Wafer processing method and device and storage medium

Abstract

The invention provides a wafer processing method, wafer processing equipment and a storage medium. The method comprises the following steps: acquiring an initial delivery scheme for processing a wafer, and acquiring state data of each device based on a plurality of devices for processing the wafer included in the initial delivery scheme; wherein each device comprises a plurality of components; based on the state data, using a fault prediction model to predict fault information of each device, the fault information being used to predict whether the device has a fault and a fault component; optimizing the initial delivery scheme according to the fault information so as to avoid using the fault component to process the wafer; and controlling each device to process the wafer according to the optimized delivery scheme. According to the method, the equipment fault can be predicted in advance, and each piece of equipment is controlled to process the wafer according to the optimized delivery scheme, so that each piece of equipment is ensured to operate with relatively large throughput, and the manufacturing efficiency of the chip is improved.

Description

Technical Field

[0001] The present application relates to the field of semiconductor technology, and in particular to a wafer processing method, device and storage medium. Background Art

[0002] In semiconductor manufacturing, the throughput of each device affects the efficiency of the entire chip manufacturing. Throughput refers to the number of wafers that pass through the device within a certain period of time. Therefore, it is very important to monitor and optimize the throughput of the device.

[0003] The current method of handling wafer throughput solutions usually reserves a longer time to deal with equipment failures, resulting in lower equipment throughput and, in turn, lower chip manufacturing efficiency. Summary of the invention

[0004] The present application provides a wafer processing method, equipment and storage medium to solve the problem that the prior art equipment usually reserves a long time to deal with equipment failures, resulting in low equipment throughput and low chip manufacturing efficiency.

[0005] In a first aspect, the present application provides a wafer processing method, comprising:

[0006] Obtaining an initial cargo transfer plan for processing wafers, and obtaining status data of each device based on multiple devices for processing wafers included in the initial cargo transfer plan; wherein each device includes multiple components;

[0007] Based on the status data, the fault prediction model is used to predict the fault information of each device. The fault information is used to predict whether the device will fail and the faulty components.

[0008] Optimize the initial delivery plan based on the failure information to avoid processing wafers with failed parts;

[0009] According to the optimized cargo transfer plan, each device is controlled to process the wafers.

[0010] In a second aspect, the present application provides a wafer processing device, comprising:

[0011] An acquisition module is used to acquire an initial cargo transfer plan for processing wafers, and based on multiple devices for processing wafers included in the initial cargo transfer plan, acquire status data of each device; wherein each device includes multiple components;

[0012] A processing module is used to predict the fault information of each device based on the state data using a fault prediction model, and the fault information is used to predict whether the device will fail and the faulty component;

[0013] The processing module is further used to optimize the initial cargo transfer plan according to the fault information to avoid using the faulty parts to process the wafer;

[0014] The control module is used to control various devices to process the wafers according to the optimized cargo transfer plan.

[0015] In a third aspect, the present application provides a wafer processing device, comprising: a processor, a memory, wherein codes are stored in the memory, and the processor runs the codes stored in the memory to execute any method as in the first aspect.

[0016] In a fourth aspect, the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement any method as in the first aspect.

[0017] In a fifth aspect, the present application provides a computer program product, comprising a computer program, which implements any one of the methods in the first aspect when executed by a processor.

[0018] The present application provides a wafer processing method, device and storage medium, obtains an initial delivery plan for processing wafers, obtains status data of each device based on multiple devices for processing wafers included in the initial delivery plan, and uses a fault prediction model to predict the fault information of each device, which can predict whether the device will fail and the faulty component. The initial delivery plan is optimized according to the fault information to avoid using faulty components to process wafers, and each device is controlled to process the wafer according to the optimized delivery plan. The present application can timely perceive or predict component failures in advance through a fault prediction model, and optimize the initial delivery plan based on the fault information, without the need for faulty components to process wafers, thereby effectively ensuring that each device operates at a higher throughput, without the need to reserve a long time to deal with equipment failures, which is beneficial to improving chip manufacturing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0020] Figure 1 A schematic diagram of a wafer processing scenario provided in an embodiment of the present application;

[0021] Figure 2 A flow chart of a wafer processing method provided in an embodiment of the present application;

[0022] Figure 3 A flow chart of a method for controlling a device to process a wafer provided in an embodiment of the present application;

[0023] Figure 4 A flow chart of a method for performing dimensionality reduction processing on a data set provided in an embodiment of the present application;

[0024] Figure 5 A schematic diagram of a wafer processing device provided in an embodiment of the present application Figure 1 ;

[0025] Figure 6 A schematic diagram of a wafer processing device provided in an embodiment of the present application Figure 2 .

[0026] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0027] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0028] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0029] Optimizing throughput means coordinating the time for each wafer to enter the component so that the equipment can operate at the maximum throughput within a given time. Monitoring equipment throughput means real-time tracking of the wafer's residence time in each component, entry and exit time, and timely detection and alarm of relevant personnel when abnormalities occur. Throughput refers to the number of wafers that pass through the equipment within a certain period of time.

[0030] In semiconductor manufacturing, the throughput of each device affects the efficiency of the entire chip manufacturing, so it is very important to monitor and optimize the throughput of the equipment. Take etching as an example. Etching is the process of removing unnecessary materials attached to the surface of the wafer by chemical or physical methods. The etching process sequence is after coating and photolithography, that is, a thin film of the material used to carve the circuit is first deposited on the wafer, and then a photoresist is deposited on it. In the process of chip production, it is often necessary to repeat the etching steps many times. If an etching device fails and causes a blockage, it will cause a lag in the entire process.

[0031] In semiconductor manufacturing, the current method of scheduling wafer shipments typically allows for a longer period of time to respond to equipment failures, resulting in lower efficiency.

[0032] The present application provides a method for processing wafers, obtaining an initial transfer plan for processing wafers, and obtaining status data of each device based on multiple devices for processing wafers included in the initial transfer plan. Based on the status data, a fault prediction model is used to predict component failures and the time required to repair faulty components in advance. After determining fault information such as component failures and the time required to repair faulty components, the initial transfer plan is optimized based on the repair information to avoid using faulty components to process wafers, and each device is controlled to process wafers with the optimized transfer plan, so that the device where the faulty component is located can also use other normal components to continue to process specific wafers, ensuring that each device operates at a higher throughput, and there is no need to reserve a long time to deal with equipment failures, effectively improving chip manufacturing efficiency.

[0033] Figure 1 A schematic diagram of a wafer processing scenario provided in an embodiment of the present application, such as Figure 1As shown, the data set is encoded, integrated, filtered, etc., and the data set is trained based on the classification algorithm to obtain a fault prediction model. An initial cargo transfer plan for processing wafers is obtained, and the status data of multiple devices currently processing wafers are obtained based on the initial cargo transfer plan. Fault prediction is performed using the fault prediction model, and fault information such as faulty components, fault probabilities, and the time required to repair faulty components are output. Based on the fault information, the initial cargo transfer plan can be optimized using an optimization algorithm to obtain a first cargo transfer plan. At this time, the user can manually judge whether the equipment or components have failed and the time required to repair the faulty components, and based on the user's judgment results, the first cargo transfer plan is optimized again using an optimization algorithm, and adjustments are made on the basis of the first cargo transfer plan to obtain a second cargo transfer plan. The second cargo transfer plan is used as the current better cargo transfer plan, and each device can be controlled to process the wafer with the second cargo transfer plan, so that each device operates with a larger throughput. At the same time, the second cargo transfer plan and the new data generated during the user's repair of the equipment or components are updated to the data pool to optimize the fault prediction model.

[0034] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0035] Figure 2 The flowchart of a wafer processing method provided in an embodiment of the present application. The method of this embodiment can be executed by a wafer processing device and can be implemented by hardware, software, or a combination of hardware and software. Figure 2 As shown, the method is as follows:

[0036] S201: Obtain an initial cargo transfer plan for processing wafers, and based on multiple devices for processing wafers included in the initial cargo transfer plan, obtain status data of each device; wherein each device includes multiple components.

[0037] In some embodiments, the equipment for processing wafers may be an etching equipment, a deposition equipment, a photolithography machine, etc. A component is a component of a device. For example, an etching equipment includes but is not limited to multiple components such as a chamber, a vacuum pump, a vacuum valve, a flow meter, and a pipeline. A deposition equipment includes but is not limited to multiple components such as a chamber, a robot arm, a vacuum pump, and a pipeline. A photolithography machine may include multiple components such as a measuring table, a laser, a beam straightener, an energy controller, and a pipeline.

[0038] In some embodiments, the initial delivery plan is a plan for processing wafers, including the processing order of the wafers, the expected time in different equipment and equipment components, and the recipe, etc. When processing wafers with different recipes, the components used may be different.

[0039] In some embodiments, the status data may include a record of the time at which the equipment processes the wafer and equipment sensor data.

[0040] The time record of the equipment processing wafers includes the time when the wafer enters each component and the time required, the time it stays in the component, and the time when it leaves the component and the time required. The equipment sensor data includes the sampling time of the sensor and the sampled value. The sampled value can be information such as temperature, pressure, pressure, humidity, etc.

[0041] In one implementation scenario, the status data of the device may be that the device is currently in a non-operating state, and is compared with the status data generated during the previous operation process, so as to predict the probability of failure when the device is turned on and the wafer is processed based on the status data of the previous operation process.

[0042] In another implementation scenario, the status data of the device may also be status data generated in real time when the device is in an operating state.

[0043] S202: Based on the status data, a fault prediction model is used to predict the fault information of each device. The fault information is used to predict whether the device will fail and the faulty component.

[0044] In some embodiments, the state data is input into the fault prediction model, and the fault prediction model can output the corresponding fault information. The fault information includes whether there will be a fault and the faulty component, and can also include the cause of the fault, the probability of the fault, the fault type, and the time required to repair the faulty component. The time required to repair the faulty component can be the product of the probability of the fault and the repair time corresponding to the fault type.

[0045] In some embodiments, the number of fault prediction models can be multiple, and different types of equipment correspond to different fault prediction models. For example, when predicting the fault information of an etching device, the state data of the etching device needs to be input into the fault prediction model for predicting the fault of the etching device. When predicting the fault of a deposition device, the state data of the deposition device needs to be input into the fault prediction model for predicting the fault of the deposition device.

[0046] S203: Optimize the initial cargo transfer plan according to the fault information to avoid using faulty components to process wafers.

[0047] In some embodiments, multiple devices or components currently processing wafers may be updated according to the fault information to obtain available components other than the faulty components. Based on the currently available components, the initial delivery plan is optimized using an optimization algorithm, wherein the optimization algorithm may be a particle swarm algorithm.

[0048] Specifically, when using the particle swarm algorithm for optimization, the order and speed of wafer delivery can be initialized, the fitness function value, individual extreme value and group extreme value can be calculated, and the speed and wafer delivery order can be updated. The calculation of the fitness function value, individual extreme value and group extreme value, as well as the update of the speed and wafer delivery order are repeated until the termination condition is met, and the optimal solution is output. Among them, the termination condition can be the shortest time required to process the wafer.

[0049] In some embodiments, since the fault information also includes the time required to repair the faulty component, the initial cargo transfer plan can be optimized based on the faulty component and the time required to repair the faulty component to avoid using the faulty component to process wafers during the time it takes to repair the faulty component.

[0050] In one implementation scenario, since there are certain differences in the components required to process wafers of different recipes, when a component of a device that processes wafers fails, the device can be set to process other types of wafers that do not require the failed component within the time required to repair the failed component. This allows the device where the failed component is located to continue to operate normally while the user is repairing the failed component, thereby ensuring the throughput of the device and improving efficiency.

[0051] In one implementation scenario, if the component of other equipment has not failed, in order to ensure the production of this type of wafer, other equipment can be ordered to process this type of wafer. Therefore, the initial delivery plan can be optimized based on the fault information to meet the wafer production needs while ensuring a high equipment throughput.

[0052] S204: According to the optimized cargo transfer plan, control each device to process the wafers.

[0053] In some embodiments, the optimized cargo transfer plan and fault information can be sent to the user for confirmation. The user confirms whether a fault has occurred and whether the time required to repair the faulty component is reasonable. Based on the result of the user's confirmation, each device is controlled to process the wafer.

[0054] The embodiment of the present application provides a method for processing wafers, obtains an initial cargo transfer plan for processing wafers, and obtains the status data of each device based on multiple devices for processing wafers included in the initial cargo transfer plan. Based on the status data, a fault prediction model is used to predict the fault information of each device, and the fault information can predict whether the device will fail and the faulty component. The initial cargo transfer plan is optimized according to the fault information to avoid using faulty components to process wafers. According to the optimized cargo transfer plan, each device is controlled to process the wafer. The method of the present application can predict equipment failures in advance based on the status data of the equipment using the fault prediction model, which is conducive to improving safety. At the same time, the present application also optimizes the initial cargo transfer plan based on the fault information, controls each device to process the wafer with the optimized cargo transfer plan, so that the device where the faulty component is located can also use other components to process the wafer, which can effectively ensure that each device operates at a larger throughput, without reserving a long time to deal with equipment failures, and further improves the manufacturing efficiency of the chip.

[0055] Figure 3 A flow chart of a method for controlling a device to process a wafer provided in an embodiment of the present application. In one or more embodiments of the present application, before controlling each device to process the wafer according to the optimized cargo transfer scheme, the following steps need to be repeatedly performed until it is manually confirmed that the fault information of each device does not need to be modified, and each device is controlled to process the wafer according to the optimized cargo transfer scheme, specifically including: Figure 3 Steps shown:

[0056] S301: Display the fault information of each device and the optimized cargo transfer plan for manual confirmation.

[0057] In one implementation scenario, the fault information of each device and the optimized cargo transfer plan can be directly displayed for manual confirmation.

[0058] In another implementation scenario, the fault information of each device and the optimized cargo transfer plan can also be sent to the user responsible for the device via email, text message, etc. for manual confirmation.

[0059] In some embodiments, when determining whether a component is faulty, the user can determine whether the component is faulty based on the probability of the component failing and the user's tracking and understanding of the equipment operation status. If the user confirms that the fault information of each device does not need to be modified, each device is controlled to process the wafer according to the optimized delivery plan.

[0060] For example, the fault information of a certain device includes three components that may fail and the time required to repair each faulty component. The user confirms that all three components are faulty and the time required to repair each faulty component is reasonable. In this case, there is no need to modify the fault information of the device. At this time, each device can be controlled to process the wafers according to the optimized delivery plan.

[0061] S302: If it is manually confirmed that the fault information of any device needs to be modified, the optimized cargo transfer plan is optimized according to the modified fault information.

[0062] As the fault information changes, the currently optimized cargo handling plan needs to be optimized again according to the changed fault information so that the equipment can operate at a higher throughput.

[0063] For example, the fault information of a certain device contains three components that may fail, but the user only confirms that one of the components is faulty. At this time, if each device is still controlled to process the wafer based on the above optimized cargo transfer plan, it cannot be guaranteed that each device operates at a large throughput. Therefore, based on the fault information confirmed by the user, the optimized cargo transfer plan can be optimized again using the optimization algorithm. For the convenience of description, the optimization of the initial cargo transfer plan can be called the first optimization, and the optimization of the optimized cargo transfer plan can be called the second optimization.

[0064] Specifically, the optimized cargo transfer plan can be used as the initial value of the optimization algorithm, and iterated until convergence, so as to obtain the cargo transfer plan after the second optimization. The optimization algorithm can be a particle swarm algorithm.

[0065] Since the second optimization is performed on the basis of the cargo transfer plan after the first optimization, it is a fine-tuning of the cargo transfer plan after the first optimization. Therefore, the time required for convergence is shorter and it is more flexible than the first randomly initialized optimization plan.

[0066] The cargo transfer plan after the second optimization and the corresponding fault information are displayed for manual confirmation. If the fault information still needs to be modified after manual confirmation, the cargo transfer plan after the second optimization can be optimized for the third time based on the fault information after the second modification, and so on, until it is manually confirmed that the fault information does not need to be modified. The cargo transfer plan corresponding to the fault information at this time is used as the final cargo transfer plan, and each device is controlled to process the wafers according to the final cargo transfer plan.

[0067] Since the second optimization and the third optimization are optimized based on the fault information of each device after the user's confirmation, that is, the cargo transfer plan after the second and third optimization is an optimization plan proposed based on the user's tracking of the current equipment operation status, it is closer to the actual operation status of the current equipment than the cargo transfer plan after the first optimization. Controlling each device to process the wafer according to the final cargo transfer plan after multiple optimizations is conducive to maximizing the throughput of each device.

[0068] In summary, when it is manually confirmed that the fault information of any device needs to be modified, the optimized cargo transfer plan is optimized according to the modified fault information until it is manually confirmed that the fault information of each device does not need to be modified, so that the second or multiple optimized cargo transfer plans are more in line with the actual operating conditions of each device, which is conducive to achieving each device running at a larger throughput and improving chip manufacturing efficiency.

[0069] In one or more embodiments of the present application, the process of establishing a fault prediction model may include:

[0070] Obtain data sets of multiple devices of the same type and build a fault prediction model based on the data sets.

[0071] In some embodiments, devices of the same type may be devices of the same model.

[0072] In one implementation scenario, the data set may include: historical status numbers of equipment, historical fault information, and historical cargo transfer plans.

[0073] Since the status data may include the time records of the equipment processing wafers and the equipment sensor data, the data set may include the following information: the historical time records of each equipment processing wafers, the historical data of the equipment sensors, the historical delivery plans, and the historical failure information.

[0074] Among them, the historical time records of wafer processing by each device include the time and time required for the wafer to enter each component, the time and time required for the wafer to leave the component, and the residence time in the component, as shown in Table 1.

[0075] Table 1 Historical time record of equipment processing wafers

[0076]

[0077] The historical data of the device sensor can be obtained from the device log. The historical data of the device sensor may include the sampling time and sampling value of the sensor. For details, please refer to Table 2. It should be noted that the values ​​in Table 2 are all examples.

[0078] Table 2 Sensor sampling time and sampling value list

[0079]

[0080] The historical fault information includes but is not limited to the real fault cause, the name of the replaced or repaired parts, and the real repair time. It should be noted that the real fault cause, the name of the replaced or repaired parts, and the real repair time in the historical fault information are all real information about the user's repair of the parts.

[0081] The historical delivery plan includes the processing order of the wafers, the time in the component, and the recipe. Among them, the recipe refers to the method followed to process the wafers, for example, in a certain component, a certain volume of a certain liquid needs to be added to process the wafer.

[0082] In some embodiments, after obtaining data sets of multiple devices of the same type, the data sets may be preprocessed to obtain information that can intuitively reflect whether the devices are working healthily. According to the logic of the semiconductor factory, the following rules may be adopted to preprocess the data sets:

[0083] (1) Since the components used in wafers corresponding to different recipes, i.e., the processing paths, may differ to a certain extent, the data of different recipes in the data set can be distinguished, i.e., the data can be marked.

[0084] (2) Since there will be certain differences in the data before and after the equipment is repaired or parts are replaced, the data before and after the repair or replacement of parts can be distinguished in the data set, that is, the data before and after the repair or replacement of parts can be marked.

[0085] (3) Since the probability of a minor fault occurring is usually higher than that of a major fault occurring, the historical time range of the required data can be determined based on the degree of component replacement to ensure the timeliness of the data. When the degree of component replacement is low, the historical time range of the required data can be a shorter time range. When the degree of component replacement is high, data of a longer time range is required.

[0086] In some embodiments, the historical time range of the data set also affects the accuracy and prediction speed of the model. When the data set is data within a shorter time range, the accuracy of its model is lower, but the prediction speed is faster. Therefore, based on the model accuracy and prediction speed, the period of the data contained in the data set, i.e., the historical time range, can be three months.

[0087] In some embodiments, after obtaining the data set, since the historical fault information contains the actual fault cause and the actual repair duration, and the actual fault cause and the actual repair duration contain text and time information, the actual fault cause and the actual repair duration can be encoded.

[0088] In one implementation scenario, since the causes of failures are complex and diverse, this application focuses on whether a component fails and the corresponding maintenance time. Therefore, in the embodiment of this application, only the actual maintenance time can be encoded. When encoding the actual maintenance time, it can be encoded according to the relationship list between the maintenance time and the corresponding code shown in Table 3. Among them, one code corresponds to one fault category.

[0089] Table 3 Relationship between maintenance time and corresponding codes

[0090] Maintenance time coding 1 day [1 0 0 0 0] 3 days [0 1 0 0 0] 5 days [0 0 1 0 0] More than 1 week and less than 1 month [0 0 0 1 0] More than 1 month [0 0 0 0 1]

[0091] In one implementation scenario, after the data set is preprocessed and encoded, the data set can be integrated using statistical learning methods, for example, statistical learning methods such as mean, variance, maximum value, minimum value, median, etc. can be used multiple times.

[0092] In one implementation scenario, after the data set is integrated, the data set can be reduced in dimension based on a feature selection algorithm. Specifically, the data set can be reduced in dimension using Fisher score, which is an effective method for feature selection.

[0093] In some embodiments, when a fault prediction model is constructed using a data set, since the data set contains data corresponding to multiple recipes, multiple neural network models can be constructed. Based on the data corresponding to any recipe, one of the neural network models is trained to finally obtain the fault prediction models corresponding to each recipe. At this time, a fault prediction model only predicts the data of a certain recipe.

[0094] In another implementation scenario, a comprehensive neural network model can be constructed. Since the data in the data set has been distinguished by different recipes, that is, the data has labels, the neural network model can be trained based on the data set to obtain a fault prediction model. At this time, the fault prediction model can predict data for multiple recipes.

[0095] It should also be noted that different types and models of equipment correspond to different fault prediction models.

[0096] In some embodiments, since the data in the data set can reflect the operating status of the equipment and the periodic information of the equipment, it needs to be updated at regular intervals to ensure the timeliness of the data. In one implementation scenario, after the user repairs the equipment or components, the time record of the equipment processing the wafer and the equipment sensor data, the real fault information and the optimized delivery plan can be obtained. The above information is used as historical data, so the above information can be updated to the data set to optimize the fault prediction model according to the data set, thereby ensuring the timeliness of the data set and improving the accuracy of fault prediction by the fault prediction model.

[0097] In summary, we can obtain data sets of multiple devices of the same type, build a neural network model based on the data sets, and use the fault prediction model to predict device or component failures. By updating the data set, the timeliness of the data set is guaranteed. At the same time, using the updated data to optimize the fault prediction model is conducive to improving the accuracy of the fault prediction model.

[0098] Figure 4 A flowchart of a method for reducing the dimension of a data set provided in an embodiment of the present application. In one or more embodiments of the present application, the process of reducing the dimension of a data set based on a feature selection algorithm may include: Figure 4 Steps shown:

[0099] S401: If the number of dimensions included in the data set is greater than a preset number threshold, obtain the eigenvalue of each dimension, and obtain the first N dimensions with the highest eigenvalues, where N is a positive integer greater than 1.

[0100] In some embodiments, when the dimension included in the data set is greater than a preset number threshold, it indicates that the data set is high-dimensional data. The dimensions included in the data set include but are not limited to temperature, humidity, pressure, and the like.

[0101] In an implementation scenario, the feature value may be a Fisher score, which is a supervised feature selection algorithm.

[0102] In another implementation scenario, if the number of repairs included in the data set is less than a preset quantity threshold, then there is no need to obtain the Fisher score of each dimension, and step S402 can be directly executed.

[0103] S402: Based on the data sets corresponding to the first N dimensions, a classification algorithm is used to classify whether each device is faulty, and a score for each category is obtained.

[0104] In some embodiments, the classification algorithm may use a random forest, which is a classifier that uses multiple trees to train and predict samples, i.e., data sets.

[0105] In some embodiments, each category corresponds to a fault code, and the score of each category can be accuracy, precision, recall, etc. One or more of the above evaluation indicators are selected as the score of each category, which can be determined according to actual conditions, and this application does not impose any restrictions on this.

[0106] S403: Remove data of any dimension in the data set to obtain the score of each current category.

[0107] In some embodiments, to confirm whether the dimension is a valid dimension, the data of the dimension may be removed from the data set. The score of each current category is obtained based on the data set after the data of the dimension is removed. The score of each category corresponding to the data set before the data of the dimension is removed is compared to determine whether the dimension is a valid dimension.

[0108] S404: If the score of any category decreases, the dimension is taken as a valid dimension, and the data of the dimension is restored in the data set.

[0109] When the data of this dimension is removed, the score of any category decreases, indicating that there is a certain correlation between the data of this dimension and the category whose score changes. Therefore, this dimension can be used as a valid dimension. Therefore, it is necessary to restore the data of this dimension in the data set.

[0110] S405: If the scores of each category have not been reduced, the dimension is regarded as an invalid dimension, and the data of the dimension is not restored.

[0111] After removing the data of this dimension, the score of each category did not decrease, indicating that the data of this dimension is not associated with each category. Therefore, deleting the data of this dimension does not affect the validity of the data set, so there is no need to restore the data of this dimension in the data set.

[0112] In summary, this application performs dimensionality reduction processing on the data set based on the Fisher score and classification algorithm, and filters out invalid data in the data set, which is conducive to simplifying the training and prediction process of the fault prediction model, and further helps to improve the training efficiency and prediction efficiency of the fault prediction model.

[0113] Based on the above embodiments, a specific embodiment is provided to describe the entire process of wafer processing in detail in stages.

[0114] In one or more embodiments of the present application, the entire process of wafer processing can be divided into the following stages: data set acquisition stage, data integration stage, fault prediction stage, online cargo arrangement stage, fault alarm stage, optimized cargo transfer plan improvement stage, fault handling stage and data set update stage. The specific process of each stage can be referred to as shown below:

[0115] (I) Dataset acquisition stage

[0116] Obtain the historical time records, historical failure information, historical data of equipment sensors, and historical delivery plans of multiple devices of the same type for processing wafers to build a data set.

[0117] (II) Data Integration Phase

[0118] The principle followed in this stage is to integrate data that can reflect the status of the equipment, filter redundant data, and reasonably classify the data to obtain data that can intuitively reflect whether the equipment is working healthily or not. Among them, data refers to the data contained in the data set.

[0119] In some embodiments, the data of different recipes can be distinguished in the data set, the data before and after the repair or replacement of parts can be distinguished, and the historical time range of the required data set can be determined according to the degree of part replacement, so as to ensure the timeliness of the data. At the same time, since the historical fault information records in the data set contain the real repair time, the real repair time can be encoded. Subsequently, the above data can be integrated using statistical learning methods, and finally the integrated data can be processed by dimensionality reduction.

[0120] (III) Fault prediction stage

[0121] The fault prediction stage includes the training and prediction process of the fault prediction model. Based on the data set after the data integration stage, the neural network model is trained to obtain the fault prediction model. The fault prediction model is a classification model and can be classified using classification algorithms, such as maximum likelihood estimation.

[0122] The status data of each device among the multiple devices currently processing the wafer is obtained, and the status data is input into the fault prediction model, which can predict the fault information of the device. The status data includes but is not limited to the time record of the device processing the wafer, the device sensor data and the current initial delivery plan. The fault information includes but is not limited to the faulty component, the cause of the fault, the probability of the fault, the fault type and the time required to repair the faulty component.

[0123] (IV) Online arrangement of cargo transfer

[0124] Based on the fault information, the available equipment or components are updated and the initial delivery plan is optimized to avoid using faulty components to process wafers. Each device is controlled to process the wafers according to the optimized delivery plan, so that the throughput of each device is at a maximum value, which is conducive to improving chip processing efficiency.

[0125] The cargo transfer plan obtained after optimization based on the initial cargo transfer plan may be referred to as the first cargo transfer plan.

[0126] (V) Fault alarm stage

[0127] The fault information and the first cargo transfer plan are sent to the user via email. While the user is warned of the fault, the user determines whether the equipment or component is faulty, needs to be shut down for maintenance, and determines the time for shutting down for maintenance.

[0128] (VI) Optimized cargo transfer plan improvement stage

[0129] According to whether the equipment or components are faulty, whether they need to be shut down for maintenance, and the time required for maintenance after confirmation by the user, the optimization algorithm is continued to be used to optimize the first cargo transfer plan. The optimization algorithm at this stage uses the first cargo transfer plan as the initial value, iterates until convergence, and obtains the second cargo transfer plan.

[0130] The second cargo transfer scheme can be used as the final optimization scheme. On the one hand, the second cargo transfer scheme is an optimization scheme proposed based on the user's tracking of the current device operating status, which is closer to the current device operating status. On the other hand, the second cargo transfer scheme is a fine-tuning of the first cargo transfer scheme, so the convergence time is shorter and it is more flexible than the randomly initialized optimization scheme.

[0131] (VII) Troubleshooting stage

[0132] The fault information and the second cargo transfer plan can be sent to the user again for confirmation. After the user confirms, each device is controlled to transfer the goods according to the corresponding second cargo transfer plan, that is, the wafers are processed. At the same time, engineers can deal with the faulty equipment or components in a timely manner.

[0133] For example, if component a of device A fails, the user can remove component a. During the period when the user is repairing component a, device A can continue to process wafers that do not require component a.

[0134] In another implementation scenario, if component b of device A fails but cannot be removed, if device A is in operation, then after device A completes processing of the wafer currently being processed, device A is stopped from continuing to operate, and engineers can then repair component b of device A.

[0135] In another implementation scenario, if component c of device A is faulty, component c can be removed and a non-faulty component c can be installed on device A. At this time, the maintenance time of the faulty component c does not affect the continued operation of device A.

[0136] It should be noted that the second cargo transfer scheme is usually applied during the user maintenance period. When the user maintenance is completed, the second cargo transfer scheme can be optimized again to ensure a larger throughput of each device.

[0137] (VIII) Dataset Update Phase

[0138] After the user repairs the equipment or components, the status data of the equipment before and after the repair, the actual fault information, and the second cargo transfer plan are updated to the data set to optimize the fault prediction model, ensure the timeliness of the data set, and then ensure the accuracy of the fault prediction model.

[0139] In summary, through the above-mentioned data set acquisition stage, data integration stage, fault prediction stage, online cargo delivery arrangement stage, fault alarm stage, optimized cargo delivery plan improvement stage, fault handling stage and data set update stage, it is possible to monitor the operating status of the equipment, obtain component failures in time or in advance, and optimize the cargo delivery plan based on the fault information, that is, the throughput of each device is optimized, ensuring that each device runs at a larger throughput, thereby effectively improving the chip manufacturing efficiency.

[0140] Figure 5 A schematic diagram of a wafer processing device provided in an embodiment of the present application Figure 1 .like Figure 5 As shown, an embodiment of the present application provides a wafer processing device 500 , which may include an acquisition module 501 , a processing module 502 and a control module 503 .

[0141] The acquisition module 501 is used to acquire an initial cargo transfer plan for processing wafers, and based on multiple devices for processing wafers included in the initial cargo transfer plan, acquire status data of each device; wherein each device includes multiple components;

[0142] A processing module 502 is used to predict the fault information of each device based on the state data using a fault prediction model, where the fault information is used to indicate whether a fault will occur and the faulty component;

[0143] The processing module 502 is further used to optimize the initial cargo transfer plan according to the fault information to avoid using the faulty component to process the wafer;

[0144] The control module 503 is used to control each device to process the wafer according to the optimized cargo transfer plan.

[0145] The device of this embodiment can be used to perform the following steps: Figures 2 to 4 The implementation principle and technical effect of the method embodiment shown are similar and will not be repeated here.

[0146] Figure 6 A schematic diagram of a wafer processing device provided in an embodiment of the present application Figure 2 .like Figure 6 As shown, an embodiment of the present application provides a wafer processing device 600 including a processor 601 and a memory 602 , wherein the processor 601 and the memory 602 are connected via a bus 603 .

[0147] In the specific implementation process, the memory 602 stores codes, and the processor 601 runs the codes stored in the memory 602 to execute the method of the above method embodiment.

[0148] The specific implementation process of the processor 601 can be found in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be repeated here.

[0149] In the above Figure 6 In the illustrated embodiment, it should be understood that the processor 601 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the invention may be directly implemented as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0150] The memory 602 may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk storage.

[0151] The bus 603 may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus 603 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus 603 in the drawings of the present application is not limited to only one bus or one type of bus.

[0152] An embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the method of the above method embodiment.

[0153] The above-mentioned computer readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0154] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0155] An embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, the method provided in any embodiment of the present application is implemented.

[0156] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0157] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A wafer processing method, characterized in that: include: Obtaining an initial cargo transfer plan for processing wafers, and based on multiple devices for processing wafers included in the initial cargo transfer plan, obtaining status data of each device; wherein each device includes multiple components; Based on the state data, using a fault prediction model to predict fault information of each device, the fault information is used to predict whether the device will fail and the faulty component; Optimizing the initial cargo transfer plan according to the fault information to avoid using the faulty component to process the wafer; According to the optimized cargo transfer plan, each device is controlled to process the wafers.

2. The method according to claim 1, characterized in that Before controlling each device to process the wafer according to the optimized cargo transfer plan, the method further includes: Repeat the following steps until it is manually confirmed that the fault information of each device does not need to be modified, and control each device to process the wafer according to the optimized delivery plan: Display the fault information of each device and the optimized cargo transfer plan for manual confirmation; If it is manually confirmed that the fault information of any device needs to be modified, the optimized cargo transfer plan is optimized according to the modified fault information.

3. The method according to claim 1, characterized in that The fault information also includes the time required to repair the faulty component; Accordingly, the initial cargo transfer plan is optimized according to the fault information, including: The initial cargo transfer plan is optimized according to the faulty parts and the time required to repair the faulty parts.

4. The method according to claim 3, characterized in that The method further comprises: A data set of multiple devices of the same type is obtained, and the fault prediction model is constructed according to the data set.

5. The method according to claim 4, characterized in that After obtaining the data sets of multiple devices of the same type, the method further includes: Based on the feature selection algorithm, the dimension reduction process is performed on the data set.

6. The method according to claim 5, characterized in that The step of performing dimensionality reduction processing on the data set based on a feature selection algorithm includes: If the number of dimensions contained in the data set is greater than a preset number threshold, obtain the eigenvalue of each dimension, and obtain the top N dimensions with the highest eigenvalues, where N is a positive integer greater than 1; Based on the data sets corresponding to the first N dimensions, a classification algorithm is used to classify whether each device is faulty or not, and the score of each category is obtained; Remove data of any dimension from the data set to obtain the score of each current category; If the score of any category decreases, the dimension is taken as a valid dimension, and the data of the dimension is restored in the data set; If the scores of each category have not been reduced, the dimension is regarded as an invalid dimension, and the data of the dimension is not restored.

7. A wafer processing device, characterized in that: include: An acquisition module, used to acquire an initial cargo transfer plan for processing wafers, and based on multiple devices for processing wafers included in the initial cargo transfer plan, acquire status data of each device; wherein each device includes multiple components; A processing module, configured to predict the fault information of each device using a fault prediction model based on the state data, wherein the fault information is used to predict whether the device will fail and the faulty component; The processing module is further used to optimize the initial cargo transfer plan according to the fault information to avoid using the faulty component to process the wafer; The control module is used to control various devices to process the wafers according to the optimized cargo transfer plan.

8. A wafer processing device, comprising: A processor and a memory, wherein the memory stores codes, and the processor runs the codes stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

10. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images