Equipment maintenance method and system, storage medium and electronic equipment

Abstract

The invention provides an equipment maintenance method, which comprises the following steps: acquiring tracking signals, and calculating statistical characteristics of each tracking signal; the statistical features are divided into features before component replacement and features after component replacement; sequencing the features before component replacement and the features after component replacement by using a Fisher criterion, and determining feature subsets with the same features; establishing a health model according to the feature subset; substituting the component application parameter of the tracking signal, and outputting a health index; the health index is used to indicate the device life and the accuracy of the device life to guide device maintenance. According to the method, the health prediction can be performed on the corresponding equipment component in a targeted manner, so that the cost increase caused by accidental or unplanned shutdown is eliminated, the production efficiency and quality of products are improved, and the consumption and purchase quantity of the component can be accurately estimated based on the health index calculated by the method. The invention also provides an equipment maintenance system, a storage medium and electronic equipment, which have the above beneficial effects.

Description

Technical Field

[0001] This application relates to the field of industrial equipment, and particularly to a device maintenance method, system, storage medium, and electronic device. Background Art

[0002] The operation and management of the semiconductor manufacturing industry have always relied on time-based maintenance strategies to maintain product production efficiency and quality. However, the cost of time-based maintenance strategies is high. Especially when the interval between component replacements is kept very short to reduce the mean time between failures rate, there may be problems such as unexpected downtime, which further increases the equipment maintenance cost. Summary of the Invention

[0003] The purpose of this application is to provide a device maintenance method, system, storage medium, and electronic device, which can reduce the equipment maintenance cost by calculating the health index of the device.

[0004] To solve the above technical problems, this application provides a device maintenance method, and the specific technical solution is as follows:

[0005] Obtain a tracking signal, determine the statistical characteristics of each tracking signal, and extract actual characteristics from the tracking signal according to the statistical characteristics;

[0006] Divide the actual characteristics into pre-component processing characteristics and post-component processing characteristics;

[0007] Use the Fisher criterion to sort the pre-component processing characteristics and the post-component processing characteristics, and determine a feature subset with the same features;

[0008] Use the feature subset as training data for iterative training to generate a health model;

[0009] Substitute the component application parameters of the tracking signal into the health model to output a health index; the health index is used to indicate the device life and the accuracy of the device life to guide device maintenance.

[0010] Optionally, after obtaining the tracking signal, before determining the statistical characteristics of each tracking signal and extracting actual characteristics from the tracking signal according to the statistical characteristics, it further includes:

[0011] Preprocess the tracking signal to determine the recipe step corresponding to the tracking signal.

[0012] Optionally, determining the statistical characteristics of each tracking signal and extracting actual characteristics from the tracking signal according to the statistical characteristics includes:

[0013] Determine the statistical features of each of the tracking signals according to the recipe steps, and extract the actual features from the tracking signals according to the statistical features; the statistical features include at least one or any combination of mean value, standard deviation, maximum value, minimum value, limit, duration of the step, maximum duration of the step, minimum duration of the step, and overflow value.

[0014] Optionally, after sorting the features of the component before processing and the features of the component after processing using the Fisher criterion, it further includes:

[0015] Delete the features with the highest rankings in the first ranking corresponding to the features of the component before replacement and the second ranking corresponding to the features of the component after replacement, respectively, to obtain the first remaining feature ranking and the second remaining feature ranking.

[0016] Optionally, the determining of the feature subset with the same features includes:

[0017] Determine the feature rankings of each process chamber, and select the feature subset with the same characteristics; the characteristics are the attribute parameters of the process chamber.

[0018] Optionally, using the feature subset as training data for iterative training to generate a health model includes:

[0019] Use the feature subset as training data for iterative calculation; during the iterative calculation process, reduce the feature dimension of the training data and retain the topological structure of the training data;

[0020] Use an unsupervised learning algorithm to construct a health model based on the training data; the health model is a regression prediction model.

[0021] Optionally, substituting the component application parameters of the tracking signal into the health model and outputting a health index includes:

[0022] According to the component application parameters corresponding to the tracking signal, respectively determine the first linear model regression equation before component cleaning and the second linear model regression equation after component cleaning;

[0023] Use the prediction interval to determine the upper and lower limits of the first linear model regression equation and the second linear model regression equation; the upper and lower limits are the health prediction limits of the device, which are used to feedback the health index of the device.

[0024] This application also provides a device maintenance system, including:

[0025] A signal acquisition module, configured to acquire tracking signals, determine the statistical features of each of the tracking signals, and extract actual features from the tracking signals according to the statistical features;

[0026] A feature classification module, configured to classify the actual features into pre-component processing features and post-component processing features;

[0027] A feature sorting module, configured to sort the pre-component processing features and the post-component processing features by using the Fisher criterion to determine a feature subset with the same features;

[0028] A model construction module, configured to iteratively train by using the feature subset as training data to generate a health model;

[0029] A device health prediction module, configured to substitute the component application parameters of the tracking signal into the health model to output a health index; the health index is used to indicate the device life and the accuracy of the device life to guide device maintenance.

[0030] The present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described above are implemented.

[0031] The present application further provides an electronic device, including a memory and a processor. A computer program is stored in the memory. When the processor calls the computer program in the memory, the steps of the method described above are implemented.

[0032] The present application provides a device maintenance method, including: obtaining a tracking signal, determining the statistical features of each tracking signal, and extracting actual features from the tracking signal according to the statistical features; classifying the actual features into pre-component processing features and post-component processing features; sorting the pre-component processing features and the post-component processing features by using the Fisher criterion to determine a feature subset with the same features; iteratively training by using the feature subset as training data to generate a health model; substituting the component application parameters of the tracking signal into the health model to output a health index; the health index is used to indicate the device life and the accuracy of the device life to guide device maintenance.

[0033] The method proposed by the present application includes steps such as data acquisition, feature classification, feature selection, and prediction model analysis. Each step can be matched with the data features of the application component. It can perform health prediction on the device components in a targeted manner to eliminate the increased costs caused by accidental or unplanned downtime, improve the production efficiency and quality of products, and can also accurately budget the consumption and procurement volume of components based on the health index calculated by the present application.

[0034] The present application further provides a device maintenance system, a storage medium, and an electronic device, which have the above beneficial effects and will not be elaborated here. Description of the Drawings

[0035] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained according to the provided accompanying drawings.

[0036] Figure 1 Flowchart of a device maintenance method provided by an embodiment of the present application;

[0037] Figure 2 Structural schematic diagram of a device maintenance system provided by an embodiment of the present application;

[0038] Figure 3 Structural schematic diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of them. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0040] Please refer to Figure 1 , Figure 1 Flowchart of a device maintenance method provided by an embodiment of the present application. The method includes:

[0041] S101: Obtain a tracking signal, determine the statistical characteristics of each tracking signal, and extract actual characteristics from the tracking signal according to the statistical characteristics;

[0042] S102: Divide the actual characteristics into pre-component processing characteristics and post-component processing characteristics;

[0043] S103: Use the Fisher criterion to sort the pre-component processing characteristics and the post-component processing characteristics, and determine a feature subset with the same features;

[0044] S104: Use the feature subset as training data for iterative training to generate a health model;

[0045] S105: Substitute the component application parameters of the tracking signal into the health model to output a health index; the health index is used to indicate the device life and the accuracy of the device life to guide device maintenance.

[0046] The tracking signal mainly includes tracking data and measurement data in the device, etc., and can also be combined with other important signals. The tracking data mainly refers to the maintenance information of the device, including but not limited to the process chamber used, the recipe, the number of steps in the recipe, and other maintenance information, etc. The process chamber refers to the device or structure used to accommodate and protect the product. It is usually made of metal, plastic or other materials, and has a specific shape and size to ensure the stability and accuracy of the product during the manufacturing process. The recipe is the manufacturing or preparation plan. The number of steps in the recipe refers to the number of operation steps required in the manufacturing process. Each step usually includes specific operation, temperature, time and other parameters to ensure the stability and accuracy of the product during the manufacturing process. The maintenance information refers to the guidelines for the maintenance and servicing activities required during the manufacturing process. For example, it includes operations such as cleaning, lubrication, inspection, etc. to ensure the normal operation and service life of the device. Other maintenance information includes but is not limited to when to replace components or when to clean, etc. The measurement data is the data obtained after statistically processing the tracking data, such as the average value, maximum value, minimum value, etc., or includes the duration of the step, the maximum duration of the step, the minimum duration of the step and the overflow value, etc., where the overflow value refers to the signal amount exceeding the threshold part.

[0047] In a feasible implementation manner, after obtaining the tracking signal, the tracking signal can also be preprocessed to determine the recipe step corresponding to the tracking signal. Usually, all potentially relevant features are extracted from the tracking signal, and then selected after performing an association evaluation with components in a comprehensive list containing all potential features. In this embodiment, the statistical features of each trace signal of the recipe step are utilized. If it is considered that the number of tracking variables may exceed 100 and the recipe may involve more than 15 steps, the number of features generated by calculating the statistical data of each tracking signal and step may be very large, but only a small part of them are important features.

[0048] Specifically, when calculating the statistical features, the statistical features of each tracking signal can be determined according to the recipe step, and the actual features can be extracted from the tracking signal according to the statistical features. The statistical features include at least one or any combination of the average value, standard deviation, maximum value, minimum value, limit, duration of the step, maximum duration of the step, minimum duration of the step and overflow value. If the device includes an electrostatic chuck, taking the electrostatic chuck as an example, the influence of the above statistical features on the electrostatic chuck is evaluated, and 4 signals with greater influence on the electrostatic chuck are retained, and the four statistical features of the average value, standard deviation, maximum value and minimum value are retained. Assuming there are 30 tracking signals, a recipe with 18 steps and 4 statistical features, a total of 30×18×4 = 2160 actual features are extracted from the tracking signals, and each actual feature contains the information of the tracking signal, the recipe step and the statistical feature.

[0049] Step S102 aims to perform feature classification of actual features, while step S103 aims to perform feature ranking, both serving the establishment of the health model in subsequent steps.

[0050] In step S102, considering that there are maintenance records and data before and after component replacement, these two classes can be used to sort the actual features using a filtering method. Taking the electrostatic chuck as an example, the data after replacing the electrostatic chuck will represent the new and healthy component state, and the data before replacing the electrostatic chuck will represent the degraded component health state.

[0051] When performing the ranking of actual features, the Fisher criterion is used for ranking. The Fisher criterion selects a comprehensive discriminant variable or projection direction such that the points of each class are as concentrated as possible, while the classes are as separated as possible. That is, it achieves the minimum within-class scatter and the maximum between-class scatter. In this embodiment, the Fisher criterion is used to rank the features of the component before processing and the features of the component after processing to determine a feature subset with the same features.

[0052] The ranking basis of this feature subset is:

[0053] θ(f1)≥θ(f2)≥...θ(f n-1 )≥θ(f n ) (1)

[0054] Specifically, they are ranked according to the ability of each feature to distinguish between the two categories of the component before processing and the component after processing. Features with larger Fisher values are ranked higher than features with lower Fisher values.

[0055] Among them, θ(·) is a standard function that measures the ability of a specific feature to distinguish between two classes. Equation (2) provides an expression for the discriminant value of the Fisher criterion, where i and j are class labels, and are the means of the k th features, and are the variances of the k th features, and the values are calculated and sorted according to equation (2).

[0056]

[0057] The actual features are divided into two categories: the features of the component before processing and the features of the component after processing. The Fisher value of each actual feature between the two categories of the component before processing and the component after processing is calculated. The larger the Fisher value, the stronger the ability of the actual feature to distinguish between these two categories.

[0058] Taking an electrostatic chuck as an example, after replacing the electrostatic chuck, a cleaning process usually needs to be performed. If the characteristic changes before and after the component processing are determined at this time, the Fisher value obtained is likely to be caused by significant changes in the corresponding characteristics before and after cleaning. However, in fact, if the cleaning is not performed, the ability of this characteristic to distinguish between before and after component replacement is not significant, that is, it is not actually a suitable characteristic for judging the health index of the component. In this regard, the solution is to perform a Fisher value sorting on the data before and after the electrostatic chuck replacement, and perform another Fisher value sorting on the data before and after the cleaning. For the sake of easy understanding and expression, they are respectively called the first sorting and the second sorting.

[0059] Thereafter, the features with the highest rankings in the first sorting corresponding to the actual features before and after component replacement and the second sorting corresponding to the actual features before and after component cleaning can be deleted respectively, to obtain the first remaining feature sorting and the second remaining feature sorting, so as to eliminate irrelevant parameters and exclude features that may overly affect the sorting due to specific reasons, in order to improve the calculation accuracy.

[0060] Compare the feature rankings of each process chamber, and select the feature subset with consistent features in the first remaining feature sorting and the second remaining feature sorting. It should be noted that the equipment may include several process chambers, which have the same structure but different functions. When selecting the feature subset, the main consideration is the consistent characteristics. The characteristics are the attribute parameters of the process chamber, including temperature, air pressure, etc. For example, the same temperature or the same gas can be determined as the feature subset with the same features. For each process chamber, calculate the Fisher value of its features before and after cleaning, and sort them according to the size of the Fisher value. Compare the feature rankings between different process chambers to determine which actual features have the same level of importance in different process chambers. Select the features with high Fisher values in different process chambers to form a feature subset with the same features. The features in this feature subset have high reliability for describing and distinguishing the performance changes of different process chambers.

[0061] Thereafter, use the feature subset as training data for iterative training to generate a health model. Specifically, use the feature subset as training data for iterative calculation. During the iterative calculation process, the purpose is to reduce the feature dimension of the training data and retain the topological structure of the training data. Specifically, an unsupervised learning algorithm can be used to construct a health model based on the training data; the health model is a regression prediction model. Here, no limitation is imposed on which unsupervised learning algorithm to use. In a feasible implementation manner, a self-organizing map health assessment algorithm can be used. The baseline data in the self-organizing map health assessment algorithm is used to train the self-organizing map, and the regions of acceptable component behavior and health status are defined.

[0062] Construct a health model based on the training data using an unsupervised learning algorithm; when the health model is a regression prediction model, it can include two processes. First, apply the unsupervised learning algorithm to evaluate the deviation degree of the current state of the component from the baseline state. Thereafter, call the regression prediction model according to this deviation degree to construct the health model. This deviation degree determines whether the regression prediction model uses a linear model or a non-linear model. When the deviation degree has a linear pattern, a linear regression prediction model can be used and methods such as the least squares method are used for fitting. When the deviation degree does not have a linear pattern, polynomial regression, exponential regression, logarithmic regression, etc. can be used to fit to obtain the health model.

[0063] When applying the self-organizing health assessment algorithm as the unsupervised learning algorithm, only the baseline data needs to be used to train the algorithm, which can highlight the key aspects of the features. Self-organizing mapping is a neural network for clustering data. When using this algorithm for health assessment, the baseline data is used to train the self-organizing mapping and define the regions of acceptable component behavior and health status. Then, when the data from the currently monitored component is presented to the trained health, the best matching unit is found between the data vector from the current monitoring system and the trained map. Finally, the distance between the best matching unit and the data vector from the monitored system is defined as the minimum quantization error (MQE), and the minimum quantization error is used to evaluate the deviation degree of the current condition from the baseline state.

[0064] In other words, the baseline data represents the system behavior in the normal or healthy state. Taking an electrostatic chuck as an example, its baseline data may include performance parameters, temperature, voltage, etc. under various operating conditions. These data provide a "health" reference framework for the self-organizing mapping algorithm.

[0065] Self-organizing mapping is an unsupervised neural network that can learn the topological structure of the data. Training the self-organizing mapping with the baseline data means that the network will learn and map out the typical behavior patterns of healthy components. When the data from the currently monitored component is input into the trained self-organizing mapping, the algorithm will search for the best matching unit. This unit is formed during the training process and represents the health state closest to the input data.

[0066] The minimum quantization error represents the difference between the input data and the best matching unit in the self-organizing mapping. A high minimum quantization error value may mean that the currently monitored component deviates significantly from the healthy state, while a low minimum quantization error value indicates that the component behavior is close to the healthy state represented by the baseline data. Therefore, the deviation degree of the current state of the process chamber from the baseline state can be evaluated through the minimum quantization error.

[0067] Since performing cleaning may affect process chamber performance parameters, even if those changes are slight, it is important to ensure that these changes do not mislead the health assessment algorithm into falsely believing that the health of the component has improved or deteriorated.

[0068] In order to eliminate the impact of cleaning on health assessment, correction factors can be introduced. These factors can be adjusted according to the performance changes before and after cleaning, ensuring that the minimum quantitative error or other health indicators can accurately reflect the true health status of the process components without being affected by maintenance operations.

[0069] Taking the electrostatic chuck as an example, for the health monitoring and life prediction of the electrostatic chuck, the key parameters of the health model corresponding to the electrostatic chuck will also change slightly after cleaning. Despite the change in function, the health status of the electrostatic chuck after cleaning predicted by the model should not change significantly. Therefore, a correction factor can also be used to ensure that the health status before and after cleaning remains constant.

[0070] It can be seen that the health model constructed in this embodiment is adapted to the corresponding device component. After obtaining the health model, the component application parameters of the tracking signal can be further substituted to output the health index. It is easy to understand that the component application parameters of different device components are different. The component application parameters refer to the usage data corresponding to the component, mainly the working time of the component and other data.

[0071] Taking the electrostatic chuck as an example, the amplification time of the transformer coupler at the top of its cavity directly affects its working life. Therefore, the health index of the electrostatic chuck presents a nonlinear curve pattern before the first cleaning. The nonlinear regression model described by formula (3) can be used to correlate the time in the amplification time x of the transformer coupler with the health index. After the first cleaning, the health of the electrostatic chuck is more directly affected, and the linear model of the health trend can be used at this time, and the health index shown in formula (4) The linear regression model is applicable to the period from the first cleaning to the replacement of the electrostatic chuck. As a supplement, after the second cleaning, a new set of linear model regression coefficients is used because the conditions before and after cleaning are slightly different, that is, b1 and b2 in formula (4) are the new regression coefficients obtained by updating formula (3) after one cleaning.

[0072]

[0073]

[0074] For any prediction of the health status of a component or system, there is a certain degree of uncertainty. In this embodiment, the lower and upper limits of the health prediction estimate can be further obtained using a prediction interval. The lower and upper limits of the prediction interval can be calculated using Equation (5), where n is the number of samples, m is the number of terms in the regression model (excluding the constant term), is the distribution of t and the degrees of freedom at a specified α level, is the average value of the time variable.

[0075] The additional term S in the prediction interval expression xx includes terms and the root mean squared error (RMSE). Here, the α level refers to the α significance level, which represents the probability of making a mistake when estimating that the population parameter falls within a certain interval; the significance level is used to reflect the likelihood of a small-probability event occurring in a single trial. The t-distribution contains only one parameter, the degrees of freedom, and the t-distribution is symmetric about the Y-axis, so the mean of the t-distribution is 0. From the graph of the probability density function of the t-distribution, it can be seen that when the degrees of freedom are small, the t-distribution is milder compared to the normal distribution; when the degrees of freedom increase, the t-distribution approaches the normal distribution.

[0076]

[0077] In other embodiments of the present application, if there is more variance in the regression fit, this will result in a larger root mean squared error and larger bounds of the prediction interval. Additionally, if predicting further into the future, the bounds of the prediction interval will also be larger. Because compared to short-term predictions, there is more uncertainty in making long-term predictions about the health status of the monitored components. Therefore, the predicted health index can also be adjusted or compensated based on the root mean squared error of the prediction interval to improve the prediction accuracy.

[0078] The method proposed in the embodiments of the present application includes steps such as data acquisition, feature partitioning, feature selection, and prediction model analysis. Each step can be matched with the data characteristics of the application component. It can specifically perform health predictions for equipment components to eliminate the increased costs caused by unexpected or unplanned downtime, improve the production efficiency and quality of products, and can also accurately estimate the consumption and procurement quantities of components based on the health index calculated in the present application. That is, after predicting the component life with reference to the health index, the component consumption status can be determined, which is conducive to achieving precise procurement of components.

[0079] Next, an equipment maintenance system provided by the embodiments of the present application will be introduced. The equipment maintenance system described below can be correspondingly referred to the equipment maintenance method described above.

[0080] See Figure 2 , Figure 2A schematic diagram of a device maintenance structure provided by an embodiment of the present application. The present application also provides a device maintenance system, including:

[0081] A signal acquisition module, configured to acquire tracking signals, determine statistical features of each of the tracking signals, and extract actual features from the tracking signals according to the statistical features;

[0082] A feature classification module, configured to classify the actual features into features before component processing and features after component processing;

[0083] A feature sorting module, configured to sort the features before component processing and the features after component processing using the Fisher criterion to determine a feature subset with the same features;

[0084] A model construction module, configured to perform iterative training using the feature subset as training data to generate a health model;

[0085] A device health prediction module, configured to substitute component application parameters of the tracking signals into the health model and output a health index; the health index is used to indicate the device life and the accuracy of the device life to guide device maintenance.

[0086] Based on the above embodiments, as a preferred embodiment, it further includes:

[0087] A signal preprocessing module, configured to preprocess the tracking signals to determine the recipe steps corresponding to the tracking signals.

[0088] Based on the above embodiments, as a preferred embodiment, the signal acquisition module is a module configured to determine statistical features of each of the tracking signals according to the recipe steps and extract actual features from the tracking signals according to the statistical features; the statistical features include at least one or any combination of average value, standard deviation, maximum value, minimum value, limit, duration of the step, maximum duration of the step, minimum duration of the step, and overflow value.

[0089] Based on the above embodiments, as a preferred embodiment, it further includes:

[0090] A feature deletion module, configured to delete the features with the highest rankings in the first sorting corresponding to the features before component replacement and the second sorting corresponding to the features after component replacement respectively to obtain a first remaining feature sorting and a second remaining feature sorting.

[0091] Based on the above embodiments, as a preferred embodiment, the feature sorting module includes:

[0092] A feature subset determination unit, configured to determine the feature rankings of each process chamber and select a feature subset with the same characteristics; the characteristics are attribute parameters of the process chamber.

[0093] Based on the above embodiments, as a preferred embodiment, the model construction module is a module for establishing a health model based on the self-organizing mapping health assessment algorithm using the feature subset; the baseline data in the self-organizing mapping health assessment algorithm is used to train the self-organizing mapping and define the regions of acceptable component behavior and health status.

[0094] Based on the above embodiments, as a preferred embodiment, the device health prediction module includes:

[0095] An equation establishment unit for respectively determining a first linear model regression equation before component cleaning and a second linear model regression equation after component cleaning according to the component application parameters corresponding to the tracking signal;

[0096] A health prediction unit for using the prediction interval to determine the upper and lower limits of the first linear model regression equation and the second linear model regression equation; the upper and lower limits are the health prediction limits of the device.

[0097] The present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed, the steps provided in the above embodiments can be implemented. The storage medium may include: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc that can store program codes.

[0098] The present application also provides an electronic device, which may include a memory and a processor. When the processor calls the computer program stored in the memory, the steps provided in the above embodiments can be implemented. Of course, the electronic device may also include various network interfaces, power supplies, and other components.

[0099] The present application also provides an electronic device, see Figure 3 , a structural diagram of an electronic device provided by an embodiment of the present application, as shown in Figure 3 , may include a processor 1410 and a memory 1420.

[0100] Among them, the processor 1410 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 1410 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 1410 may also include a main processor and a coprocessor. The main processor is a processor used to process data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 1410 may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 1410 may further include an AI (Artificial Intelligence) processor, and the AI processor is used to process computational operations related to machine learning.

[0101] The memory 1420 may include one or more computer-readable storage media, and the computer-readable storage media may be non-transitory. The memory 1420 may further include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash storage devices. In this embodiment, the memory 1420 is at least used to store the following computer program 1421. After the computer program is loaded and executed by the processor 1410, it can implement the relevant steps in the network traffic collection method of the electronic device application executed by the electronic device side disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 1420 may further include an operating system 1422 and data 1423, etc., and the storage method may be transient storage or permanent storage. Among them, the operating system 1422 may include Windows, Linux, Android, etc.

[0102] In some embodiments, the electronic device may further include a display screen 1430, an input / output interface 1440, a communication interface 1450, a sensor 1460, a power supply 1470, and a communication bus 1480.

[0103] Of course, Figure 3 The structure of the electronic device shown does not constitute a limitation on the electronic device in the embodiments of the present application. In practical applications, the electronic device may include more or fewer components than Figure 3 shown, or combine certain components.

[0104] The various embodiments in the specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the system provided in the embodiment, since it corresponds to the method provided in the embodiment, the description is relatively simple. For the relevant parts, reference can be made to the description in the method section.

[0105] In this article, specific examples are used to elaborate on the principles and implementation manners of the present application. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present application. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the principle of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the protection scope of the claims of the present application.

[0106] It should also be noted that in this specification, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.

Claims

1. A device maintenance method, characterized in that, Including: Obtain a tracking signal, determine the statistical characteristics of each of the tracking signals, and extract actual characteristics from the tracking signals according to the statistical characteristics; Divide the actual characteristics into characteristics before component processing and characteristics after component processing; Use the Fisher criterion to sort the characteristics before component processing and the characteristics after component processing, and determine a subset of characteristics with the same characteristics; Use the subset of characteristics as training data for iterative training to generate a health model; Substitute the component application parameters of the tracking signal into the health model to output a health index; the health index is used to indicate the device life and the accuracy of the device life to guide device maintenance.

2. The device maintenance method according to claim 1, wherein After obtaining the tracking signal and before determining the statistical characteristics of each of the tracking signals and extracting actual characteristics from the tracking signals according to the statistical characteristics, it further includes: Preprocess the tracking signal to determine the recipe step corresponding to the tracking signal.

3. The device maintenance method according to claim 2, characterized in that The determining the statistical characteristics of each of the tracking signals and extracting actual characteristics from the tracking signals according to the statistical characteristics includes: Determine the statistical characteristics of each of the tracking signals according to the recipe step, and extract actual characteristics from the tracking signals according to the statistical characteristics; the statistical characteristics include at least one or any combination of average value, standard deviation, maximum value, minimum value, limit, duration of the step, maximum duration of the step, minimum duration of the step, and overflow value.

4. The equipment maintenance method according to claim 1, wherein, After sorting the characteristics before component processing and the characteristics after component processing using the Fisher criterion, it further includes: Delete the characteristics with the highest rankings in the first ranking corresponding to the characteristics before component replacement and the second ranking corresponding to the characteristics after component replacement respectively to obtain a first remaining feature ranking and a second remaining feature ranking.

5. The device maintenance method according to claim 4, characterized in that The determining the subset of characteristics with the same characteristics includes: Determine the feature rankings of each process chamber, and select a subset of characteristics with the same characteristics; the characteristics are the attribute parameters of the process chamber.

6. The device maintenance method according to claim 1, characterized in that, Using the subset of characteristics as training data for iterative training to generate a health model includes: Use the subset of characteristics as training data for iterative calculation; during the iterative calculation process, reduce the feature dimension of the training data and retain the topological structure of the training data; Use an unsupervised learning algorithm to construct a health model based on the training data; the health model is a regression prediction model.

7. The device maintenance method according to claim 1, wherein, Substituting the component application parameters of the tracking signal into the health model to output a health index includes: According to the component application parameters corresponding to the tracking signal, respectively determine a first linear model regression equation before component cleaning and a second linear model regression equation after component cleaning; Use a prediction interval to determine the upper and lower limits of the first linear model regression equation and the second linear model regression equation; the upper and lower limits are the health prediction limits of the device and are used to feedback the health index of the device.

8. An equipment maintenance system, characterized in that, Including: A signal acquisition module, configured to obtain a tracking signal, determine the statistical characteristics of each of the tracking signals, and extract actual characteristics from the tracking signals according to the statistical characteristics; A feature classification module, configured to divide the actual characteristics into characteristics before component processing and characteristics after component processing; A feature sorting module, configured to sort the features before component processing and the features after component processing by using the Fisher criterion, and determine a feature subset with the same features; A model construction module, configured to iteratively train using the feature subset as training data to generate a health model; A device health prediction module, configured to substitute the component application parameters of the tracking signal into the health model and output a health index; the health index is used to indicate the device life and the accuracy of the device life to guide device maintenance.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-7.

10. An electronic device, characterized in that, It includes a memory and a processor. A computer program is stored in the memory. When the processor calls the computer program in the memory, it implements the steps of the method according to any one of claims 1-7.

Images