Semiconductor production equipment state prediction method, system and equipment and computer readable storage medium

Abstract

The invention discloses a semiconductor production equipment state prediction method, system and equipment and a computer readable storage medium, and the method comprises the steps: obtaining the operation state data of semiconductor production equipment and a time node corresponding to the operation state data, and determining training data according to the operation state data and the time node; based on the training data, training an xgboost algorithm model, and determining a model evaluation value corresponding to the xgboost algorithm model; if the model evaluation value is greater than a first preset threshold value, finishing training the xgboost algorithm model to obtain an equipment state prediction model; and based on the equipment state prediction model, predicting the operation state of the semiconductor production equipment to obtain a prediction result of the operation state of the semiconductor production equipment. The running state of the semiconductor production equipment can be predicted in advance according to the real-time running state of the semiconductor production equipment, and the technical problem that the running state of the semiconductor production equipment cannot be judged in time in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of equipment maintenance technology, and in particular to a method, system, device, and computer-readable storage medium for predicting the condition of semiconductor manufacturing equipment. Background Technology

[0002] With the increasing sophistication of industrial production data acquisition technology, predictive maintenance of industrial production equipment has become a hot topic in both industry and academia. Generally speaking, predictive maintenance of industrial production equipment refers to using real-time industrial production data, employing data analysis techniques to analyze the operating status of industrial production equipment, and further predicting potential failures and threats, thereby improving the reliability of industrial equipment. Semiconductor factories have extremely high requirements for process control during processing; even minor errors can lead to wafer scrapping. Therefore, once a failure occurs, it will cause certain economic losses, and if such failures are not maintained in a timely manner, they may have long-term effects on the equipment, resulting in high repair costs. Traditional methods for judging equipment operating status are mostly based on a series of complex process controls and years of maintenance personnel's experience. These methods rely heavily on the experience of maintenance personnel, but the number of maintenance personnel is usually limited, and the number of semiconductor production equipment is large. Relying solely on maintenance personnel makes it difficult to oversee numerous production devices and to promptly judge the operating status of semiconductor production equipment.

[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this invention is to provide a method, system, device, and computer-readable storage medium for predicting the status of semiconductor manufacturing equipment, aiming to solve the technical problem that existing technologies cannot determine the operating status of semiconductor manufacturing equipment in a timely manner.

[0005] To achieve the above objectives, the present invention provides a method for predicting the state of semiconductor manufacturing equipment, the method comprising the following steps: Acquire the operating status data of the semiconductor manufacturing equipment and the corresponding time nodes of the operating status data, and determine training data based on the operating status data and the time nodes, wherein the time node is the acquisition time of the operating status data; Based on the training data, the xgboost algorithm model is trained, and the model evaluation value corresponding to the xgboost algorithm model is determined. If the model evaluation value is greater than the first preset threshold, the training of the xgboost algorithm model is completed, and the device status prediction model is obtained. Based on the equipment status prediction model, the operating status of the semiconductor manufacturing equipment is predicted, and the estimated result of the operating status of the semiconductor manufacturing equipment is obtained.

[0006] Optionally, the step of training the XGBoost algorithm model based on the training data and determining the model evaluation value corresponding to the XGBoost algorithm model includes: Feature filtering is performed on the training data to obtain new training data; The new training data is used as the training data, and the following steps are performed: the xgboost algorithm model is trained based on the training data, and the model evaluation value corresponding to the xgboost algorithm model is determined.

[0007] Optionally, the step of performing feature filtering on the training data to obtain new training data includes: Based on the training data, determine the Pearson correlation coefficient of the training data, wherein the Pearson correlation coefficient represents the correlation between any two training data points; The training data corresponding to the Pearson correlation coefficient being greater than the second preset threshold are selected and used as new training data.

[0008] Optionally, the formula for determining the Pearson correlation coefficient of the training data based on the training data is:

[0009] Wherein, X represents any training data item in the training data, and Y represents another arbitrary training data item in the training data.

[0010] Optionally, the step of determining the training data based on the running status data and the time node includes: The operating status data and the corresponding time points are stored in the database; Obtain first running status data within a preset time period from the running status data, wherein the first running status data is any running status data, and the preset time period is any time period; Based on the running status data and the time node, determine the second running status data that is after the preset time period and is every preset time interval from the first running status data; Training data is determined based on the first running status data and the second running status data.

[0011] Optionally, the step of determining the training data based on the first running state data and the second running state data includes: The first operating status data is statistically analyzed to obtain the first statistical result; Perform a Fourier transform on the first operating state data to obtain the frequency domain information corresponding to the first operating state data; The second operating status data is statistically analyzed to obtain a second statistical result, and the equipment operating result is determined based on the second statistical result; The first statistical result and the frequency domain information are used as features in the training data, and the device operation results are used as labels in the training data.

[0012] Optionally, the step of using the first statistical result and the frequency domain information as features in the training data, and using the device operation result as a label in the training data, includes: Delete duplicate and abnormal data in the first statistical results and the frequency domain information, and fill in the missing data in the first statistical results and the frequency domain information to obtain the first data processing result; Delete duplicate and abnormal data in the second statistical result, and fill in the missing data in the second statistical result to obtain the second data processing result corresponding to the second statistical result; The first data processing result is used as a feature in the training data, and the second data processing result is used as a label in the training data.

[0013] Furthermore, to achieve the above objectives, the present invention also provides a semiconductor manufacturing equipment status prediction system, the semiconductor manufacturing equipment status prediction system comprising: The acquisition module is used to acquire the operating status data of the semiconductor manufacturing equipment and the time nodes corresponding to the operating status data, and to determine training data based on the operating status data and the time nodes, wherein the time node is the acquisition time of the operating status data. The determination module is used to train the xgboost algorithm model based on the training data and determine the model evaluation value corresponding to the xgboost algorithm model. The determination module is used to determine the status prediction model of the device if the model evaluation value is greater than the first preset threshold. The prediction module is used to predict the operating status of the semiconductor manufacturing equipment based on the equipment status prediction model, and obtain the estimated result of the operating status of the semiconductor manufacturing equipment.

[0014] In addition, to achieve the above objectives, the present invention also provides an operating state prediction device, the operating state prediction device comprising: a memory, a processor, and an operating state prediction program stored in the memory and executable on the processor, wherein the operating state prediction program, when executed by the processor, implements the steps of the semiconductor manufacturing equipment state prediction method as described above.

[0015] In addition, to achieve the above objectives, the present invention also provides a computer-readable storage medium storing a running state prediction program, which, when executed by a processor, implements the steps of the semiconductor manufacturing equipment state prediction method described above.

[0016] This invention acquires the operating status data of semiconductor manufacturing equipment and the corresponding time points of the operating status data. Training data is determined based on the operating status data and the time points, where the time point is the time when the operating status data is collected. Based on the training data, an XGBoost algorithm model is trained to determine the model evaluation value. If the model evaluation value is greater than a first preset threshold, the training of the XGBoost algorithm model is complete, resulting in an equipment status prediction model. Based on the equipment status prediction model, the operating status of the semiconductor manufacturing equipment is predicted to obtain an estimated result of the operating status of the semiconductor manufacturing equipment. In this embodiment, the operating status data of the semiconductor manufacturing equipment is collected to train the XGBoost algorithm model to construct an equipment status prediction model. The real-time operating status of the semiconductor manufacturing equipment is predicted based on the trained equipment status prediction model. This enables real-time prediction of changes in the operating status of factory equipment and allows for advance prediction of the operating status of the semiconductor manufacturing equipment based on its real-time operating status, solving the technical problem in the prior art of not being able to timely determine the operating status of semiconductor manufacturing equipment. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the hardware operating environment operation status prediction device involved in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the first embodiment of the semiconductor manufacturing equipment status prediction method of the present invention; Figure 3 This is a flowchart illustrating the second embodiment of the semiconductor manufacturing equipment status prediction method of the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] like Figure 1 As shown, Figure 1 This is a schematic diagram of the hardware operating environment operation status prediction device involved in the embodiments of the present invention.

[0021] The device for predicting the operating status in this embodiment of the invention can be a PC, or a smartphone, tablet computer, e-book reader, MP3 (Moving Picture Experts Group Audio Layer III) player, MP4 (Moving Picture Experts Group Audio Layer IV) player, portable computer, or other portable terminal device with display function.

[0022] like Figure 1 As shown, the operating status prediction device may include: a processor 1001, such as a CPU; a network interface 1004; a user interface 1003; a memory 1005; and a communication bus 1002. The communication bus 1002 is used to establish communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0023] Optionally, the operational status prediction device may also include a camera, RF (Radio Frequency) circuitry, sensors, audio circuitry, a WiFi module, and so on. Among these, sensors may include light sensors, motion sensors, and other sensors.

[0024] Those skilled in the art will understand that Figure 1 The operational status prediction device structure shown does not constitute a limitation on the operational status prediction device, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0025] like Figure 1As shown, the memory 1005, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a running status prediction program.

[0026] exist Figure 1 In the operation status prediction device shown, the network interface 1004 is mainly used to connect to the backend server and communicate data with the backend server; the user interface 1003 is mainly used to connect to the client (user end) and communicate data with the client; and the processor 1001 can be used to call the operation status prediction program stored in the memory 1005.

[0027] In this embodiment, the operation status prediction device includes: a memory 1005, a processor 1001, and an operation status prediction program stored in the memory 1005 and executable on the processor 1001. When the processor 1001 calls the operation status prediction program stored in the memory 1005, it performs the following operations: Acquire the operating status data of the semiconductor manufacturing equipment and the corresponding time nodes of the operating status data, and determine training data based on the operating status data and the time nodes, wherein the time node is the acquisition time of the operating status data; Based on the training data, the xgboost algorithm model is trained, and the model evaluation value corresponding to the xgboost algorithm model is determined. If the model evaluation value is greater than the first preset threshold, the training of the xgboost algorithm model is completed, and the device status prediction model is obtained. Based on the equipment status prediction model, the operating status of the semiconductor manufacturing equipment is predicted, and the estimated result of the operating status of the semiconductor manufacturing equipment is obtained.

[0028] Furthermore, the processor 1001 can call the runtime prediction program stored in the memory 1005 and also perform the following operations: Feature filtering is performed on the training data to obtain new training data; The new training data is used as the training data, and the following steps are performed: the xgboost algorithm model is trained based on the training data, and the model evaluation value corresponding to the xgboost algorithm model is determined.

[0029] Furthermore, the processor 1001 can call the runtime prediction program stored in the memory 1005 and also perform the following operations: Based on the training data, determine the Pearson correlation coefficient of the training data, wherein the Pearson correlation coefficient represents the correlation between any two training data points; The training data corresponding to the Pearson correlation coefficient being greater than the second preset threshold are selected and used as new training data.

[0030] Furthermore, the processor 1001 can call the runtime prediction program stored in the memory 1005 and also perform the following operations: The operating status data and the corresponding time points are stored in the database; Obtain first running status data within a preset time period from the running status data, wherein the first running status data is any running status data, and the preset time period is any time period; Based on the running status data and the time node, determine the second running status data that is after the preset time period and is every preset time interval from the first running status data; Training data is determined based on the first running status data and the second running status data.

[0031] Furthermore, the processor 1001 can call the runtime prediction program stored in the memory 1005 and also perform the following operations: The first operating status data is statistically analyzed to obtain the first statistical result; Perform a Fourier transform on the first operating state data to obtain the frequency domain information corresponding to the first operating state data; The second operating status data is statistically analyzed to obtain a second statistical result, and the equipment operating result is determined based on the second statistical result; The first statistical result and the frequency domain information are used as features in the training data, and the device operation results are used as labels in the training data.

[0032] Furthermore, the processor 1001 can call the runtime prediction program stored in the memory 1005 and also perform the following operations: Delete duplicate and abnormal data in the first statistical results and the frequency domain information, and fill in the missing data in the first statistical results and the frequency domain information to obtain the first data processing result; Delete duplicate and abnormal data in the second statistical result, and fill in the missing data in the second statistical result to obtain the second data processing result corresponding to the second statistical result; The first data processing result is used as a feature in the training data, and the second data processing result is used as a label in the training data.

[0033] This invention also provides a method for predicting the status of semiconductor manufacturing equipment, referring to... Figure 2 , Figure 2This is a flowchart illustrating the first embodiment of the semiconductor manufacturing equipment status prediction method of the present invention.

[0034] In this embodiment, the semiconductor manufacturing equipment status prediction method includes the following steps: Step S10: Obtain the operating status data of the semiconductor manufacturing equipment and the time node corresponding to the operating status data; determine the training data based on the operating status data and the time node, wherein the time node is the collection time of the operating status data. In this embodiment, the training data includes positive and negative samples. Normal operating status data is used as positive samples in the training data, while abnormal operating status data is used as negative samples. The XGBoost algorithm model is a classification tree model containing multiple decision trees. Before training the XGBoost algorithm model, operating status data of the semiconductor manufacturing equipment is collected, and the collection time is recorded, serving as the time node for the operating status data. After collecting the operating status data of the semiconductor manufacturing equipment, it should be noted that the operating status data needs to be analyzed and processed before it can be used as training data for the XGBoost algorithm model. Specifically, the operating status data is analyzed and processed based on the operating status data and the time nodes to determine the training data.

[0035] Furthermore, the operation status data of the semiconductor manufacturing equipment is collected by sensors, and the sensors also record the time nodes of the operation status data at the same time. The operation status data and the corresponding time nodes are stored in the database so that training data can be obtained from the database to train the XGBoost algorithm model. The operation status data includes temperature data, circuit status data or tool alarm data of the semiconductor manufacturing equipment at each stage of operation.

[0036] Step S20: Based on the training data, train the xgboost algorithm model and determine the model evaluation value corresponding to the xgboost algorithm model; In this embodiment, during the training of the XGBoost algorithm model based on training data, the training data is input into the XGBoost algorithm model to train it using the training data corresponding to the running state data. During training, the model evaluation value is calculated after each round of training. The model evaluation value represents the training effect of the XGBoost algorithm model; generally, the higher the model evaluation value, the better the training effect. The model evaluation value can be the accuracy, precision, recall, or the harmonic mean of precision and recall of the XGBoost algorithm model. Precision is the proportion of positive samples predicted as positive by the XGBoost algorithm model; recall is the proportion of positive samples predicted as positive among the actual positive samples; and accuracy is the proportion of correctly predicted samples in the XGBoost algorithm model's prediction results. The calculation methods for precision, recall, and the harmonic mean of precision and recall are as follows: Precision: P = TP / (TP + FP) Recall rate: R = TP / (TP + TN) The harmonic mean of precision and recall: F1, where 2 / F1 = 1 / P + 1 / R.

[0037] Furthermore, the XGBoost algorithm (eXtreme Gradient Boosting) is a machine learning algorithm, essentially a classification tree model containing multiple decision trees. It boasts advantages such as speed, high performance, ability to handle large-scale data, support for multiple languages, and support for custom loss functions. The training process for the XGBoost model involves first continuously adjusting the model parameters to find the optimal solution. Then, k-fold cross-validation is used to train the model. The trained model parameters are then ranked by importance, and the most important parameters are selected for remodeling. This process is repeated to solve the objective function, finding the optimal solution for each round and continuously updating the model parameters until training is complete, resulting in the trained XGBoost model.

[0038] Step S30: If the model evaluation value is greater than the first preset threshold, the training of the xgboost algorithm model is completed, and the device status prediction model is obtained. In this embodiment, during the training of the XGBoost algorithm model, the model evaluation value of the XGBoost algorithm model is calculated after each round of training, and it is determined whether the model evaluation value is greater than a first preset threshold. It should be noted that if the model evaluation value is greater than the first preset threshold, it means that the XGBoost algorithm model training is complete, and the training of the XGBoost algorithm model is stopped to obtain the device status prediction model; if the model evaluation value is less than or equal to the first preset threshold, it means that the current XGBoost algorithm model training is not complete, and the XGBoost algorithm model training needs to continue until the model evaluation value is greater than the first preset threshold.

[0039] Step S40: Based on the equipment status prediction model, predict the operating status of the semiconductor manufacturing equipment to obtain the estimated result of the operating status of the semiconductor manufacturing equipment.

[0040] In this embodiment, after obtaining the equipment status prediction model, it is possible to predict whether any of the production equipment in the factory will malfunction based on this model. Specifically, real-time operating status data of the semiconductor production equipment is collected and input into the equipment status prediction model corresponding to the already trained XGBoost algorithm model. This allows the equipment status prediction model to determine the estimated operating status of the semiconductor production equipment based on the real-time operating status data, and to plot the operating status data curve of the semiconductor production equipment based on the estimated results. This enables real-time prediction of changes in the operating status of the semiconductor production equipment, and combines historical maintenance data and industry-specific methods to perform predictive maintenance of the semiconductor production equipment.

[0041] The semiconductor manufacturing equipment status prediction method proposed in this embodiment acquires the operating status data of the semiconductor manufacturing equipment and the corresponding time nodes. Training data is determined based on the operating status data and the time nodes, where the time node is the time when the operating status data is collected. Based on the training data, an XGBoost algorithm model is trained to determine the model evaluation value. If the model evaluation value is greater than a first preset threshold, the training of the XGBoost algorithm model is complete, resulting in an equipment status prediction model. Based on the equipment status prediction model, the operating status of the semiconductor manufacturing equipment is predicted to obtain an estimated result. In this embodiment, the operating status data of the semiconductor manufacturing equipment is collected to train the XGBoost algorithm model to construct an equipment status prediction model. The real-time operating status of the semiconductor manufacturing equipment is predicted based on the trained equipment status prediction model. This method can predict changes in the operating status of factory equipment in real time and can predict the operating status of semiconductor manufacturing equipment in advance based on its real-time operating status, solving the technical problem in the prior art of not being able to determine the operating status of semiconductor manufacturing equipment in a timely manner.

[0042] Based on the first embodiment, a second embodiment of the semiconductor manufacturing equipment status prediction method of the present invention is proposed, referring to... Figure 3 In this embodiment, step S20 includes: Step S21: Perform feature filtering on the training data to obtain new training data; Step S22: Use the new training data as the training data and perform the following steps: train the xgboost algorithm model based on the training data and determine the model evaluation value corresponding to the xgboost algorithm model.

[0043] In this embodiment, before training the XGBoost algorithm model, feature filtering can be performed on the training data to select training data with high feature relevance and remove training data with low feature relevance. Then, the XGBoost algorithm model is trained based on the training data with high feature relevance, thereby improving the model training effect. Specifically, feature filtering is performed on the training data to further filter out new training data. Then, the XGBoost algorithm model is trained based on the newly selected training data until the model evaluation value exceeds a preset threshold, at which point training of the XGBoost algorithm model stops.

[0044] Furthermore, the step of performing feature filtering on the training data to obtain new training data includes: Step S211: Determine the Pearson correlation coefficient of the training data based on the training data, wherein the Pearson correlation coefficient represents the correlation between any two training data points; Step S212: Select the training data whose Pearson correlation coefficient is greater than the second preset threshold, and use the training data whose Pearson correlation coefficient is greater than the second preset threshold as new training data.

[0045] In this embodiment, the specific process of feature selection for training data is as follows: Based on the training data, the Pearson correlation coefficient is calculated for any two training data points. This coefficient is used to select training data, where the Pearson correlation coefficient represents the correlation between any two training data points. Then, based on the Pearson correlation coefficient between any two training data points, training data with a Pearson correlation coefficient greater than a second preset threshold are selected and used as new training data. This embodiment selects training data by calculating the correlation between any two training data points, and selects training data with higher correlation, so that the model can be trained based on the new training data, thereby further improving the model's training effect.

[0046] Furthermore, after iterating the XGBoost algorithm model a preset number of times, feature filtering can be performed on the training data. This allows for further training of the XGBoost algorithm model only after it has achieved a certain level of performance. This avoids the calculated Pearson correlation coefficient being too unreliable, thereby improving the efficiency and effectiveness of model training.

[0047] Furthermore, the formula for calculating the Pearson correlation coefficient between any two training data points is as follows:

[0048] Where X represents any training data item in the training data, and Y represents another arbitrary training data item in the training data. i For each feature included in X, Y i Let Y be the features included in Y.

[0049] Furthermore, the step of determining the training data based on the running status data and the time node includes: Step S11: Store the running status data and the corresponding time nodes in the database; Step S12: Obtain first running status data within a preset time period from the running status data, wherein the first running status data is any running status data and the preset time period is any time period; Step S13: Based on the running status data and the time node, determine the second running status data that is after the preset time period and is every preset time interval from the first running status data; Step S14: Determine training data based on the first running status data and the second running status data.

[0050] In this embodiment, when collecting the operating status data of the semiconductor manufacturing equipment, the time node of the operating status data is recorded simultaneously, and the operating status data and the corresponding time node are stored in the database. This allows the operating status data to be collected in the database so that training data can be obtained from the database to train the xgboost algorithm model, and the massive amount of operating status data stored in the database can be analyzed and processed.

[0051] After storing the operational status data and corresponding time points in the database, this operational status data is unprocessed. After collecting the operational status data into the database, it needs to be processed. Specifically, operational status data within a preset time period t1 is obtained to obtain first operational status data. The first operational status data can be any operational status data, and the preset time period can be any time period. Since each operational status data corresponds to a unique time point, operational status data after the preset time period and occurring every preset time period t2 from the first operational status data can be determined to obtain second operational status data. The determined second operational status data is used to determine the device operation result corresponding to the first operational status data (including device alarm generation and normal device operation). After determining the first operational status data and the corresponding second operational status data, the statistical results corresponding to the first operational status data and the device operation results corresponding to the second operational status data are determined. The statistical results corresponding to the first operational status data are used as features of the training data, and the device operation results corresponding to the second operational status data are used as labels of the training data to obtain training data.

[0052] Further, the step of determining the training data based on the first operating state data and the second operating state data includes: Step S141: Statistically analyze the first operating status data to obtain a first statistical result; Step S142: Perform a Fourier transform on the first running state data to obtain the frequency domain information corresponding to the first running state data; Step S143: Statistically analyze the second operating status data to obtain a second statistical result, and determine the equipment operating result based on the second statistical result; Step S144: Use the first statistical result and the frequency domain information as features in the training data, and use the device operation result as a label in the training data.

[0053] In this embodiment, after obtaining the first operating state data and the second operating state data, statistics are performed on the first and second operating state data respectively. This may include calculating the maximum, minimum, average, standard deviation, or slope of the two data to obtain a first statistical result corresponding to the first operating state data and a second statistical result corresponding to the second operating state data. Furthermore, a Fourier transform is performed on the first operating state data to obtain information in the frequency domain, thus obtaining the corresponding frequency domain information. This frequency domain information is then added to the features of the training data to improve the model training effect. After determining the first statistical result and the corresponding second statistical result, the device operating result is determined based on the second statistical result. The first statistical result and the frequency domain information corresponding to the first operating state are used as features in the training data, and the device operating result is used as a label in the training data. The device operating result includes whether the device generates an alarm or operates normally.

[0054] Further, the step of using the first statistical result and the frequency domain information as features in the training data, and using the device operation result as a label in the training data, includes: Step S1441: Delete duplicate and abnormal data in the first statistical results and the frequency domain information, and complete the missing data in the first statistical results and the frequency domain information to obtain the first data processing result; Step S1442: Delete duplicate and abnormal data in the second statistical result, and fill in the missing data in the second statistical result to obtain the second data processing result corresponding to the second statistical result; Step S1443: Use the first data processing result as a feature in the training data and the second data processing result as a label in the training data.

[0055] In this embodiment, after determining the first statistical result and frequency domain information corresponding to the first operating state data, the duplicate data, missing data, and abnormal data in the first statistical result and frequency domain information corresponding to the first operating state data can be further processed. After processing the duplicate data, missing data, and abnormal data, the processed first statistical result and frequency domain information are used as features in the training data. Specifically, for duplicate data and abnormal data in the first statistical result and frequency domain information, the duplicate data and abnormal data are deleted; for missing data in the first statistical result and frequency domain information, the missing data is filled in. Further, similar data corresponding to the missing data can be calculated, and then the similar data can be filled in accordingly. The same applies to the second operating state data.

[0056] The semiconductor manufacturing equipment status prediction method proposed in this embodiment obtains new training data by performing feature filtering on the training data; the new training data is then used as the training data, and the following steps are performed: training the XGBoost algorithm model based on the training data, and determining the model evaluation value corresponding to the XGBoost algorithm model. In this embodiment, before training the XGBoost algorithm model, feature filtering is performed on the training data to select training data with high feature correlation and remove training data with low feature correlation. Then, the XGBoost algorithm model is trained based on the training data with high feature correlation, thereby improving the model training effect of the XGBoost algorithm model.

[0057] Furthermore, embodiments of the present invention also propose a semiconductor manufacturing equipment status prediction system, the semiconductor manufacturing equipment status prediction system comprising: The acquisition module is used to acquire the operating status data of the semiconductor manufacturing equipment and the time nodes corresponding to the operating status data, and to determine training data based on the operating status data and the time nodes, wherein the time node is the acquisition time of the operating status data. The determination module is used to train the xgboost algorithm model based on the training data and determine the model evaluation value corresponding to the xgboost algorithm model. The determination module is used to determine the status prediction model of the device if the model evaluation value is greater than the first preset threshold. The prediction module is used to predict the operating status of the semiconductor manufacturing equipment based on the equipment status prediction model, and obtain the estimated result of the operating status of the semiconductor manufacturing equipment.

[0058] Furthermore, the determining module is also used for: Feature filtering is performed on the training data to obtain new training data; The new training data is used as the training data, and the following steps are performed: the xgboost algorithm model is trained based on the training data, and the model evaluation value corresponding to the xgboost algorithm model is determined.

[0059] Furthermore, the determining module is also used for: Based on the training data, determine the Pearson correlation coefficient of the training data, wherein the Pearson correlation coefficient represents the correlation between any two training data points; The training data corresponding to the Pearson correlation coefficient being greater than the second preset threshold are selected and used as new training data.

[0060] Furthermore, the formula for determining the Pearson correlation coefficient of the training data based on the training data is as follows:

[0061] Wherein, X represents any training data item in the training data, and Y represents another arbitrary training data item in the training data.

[0062] Furthermore, the acquisition module is also used for: The operating status data and the corresponding time points are stored in the database; Obtain first running status data within a preset time period from the running status data, wherein the first running status data is any running status data, and the preset time period is any time period; Based on the running status data and the time node, determine the second running status data that is after the preset time period and is every preset time interval from the first running status data; Training data is determined based on the first running status data and the second running status data.

[0063] Furthermore, the acquisition module is also used for: The first operating status data is statistically analyzed to obtain the first statistical result; Perform a Fourier transform on the first operating state data to obtain the frequency domain information corresponding to the first operating state data; The second operating status data is statistically analyzed to obtain a second statistical result, and the equipment operating result is determined based on the second statistical result; The first statistical result and the frequency domain information are used as features in the training data, and the device operation results are used as labels in the training data.

[0064] Furthermore, the acquisition module is also used for: Delete duplicate and abnormal data in the first statistical results and the frequency domain information, and fill in the missing data in the first statistical results and the frequency domain information to obtain the first data processing result; Delete duplicate and abnormal data in the second statistical result, and fill in the missing data in the second statistical result to obtain the second data processing result corresponding to the second statistical result; The first data processing result is used as a feature in the training data, and the second data processing result is used as a label in the training data.

[0065] Furthermore, embodiments of the present invention also propose a computer-readable storage medium storing a running state prediction program, which, when executed by a processor, implements the steps of the semiconductor manufacturing equipment state prediction method as described in any of the above embodiments.

[0066] The specific embodiments of the computer-readable storage medium of the present invention are basically the same as the embodiments of the semiconductor manufacturing equipment state prediction method described above, and will not be described in detail here.

[0067] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0068] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0069] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0070] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for predicting the status of semiconductor manufacturing equipment, characterized in that, The semiconductor manufacturing equipment status prediction method includes the following steps: Acquire the operating status data of the semiconductor manufacturing equipment and the corresponding time nodes of the operating status data, and determine training data based on the operating status data and the time nodes, wherein the time node is the acquisition time of the operating status data; Based on the training data, the xgboost algorithm model is trained, and the model evaluation value corresponding to the xgboost algorithm model is determined. If the model evaluation value is greater than the first preset threshold, the training of the xgboost algorithm model is completed, and the device status prediction model is obtained. Based on the equipment status prediction model, the operating status of the semiconductor manufacturing equipment is predicted, and the estimated result of the operating status of the semiconductor manufacturing equipment is obtained.

2. The semiconductor manufacturing equipment status prediction method as described in claim 1, characterized in that, The step of training the XGBoost algorithm model based on the training data and determining the model evaluation value corresponding to the XGBoost algorithm model includes: Feature filtering is performed on the training data to obtain new training data; The new training data is used as the training data, and the following steps are performed: the xgboost algorithm model is trained based on the training data, and the model evaluation value corresponding to the xgboost algorithm model is determined.

3. The semiconductor manufacturing equipment status prediction method as described in claim 2, characterized in that, The step of performing feature filtering on the training data to obtain new training data includes: Based on the training data, determine the Pearson correlation coefficient of the training data, wherein the Pearson correlation coefficient represents the correlation between any two training data points; The training data corresponding to the Pearson correlation coefficient being greater than the second preset threshold are selected and used as new training data.

4. The semiconductor manufacturing equipment status prediction method as described in claim 3, characterized in that, The formula for determining the Pearson correlation coefficient of the training data based on the training data is as follows: Wherein, X represents any training data item in the training data, and Y represents another arbitrary training data item in the training data.

5. The semiconductor manufacturing equipment status prediction method according to any one of claims 1 to 4, characterized in that, The step of determining training data based on the running status data and the time node includes: The operating status data and the corresponding time points are stored in the database; Obtain first running status data within a preset time period from the running status data, wherein the first running status data is any running status data, and the preset time period is any time period; Based on the running status data and the time node, determine the second running status data that is after the preset time period and is every preset time interval from the first running status data; Training data is determined based on the first running status data and the second running status data.

6. The semiconductor manufacturing equipment status prediction method as described in claim 5, characterized in that, The step of determining the training data based on the first operating state data and the second operating state data includes: The first operating status data is statistically analyzed to obtain the first statistical result; Perform a Fourier transform on the first operating state data to obtain the frequency domain information corresponding to the first operating state data; The second operating status data is statistically analyzed to obtain a second statistical result, and the equipment operating result is determined based on the second statistical result; The first statistical result and the frequency domain information are used as features in the training data, and the device operation results are used as labels in the training data.

7. The semiconductor manufacturing equipment status prediction method as described in claim 6, characterized in that, The step of using the first statistical result and the frequency domain information as features in the training data, and using the device operation result as a label in the training data, includes: Delete duplicate and abnormal data in the first statistical results and the frequency domain information, and fill in the missing data in the first statistical results and the frequency domain information to obtain the first data processing result; Delete duplicate and abnormal data in the second statistical result, and fill in the missing data in the second statistical result to obtain the second data processing result corresponding to the second statistical result; The first data processing result is used as a feature in the training data, and the second data processing result is used as a label in the training data.

8. A semiconductor manufacturing equipment status prediction system, characterized in that, The semiconductor manufacturing equipment status prediction system includes: The acquisition module is used to acquire the operating status data of the semiconductor manufacturing equipment and the time nodes corresponding to the operating status data, and to determine training data based on the operating status data and the time nodes, wherein the time node is the acquisition time of the operating status data. The determination module is used to train the xgboost algorithm model based on the training data and determine the model evaluation value corresponding to the xgboost algorithm model. The determination module is used to determine the status prediction model of the device if the model evaluation value is greater than the first preset threshold. The prediction module is used to predict the operating status of the semiconductor manufacturing equipment based on the equipment status prediction model, and obtain the estimated result of the operating status of the semiconductor manufacturing equipment.

9. A semiconductor manufacturing equipment status prediction device, characterized in that, The semiconductor manufacturing equipment status prediction device includes: a memory, a processor, and a running status prediction program stored in the memory and executable on the processor. When the running status prediction program is executed by the processor, it implements the steps of the semiconductor manufacturing equipment status prediction method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a running state prediction program, which, when executed by a processor, implements the steps of the semiconductor manufacturing equipment state prediction method as described in any one of claims 1 to 7.

Images