Wafer yield control method, device, equipment and medium based on factory data

Abstract

The present application provides a wafer yield control method, device, equipment and medium based on factory data. The method includes: obtaining basic factory data of the wafer currently being manufactured; wherein the basic factory data includes at least one of water parameters, electricity parameters and gasification parameters in the wafer manufacturing process; determining key factory data associated with wafer yield from the basic factory data based on a pre-built machine learning model; the key factory data is input into the machine learning model to determine whether the estimated yield of the wafer currently being manufactured reaches a preset threshold; if the preset threshold is not reached, an instruction to adjust the key factory data is generated. This technical solution controls factory data, improves wafer yield, and avoids the problem of low yield in wafer manufacturing and the inability to analyze the causes that affect wafer yield.

Description

Technical Field

[0001] The present application relates to the field of semiconductor manufacturing technology, and in particular to a wafer yield control method, device, equipment and medium based on factory data. Background Art

[0002] In recent years, with the rapid development of science and technology, the semiconductor manufacturing industry has also developed rapidly. As the basic material for making silicon semiconductor chips, the manufacturing technology of wafers has become one of the foundations and core competitiveness that restrict the development of semiconductors.

[0003] The wafer manufacturing process is very complicated, including thin film deposition, photoresist coating, photolithography development, etching, measurement, cleaning, ion implantation and many other processes. Wafer manufacturers provide a good production environment and water, electricity and gasification supply for wafer manufacturing. It mainly includes clean room system, chilled water system, process cooling water system, air compression system, vacuum system, bulk gas system, special gas system, hot water system, pure water system, wastewater system, waste gas system, chemical system and power monitoring system, etc.

[0004] However, there is no analysis of the impact of factory data on yield. For example, different gas parameters of a certain machine production equipment may lead to different wafer yields. Therefore, it is necessary to find an intelligent method to mine the attribution of factory data to yield, which is a technical problem that technicians in this field need to solve urgently. Summary of the invention

[0005] The purpose of this application is to provide a wafer yield control method, device, equipment and medium based on factory data, at least to solve the problem of analyzing whether factory data has an impact on wafer yield. The purpose of this application is to provide a new wafer yield control method based on factory data. This method uses a pre-trained machine learning model to mine key factory data related to wafer yield from conventional factory data, and predict whether the yield of the currently manufactured wafers meets the standards based on the key factory data, so as to achieve control of factory data, improve wafer yield, and avoid the problem of low yield in wafer manufacturing and inability to analyze the causes that affect wafer yield.

[0006] To achieve the above objectives, some embodiments of the present application provide the following aspects:

[0007] In a first aspect, some embodiments of the present application further provide a wafer yield control method based on factory data, the method comprising:

[0008] Obtaining basic factory data of the wafer currently being manufactured; wherein the basic factory data includes at least one of water class parameters, electrical class parameters, and gasification class parameters in the wafer manufacturing process;

[0009] Determine key fab data associated with wafer yield from basic fab data based on pre-built machine learning models;

[0010] Inputting the key factory data into the machine learning model to determine whether the estimated yield of the wafers currently being manufactured reaches a preset threshold;

[0011] If the preset threshold is not reached, an instruction to adjust the key factory affairs data is generated.

[0012] In a second aspect, some embodiments of the present application further provide a wafer yield control device based on factory service data, the device comprising:

[0013] A basic factory data acquisition module is used to acquire basic factory data of the wafer currently being manufactured; wherein the basic factory data includes at least one of water class parameters, electrical class parameters and gasification class parameters in the wafer manufacturing process;

[0014] A key factory service data determination module, used to determine key factory service data associated with wafer yield from basic factory service data based on a pre-built machine learning model;

[0015] A wafer yield prediction module, used to input the key factory data into the machine learning model to determine whether the estimated yield of the wafer currently being manufactured reaches a preset threshold;

[0016] The factory affairs data adjustment module is used to generate an instruction to adjust the key factory affairs data if the preset threshold is not reached.

[0017] In a third aspect, some embodiments of the present application further provide a computer device, the device comprising:

[0018] one or more processors; and

[0019] A memory storing computer program instructions, wherein when the computer program instructions are executed, the processor executes the wafer yield control method based on fab data as described above.

[0020] In a fourth aspect, some embodiments of the present application further provide a computer-readable medium having computer program instructions stored thereon, wherein the computer program instructions can be executed by a processor to implement the wafer yield control method based on factory data as described above.

[0021] Compared with the prior art, the solution provided in the embodiment of the present application obtains the basic factory data of the wafers currently being manufactured; wherein the basic factory data includes at least one of the water class parameters, electrical class parameters, and gasification class parameters in the wafer manufacturing process; based on a pre-built machine learning model, key factory data associated with the wafer yield is determined from the basic factory data; the key factory data is input into the machine learning model to determine whether the estimated yield of the wafers currently being manufactured reaches a preset threshold; if the preset threshold is not reached, an instruction to adjust the key factory data is generated. This solution improves the wafer yield by controlling the factory data, avoiding the problem of low yield in wafer manufacturing and the inability to analyze the causes that affect the wafer yield. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a flow chart of a wafer yield control method based on factory data provided in the first embodiment of the present application;

[0023] Figure 2 It is a flowchart of the training process of the machine learning model provided in Example 2 of the present application;

[0024] Figure 3 It is a flow chart of a wafer yield control method based on factory data provided in the third embodiment of the present application;

[0025] Figure 4 is a structural schematic diagram of a wafer yield control device based on factory data provided in the fourth embodiment of the present application;

[0026] Figure 5 It is a structural diagram of the computer device provided in Example 5 of the present application. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0028] The technical solution provided by the embodiment of the present application is described in detail below through specific embodiments and their application scenarios in conjunction with the accompanying drawings.

[0029] Embodiment 1

[0030] Figure 1 1 is a flow chart of a wafer yield control method based on factory data provided in the first embodiment of the present application. Figure 1As shown, the specific steps include: The process includes:

[0031] Step S101, obtaining basic factory data of the wafer currently being manufactured; wherein the basic factory data includes at least one of water class parameters, electrical class parameters and gasification class parameters in the wafer manufacturing process.

[0032] Wafer refers to the silicon wafer used to make silicon semiconductor circuits. Its raw material is silicon. High-purity polysilicon is dissolved and doped with silicon crystal seeds, and then slowly pulled out to form cylindrical single crystal silicon. After grinding, polishing and slicing, the silicon crystal rod is formed into a silicon wafer, that is, a wafer. Currently, the wafer production lines on the market are mainly 8 inches and 12 inches. The main processing methods for wafers are sheet processing and batch processing, that is, processing one or more wafers at the same time.

[0033] During the wafer manufacturing process, various basic factory data are involved, and each basic factory data is regulated to ensure the quality of each manufacturing process. Good machine parameter data is conducive to controlling the production of excellent wafer products.

[0034] In this solution, basic factory data includes at least one of water class parameters, electricity class parameters and gasification class parameters in the wafer manufacturing process. Among them, water class parameters may include chemical agents, waste chemicals, sludge outsourcing treatment results, water quality testing, water balance, chemical consumption analysis, water consumption analysis, water quality index analysis, water system electricity and gas energy consumption analysis, etc. Electricity class parameters may include power consumption of various systems or equipment, such as working voltage, working current, power, power consumption per unit cycle and other data. Gasification class parameters may include gas balance monitoring and analysis, gas cylinder replacement monitoring and maintenance, gas consumption monitoring and statistics, gas detector replacement and maintenance, chemical acid barrel replacement and maintenance analysis, chemical consumption statistical analysis, chemical filter replacement and maintenance, gasification system electricity consumption statistical analysis, and gasification system water consumption statistics and analysis.

[0035] Step S102, determining key factory data associated with wafer yield from basic factory data based on a pre-built machine learning model.

[0036] In this solution, a pre-trained machine learning model is used to screen basic factory data to obtain key factory data.

[0037] Understandably, machine learning models can better mine the impact of various basic factory data on product yield. For example, different gas parameters of a certain machine production equipment may lead to different wafer yields, so it is necessary to find an intelligent method to mine the attribution of factory data to yield.

[0038] Among them, machine learning simulation includes but is not limited to regression tree models. Machine learning simulation can be a model obtained by learning various basic factory data and wafer yield related data collected historically. The model can obtain the current wafer yield after inputting the basic factory data of the current wafer manufacturing. In this solution, a part of the algorithm in the machine learning model can be used to filter out key factory data from the basic factory data. Among them, key factory data is the basic factory data that has a direct impact on the wafer yield.

[0039] In this solution, the LightGBM model can be used. LightGBM (Light Gradient Boosting Machine) is Microsoft's open source distributed high-performance Gradient Boosting framework that uses a decision tree learning algorithm. The following describes the optimization of this framework:

[0040] Speed ​​and memory optimization:

[0041] Many boosting tools use a pre-sorted algorithm (approximate histogram algorithm) (such as the default algorithm in XGBoost) for decision tree learning. This is a simpler solution, but not easy to optimize. LightGBM uses a histogram-based algorithm that stores continuous feature values ​​into discrete bins. This can speed up training and reduce memory usage.

[0042] Optimizing for sparse features:

[0043] For sparse features, only O(2*number of non-zero samples) time complexity is required to construct the histogram.

[0044] Optimize the tree growing strategy to improve accuracy.

[0045] Optimization of network communication:

[0046] In the parallel learning of LightGBM, it only needs to use some aggregate communication algorithms, such as "All reduce", "All gather" and "Reduce scatter". LightGBM implements the most advanced state-of-art algorithms. These aggregate communication algorithms can provide better performance than point-to-point communication.

[0047] Optimization in parallel learning:

[0048] LightGBM provides optimizations for the following parallel learning algorithms: feature parallelism, data parallelism, and voting parallelism.

[0049] Step S103, inputting the key factory data into the machine learning model to determine whether the estimated yield of the currently manufactured wafers reaches a preset threshold.

[0050] Among them, after determining the key factory data, the values ​​of the key factory data of the currently manufactured wafers can be input into the machine learning model to determine the estimated yield of the currently manufactured wafers based on the output results of the machine learning model. It can be understood that the estimated wafer yield here is based on the key factory data currently input. On the one hand, in the process of wafer production, an assembly line operation is adopted. Generally, other data will not change in the continuous manufacturing process of multiple wafers, so the estimation based on key factory data will have a very high accuracy.

[0051] In this solution, the preset threshold value may be a value set according to demand, for example, the yield rate needs to be above 95% so that the manufacturing of the current batch of wafers can meet the production demand. Then the preset threshold value here may be set to 95%.

[0052] After obtaining the estimated yield of the currently manufactured wafer, it can be identified whether the obtained estimated yield reaches or exceeds 95%. For example, if the estimated yield reaches 97% or exceeds 95%, it can be determined that the preset threshold is reached.

[0053] Step S104: if the preset threshold is not reached, an instruction to adjust the key factory affairs data is generated.

[0054] Here, if the estimated yield of the currently manufactured wafers output by the machine learning model does not reach the preset threshold, for example, the estimated yield of the currently manufactured wafers is 92%, which is less than the preset threshold of 95%, then an instruction to adjust the key factory data can be generated.

[0055] For example, if the estimated wafer yield obtained based on the key factory data of the current wafer manufacturing is lower than the preset threshold, the current key factory data can be adjusted. The adjustment instruction here can be for one of the key factory data or for multiple key factory data, and can be adjusted from the original value to the target value. For example, the nitrogen flow rate is increased from the original flow value to twice the original value.

[0056] This technical solution provides a new wafer yield control method based on factory data. This method uses a pre-trained machine learning model to mine key factory data related to wafer yield from conventional factory data, and predicts whether the yield of the currently manufactured wafers meets the standard based on the key factory data, so as to achieve control of factory data, improve wafer yield, and avoid the problem of low yield in wafer manufacturing and inability to analyze the causes that affect wafer yield.

[0057] In another embodiment, after generating an instruction to adjust the key plant service data if the preset threshold is not reached, the method further includes:

[0058] After adjusting the key factory affairs data, inputting the adjusted updated factory affairs data into the machine learning model;

[0059] Identify whether the predicted yield of the output of the machine learning model reaches a preset threshold;

[0060] If the predicted yield does not reach a preset threshold, an alarm message is generated.

[0061] It is understandable that after the adjustment instruction is generated, the key factory data can be adjusted to obtain updated factory data. The updated factory data is input into the machine learning model again. At this time, if the predicted yield still does not reach the preset threshold, an alarm message is generated. Generally, the yield of the current wafer manufacturing may have been affected by other factors, and no matter how the key factory data is adjusted, it cannot reach the preset threshold.

[0062] This arrangement of the technical solution can improve the yield of wafer manufacturing. And in the process of improving the yield, if it is found that the preset threshold is still not met, an alarm message can be generated to ensure the yield problem of the current wafer manufacturing through manual intervention. In addition, the user can also be informed of which key factory data adjustments have been made based on the alarm, so that the staff can check whether the above adjustments are reasonable. For example, if the staff cancels the adjustment of a key factory data, it means that there is a certain irrationality in the adjustment of the key factory data, and the machine learning model needs to be updated based on this to ensure the accuracy of the output results of the machine learning model.

[0063] In another embodiment, after inputting the key factory service data into the machine learning model to determine whether the estimated yield of the wafer currently being manufactured reaches a preset threshold, the method further includes:

[0064] If the preset threshold is reached, the wafers are manufactured according to the current key factory data, and the yield of the currently manufactured wafers is counted as updated sample data;

[0065] The machine learning model is iteratively updated according to the updated sample data.

[0066] In one case, if the estimated yield of the currently manufactured wafers reaches the preset threshold, the wafer manufacturing can continue according to the current factory data, and the yield of this batch of wafers can be calculated. The yield of this batch of wafers and the current key factory data are used as updated sample data, and the machine learning model is further iterated and updated, which can make the output results of the machine learning model more accurate.

[0067] Through such a setting, this solution can gradually iterate and update the machine learning model over time. Not only does the sample data used not need to be manually produced and has higher accuracy, but it can also better reflect the wafer manufacturing status of the current equipment. After iteration, the estimated results output by the machine learning model will be more accurate and in line with actual production needs.

[0068] In another embodiment, the machine learning model is a LightGBM regression prediction model.

[0069] Among them, LightGBM (LGB) can be called an optimized and enhanced version of xgboost. Through the optimization of tree growth strategies, histogram algorithms, GOSS, parallel computing, etc., it has gradually become a magic weapon in various machine learning modeling competitions. In actual engineering projects, its processing of categorical features and single sample weighting is very suitable for some financial data modeling scenarios.

[0070] The data set has undergone data cleaning and preprocessing in the early stage, including missing value processing, outlier processing, standardization processing, correlation analysis and other steps.

[0071] Setting of LGB core hyperparameters:

[0072] task, specify the task, optional train, predict, etc.

[0073] boosting_type, boosting method, optional gbdt, rf, etc.

[0074] objective, target (function), if it is a regression task, it can be l2, l1, huber, quantile, etc.; if it is a classification task, it can be binary, multiclass, etc.

[0075] max_depth, the maximum depth of the tree, is an effective means of controlling overfitting.

[0076] num_leaves, the maximum number of leaf nodes in the tree.

[0077] feature_fraction, the random sampling rate of the feature, refers to

[0078] bagging_fraction, random sampling rate of samples

[0079] bagging_freq, whether to enable bagging and set the iteration rounds. If enabled, the random sampling of the above features and samples needs to be set.

[0080] learning_rate, learning rate;

[0081] lambda_l1, L1 regularization;

[0082] lambda_l2, L2 regularization;

[0083] Generally, regression models use error indicators such as MSE, MAE, MAPE, and MedianAE to evaluate model effects.

[0084] The objective function and the metric function are crucial in the GBM model. The objective function affects the learning of the model, while the early_stopping in the callback is controlled by the metric function. Choosing a suitable combination of the objective function and the metric function based on the distribution of the data and the final goal has a great impact on the model effect.

[0085] Other important core hyperparameters are adjusted through the grid search method to select the optimal parameter combination.

[0086] By using the LightGBM regression prediction model, this solution can more accurately predict the wafer yield based on factory data and improve the control accuracy of factory data during the wafer manufacturing process.

[0087] Embodiment 2

[0088] Figure 2 Schematic diagram of the training process of the machine learning model provided in Example 2 of the present application. Figure 2 As shown, the specific steps include:

[0089] Step S201, obtaining historical factory data of historically manufactured wafers and historical yield of wafer manufacturing.

[0090] Among them, the historical factory data can be the factory data collected in the past wafer manufacturing process. The historical yield of wafer manufacturing can be the yield data of each batch of wafers in the historical manufacturing process. It can be known that the yield data can be the ratio of the number of qualified wafers to the number of all wafers produced in the current batch.

[0091] Step S202, performing data preprocessing and feature engineering processing on the historical factory data to obtain sample data; and using the historical yield as a label for the sample data.

[0092] Among them, data preprocessing is used to improve the quality of sample data and make the sample data more suitable for analysis and model training. Common data preprocessing methods include data cleaning, data conversion, data integration, data reduction, etc. For example, duplicate data, abnormal data, and missing data in sample data can be processed to clean the sample data.

[0093] Feature engineering is used to extract features from sample data through data mining technology. Common feature engineering processing methods include standardization, interval scaling, normalization, quantitative feature binarization, qualitative feature dummy coding, missing value processing, and data conversion, etc.

[0094] The advantage of this arrangement of the technical solution is that it improves the standardization of sample data, accelerates the convergence speed of model training, improves training efficiency, and makes the training effect of the model faster and more accurate.

[0095] Step S203: input the labeled sample data into the initial model, train the initial model, and obtain a machine learning model.

[0096] The training method can be supervised training or unsupervised training.

[0097] Specifically, the labeled sample data can be arranged in the order of time nodes and input into the initial model, the initial model can be trained, and the parameters of the initial model can be adjusted in each training round until the training is completed to obtain the machine learning model.

[0098] During the training process, the training end condition may be that the prediction accuracy of the machine learning model is high enough, for example, the prediction accuracy exceeds 80%. The training end condition may also be that the number of training times reaches a preset number of times. For example, the preset number of times is 20,000 times. Those skilled in the art may set a preset threshold or preset number of times according to actual conditions so that the machine learning model will not overfit while learning the factors affecting the wafer yield of the factory data.

[0099] The advantage of this technical solution is that by providing a training process for the machine learning model, an accurate machine learning model can be obtained. The control of the number and authenticity of samples can improve the training speed and accuracy of the machine learning model.

[0100] Embodiment 3

[0101] Figure 3: is a flow chart of a wafer yield control method based on factory data provided in the third embodiment of the present application. Figure 3 As shown, the specific steps include:

[0102] Step S301, obtaining basic factory data of the wafer currently being manufactured; wherein the basic factory data includes at least one of water class parameters, electrical class parameters, and gasification class parameters in the wafer manufacturing process;

[0103] Step S302, determining the importance index of each basic factory data and the wafer yield based on the first algorithm included in the machine learning model;

[0104] In this solution, the machine learning model may include a portion of an algorithm for filtering out importance indicators related to wafer yield from basic factory data. It can be understood that the importance indicators can be used to characterize the importance of the impact of the factory data on the wafer yield.

[0105] Step S303, determining a correlation index between each basic factory service data and the wafer yield based on a second algorithm included in the machine learning model;

[0106] In this solution, the machine learning model may include another part of the algorithm for filtering out correlation indicators related to wafer yield from basic factory data. It can be understood that the correlation indicator can be used to characterize the correlation of the impact of the factory data on the wafer yield. For example, an increase or decrease in the value will have a positive or negative correlation with the wafer yield.

[0107] Step S304: determining key plant affairs data from basic plant affairs data according to the importance index and / or the relevance index.

[0108] In this solution, whether the basic factory data is key factory data can be determined based on the importance index and / or the correlation index. For example, if the importance index is high enough, it can be determined as key factory data, or if the correlation index shows a positive correlation or negative correlation, it can be determined as key factory data, or if both the importance index is high enough and the correlation index shows a positive correlation, it can be determined as key factory data.

[0109] Step S305, inputting the key factory data into the machine learning model to determine whether the estimated yield of the wafer currently manufactured reaches a preset threshold;

[0110] Step S306: if the preset threshold is not reached, an instruction to adjust the key factory affairs data is generated.

[0111] Based on the above technical solutions, this technical solution provides a specific method of screening out key factory data from basic factory data based on a machine learning model. Through such a setting, interference with factory data that has little or no impact on wafer yield can be avoided. Through key factory data, accurate data can be better provided for wafer yield estimation and factory data adjustment.

[0112] In one embodiment, determining key plant affairs data from basic plant affairs data according to the importance index and / or the relevance index includes:

[0113] If the importance index of the current basic factory affairs data exceeds the importance threshold, the current basic factory affairs data is determined to be key factory affairs data;

[0114] or,

[0115] If the correlation index of the current basic factory affairs data exceeds the positive correlation threshold, the current basic factory affairs data is determined to be key factory affairs data, and is positively correlated; if the correlation index of the current basic factory affairs data is less than the negative correlation threshold, the current basic factory affairs data is determined to be key factory affairs data, and is negatively correlated;

[0116] or,

[0117] If the importance index of the current basic factory affairs data exceeds the importance threshold, and the relevance index of the current basic factory affairs data exceeds the positive relevance threshold, the current basic factory affairs data is determined to be the key factory affairs data; or, if the importance index of the current basic factory affairs data exceeds the importance threshold, and the relevance index of the current basic factory affairs data is less than the negative relevance threshold, the current basic factory affairs data is determined to be the key factory affairs data.

[0118] The correlation may include being greater than a positive correlation threshold or being less than a negative correlation threshold, both of which may reflect the degree of correlation between the current factory data and the wafer yield.

[0119] Here, the importance value range can be 0 to 1, for example, the importance threshold can be 0.8, and the correlation can be positive or negative, and the value range can be -1 to 1, where -1 to 0 is negative correlation and 0 to 1 is positive correlation. The two can be combined to find key factory data. Two indicators can be output based on one model. For example, the correlation indicator can generally be output with a small algorithm in the data processing and feature engineering stage, and the importance indicator is generally output in the model training stage.

[0120] This scheme is set up so that the impact of factory data on wafer yield can be better reflected, and key factory data can be determined more accurately, providing a precise control range for subsequent wafer yield prediction and factory data adjustment.

[0121] This scheme also provides a preferred implementation scheme, which is as follows:

[0122] The first step is to collect data:

[0123] Feature data:

[0124] Contains water, electricity, gas and other data of each plant equipment, and its attribute data is the respective parameter values, etc. Historical target data: can be defective product count or yield data.

[0125] Examples of water courses: chemicals / waste chemicals, sludge outsourcing, water quality testing, water balance, chemical consumption analysis, water consumption analysis, water quality index analysis, water system electricity and gas consumption analysis, etc.

[0126] Examples of gasification courses: gas balance monitoring and analysis, gas cylinder replacement monitoring and maintenance, gas consumption monitoring and statistics, gas detector replacement and maintenance, chemical acid barrel replacement and maintenance analysis, chemical consumption statistics and analysis, chemical filter replacement and maintenance, gasification system power consumption statistics and analysis, gasification system water consumption statistics and analysis, etc.

[0127] The second step is to perform feature engineering on the collected data, convert the raw data into feature representations that can be used by machine learning algorithms, and perform mining extraction and feature engineering, such as text data encoding, feature normalization or standardization operations. Set thresholds for the collected historical wafer yield data, and issue an alarm if the threshold range is exceeded.

[0128] The third step is to build a regression algorithm tree model such as lightgbm, perform Bayesian optimization on the hyperparameters, find the optimal model hyperparameters for training, and save the trained model.

[0129] The fourth step is to output the importance ranking of factory characteristics and yield based on the tree model. Through correlation analysis, output the correlation coefficient ranking of factory characteristics and yield. Perform attribution analysis based on importance indicators and correlation indicators, explain the correlation between specific factory factors and yield, determine which factory factors play a key role in improving or reducing the yield, and find out the key factory factors that affect the yield.

[0130] The fifth step is to predict the wafers in production, based on the key yield factors output in the previous step. If the yield of the current wafer is predicted to be lower than the threshold obtained previously, the factory factors are adjusted. Similarly, the relevant factory data features are processed, the trained model is called, and the wafer yield regression prediction value is output.

[0131] The sixth step is to determine the yield threshold based on the wafer yield output after adjustment. If it is lower than the threshold, the machine engineer will be notified.

[0132] The seventh step is to iterate and optimize the model regularly based on the newly added historical task data to continuously improve the model's effectiveness.

[0133] This technical solution outputs the importance index and correlation index of the factory-level features through a machine learning model, which can help determine which factory-level factors have a greater impact on the yield. Combined with the determination of the machine and time window for yield analysis, cross-analyze the factory-level and yield-related factors, and attribute and define possible related factors for abnormal processing and accumulate abnormal data features. Through the prediction model, based on different factory data, the future wafer yield situation can be simulated, and predictions and intervention adjustments can be made in advance. Through the prediction model, for real-time wafer production, combined with current factory data, the yield threshold judgment can be made and timely alarms can be issued. Closely synchronizing factory-level data with production equipment and factory prediction model systems can play an important role in improving production efficiency and yield.

[0134] Embodiment 4

[0135] Figure 4 Schematic diagram of the structure of the wafer yield control device based on factory data provided in the fourth embodiment of the present application. Figure 4 As shown, specifically including the following:

[0136] The basic factory data acquisition module 410 is used to acquire the basic factory data of the wafer currently being manufactured; wherein the basic factory data includes at least one of the water class parameters, electrical class parameters and gasification class parameters in the wafer manufacturing process;

[0137] A key factory service data determination module 420, configured to determine key factory service data associated with wafer yield from basic factory service data based on a pre-built machine learning model;

[0138] The wafer yield prediction module 430 is used to input the key factory data into the machine learning model to determine whether the estimated yield of the wafer currently manufactured reaches a preset threshold;

[0139] The factory affairs data adjustment module 440 is used to generate an instruction to adjust the key factory affairs data if the preset threshold is not reached.

[0140] The wafer yield control device based on factory data in the embodiment of the present application can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, an in-vehicle electronic device, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc., and the non-mobile electronic device can be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc., which is not specifically limited in the embodiment of the present application.

[0141] The wafer yield control device based on factory data in the embodiment of the present application can be a device with an operating system. The operating system can be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0142] The wafer yield control device based on factory data provided in the embodiment of the present application can implement each process implemented in the above method embodiment. To avoid repetition, it will not be described here.

[0143] Embodiment 5

[0144] In addition, the embodiment of the present application also provides a computer device, Figure 5 Schematic diagram of the structure of the computer device provided in Example 5 of the present application. Figure 5 As shown, the device includes a memory 51 for storing computer-readable instructions and a processor 52 for executing the computer-readable instructions, wherein when the computer-readable instructions are executed by the processor, the processor is triggered to execute the method described.

[0145] The methods and / or embodiments in the embodiments of the present application may be implemented as computer software programs. For example, the embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes program code for executing the method shown in the flowchart. When the computer program is executed by the processing unit, the above functions defined in the method of the present application are executed.

[0146] It should be noted that the computer-readable medium described in the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device.

[0147] In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0148] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0149] The flow chart or block diagram in the accompanying drawings shows the possible architecture, function and operation of the equipment, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated system for hardware that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0150] As another aspect, the embodiments of the present application further provide a computer-readable medium, which may be included in the device described in the above embodiments; or may exist independently without being assembled into the device. The above computer-readable medium carries one or more computer-readable instructions, which may be executed by a processor to implement the steps of the methods and / or technical solutions of the above multiple embodiments of the present application.

[0151] In a typical configuration of the present application, the terminal and the equipment of the service network each include one or more processors (CPU), input / output interface, network interface and memory.

[0152] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0153] Computer readable media include permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, modules of programs or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

[0154] In addition, an embodiment of the present application further provides a computer program, which is stored in a computer device, so that the computer device executes the method for controlling code execution.

[0155] It should be noted that the present application can be implemented in software and / or a combination of software and hardware, for example, can be implemented using an application specific integrated circuit (ASIC), a general purpose computer or any other similar hardware device. In certain embodiments, the software program of the present application can be executed by a processor to implement the above steps or functions. Similarly, the software program of the present application (including related data structures) can be stored in a computer-readable recording medium, for example, a RAM memory, a magnetic or optical drive or a floppy disk and similar devices. In addition, some steps or functions of the present application can be implemented using hardware, for example, as a circuit that cooperates with a processor to perform each step or function.

[0156] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or basic features of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present application is limited by the attached claims rather than the above description, so it is intended to include all changes that fall within the meaning and scope of the equivalent elements of the claims in the present application. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the device claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

Claims

1. A wafer yield control method based on factory data, characterized in that: The method comprises: Obtaining basic factory data of the wafer currently being manufactured; wherein the basic factory data includes at least one of water class parameters, electrical class parameters, and gasification class parameters in the wafer manufacturing process; Determine key fab data associated with wafer yield from basic fab data based on pre-built machine learning models; Inputting the key factory data into the machine learning model to determine whether the estimated yield of the wafers currently being manufactured reaches a preset threshold; If the preset threshold is not reached, an instruction to adjust the key factory affairs data is generated; The training process of the machine learning model includes: Obtain historical factory data of historically manufactured wafers and historical yield of wafer manufacturing; Performing data preprocessing and feature engineering processing on the historical factory data to obtain sample data; and using the historical yield as a label for the sample data; Inputting labeled sample data into an initial model, training the initial model, and obtaining a machine learning model; Among them, the key factory data associated with wafer yield is determined from the basic factory data based on the pre-built machine learning model, including: Determine the importance index of each basic factory data and the wafer yield based on a first algorithm included in the machine learning model; Determine a correlation index between each basic factory service data and the wafer yield based on a second algorithm included in the machine learning model; Determining key plant affairs data from basic plant affairs data according to the importance index and / or the relevance index; Wherein, according to the importance index and / or the relevance index, key plant affairs data is determined from the basic plant affairs data, including: If the importance index of the current basic factory affairs data exceeds the importance threshold, the current basic factory affairs data is determined to be key factory affairs data; or, If the correlation index of the current basic factory affairs data exceeds the positive correlation threshold, the current basic factory affairs data is determined to be key factory affairs data, and is positively correlated; if the correlation index of the current basic factory affairs data is less than the negative correlation threshold, the current basic factory affairs data is determined to be key factory affairs data, and is negatively correlated; or, If the importance index of the current basic factory affairs data exceeds the importance threshold, and the relevance index of the current basic factory affairs data exceeds the positive relevance threshold, the current basic factory affairs data is determined to be the key factory affairs data; or, if the importance index of the current basic factory affairs data exceeds the importance threshold, and the relevance index of the current basic factory affairs data is less than the negative relevance threshold, the current basic factory affairs data is determined to be the key factory affairs data.

2. The method according to claim 1, characterized in that After generating an instruction to adjust the key plant service data if the preset threshold is not reached, the method further includes: After adjusting the key factory affairs data, inputting the adjusted updated factory affairs data into the machine learning model; Identify whether the predicted yield of the output of the machine learning model reaches a preset threshold; If the predicted yield does not reach a preset threshold, an alarm message is generated.

3. The method according to claim 1, characterized in that After inputting the key factory data into the machine learning model to determine whether the estimated yield of the wafer currently being manufactured reaches a preset threshold, the method further includes: If the preset threshold is reached, the wafers are manufactured according to the current key factory data, and the yield of the currently manufactured wafers is counted as updated sample data; The machine learning model is iteratively updated according to the updated sample data.

4. The method according to any one of claims 1 to 3, characterized in that The machine learning model is a LightGBM regression prediction model.

5. A wafer yield control device based on factory data, characterized in that: The device comprises: A basic factory data acquisition module is used to acquire basic factory data of the wafer currently being manufactured; wherein the basic factory data includes at least one of water class parameters, electrical class parameters and gasification class parameters in the wafer manufacturing process; A key factory service data determination module, used to determine key factory service data associated with wafer yield from basic factory service data based on a pre-built machine learning model; A wafer yield prediction module, used to input the key factory data into the machine learning model to determine whether the estimated yield of the wafer currently being manufactured reaches a preset threshold; A factory affairs data adjustment module, used for generating an instruction to adjust the key factory affairs data if the preset threshold is not reached; The training process of the machine learning model includes: Obtain historical factory data of historically manufactured wafers and historical yield of wafer manufacturing; Performing data preprocessing and feature engineering processing on the historical factory data to obtain sample data; and using the historical yield as a label for the sample data; Inputting labeled sample data into an initial model, training the initial model, and obtaining a machine learning model; The key factory affairs data determination module is specifically used for: Determine the importance index of each basic factory data and the wafer yield based on a first algorithm included in the machine learning model; Determine a correlation index between each basic factory service data and the wafer yield based on a second algorithm included in the machine learning model; Determining key plant affairs data from basic plant affairs data according to the importance index and / or the relevance index; The key factory affairs data determination module is specifically used for: If the importance index of the current basic factory affairs data exceeds the importance threshold, the current basic factory affairs data is determined to be key factory affairs data; or, If the correlation index of the current basic factory affairs data exceeds the positive correlation threshold, the current basic factory affairs data is determined to be key factory affairs data, and is positively correlated; if the correlation index of the current basic factory affairs data is less than the negative correlation threshold, the current basic factory affairs data is determined to be key factory affairs data, and is negatively correlated; or, If the importance index of the current basic factory affairs data exceeds the importance threshold, and the relevance index of the current basic factory affairs data exceeds the positive relevance threshold, the current basic factory affairs data is determined to be the key factory affairs data; or, if the importance index of the current basic factory affairs data exceeds the importance threshold, and the relevance index of the current basic factory affairs data is less than the negative relevance threshold, the current basic factory affairs data is determined to be the key factory affairs data.

6. A computer device, characterized in that: The device comprises: One or more processors; and a memory storing computer program instructions, wherein when the computer program instructions are executed, the processor executes the wafer yield control method based on fab data as described in any one of claims 1 to 4.

7. A computer-readable medium, characterized in that Computer program instructions are stored thereon, and the computer program instructions can be executed by a processor to implement the wafer yield control method based on factory data as described in any one of claims 1-4.

Images