Semiconductor device detection method, processor and semiconductor device

Abstract

The present application provides a semiconductor device detection method, a processor and a semiconductor device. The method includes: the processor obtains multiple historical data, the multiple historical data at least includes the size parameters, process parameters and total number of processed wafers of the historical wafer and the wear degree data of the equipment; the processor constructs a model of the wear degree of the equipment according to the multiple historical data, and uses the model of the wear degree of the equipment and the size parameters and process parameters of the target wafer to obtain the degree of damage of the equipment used to make the target wafer. In this method, when the processor obtains the historical data, there is no need to manually set standards or labels for the historical data, so the historical data obtained by the processor can truly reflect the state of the semiconductor device, and then the processor constructs a model of the wear degree of the equipment according to the historical data. According to the model of the wear degree of the equipment and the target data, the degree of damage of the equipment used to make the target wafer can be accurately predicted.

Description

Technical Field

[0001] The present application relates to the semiconductor field, and more specifically, to a semiconductor device detection method, a processor, a computer-readable storage medium, and a semiconductor device. Background Art

[0002] In existing semiconductor manufacturing systems, most semiconductor equipment status detection methods are based on statistical process control (SPC) and supervised machine learning (ML) models. SPC detects the status of semiconductor equipment by setting standards for sensors and key process parameters, and ML models use neural networks and deep learning techniques to identify the status of semiconductor equipment. The former method lacks the ability to utilize large amounts of data, and the resulting semiconductor equipment status is inaccurate, while the latter method is highly dependent on labeled data and requires manual labeling, which will filter out a large amount of useful data and cannot truly reflect the status of semiconductor equipment. Therefore, the status monitoring methods in the prior art cannot accurately predict the degree of wear of semiconductor equipment.

[0003] Therefore, there is an urgent need for a method that can accurately predict the degree of wear of semiconductor devices.

[0004] The above information disclosed in the background technology section is only used to enhance the understanding of the background technology of the technology described in this article. Therefore, the background technology may contain certain information that does not form the prior art known in this country for those skilled in the art. Summary of the invention

[0005] The main purpose of the present application is to provide a semiconductor device detection method, a processor, a computer-readable storage medium and a semiconductor device to solve the problem in the prior art that there is a lack of methods that can accurately predict the degree of wear of semiconductor devices.

[0006] According to one aspect of an embodiment of the present invention, a method for detecting semiconductor equipment is provided, comprising: a processor acquires a plurality of historical data, the plurality of historical data at least comprising a plurality of first historical data, a plurality of second historical data and a plurality of third historical data, wherein the first historical data comprises size parameters of historical wafers and process parameters when manufacturing the historical wafers, the second historical data comprises a total number of processed wafers when manufacturing each of the historical wafers, and the third historical data is data on the degree of equipment wear detected, the third historical data corresponding one-to-one to the first historical data; the processor constructs a model of the degree of equipment wear according to the plurality of historical data, and determines the degree of damage of equipment for manufacturing a target wafer by using the model of the degree of equipment wear and target data, the target data comprising the size parameters of the target wafer and process parameters when manufacturing the target wafer.

[0007] Optionally, the processor constructs a model of the degree of equipment wear based on the multiple historical data, including: performing feature extraction on the multiple first historical data to obtain feature data; performing at least correlation analysis on the feature data and the second historical data to obtain training data; using the training data and the corresponding third historical data to train an initial model, updating initial parameters in the initial model to target parameters, and obtaining a model of the degree of equipment wear, wherein the initial model is the product of the Hotelling-T square value calculation formula and the initial parameters.

[0008] Optionally, feature extraction is performed on the plurality of first historical data to obtain feature data, including: extracting time domain feature data and frequency domain feature data of the same parameter in the plurality of first historical data; and normalizing the time domain feature data and the frequency domain feature data to obtain the feature data.

[0009] Optionally, at least a correlation analysis is performed on the feature data and the second historical data to obtain training data, including: performing a correlation analysis on each feature data and the corresponding second historical data to obtain multiple correlation coefficients; when the correlation coefficient is less than 0, calculating the difference between 1 and the feature data corresponding to the correlation coefficient less than 0 to obtain a corrected correlation coefficient; filtering out the feature data corresponding to the corrected correlation coefficient greater than a first threshold to obtain the training data.

[0010] Optionally, at least a correlation analysis is performed on the feature data and the second historical data to obtain training data, including: performing at least a correlation analysis on the feature data and the second historical data, filtering out correlated data, and obtaining first training data; performing redundancy analysis on the correlated data, filtering out the corresponding correlated data having a correlation coefficient greater than a second threshold, and obtaining second training data, wherein the first training data and the second training data constitute the training data.

[0011] Optionally, the processor includes a first sub-processor and a second sub-processor, and the processor builds a model of the degree of equipment wear based on the multiple historical data, and uses the model of the degree of equipment wear and the target data to determine the degree of damage of the equipment for producing the target wafer, including: the first sub-processor transmits the target parameters to the second sub-processor; the second sub-processor determines the degree of damage of the equipment for producing the target wafer based on the target data, the target parameters and the initial model.

[0012] Optionally, the method further includes: the second sub-processor issuing an alarm when the degree of damage to the equipment for manufacturing the target wafer is greater than or equal to a third threshold.

[0013] Optionally, the method also includes: when the degree of damage of the equipment for producing the target wafer is greater than or equal to the third threshold, determining the corresponding target data based on a model of the degree of damage of the equipment for producing the target wafer and the degree of wear of the equipment; and determining the corresponding fault type based on the corresponding target data.

[0014] According to another aspect of an embodiment of the present invention, a processor is also provided, including: an acquisition unit, used for the processor to acquire multiple historical data, the multiple historical data at least include multiple first historical data, multiple second historical data and multiple third historical data, wherein the first historical data include size parameters of historical wafers and process parameters when making the historical wafers, the second historical data include the total number of processed wafers when making each of the historical wafers, and the third historical data is detected equipment wear degree data, and the third historical data corresponds one-to-one with the first historical data; a determination unit, used for the processor to build a model of equipment wear degree based on the multiple historical data, and use the equipment wear degree model and target data to determine the damage degree of the equipment for making the target wafer, the target data include size parameters of the target wafer and process parameters when making the target wafer.

[0015] According to another aspect of the embodiments of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein the program executes any one of the methods described.

[0016] According to another aspect of an embodiment of the present invention, a semiconductor device is provided, comprising: one or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include methods for executing any one of the methods described.

[0017] In an embodiment of the present invention, the processor first obtains a plurality of historical data, the plurality of historical data at least including the size parameters, process parameters and total number of processed wafers of the historical wafer and the wear degree data of the equipment, and then the processor constructs a model of the wear degree of the equipment according to the plurality of historical data, and uses the model of the wear degree of the equipment and the size parameters and process parameters of the target wafer to obtain the degree of damage of the equipment for making the target wafer. In this method, when the processor obtains the historical data, it is not necessary to manually set standards or labels for the historical data, so the historical data obtained by the processor can truly reflect the state of the semiconductor equipment, and then the processor constructs a model of the wear degree of the equipment according to the historical data, so that the constructed model of the wear degree of the equipment is more accurate, that is, it can accurately reflect the wear degree of the semiconductor equipment, and finally the processor can accurately predict the degree of damage of the equipment for making the target wafer according to the model of the wear degree of the equipment and the target data, thereby solving the problem of the lack of a method for accurately predicting the wear degree of semiconductor equipment in the prior art. Subsequently, according to the detection results, the semiconductor equipment can be repaired in time to avoid causing greater losses in subsequent semiconductor production. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings constituting part of the present application are used to provide a further understanding of the present application. The exemplary embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0019] Figure 1 A schematic flow chart of a method for detecting a semiconductor device according to an embodiment of the present application is shown;

[0020] Figure 2 A schematic diagram of the structure of a processor according to an embodiment of the present application is shown;

[0021] Figure 3 shows a heat map of feature data before processing according to an embodiment of the present application;

[0022] Figure 4 shows a heat map of processed feature data according to an embodiment of the present application;

[0023] Figure 5 A graph showing the degree of damage of equipment for manufacturing a target wafer according to an embodiment of the present application;

[0024] Figure 6 A schematic flow chart of a method for detecting a semiconductor device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0025] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0026] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0028] As mentioned in the background technology, the prior art lacks a method that can accurately predict the degree of wear of semiconductor equipment. In order to solve the above problem, in a typical embodiment of the present application, a semiconductor equipment detection method, a processor, a computer-readable storage medium and a semiconductor equipment are provided.

[0029] According to an embodiment of the present application, a method for detecting a semiconductor device is provided.

[0030] Figure 1 FIG. 1 is a flow chart of a method for detecting a semiconductor device according to an embodiment of the present application. Figure 1 As shown, the method comprises the following steps:

[0031] Step S101, the processor obtains a plurality of historical data, the plurality of historical data at least including a plurality of first historical data, a plurality of second historical data and a plurality of third historical data, wherein the first historical data includes size parameters of historical wafers and process parameters when manufacturing the historical wafers, the second historical data includes the total number of processed wafers when manufacturing each of the historical wafers, the third historical data is the detected equipment wear degree data, and the third historical data corresponds to the first historical data one by one;

[0032] In step S102, the processor constructs a model of equipment wear based on the multiple historical data, and determines the degree of damage of equipment used to manufacture a target wafer using the model of equipment wear and target data, wherein the target data includes size parameters of the target wafer and process parameters used to manufacture the target wafer.

[0033] In the above method, the processor first obtains multiple historical data, and the multiple historical data at least include the size parameters, process parameters and total number of processed wafers of the historical wafer and the wear degree data of the equipment. The processor then constructs a model of the wear degree of the equipment based on the multiple historical data, and uses the model of the wear degree of the equipment and the size parameters and process parameters of the target wafer to obtain the degree of damage of the equipment for making the target wafer. In this method, when the processor obtains historical data, it is not necessary to manually set standards or labels for the historical data, so the historical data obtained by the processor can truly reflect the state of the semiconductor equipment. Then the processor constructs a model of the wear degree of the equipment based on the historical data, so that the constructed model of the wear degree of the equipment is more accurate, that is, it can accurately reflect the wear degree of the semiconductor equipment. Finally, the processor can accurately predict the degree of damage of the equipment for making the target wafer based on the model of the wear degree of the equipment and the target data, thereby solving the problem of the lack of methods that can accurately predict the wear degree of semiconductor equipment in the prior art. Subsequently, according to the detection results, the semiconductor equipment can be repaired in time to avoid greater losses in subsequent semiconductor production.

[0034] Specifically, the third historical data is the equipment wear data detected during preventive maintenance (PM), that is, engineers regularly disassemble semiconductor equipment and then perform relevant tests to obtain equipment wear data of semiconductor equipment. The size parameters of the historical wafers are the size parameters in the historical wafer manufacturing process, such as the thickness of the wafer. The process parameters when manufacturing the historical wafers are the process data collected by the machine sensor at each process step when manufacturing the wafer. The first process data and the second process data can be temperature, humidity and / or pressure, etc.

[0035] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0036] In one embodiment of the present application, the processor constructs a model of the degree of equipment wear based on the multiple historical data, including: performing feature extraction on the multiple first historical data to obtain feature data; performing at least correlation analysis on the feature data and the second historical data to obtain training data; training an initial model using the training data and the corresponding third historical data, updating the initial parameters in the initial model to target parameters, and obtaining the model of the degree of equipment wear, wherein the initial model is the product of the Hotelling-T square value calculation formula and the initial parameters. In this embodiment, feature extraction is performed on multiple first historical data, such as peak value, root mean square, frequency and other parameters, and then a correlation analysis is performed on the feature data and the total number of processed wafers when each historical wafer is produced, and duplicate data is deleted to obtain training data, so that the training data can more accurately represent the current state of the semiconductor device. The initial model is then trained using the training data and the corresponding third historical data. The initial model is the product of the Hotelling-T square value calculation formula and the initial parameters. The initial parameters in the initial model are updated to the target parameters to obtain a model of the degree of equipment wear. Because the training data can more accurately represent the current state of the semiconductor device, the model of the degree of equipment wear obtained by training is also more accurate, thereby further improving the accuracy of the prediction.

[0037] Specifically, the above Hotelling-T square value calculation formula is: In which, let X1,…,X N Is independent N m (μ,∑) random vector, S is from N m The sample mean and sample covariance matrix of (μ,∑), m, n are the degrees of freedom, and μ is the population mean.

[0038] In order to obtain a more accurate model and thus improve the accuracy of the prediction, in another embodiment of the present application, feature extraction is performed on multiple of the above-mentioned first historical data to obtain feature data, including: extracting time domain feature data and frequency domain feature data of the same parameter in the multiple of the above-mentioned first historical data, for example, the first historical data is multiple pressures and multiple temperatures within a period of time when the above-mentioned historical wafer is manufactured, and time domain feature data and frequency domain feature data are extracted for the multiple pressures and multiple temperatures respectively; the above-mentioned time domain feature data and the above-mentioned frequency domain feature data are normalized to obtain the above-mentioned feature data.

[0039] In another embodiment of the present application, at least correlation analysis is performed on the above-mentioned feature data and the above-mentioned second historical data to obtain training data, including: performing correlation analysis on each of the above-mentioned feature data and the corresponding second historical data to obtain multiple correlation coefficients; when the above-mentioned correlation coefficient is less than 0, calculating the difference between 1 and the above-mentioned feature data corresponding to the correlation coefficient less than 0 to obtain the corrected correlation coefficient; filtering out the above-mentioned feature data corresponding to the above-mentioned corrected correlation coefficient greater than the first threshold to obtain the above-mentioned training data. In this embodiment, correlation analysis is performed on the feature data and the corresponding second historical data to obtain multiple correlation coefficients, and the above-mentioned correlation coefficient corresponds to the above-mentioned feature data one by one. By performing correlation analysis on the feature data and the number of historical processed wafers, it is possible to avoid setting labels for the historical data, thereby making the model more accurate. The range of the above-mentioned correlation coefficient is between -1 and 1. In order to facilitate subsequent processing, the correlation coefficient less than 0 is corrected, and the difference between 1 and the above-mentioned feature data corresponding to the correlation coefficient less than 0 is calculated to obtain the corrected correlation coefficient. The larger the correlation coefficient, the more repeated the feature data is, the more repeated data is, and the lower the accuracy of the model is, so it is necessary to filter out the data greater than the first threshold and delete the repeated data, thereby further improving the accuracy of the prediction model.

[0040] In a specific embodiment of the present application, the range of the first threshold is 0.5-0.7, so that most of the repeated data can be filtered out. Of course, in actual applications, the first threshold can also be other values, and those skilled in the art can choose according to actual conditions.

[0041] In another embodiment of the present application, at least correlation analysis is performed on the above-mentioned feature data and the above-mentioned second historical data to obtain training data, including: at least correlation analysis is performed on the above-mentioned feature data and the above-mentioned second historical data, and relevant data is filtered out to obtain first training data; redundancy analysis is performed on the above-mentioned relevant data, and the corresponding relevant data with a correlation coefficient greater than a second threshold is filtered out to obtain second training data, and the above-mentioned first training data and the second training data constitute the above-mentioned training data. In this embodiment, after the correlation analysis is performed on the feature data and the second historical data, the relevant data is filtered out, but there is still some useful data in the relevant data, so the relevant data is analyzed by redundancy analysis, and the redundancy analysis can accurately filter out repeated data in the relevant data, thereby further improving the accuracy of the model. Figure 3 It shows the correlation before feature data processing. The darker the color, the smaller the correlation and the less data duplication. The lighter the color, the greater the correlation and the more data duplication. Figure 4 It shows the correlation after processing the feature data. It can be seen from the figure that Figure 4 The light-colored part is significantly less than Figure 3 The light-colored part in the data, so after processing, there are fewer duplicate data, so the authenticity of the data is significantly improved, thereby improving the accuracy of the model.

[0042] In a specific embodiment of the present application, the second threshold is 0.8, so that repeated data can be accurately filtered out. Similarly, in actual applications, the second threshold can also be other values, and those skilled in the art can select according to actual conditions.

[0043] In another specific embodiment of the present application, the above-mentioned related data is the above-mentioned characteristic data corresponding to the above-mentioned corrected correlation coefficient being greater than or equal to the fourth threshold, and the above-mentioned fourth threshold is less than the above-mentioned first threshold, wherein the above-mentioned fourth threshold can be 0.4. Because the use of redundant analysis can accurately filter out duplicate data and retain useful data, the range of related data is increased, thereby making the model more accurate.

[0044] In order to further improve the accuracy of the prediction, in another embodiment of the present application, the processor includes a first sub-processor and a second sub-processor, and the processor builds a model of the degree of equipment wear according to the plurality of historical data, and uses the model of the degree of equipment wear and the target data to determine the degree of damage of the equipment for making the target wafer, including: the first sub-processor transmits the target parameter to the second sub-processor; the second sub-processor determines the degree of damage of the equipment for making the target wafer according to the target data, the target parameter and the initial model. The degree of damage of the equipment for making the target wafer is between 0 and 1.

[0045] In another embodiment of the present application, the method further includes: the second sub-processor issues an alarm when the degree of damage to the equipment for manufacturing the target wafer is greater than or equal to a third threshold. In this embodiment, when the degree of damage to the equipment for manufacturing the target wafer is greater than or equal to the third threshold, the second sub-processor issues an alarm, indicating that the semiconductor equipment has a high degree of wear and tear and is very prone to danger, so that the semiconductor equipment can be repaired in time to avoid greater losses in subsequent semiconductor manufacturing.

[0046] In a specific embodiment of the present application, the third threshold is 0.7. Figure 5 The curve shows the damage degree of the equipment for making the target wafer. The horizontal line is the third threshold. When the damage degree of the equipment for making the target wafer is greater than or equal to 0.7, the second sub-processor sends out an alarm to remind the engineer that the equipment needs to be repaired.

[0047] In another specific embodiment of the present application, each time a target wafer is produced, the degree of damage to the equipment used to produce the target wafer is determined. When the degree of damage is less than a third threshold, the next target wafer is produced. When the degree of damage is greater than or equal to the third threshold, the second sub-processor sends out an alarm. Engineers can check whether the semiconductor equipment has actually failed based on the alarm. If a failure has occurred, the semiconductor equipment will be repaired. If not, the next target wafer will be produced.

[0048] In another embodiment of the present application, the method further includes: when the degree of damage of the equipment for manufacturing the target wafer is greater than or equal to the third threshold, determining the corresponding target data according to the model of the degree of damage of the equipment for manufacturing the target wafer and the degree of wear of the equipment; and determining the corresponding fault type according to the corresponding target data. In this embodiment, when the degree of damage of the equipment for manufacturing the target wafer is relatively high, the corresponding target data can be found according to the model, and then the target data can be analyzed to find the fault type of the equipment, so as to facilitate engineers to repair the semiconductor equipment.

[0049] The embodiment of the present application further provides a processor, and it should be noted that the processor of the embodiment of the present application can be used to execute the detection method for semiconductor devices provided in the embodiment of the present application. The processor provided in the embodiment of the present application is introduced below.

[0050] Figure 2 is a schematic diagram of a processor according to an embodiment of the present application. Figure 2 As shown, the processor includes:

[0051] The acquisition unit 10 is used for the processor to acquire multiple historical data, wherein the multiple historical data at least include multiple first historical data, multiple second historical data and multiple third historical data, wherein the first historical data include size parameters of historical wafers and process parameters when making the historical wafers, the second historical data include the total number of processed wafers when making each of the historical wafers, and the third historical data is the detected equipment wear degree data, and the third historical data corresponds to the first historical data one by one;

[0052] The first determination unit 20 is used for the above-mentioned processor to build a model of the degree of equipment wear based on the above-mentioned multiple historical data, and use the above-mentioned model of the degree of equipment wear and target data to determine the degree of damage of the equipment used to manufacture the target wafer. The above-mentioned target data includes the size parameters of the target wafer and the process parameters when manufacturing the above-mentioned target wafer.

[0053] The above-mentioned processor includes an acquisition unit and a first determination unit, wherein the acquisition unit is used for the processor to acquire multiple historical data, the multiple historical data at least including the size parameters, process parameters and total number of processed wafers of the historical wafer and the wear degree data of the equipment, and the first determination unit is used for the processor to construct a model of the wear degree of the equipment according to the multiple historical data, and use the model of the wear degree of the equipment and the size parameters and process parameters of the target wafer to obtain the degree of damage of the equipment for making the target wafer. In the device, when the processor acquires the historical data, it is not necessary to manually set standards or labels for the historical data, so the historical data acquired by the processor can truly reflect the state of the semiconductor equipment, and then the processor constructs a model of the wear degree of the equipment according to the historical data, so that the constructed model of the wear degree of the equipment is more accurate, that is, it can accurately reflect the wear degree of the semiconductor equipment, and finally the processor can accurately predict the degree of damage of the equipment for making the target wafer according to the model of the wear degree of the equipment and the target data, thereby solving the problem of the lack of a method for accurately predicting the wear degree of semiconductor equipment in the prior art. Subsequently, according to the detection results, the semiconductor equipment can be repaired in time to avoid greater losses in subsequent semiconductor production.

[0054] Specifically, the third historical data is the equipment wear data detected during preventive maintenance (PM), that is, engineers regularly disassemble semiconductor equipment and then perform relevant tests to obtain equipment wear data of semiconductor equipment. The size parameters of the historical wafers are the size parameters in the historical wafer manufacturing process, such as the thickness of the wafer. The process parameters when manufacturing the historical wafers are the process data collected by the machine sensor at each process step when manufacturing the wafer. The first process data and the second process data can be temperature, humidity and / or pressure, etc.

[0055] In one embodiment of the present application, the first determination unit includes a first obtaining module, a second obtaining module and a third obtaining module, wherein the first obtaining module is used to perform feature extraction on the plurality of the above-mentioned first historical data to obtain feature data; the second obtaining module is used to perform at least correlation analysis on the above-mentioned feature data and the above-mentioned second historical data to obtain training data; the third obtaining module is used to train an initial model using the above-mentioned training data and the corresponding above-mentioned third historical data, update the initial parameters in the above-mentioned initial model to the target parameters, and obtain the model of the degree of wear of the above-mentioned equipment, wherein the above-mentioned initial model is the product of the Hotelling-T square value calculation formula and the initial parameters. In this embodiment, feature extraction is performed on multiple first historical data, such as peak value, root mean square, frequency and other parameters, and then a correlation analysis is performed on the feature data and the total number of processed wafers when each historical wafer is produced, and duplicate data is deleted to obtain training data, so that the training data can more accurately represent the current state of the semiconductor device. The initial model is then trained using the training data and the corresponding third historical data. The initial model is the product of the Hotelling-T square value calculation formula and the initial parameters. The initial parameters in the initial model are updated to the target parameters to obtain a model of the degree of equipment wear. Because the training data can more accurately represent the current state of the semiconductor device, the model of the degree of equipment wear obtained by training is also more accurate, thereby further improving the accuracy of the prediction.

[0056] Specifically, the above Hotelling-T square value calculation formula is: In which, let X1,…,X N Is independent N m (μ,∑) random vector, S is from N m The sample mean and sample covariance matrix of (μ,∑), m, n are the degrees of freedom, and μ is the population mean.

[0057] In order to obtain a more accurate model and thus improve the accuracy of the prediction, in another embodiment of the present application, the first obtaining module includes an extraction submodule and a first obtaining submodule, wherein the extraction submodule is used to extract time domain feature data and frequency domain feature data of the same parameter in the above-mentioned first historical data. For example, the first historical data is multiple pressures and multiple temperatures within a period of time when the above-mentioned historical wafer is manufactured, and time domain feature data and frequency domain feature data are extracted for the multiple pressures and multiple temperatures respectively; the first obtaining submodule is used to normalize the above-mentioned time domain feature data and the above-mentioned frequency domain feature data to obtain the above-mentioned feature data.

[0058] In another embodiment of the present application, the second obtaining module includes a second obtaining sub-module, a calculation sub-module and a first filtering sub-module, wherein the second obtaining sub-module is used to perform correlation analysis on each of the above-mentioned feature data and the corresponding second historical data to obtain multiple correlation coefficients; the calculation sub-module is used to calculate the difference between 1 and the feature data corresponding to the above-mentioned correlation coefficient being less than 0 when the above-mentioned correlation coefficient is less than 0, to obtain the corrected correlation coefficient; the first filtering sub-module is used to filter out the feature data corresponding to the above-mentioned feature data whose corrected correlation coefficient is greater than the first threshold value to obtain the above-mentioned training data. In this embodiment, a correlation analysis is performed on the feature data and the corresponding second historical data to obtain multiple correlation coefficients, and the above correlation coefficients correspond to the above feature data one by one. By performing a correlation analysis on the feature data and the number of historical processed wafers, it is possible to avoid setting labels for the historical data, thereby making the model more accurate. The range of the above correlation coefficient is between -1 and 1. In order to facilitate subsequent processing, the correlation coefficient less than 0 is corrected, and the difference between 1 and the feature data corresponding to the above correlation coefficient less than 0 is calculated to obtain the corrected correlation coefficient. The larger the correlation coefficient, the more repeated the feature data is, and the more repeated data is, the lower the accuracy of the model becomes. Therefore, it is necessary to filter out data greater than the first threshold value and delete duplicate data, thereby further improving the accuracy of the prediction model.

[0059] In a specific embodiment of the present application, the range of the first threshold is 0.5-0.7, so that most of the repeated data can be filtered out. Of course, in actual applications, the first threshold can also be other values, and those skilled in the art can choose according to actual conditions.

[0060] In another embodiment of the present application, the second obtaining module includes a second filtering submodule and a third filtering submodule, wherein the second filtering submodule is used to perform at least a correlation analysis on the above-mentioned feature data and the above-mentioned second historical data, filter out the relevant data, and obtain the first training data; the third filtering submodule is used to perform a redundancy analysis on the above-mentioned relevant data, filter out the corresponding relevant data whose correlation coefficient is greater than the second threshold, and obtain the second training data, and the above-mentioned first training data and the second training data constitute the above-mentioned training data. In this embodiment, after the correlation analysis is performed on the feature data and the second historical data, the relevant data is filtered out, but there is still some useful data in the relevant data, so the relevant data is analyzed by redundancy analysis, and the redundancy analysis can accurately filter out the repeated data in the relevant data, thereby further improving the accuracy of the model. Figure 3 It shows the correlation before feature data processing. The darker the color, the smaller the correlation and the less data duplication. The lighter the color, the greater the correlation and the more data duplication. Figure 4 It shows the correlation after processing the feature data. It can be seen from the figure that Figure 4 The light-colored part is significantly less than Figure 3 The light-colored part in the data, so after processing, there are fewer duplicate data, so the authenticity of the data is significantly improved, thereby improving the accuracy of the model.

[0061] In a specific embodiment of the present application, the second threshold is 0.8, so that repeated data can be accurately filtered out. Similarly, in actual applications, the second threshold can also be other values, and those skilled in the art can select according to actual conditions.

[0062] In another specific embodiment of the present application, the above-mentioned related data is the above-mentioned characteristic data corresponding to the above-mentioned corrected correlation coefficient which is less than or equal to the fourth threshold, and the above-mentioned fourth threshold is less than the above-mentioned first threshold, wherein the above-mentioned fourth threshold can be 0.4. Because the use of redundant analysis can accurately filter out duplicate data and retain useful data, the range of related data is increased, thereby making the model more accurate.

[0063] In order to further improve the accuracy of the prediction, in another embodiment of the present application, the processor includes a first sub-processor and a second sub-processor, the first sub-processor includes a transmission unit, and the second sub-processor includes a calculation unit, wherein the transmission unit is used for the first sub-processor to transmit the target parameter to the second sub-processor; the calculation unit is used for the second sub-processor to determine the degree of damage of the equipment for making the target wafer according to the target data, the target parameter and the initial model. The degree of damage of the equipment for making the target wafer is between 0 and 1.

[0064] In another embodiment of the present application, the second sub-processor further includes a control unit, wherein the control unit is used for the second sub-processor to issue an alarm when the degree of damage to the equipment for manufacturing the target wafer is greater than or equal to a third threshold. In this embodiment, when the degree of damage to the equipment for manufacturing the target wafer is greater than or equal to the third threshold, the second sub-processor issues an alarm, indicating that the semiconductor equipment has a high degree of wear and tear and is very prone to danger, so that the semiconductor equipment can be repaired in time to avoid greater losses in subsequent semiconductor manufacturing.

[0065] In a specific embodiment of the present application, the third threshold is 0.7. Figure 5 The curve shows the damage degree of the equipment for making the target wafer. The horizontal line is the third threshold. When the damage degree of the equipment for making the target wafer is greater than or equal to 0.7, the second sub-processor sends out an alarm to remind the engineer that the equipment needs to be repaired.

[0066] In another specific embodiment of the present application, each time a target wafer is produced, the degree of damage to the equipment used to produce the target wafer is determined. When the degree of damage is less than a third threshold, the next target wafer is produced. When the degree of damage is greater than or equal to the third threshold, the second sub-processor sends out an alarm. Engineers can check whether the semiconductor equipment has actually failed based on the alarm. If a failure has occurred, the semiconductor equipment will be repaired. If not, the next target wafer will be produced.

[0067] In another embodiment of the present application, the processor further includes a second determination unit and a third determination unit, wherein the second determination unit is used to determine the corresponding target data according to the model of the damage degree of the equipment for making the target wafer and the wear degree of the equipment when the damage degree of the equipment for making the target wafer is greater than or equal to the third threshold; the third determination unit is used to determine the corresponding fault type according to the corresponding target data. In this embodiment, when the damage degree of the equipment for making the target wafer is relatively high, the corresponding target data can be found according to the model, and then the target data can be analyzed to find the fault type of the equipment, so as to facilitate engineers to repair the semiconductor equipment.

[0068] The above-mentioned processor includes a processor and a memory. The above-mentioned acquisition unit, first determination unit, second determination unit, third determination unit, transmission unit, calculation unit and control unit are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0069] The processor includes a kernel, which calls the corresponding program unit from the memory. One or more kernels can be set, and the wear degree of semiconductor devices can be accurately predicted by adjusting kernel parameters.

[0070] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0071] An embodiment of the present invention provides a storage medium on which a program is stored. When the program is executed by a processor, the above-mentioned Z method is implemented.

[0072] An embodiment of the present invention also provides a semiconductor device, comprising one or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include methods for executing any one of the methods.

[0073] The semiconductor device mentioned above includes one or more processors, memory and one or more programs, and the one or more programs include methods for executing any of the above methods. In this method, when the processor obtains historical data, it is not necessary to manually set standards or labels for the historical data. Therefore, the historical data obtained by the processor can truly reflect the state of the semiconductor device. Then the processor constructs a model of the degree of wear of the device based on the historical data, so that the constructed model of the degree of wear of the device is more accurate, that is, it can accurately reflect the degree of wear of the semiconductor device. Finally, the processor can accurately predict the degree of damage of the device for making the target wafer based on the model of the degree of wear of the device and the target data, thereby solving the problem of the lack of methods that can accurately predict the degree of wear of semiconductor equipment in the prior art. Subsequently, the semiconductor equipment can be repaired in a timely manner according to the test results to avoid greater losses in subsequent semiconductor production.

[0074] An embodiment of the present invention provides a device, the device including a processor, a memory, and a program stored in the memory and executable on the processor, and when the processor executes the program, at least the following steps are implemented:

[0075] Step S101, the processor obtains a plurality of historical data, the plurality of historical data at least including a plurality of first historical data, a plurality of second historical data and a plurality of third historical data, wherein the first historical data includes size parameters of historical wafers and process parameters when manufacturing the historical wafers, the second historical data includes the total number of processed wafers when manufacturing each of the historical wafers, the third historical data is the detected equipment wear degree data, and the third historical data corresponds to the first historical data one by one;

[0076] In step S102, the processor constructs a model of equipment wear degree based on the plurality of historical data, and determines the degree of damage of equipment for manufacturing a target wafer using the model of equipment wear degree and target data, wherein the target data includes size parameters of the target wafer and process parameters when manufacturing the target wafer. The equipment in this article may be a server, a PC, a PAD, a mobile phone, etc.

[0077] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program for initializing at least the following method steps:

[0078] Step S101, the processor obtains a plurality of historical data, the plurality of historical data at least including a plurality of first historical data, a plurality of second historical data and a plurality of third historical data, wherein the first historical data includes size parameters of historical wafers and process parameters when manufacturing the historical wafers, the second historical data includes the total number of processed wafers when manufacturing each of the historical wafers, the third historical data is the detected equipment wear degree data, and the third historical data corresponds to the first historical data one by one;

[0079] In step S102, the processor constructs a model of equipment wear based on the multiple historical data, and determines the degree of damage of equipment used to manufacture a target wafer using the model of equipment wear and target data, wherein the target data includes size parameters of the target wafer and process parameters used to manufacture the target wafer.

[0080] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the following will be described in conjunction with specific embodiments:

[0081] Example

[0082] The flow chart of the semiconductor device detection method is as follows Figure 6 As shown, the method comprises the following steps:

[0083] Obtaining equipment wear degree data detected when the semiconductor equipment is undergoing preventive maintenance, that is, third historical data, the first sub-processor trains an initial model based on the first historical data, the second historical data, and the third historical data, the initial model being the product of a Hotelling-T square value calculation formula and initial parameters, updating the initial parameters to target parameters, and then transmitting the target parameters to the second sub-processor;

[0084] The second sub-processor determines the degree of damage of the equipment used to make the first target wafer according to the target data, target parameters and initial model obtained after making the first target wafer. If the degree of damage is less than the third threshold, the equipment does not have the risk of failure. Then the second target wafer is continued to be made. Then the degree of damage of the equipment used to make the second wafer is determined. If the degree of damage is less than the third threshold, the equipment does not have the risk of failure. Then the third target wafer is continued to be made. Each time a target wafer is made, the degree of damage of the equipment used to make the target wafer is determined. When all wafers are made and there is no risk of equipment failure, the equipment is disassembled to obtain the third historical data.

[0085] If the damage level of the equipment used to make the first target wafer is greater than or equal to the third threshold, the engineer will observe whether the equipment actually fails. If so, the engineer will maintain the equipment. If not, the engineer will make the next target wafer.

[0086] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0087] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the above-mentioned units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0088] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0089] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0090] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the above-mentioned methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0091] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0092] 1) The semiconductor equipment detection method of the present application, the processor first obtains multiple historical data, the multiple historical data at least includes the size parameters, process parameters and total number of processed wafers of the historical wafer and the wear degree data of the equipment, the processor then constructs a model of the wear degree of the equipment according to the multiple historical data, and uses the model of the wear degree of the equipment and the size parameters and process parameters of the target wafer to obtain the degree of damage of the equipment for making the target wafer. In this method, when the processor obtains the historical data, it is not necessary to manually set standards or labels for the historical data, so the historical data obtained by the processor can truly reflect the state of the semiconductor equipment, and then the processor constructs a model of the wear degree of the equipment according to the historical data, so that the constructed model of the wear degree of the equipment is more accurate, that is, it can accurately reflect the wear degree of the semiconductor equipment, and finally the processor can accurately predict the degree of damage of the equipment for making the target wafer according to the model of the wear degree of the equipment and the target data, thereby solving the problem of the lack of methods for accurately predicting the wear degree of semiconductor equipment in the prior art. Subsequently, according to the detection results, the semiconductor equipment can be repaired in time to avoid greater losses in subsequent semiconductor production.

[0093] 2) The processor of the present application includes an acquisition unit and a determination unit, wherein the acquisition unit is used for the processor to acquire multiple historical data, the multiple historical data at least including the size parameters, process parameters and total number of processed wafers of the historical wafer and the wear degree data of the equipment, and the determination unit is used for the processor to construct a model of the wear degree of the equipment according to the multiple historical data, and use the model of the wear degree of the equipment and the size parameters and process parameters of the target wafer to obtain the degree of damage of the equipment for making the target wafer. In the device, when the processor acquires the historical data, it is not necessary to manually set standards or labels for the historical data, so the historical data acquired by the processor can truly reflect the state of the semiconductor equipment, and then the processor constructs a model of the wear degree of the equipment according to the historical data, so that the constructed model of the wear degree of the equipment is more accurate, that is, it can accurately reflect the wear degree of the semiconductor equipment, and finally the processor can accurately predict the degree of damage of the equipment for making the target wafer according to the model of the wear degree of the equipment and the target data, thereby solving the problem of the lack of a method for accurately predicting the wear degree of semiconductor equipment in the prior art. Subsequently, according to the test results, the semiconductor equipment can be repaired in time to avoid greater losses in subsequent semiconductor production.

[0094] 3) The semiconductor device of the present application includes one or more processors, memory and one or more programs, and the one or more programs include methods for executing any of the above methods. In this method, when the processor obtains historical data, it is not necessary to manually set standards or labels for the historical data. Therefore, the historical data obtained by the processor can truly reflect the state of the semiconductor device. Then the processor constructs a model of the degree of wear of the device based on the historical data, so that the constructed model of the degree of wear of the device is more accurate, that is, it can accurately reflect the degree of wear of the semiconductor device. Finally, the processor can accurately predict the degree of damage of the equipment for making the target wafer based on the model of the degree of wear of the equipment and the target data, thereby solving the problem of the lack of methods that can accurately predict the degree of wear of semiconductor equipment in the prior art. Subsequently, the semiconductor equipment can be repaired in a timely manner according to the test results to avoid greater losses in subsequent semiconductor production.

[0095] The above are only preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for detecting a semiconductor device, characterized in that: include: The processor acquires a plurality of historical data, the plurality of historical data at least comprising a plurality of first historical data, a plurality of second historical data and a plurality of third historical data, wherein the first historical data comprises size parameters of historical wafers and process parameters when the historical wafers are manufactured, the second historical data comprises a total number of processed wafers when each of the historical wafers is manufactured, and the third historical data is detected equipment wear degree data, and the third historical data corresponds one-to-one to the first historical data; The processor constructs a model of equipment wear degree based on the plurality of historical data, and determines the degree of damage of equipment for manufacturing a target wafer using the model of equipment wear degree and target data, wherein the target data includes size parameters of the target wafer and process parameters when manufacturing the target wafer. The processor constructs a model of the degree of equipment wear according to the plurality of historical data, including: Extracting features from the plurality of first historical data to obtain feature data; Perform at least a correlation analysis on the feature data and the second historical data to obtain training data; The training data and the corresponding third historical data are used to train an initial model, and the initial parameters in the initial model are updated to target parameters to obtain a model of the degree of wear of the equipment, wherein the initial model is the product of the Hotelling-T square value calculation formula and the initial parameters. Performing at least a correlation analysis on the feature data and the second historical data to obtain training data includes: Performing correlation analysis on each of the characteristic data and the corresponding second historical data to obtain a plurality of correlation coefficients; When the correlation coefficient is less than 0, the difference between 1 and the characteristic data corresponding to the correlation coefficient being less than 0 is calculated to obtain a corrected correlation coefficient; The feature data corresponding to the corrected correlation coefficient being greater than a first threshold is filtered out to obtain the training data.

2. The method according to claim 1, characterized in that Extracting features from the plurality of first historical data to obtain feature data includes: Extracting time domain feature data and frequency domain feature data of the same parameter from a plurality of the first historical data; The time domain feature data and the frequency domain feature data are normalized to obtain the feature data.

3. The method according to claim 1, characterized in that Performing at least a correlation analysis on the feature data and the second historical data to obtain training data includes: Perform at least a correlation analysis on the feature data and the second historical data, filter out the relevant data, and obtain first training data, wherein the relevant data is the feature data corresponding to the feature data whose corrected correlation coefficient is greater than or equal to a fourth threshold; Redundancy analysis is performed on the relevant data, and the corresponding relevant data having a correlation coefficient greater than a second threshold is filtered out to obtain second training data, wherein the first training data and the second training data constitute the training data.

4. The method according to claim 1, characterized in that: The processor includes a first sub-processor and a second sub-processor, and the processor builds a model of equipment wear degree according to the plurality of historical data, and uses the model of equipment wear degree and target data to determine the damage degree of equipment for manufacturing a target wafer, including: The first sub-processor transmits the target parameter to the second sub-processor; The second sub-processor determines the degree of damage of the equipment for manufacturing the target wafer according to the target data, the target parameters and the initial model.

5. The method according to claim 4, characterized in that The method further comprises: The second sub-processor issues an alarm when the degree of damage to the equipment for manufacturing the target wafer is greater than or equal to a third threshold.

6. The method according to claim 5, characterized in that The method further comprises: In a case where the degree of damage of the equipment for manufacturing the target wafer is greater than or equal to the third threshold, determining the corresponding target data according to a model of the degree of damage of the equipment for manufacturing the target wafer and the degree of wear of the equipment; According to the corresponding target data, a corresponding fault type is determined.

7. A processor, characterized in that: include: An acquisition unit, configured to acquire a plurality of historical data by a processor, wherein the plurality of historical data at least includes a plurality of first historical data, a plurality of second historical data, and a plurality of third historical data, wherein the first historical data includes size parameters of historical wafers and process parameters when the historical wafers are manufactured, the second historical data includes a total number of processed wafers when each of the historical wafers is manufactured, and the third historical data is detected equipment wear degree data, and the third historical data corresponds to the first historical data one by one; a determination unit, configured for the processor to construct a model of equipment wear degree according to the plurality of historical data, and to determine the degree of damage of equipment for manufacturing a target wafer using the model of equipment wear degree and target data, wherein the target data includes size parameters of the target wafer and process parameters when manufacturing the target wafer, The determination unit includes a first obtaining module, a second obtaining module and a third obtaining module, wherein the first obtaining module is used to perform feature extraction on the plurality of the first historical data to obtain feature data; the second obtaining module is used to perform at least correlation analysis on the feature data and the second historical data to obtain training data; the third obtaining module is used to train an initial model using the training data and the corresponding third historical data, update the initial parameters in the initial model to the target parameters, and obtain the model of the degree of equipment wear, wherein the initial model is the product of the Hotelling-T square value calculation formula and the initial parameters, The second obtaining module includes a second obtaining submodule, a calculation submodule and a first filtering submodule, wherein the second obtaining submodule is used to perform correlation analysis on each of the feature data and the corresponding second historical data to obtain multiple correlation coefficients; the calculation submodule is used to calculate the difference between 1 and the feature data corresponding to the correlation coefficient being less than 0 when the correlation coefficient is less than 0 to obtain the corrected correlation coefficient; the first filtering submodule is used to filter out the feature data corresponding to the feature data whose corrected correlation coefficient is greater than a first threshold to obtain the training data.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method of any one of claims 1 to 6.

9. A semiconductor device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include methods for executing any one of claims 1 to 6.

Images