategy of the knowledge base.

5. A multi-chamber liquid intelligent circulation control method for wafer cleaning equipment according to any one of claims 1 to 3, characterized in that: The step S6 comprises: Obtain the product model information of the wafer to be processed by the equipment, and obtain the cleaning process information corresponding to the current product model based on the pre-established product model and cleaning process database; Extracting the standard value of each cleaning chamber liquid quality parameter corresponding to the current product model from the cleaning process information, and determining the threshold interval of each cleaning chamber liquid quality parameter in combination with the preset quality parameter floating range; By real-time monitoring the current value of the quality parameter of the drug solution in each cleaning cavity, it is determined whether the current value of the drug solution quality parameter is within the threshold range, and if it exceeds the corresponding threshold range, a drug solution quality parameter adjustment signal is generated; According to the liquid medicine quality parameter adjustment signal, a corresponding cleaning cavity liquid medicine circulation mode adjustment instruction is obtained, and by executing the liquid medicine circulation mode adjustment instruction, a new state of the corresponding cleaning cavity liquid medicine circulation mode is obtained; Based on the new state of the liquid medicine circulation mode, a target value of a liquid medicine circulation parameter corresponding to the cleaning chamber is obtained, and an adjustment value of the liquid medicine circulation parameter is obtained by adjusting the operating state of the liquid medicine circulation system component; Determine whether the deviation between the adjustment value of the liquid medicine circulation parameter and the target value of the liquid medicine circulation parameter is within a preset range, and if the deviation exceeds the preset range, generate a liquid medicine circulation parameter fine-tuning instruction, and fine-tune the operating state of the liquid medicine circulation system component according to the fine-tuning instruction; By collecting the liquid medicine quality parameter sensor data, it is determined whether the liquid medicine quality parameter reaches the preset quality standard. If not, the process returns to the step of real-time monitoring the current value of the liquid medicine quality parameter of each cleaning cavity, and the process is executed in a loop until the liquid medicine quality parameter reaches the standard.

6. A multi-chamber liquid intelligent circulation control method for wafer cleaning equipment according to any one of claims 1 to 3, characterized in that: The step S7 comprises: Acquire real-time operation status data of each cavity of the wafer cleaning equipment, wherein the real-time operation status data includes wafer cleaning progress and liquid circulation parameters; Transmitting the real-time operating status data to a central controller; The central controller determines whether the wafer cleaning of each chamber is completed according to the received real-time operation status data; If a cavity is cleaned, an idle cavity is selected from the pre-configured idle cavity list as the next target cavity to be switched; Obtain process requirement information of the wafer to be cleaned; Retrieving a matching liquid medicine circulation control strategy from a pre-built knowledge base according to the process requirement information; If a matching drug liquid circulation control strategy is found, the drug liquid circulation parameters in the strategy are used as preset parameters; If no matching liquid medicine circulation control strategy is found, a new set of liquid medicine circulation parameters is generated as preset parameters according to the process requirement information; Setting the liquid medicine circulation parameters of the target cavity to the preset parameters, and starting the liquid medicine circulation system of the cavity to pre-treat the internal environment of the target cavity; After the pretreatment of the target chamber is completed, the conveying system of the wafer cleaning equipment is switched to the target chamber, the wafer to be cleaned is transferred to the target chamber, and the cleaning process is started; The central controller collects the operation status data of the target cavity in real time, wherein the operation status data includes cleaning progress, liquid medicine temperature, liquid medicine flow rate and liquid medicine concentration; Comparing the collected operating status data with the preset parameters; If the running status data deviates from the preset parameters beyond the allowable range, an alarm is triggered, and the liquid circulation parameters of the target cavity are adjusted according to the degree of deviation; When the wafers in the target cavity are cleaned, the cleaned wafers are transferred out of the cavity, and the cavity is marked as an idle state.

7. A multi-chamber liquid intelligent circulation control method for wafer cleaning equipment according to any one of claims 1 to 3, characterized in that: The step S8 comprises: Acquire real-time liquid circulation data of multiple cleaning chambers, wherein the real-time liquid circulation data includes liquid quality parameters, circulation pump flow rate, and filter device parameters; Transmitting the real-time drug liquid circulation data to an intelligent control system; The intelligent control system determines whether the quality parameter of the drug solution in each cavity deviates from a preset threshold value according to a preset cavity knowledge base; If the liquid medicine quality parameter deviates from the preset threshold, the optimization control algorithm of the corresponding cavity is triggered; The optimization control algorithm calculates the optimized flow rate of the liquid medicine circulation pump and the parameters of the filter device according to the deviation degree of the liquid medicine quality parameters and historical data by using a machine learning method; Transmitting the optimized liquid medicine circulation pump flow rate and filter device parameters to the control unit of the corresponding cavity; After receiving the optimized flow rate of the liquid medicine circulation pump and the parameters of the filter device, the control unit dynamically adjusts the flow rate of the liquid medicine circulation pump and the working parameters of the filter device, and the intelligent control system dynamically generates an optimal scheduling strategy for multi-cavity collaborative cleaning; The optimal scheduling strategy includes reasonably allocating the cleaning tasks and the use time of the chemical solution of each cavity, so as to improve the overall efficiency of wafer cleaning and the utilization rate of the chemical solution.