Product quality monitoring method and monitoring system for semiconductor manufacturing factory

Abstract

The invention discloses a product quality monitoring method and system for a semiconductor manufacturing factory, and the method comprises the steps: obtaining target data, the target data comprises environmental data obtained by monitoring various environmental parameters of the semiconductor manufacturing factory and quality data obtained by monitoring quality parameters of products manufactured by the semiconductor manufacturing factory; performing partial correlation analysis based on the environment data and the quality data to determine key environment parameters related to the quality parameters from the various environment parameters; fitting a regression equation between the key environment parameters and the quality parameters based on the environment data and the quality data; and predicting the quality parameters according to the regression equation and / or controlling the quality parameters by adjusting the key environment parameters. According to the invention, the method can achieve the precise monitoring of the quality of a product manufactured by a semiconductor manufacturing factory.

Description

Technical Field

[0001] The present application relates to the field of semiconductor manufacturing technology, and in particular to a product quality monitoring method and monitoring system for a semiconductor manufacturing plant. Background Art

[0002] As semiconductor manufacturing processes become more precise, the environmental requirements for semiconductor manufacturing plants (Fabration, abbreviated as FAB) are also getting higher and higher. The quality of products produced by high-precision processes such as 28nm, 40nm and 55nm is very sensitive to environmental changes. Major semiconductor manufacturers have begun to pay attention to the impact of environmental changes on product quality. Currently, there are frequent cases of environmental anomalies affecting product quality. Through the analysis of these cases, it is found that there are multiple environmental parameters that interact with each other to jointly affect product quality. However, there is currently a lack of technology to analyze the interaction of multiple environmental parameters, so it is impossible to accurately monitor product quality based on environmental parameters. Summary of the invention

[0003] In view of the above problems, the present application provides a product quality monitoring method and a monitoring system for a semiconductor manufacturing plant, aiming to achieve accurate monitoring of the quality of products manufactured in the semiconductor manufacturing plant based on environmental parameters within the semiconductor manufacturing plant.

[0004] According to a first aspect of the present invention, there is provided a product quality monitoring method for a semiconductor manufacturing plant, comprising:

[0005] Acquiring target data, wherein the target data includes environmental data obtained by monitoring various environmental parameters of a semiconductor manufacturing plant and quality data obtained by monitoring quality parameters of products manufactured by the semiconductor manufacturing plant;

[0006] Conduct partial correlation analysis based on environmental data and quality data to identify key environmental parameters related to quality parameters from a variety of environmental parameters;

[0007] Fitting a regression equation between the key environmental parameters and the quality parameters based on the environmental data and the quality data;

[0008] The quality parameters are predicted according to the regression equation and / or controlled by adjusting the key environmental parameters.

[0009] Optionally, the monitoring frequency of the environmental parameters is lower than the monitoring frequency of the quality parameters, and the matched environmental data and quality data are used to perform the partial correlation analysis and fit the regression equation;

[0010] The product quality monitoring method further includes: extracting a first quantity of quality data having a closest interval to the monitoring time of the environmental data, and determining a mean value of the extracted first quantity of quality data as the quality data matching the environmental data.

[0011] Optionally, the monitoring frequencies of the multiple environmental parameters are not completely the same and the monitoring frequency of the first environmental parameter is the lowest, and the first environmental parameter is monitored to obtain the first environmental data;

[0012] Extracting a first quantity of quality data that is closest to the monitoring time of the environmental data includes: extracting the first quantity of quality data that is closest to the monitoring time of the first environmental data.

[0013] Optionally, the multiple environmental parameters include the first environmental parameter and a second environmental parameter with a monitoring frequency higher than that of the first environmental parameter, and the second environmental parameter is monitored to obtain second environmental data;

[0014] The product quality monitoring method further includes: extracting a second quantity of second environmental data having a closest monitoring time interval to the first environmental data, and determining the average of the second quantity of second environmental data extracted as the second environmental data matching the first environmental data.

[0015] Optionally, extracting the first quantity of quality data that is closest to the monitoring time of the first environment data includes: extracting quality data that is not more than a first time period apart from the monitoring time of the first environment data, the first time period gradually increasing from a first initial time period until the first quantity of quality data is extracted or increasing to a first time period threshold if the first quantity of quality data is not extracted;

[0016] Extracting a second quantity of second environment data that is closest to the monitoring time of the first environment data, including: extracting second environment data that is no more than a second time interval away from the monitoring time of the first environment data, the second time interval gradually increasing from a second initial time interval until the second quantity of second environment data is extracted or increasing to a second time interval threshold if the second quantity of second environment data is not extracted;

[0017] The first initial duration and the first duration threshold are determined according to the monitoring frequency of the quality parameter and are each negatively correlated with the quality parameter monitoring frequency. The second initial duration and the second duration threshold are determined according to the monitoring frequency of the second environmental parameter and are each negatively correlated with the second environmental parameter monitoring frequency.

[0018] Optionally, the first quantity is determined according to the total amount of quality data obtained by monitoring the quality parameter and there is a positive correlation between the two;

[0019] The second quantity is determined according to the total amount of second environmental data obtained by monitoring the second environmental parameter, and there is a positive correlation between the two.

[0020] Optionally, the product quality monitoring method further includes:

[0021] Determining whether the total amount of first environmental data obtained by monitoring the first environmental parameter is not less than a sample size threshold;

[0022] When the total amount of the first environmental data obtained by monitoring is not less than the sample size threshold, the first quantity of quality data closest to the monitoring time of the first environmental data is extracted, and the second quantity of second environmental data closest to the monitoring time of the first environmental data is extracted.

[0023] Optionally, the product quality monitoring method further includes executing the following steps when the total amount of the first environmental data obtained through monitoring is less than the sample size threshold:

[0024] Extracting the first quantity of quality data and the second quantity of second environment data that are closest to the monitoring time of the first environment data within a period before the monitoring time of the first environment data, so as to obtain a first group of quality data and a first group of second environment data corresponding to the first environment data;

[0025] Extracting the first quantity of quality data and the second quantity of second environment data that are closest to the monitoring time of the first environment data within a period after the monitoring time of the first environment data, so as to obtain a second group of quality data and a second group of second environment data corresponding to the first environment data;

[0026] The respective averages of the first set of quality data and the first set of second environmental data corresponding to the first environmental data and the first environmental data are taken as a set of matching environmental data and quality data, and the respective averages of the second set of quality data and the second set of second environmental data corresponding to the first environmental data and the first environmental data are taken as another set of matching environmental data and quality data.

[0027] Optionally, the plurality of environmental parameters include parameters characterizing environmental characteristics and waiting times between different process steps in a product manufacturing process.

[0028] According to a second aspect of the present invention, there is provided a product quality monitoring system for a semiconductor manufacturing plant, comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program implements any one of the product quality monitoring methods described in the first aspect when executed by the processor.

[0029] The unexpected technical effects of this application are:

[0030] The product quality monitoring method for a semiconductor manufacturing plant provided in the present application includes: obtaining target data, the target data including environmental data obtained by monitoring multiple environmental parameters of the semiconductor manufacturing plant and quality data obtained by monitoring the quality parameters of the products manufactured by the semiconductor manufacturing plant; performing partial correlation analysis based on the environmental data and the quality data to determine the key environmental parameters related to the quality parameters from the multiple environmental parameters; fitting the regression equation between the key environmental parameters and the quality parameters based on the environmental data and the quality data; predicting the quality parameters according to the regression equation and / or controlling the quality parameters by adjusting the key environmental parameters. Therefore, the present application accurately determines the influence of the interaction of multiple environmental parameters on the quality parameters based on big data through partial correlation analysis and fitting regression equations, thereby enabling accurate monitoring of the quality of products manufactured by semiconductor manufacturing plants.

[0031] Furthermore, the monitoring frequency of environmental parameters is lower than that of quality parameters, and the matching environmental data and quality data are used for partial correlation analysis and fitting regression equations. The product quality monitoring method provided by the present application also includes: extracting the first quantity of quality data that is closest to the monitoring time interval of the environmental data, and determining the mean of the extracted first quantity of quality data as the quality data that matches the environmental data, thereby solving the technical problem that the environmental data and quality data cannot be matched due to the low monitoring frequency of environmental parameters. In addition, the quality data that is determined to match the environmental data is more reasonable, so that the analysis of the impact of the interaction of multiple environmental parameters on the quality parameters can be more accurate.

[0032] Furthermore, the multiple environmental parameters include a first environmental parameter and a second environmental parameter with a monitoring frequency higher than the first environmental parameter, and the second environmental parameter is monitored to obtain second environmental data. The product quality monitoring method provided in the present application also includes: extracting a second number of second environmental data with a monitoring time interval closest to the first environmental data, and determining the average of the second number of second environmental data extracted as the second environmental data matching the first environmental data, thereby solving the technical problem of matching difficulties caused by different monitoring frequencies of various environmental parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The above and other objects, features and advantages of the present application will become more apparent through the following description of the embodiments of the present application with reference to the accompanying drawings, in which:

[0034] Figure 1 Two exemplary environmental parameters are respectively monitored and obtained as environmental data;

[0035] Figure 2 The relationship between two exemplary environmental parameters and quality data is shown;

[0036] Figure 3 A flow chart showing a product quality monitoring method according to an embodiment of the present application;

[0037] Figure 4 A schematic diagram showing a regression equation fitted according to an embodiment of the present application expressed in a three-dimensional coordinate system;

[0038] Figure 5 A schematic diagram showing an exemplary method of determining quality data matching environmental data according to the present application;

[0039] Figure 6 A schematic diagram showing an exemplary method of determining quality data and second environment data matching first environment data according to the present application;

[0040] Figure 7 A structural block diagram of a product quality monitoring system according to an embodiment of the present application is shown.

[0041] Explanation of reference numerals: 1300 - product quality monitoring system; 1310 - memory; 1320 - processor; 1330 - power supply component; 1340 - network interface; 1350 - input and output interface. DETAILED DESCRIPTION

[0042] In order to facilitate the understanding of the present application, the present application will be described more comprehensively with reference to the relevant drawings below. The preferred embodiments of the present application are given in the drawings, but the present application can be implemented in different forms, and the present application is not limited to the embodiments described herein. The purpose of providing these embodiments herein is to make the understanding of the disclosure of the present application more thorough and comprehensive.

[0043] In the semiconductor field, the environmental parameters of a semiconductor manufacturing plant refer to a series of environmental parameters related to production within a semiconductor manufacturing plant. The following are some common environmental parameters:

[0044] 1) Temperature and humidity will affect material properties, chemical reaction rates and the stability of manufacturing equipment to a certain extent;

[0045] 2) Cleanroom grade is a measure of the number of particles in the air, such as ISO 4, which requires less than 352 particles greater than or equal to 0.5 microns per cubic meter. The emphasis on cleanroom grade is to prevent pollutants in the air from affecting the performance of products manufactured in semiconductor manufacturing plants;

[0046] 3) Chemical purity, such as the purity of fluorine and chlorine in semiconductor manufacturing plants. Generally, chemicals used in semiconductor manufacturing must have extremely high purity to avoid impurities in the chemicals affecting the performance of products manufactured by the semiconductor manufacturing plant;

[0047] 4) The air pressure and electromagnetic interference level within the factory, which may affect the stability of manufacturing equipment such as ion implanters and the performance of products manufactured by semiconductor manufacturing plants.

[0048] The environmental parameters listed above are all parameters that characterize environmental characteristics, but it should be understood that these are only a part of the parameters that characterize environmental characteristics and do not represent a limitation on all types of parameters that characterize environmental characteristics.

[0049] In the semiconductor manufacturing process, semiconductor devices are usually produced in batches (lots). If the production of semiconductor devices involves multiple process steps, a batch of intermediate structures manufactured in the previous process step needs to enter the next process step in sequence, so an intermediate structure has a waiting stage between different process steps. The duration of the waiting stage is the waiting time (Queue time, referred to as Q-Time). In the embodiment of the present application, the environmental parameters can also include the waiting time between different process steps in the product manufacturing process. This is because the intermediate structures in the semiconductor manufacturing process will contact the environment of the semiconductor manufacturing plant during the waiting stage between different processes. The size of the waiting time may affect the performance of the final manufactured product to varying degrees. Therefore, using the waiting time between different process steps in the product manufacturing process as an environmental parameter to analyze the factors affecting product quality will make the analysis results more accurate.

[0050] For different semiconductor devices, the environmental parameters that affect product quality may be different, so both the testing party and the quality party hope to be able to trace and monitor the impact of environmental parameter fluctuations on quality parameters. The quality parameters mentioned here refer to the parameters that characterize the quality of products manufactured by semiconductor manufacturing plants, which can be the number of defective semiconductor devices in semiconductor devices produced in batches or the yield of semiconductor devices.

[0051] Currently, there are two main analysis schemes for environmental parameters of semiconductor manufacturing plants. One is to monitor the environmental parameters of semiconductor manufacturing plants to detect abnormal environmental data, and the other is to monitor the quality parameters of products manufactured by semiconductor manufacturing plants to detect abnormal quality data.

[0052] Specifically, the former analysis scheme mainly uses the SPC (Statistical Process Control) system to monitor environmental parameter fluctuations, and issues an SPC alarm when the environmental parameters fluctuate to the point where the corresponding environmental data obtained by monitoring is outside the preset range. Figure 1As shown in Figure (a), the environmental parameter temperature (Tem) is monitored at different times (Time) and an SPC alarm is issued in a timely manner when the temperature data obtained by monitoring is less than 22.2°C or greater than 23.8°C. The temperature data A1, A2 and A3 shown in Figure (a) are the temperature data to be alarmed; for example, Figure 1 As shown in Figure (b), the relative humidity (RH) environmental parameter is monitored at different times and an SPC alarm is issued in a timely manner when the relative humidity data obtained from the monitoring is less than 40% or greater than 46%. The relative humidity data B1 and B2 shown in Figure (b) are the relative humidity data for which an alarm is issued. Figure 1 The horizontal axes of Figures (a) and (b) in the figure represent time. The horizontal axis of Figure (a) indicates some monitoring moments, but it should be noted that the temperature is not monitored at a constant interval; the horizontal axis of Figure (b) does not indicate the monitoring moment of relative humidity. The relative humidity data shown in Figure (b) is obtained by monitoring twice a week. The purpose of SPC alarm is to timely identify and respond to abnormal conditions in the semiconductor manufacturing process and prevent the occurrence and spread of product defects by issuing warning information. In the former analysis scheme, after the SPC alarm, the cause of the fluctuation of environmental parameters should be analyzed by adding test steps based on the alarm information, and then improvement plans should be proposed based on the analyzed fluctuation causes to ensure that the environmental parameters are restored to the preset range, and experience should be summarized and measures should be taken to prevent fluctuations in environmental parameters. This analysis scheme does not correlate the environmental parameters with the quality parameters of the product. The whole process does not know whether the fluctuation of environmental parameters affects the product quality and what kind of impact it has, and there is a lack of early alarm mechanism.

[0053] The latter analysis scheme is that if the abnormality of quality data is determined to be caused by the environment of the semiconductor manufacturing plant, then the environmental data and quality data are analyzed by adding test steps to determine the environmental parameters that cause the abnormal quality data, and then the determined environmental parameters are adjusted so that the environment of the semiconductor manufacturing plant is improved in a manner that meets the requirements of quality assessment (QA) until it is confirmed that the product quality has returned to normal. The abnormal quality data refers to the quality data being outside the preset range, for example, the number of defective semiconductor devices in semiconductor devices produced in batches is greater than the defect threshold. This analysis scheme requires manual analysis of environmental data and quality data, which is inefficient; and the analysis method is single, lacking technology for analyzing the interaction of multiple environmental parameters.

[0054] At present, there are frequent cases where environmental abnormalities affect product quality. Figure 2The following is the relevant data in a case, where (a) shows the number Q of defective semiconductor devices in the semiconductor devices produced in batches through multiple black points in a two-dimensional coordinate system. def The relationship between the parameter Environment and the environmental characteristics, the figure shows the quantity Q def As the parameter Environment increases at the slope shown by the dotted line, a value of the parameter Environment will correspond to the quantity Q def The multiple values ​​of def It is not uniquely determined by the parameter Environment; (b) The figure shows the quantity Q through multiple black dots in a two-dimensional coordinate system def The relationship between the waiting time Q-Time and the number Q def As the waiting time Q-Time increases at the slope shown by the dotted line, a value of the waiting time Q-Time will correspond to the quantity Q def The multiple values ​​of def is not uniquely determined by the waiting time Q-Time. Figure 2 The cases shown show that it is of great significance to analyze the interaction of multiple environmental parameters.

[0055] Although the latter analysis scheme mentioned above analyzes the impact of environmental parameters on product quality, it is unable to analyze the interaction of multiple environmental parameters. Therefore, there is a technical problem of inaccurate analysis for cases where the interaction of multiple environmental parameters jointly affects product quality. As a result, it is impossible to accurately predict product quality based on environmental parameters according to the analysis results, nor is it possible to achieve precise control of product quality by adjusting environmental parameters.

[0056] In view of this, an embodiment of the present application provides a product quality monitoring method for a semiconductor manufacturing plant. The product quality monitoring method performs partial correlation analysis based on environmental data and quality data to determine key environmental parameters related to product quality parameters from a variety of environmental parameters, and then fits a regression equation between the key environmental parameters and the quality parameters based on the environmental data and the quality data. Therefore, according to the regression equation, a fast and accurate analysis can be performed when an abnormality occurs in the environmental data or the quality data, thereby ultimately achieving precise monitoring of the quality of products manufactured by the semiconductor manufacturing plant.

[0057] Figure 3 FIG. 1 is a flow chart of a method for monitoring product quality in a semiconductor manufacturing plant provided by an embodiment of the present application. Figure 3 As shown, the product quality monitoring method includes:

[0058] Step S110 , acquiring target data, where the target data includes environmental data obtained by monitoring various environmental parameters of the semiconductor manufacturing plant and quality data obtained by monitoring quality parameters of products manufactured by the semiconductor manufacturing plant.

[0059] Step S120 , performing partial correlation analysis based on the environmental data and the quality data to determine key environmental parameters related to the quality parameters from the multiple environmental parameters.

[0060] Step S130 , fitting a regression equation between key environmental parameters and quality parameters based on the environmental data and the quality data.

[0061] Step S140, predicting the quality parameters according to the regression equation and / or controlling the quality parameters by adjusting the key environmental parameters.

[0062] It should be noted that the above-mentioned prediction of quality parameters according to the regression equation is usually aimed at the situation where the environmental parameters fluctuate, that is, when the environmental parameters fluctuate, the actual values ​​of the environmental parameters are obtained in time to analyze whether the quality parameters are affected according to the regression equation, and when the quality parameters are affected, determine what kind of impact the quality parameters have. This is a process of tracing cause and effect, which can provide early warning and thus help to eliminate hidden dangers in time; the above-mentioned control of quality parameters by adjusting key environmental parameters according to the regression equation is usually aimed at the situation where the quality parameters fluctuate, that is, when the quality parameters fluctuate and make the quality data abnormal, the key environmental parameters are adjusted to ensure that the quality of the manufactured products meets the requirements. This process traces cause and effect and improves the quality of subsequent products by promptly solving the problems that lead to abnormal quality data.

[0063] In a parameter Environment and waiting time Q-Time interact to jointly affect the quantity Q def In the example, the expression of the correlation coefficient matrix Coe calculated by partial correlation analysis is shown in the following formula (1), where the value of the i-th row and j-th column represents the correlation coefficient between the i-th parameter and the j-th parameter, the first parameter is the parameter Environment, the second parameter is the waiting time Q-Time, and the third parameter is the quantity Q def In addition, the relationship between the parameter Environment and the quantity Q is calculated through partial correlation analysis. def The first-order partial correlation coefficient is 0.5860846986923319, and the waiting time Q-Time is related to the number Q def The first-order partial correlation coefficient is 0.473309053694991.

[0064] Coe= (1)

[0065] It should be noted that the correlation coefficient reflects the total correlation between two variables without controlling other variables, including direct and indirect relationships; the first-order partial correlation coefficient reflects only the direct relationship between two variables by controlling other variables. The value range of the correlation coefficient and the first-order partial correlation coefficient is between -1 and 1, where -1 indicates a complete negative correlation, 1 indicates a complete positive correlation, and 0 indicates no correlation.

[0066] The above formula (1) combined with the calculated first-order partial correlation coefficient can determine: the parameter Environment and the quantity Q as the quality parameter def There is a direct and indirect relationship between the waiting time Q-Time and the quantity Q as a quality parameter. def There is also a direct and indirect relationship between the two. Since the parameters Environment and waiting time Q-Time are respectively related to the quantity Q def Therefore, the parameters Environment and waiting time Q-Time are related to the quantity Q def Related key environmental parameters. In this application, key environmental parameters are related to quality parameters, which means that changes in the values ​​of key environmental parameters will lead to changes in quality parameters.

[0067] For the example of expressing the correlation coefficient matrix Coe by the above formula (1), the quantity Q def As the dependent variable, the parameter Environment as an independent variable and the waiting time Q-Time as another independent variable, the regression equation shown in the following formula (2) is fitted by the polynomial regression algorithm.

[0068] Q def =0.729-23.823×Environment+0.303×Q-Time+279.603×Environment 2 +4.632×(Environment×Q-Time)-0.003×Q-Time 2 (2)

[0069] for Figure 2 In the case shown, the parameter Environment interacts with the waiting time Q-Time to jointly affect the quantity Q def The fitted regression equation is expressed in the three-dimensional coordinate system as Figure 4 The surface shown. Figure 4 The set of values ​​of the parameters Environment and the waiting time Q-Time only corresponds to the quantity Q defThis means that the fitted regression equation accurately represents the interaction between the parameter Environment and the waiting time Q-Time and the quantity Q def impact.

[0070] It should be understood that the regression equation shown in formula (2) above can only analyze the parameters Environment and the waiting time Q-Time and the quantity Q def Therefore, the regression equation can only be used for the case where the parameter Environment and the waiting time Q-Time fluctuate. def predictions, and for the quantity Q def In the event of an abnormality, determine how the parameter Environment and the waiting time Q-Time need to be adjusted. Based on this, the various environmental parameters described in step S110 are not necessarily all environmental parameters of a semiconductor manufacturing plant. In other words, the various environmental parameters described in step S110 may be some environmental parameters that are prone to fluctuations, so that the purpose of simplifying the monitoring calculation cost is achieved by reducing the number of environmental parameters. For non-fluctuating environmental parameters, even if these environmental parameters will affect the quality parameters when they fluctuate, the impact of taking a certain constant value on the quality parameters in a non-fluctuating scenario can be fitted to the constant term of the regression equation. In practice, the various environmental parameters in step S110 can be set by the user, and the monitoring data obtained by monitoring the environmental parameters can be stored in a database, and the environmental data corresponding to the required environmental parameters can be directly obtained from the database when the product quality monitoring system executes the product quality monitoring method.

[0071] The execution result of the above-mentioned step S140 can be to generate prediction information when predicting the quality parameters and further output more intuitive prediction information in the form of an analysis report, and the prediction information includes the value of the quality parameters; it can also be to generate control information when controlling the quality parameters, and the control information includes the value of the increase or decrease of the key environmental parameters. The control information can be directly output to the control device of the key environmental parameters so that the control device can quickly adjust the key environmental parameters. The control information can also be output in the form of an analysis report to inform the user how to adjust the key environmental parameters.

[0072] Due to insufficient monitoring equipment or sampling manpower, the monitoring frequency of environmental parameters is usually low, for example, the monitoring frequency of temperature and humidity is twice a week, and the monitoring frequency of indoor total volatile organic compounds (Total Volatile Organic Compounds, referred to as TVOC) is once a week, while quality parameters are usually counted after a batch of products are generated, so the monitoring frequency of environmental parameters is lower than the monitoring frequency of quality parameters, so that quality data is monitored at the same time but there is no monitored environmental data, so environmental data and quality data will be difficult to match due to the sparse environmental data caused by the low monitoring frequency of environmental parameters, and partial correlation analysis and fitting regression equations require matching environmental data and quality data. In view of this, an optional embodiment uses the KNN (K-NearestNeighbors, K nearest neighbors) algorithm to solve the technical problem that environmental data and quality data cannot match at the monitoring time and cannot be analyzed. Specifically, the above-mentioned product quality monitoring method also includes: extracting the first number of quality data that is closest to the monitoring time interval of the environmental data, and determining the mean of the extracted first number of quality data as the quality data matching the environmental data. In the case of a small fluctuation of an environmental parameter, the mean of the first quantity of quality data extracted above can avoid the disturbance of the quality data caused by factors other than the environmental parameter; in the case of a large fluctuation of an environmental parameter, the mean of the first quantity of quality data extracted above for a certain environmental data can be regarded as the mean of multiple quality data caused by multiple environmental data fluctuating around the environmental data, and is therefore closer to the quality data caused by the environmental data. The method for determining quality data matching environmental data provided in the embodiment of the present application is more reasonable, and thus can make the analysis of the impact of the interaction of multiple environmental parameters on the quality parameters more accurate.

[0073] Figure 5 FIG. 1 is a schematic diagram of an exemplary method for determining quality data matching environmental data, wherein the environmental data is data obtained by monitoring a parameter Environment that characterizes environmental characteristics and is indicated by black dots in the figure, and the quality data is the number Q of defective semiconductor devices in semiconductor devices produced in batches. def The obtained data are indicated by black triangles in the figure. Figure 5 It shows that a total of five environmental data were monitored from 0:00 on September 3, 2023 to 0:00 on September 12, 2023 as shown on the horizontal axis. There is an environmental data marked with a black dot in each dotted box in the figure, and there are one or more dotted boxes in each dotted box. The quality data marked with black triangles in the dotted box is the first quantity of quality data extracted for the environmental data in the same dotted box.

[0074] Since different types of environmental parameters in a semiconductor manufacturing plant often involve different monitoring layouts and detection points, the detection frequencies of different types of environmental parameters may be different. For an example in which the monitoring frequencies of multiple environmental parameters are not exactly the same, if the monitoring frequency of the first environmental parameter among the multiple environmental parameters is the lowest and the first environmental data is obtained by monitoring the first environmental parameter, then the above-mentioned extraction of the first quantity of quality data with the closest interval to the monitoring time of the environmental data may include: extracting the first quantity of quality data with the closest interval to the monitoring time of the first environmental data, that is, matching the environmental data with the quality data based on the first environmental data.

[0075] The multiple environmental parameters include the above-mentioned first environmental parameter and a second environmental parameter whose monitoring frequency is higher than that of the first environmental parameter, and the second environmental parameter is monitored to obtain the second environmental data. It should be understood that the second environmental parameter in the embodiment of the present application can be one or more. The product quality monitoring method provided in the embodiment of the present application can further include: extracting the second number of second environmental data with the closest interval to the monitoring time of the first environmental data, and determining the mean of the extracted second number of second environmental data as the second environmental data matching the first environmental data, so that for the example in which the monitoring frequencies of the multiple environmental parameters are not exactly the same, the technical problem of the difficulty in matching the environmental parameters due to different monitoring frequencies is solved by reusing the KNN algorithm.

[0076] The above-mentioned extraction of the first quantity of quality data whose interval with the monitoring time of the first environment data is closest may include: extracting quality data whose interval with the monitoring time of the first environment data is not greater than the first time length; the above-mentioned extraction of the second quantity of second environment data whose interval with the monitoring time of the first environment data is closest may include: extracting the second environment data whose interval with the monitoring time of the first environment data is not greater than the second time length.

[0077] If the time when the first environment data is monitored is Time2, and the time when the quality data or the second environment data is monitored is Time1, then the extracted quality data and the second environment data need to satisfy the relationship shown in the following formula (3), wherein, when Time1 represents the time when the quality data is monitored, Time0 is the first time length, and the first time length gradually increases from the first initial time length until the first amount of quality data is extracted or increases to the first time length threshold when the first amount of quality data is not extracted, so that the first time length can be dynamically adjusted to make the extracted quality data as close to the monitoring time of the first environment data as possible among the quality data that have not been extracted; when Time1 represents the time when the second environment data is monitored, Time0 is the second time length, and the above-mentioned second time length gradually increases from the second initial time length until the second amount of second environment data is extracted or increases to the second time length threshold when the second amount of second environment data is not extracted, so that the second time length can be dynamically adjusted to make the extracted second environment data as close to the monitoring time of the first environment data as possible among the second environment data that have not been extracted.

[0078] |Time2-Time1|≤Time0(3)

[0079] The above-mentioned first initial duration and first duration threshold are determined according to the monitoring frequency of the quality parameters and are negatively correlated with the monitoring frequency of the quality parameters, that is, the higher the monitoring frequency of the quality parameters, the smaller the first initial duration and the first duration threshold. The reason why the higher the monitoring frequency of the quality parameters, the smaller the first initial duration is that: if the first initial duration is less than the duration set according to the monitoring frequency of the quality parameters, it is necessary to increase the first duration to query the quality data multiple times, and the efficiency of extracting the first quantity of quality is low; if the first initial duration is greater than the duration set according to the monitoring frequency of the quality parameters, it is very likely that the total amount of quality data whose interval with the monitoring time of the first environmental data is not greater than the first initial duration is already greater than the first quantity, and it is impossible to ensure that the first quantity of quality data with the closest interval with the monitoring time of the first environmental data is extracted. The reason why the higher the monitoring frequency of the quality parameters, the smaller the first time threshold is that: if the first time threshold is smaller than the time set according to the monitoring frequency of the quality parameters, it may be possible to extract only quality data less than the first quantity; if the first time threshold is larger than the time set according to the monitoring frequency of the quality parameters, then if the adjustment step of the first time is slightly larger, the extracted quality data may not be the quality data with the closest monitoring time interval to the first environment data among the quality data that have not been extracted.

[0080] The above-mentioned second initial duration and second duration threshold are determined according to the monitoring frequency of the second environmental parameter and are each negatively correlated with the monitoring frequency of the second environmental parameter. Such a setting is conducive to efficiently extracting the second quantity of second environmental data. The specific reasons can be understood by referring to the above-mentioned description of the quality data, and will not be repeated here.

[0081] The first quantity can be determined based on the total amount of quality data obtained by monitoring the quality parameters and there is a positive correlation between the two. The second quantity can be determined based on the total amount of second environment data obtained by monitoring the second environment parameters and there is a positive correlation between the two. It should be understood that a large total amount of quality data obtained by monitoring the quality parameter in the same time period indicates a high monitoring frequency of the quality parameter. Similarly, a large total amount of second environmental data obtained by monitoring the second environmental parameter in the same time period indicates a high monitoring frequency of the second environmental parameter. Therefore, the method of determining the first quantity and the second quantity here further considers the monitoring frequency of the quality parameter and the monitoring frequency of the second environmental parameter, thereby combining with the above-mentioned method of determining the first initial time length and the first time length threshold and the method of determining the second initial time length and the second time length threshold so that in order to extract the first quantity of quality data and the second quantity of second environmental data, there is no need to set a larger first time length threshold and second time length threshold. This is because if the first time length threshold and the second time length threshold are set too large, quality data and second environmental data that are too far away from the monitoring time of the first environmental data will be extracted, and the quality data that are too far away from the monitoring time of the first environmental data may no longer have any correlation with the first environmental data. In addition, the second environmental data that is too far away from the monitoring time of the first environmental data may no longer have any correlation with the first environmental data in terms of its impact on the quality data.

[0082] It should be understood that the first number and the second number can each be regarded as the K value in the KNN algorithm, and the above process of determining the first number and the second number is the process of dynamically adjusting the K value according to the number of samples.

[0083] In some examples, the product quality monitoring method provided by the embodiments of the present application may also include: determining whether the total amount of first environmental data obtained by monitoring the first environmental parameter is not less than the sample size threshold; when the total amount of the first environmental data obtained by monitoring is not less than the sample size threshold, executing the above-mentioned step of extracting the first quantity of quality data with the closest interval to the monitoring time of the first environmental data, and executing the above-mentioned step of extracting the second quantity of second environmental data with the closest interval to the monitoring time of the first environmental data.

[0084] Specifically, the above-mentioned sample size threshold can be a preset value used to measure whether the sample quantity is insufficient in the process of determining the matching quality data of the first environment data and the matching second environment data. If the total amount of the first environment data is less than the sample size threshold, it indicates that there is a problem of insufficient sample quantity.

[0085] Furthermore, the product quality monitoring method provided by the embodiment of the present application may also include performing the following steps when the total amount of the first environmental data obtained by monitoring is less than the sample size threshold: extracting the first quantity of quality data and the second quantity of second environmental data that are closest to the monitoring time of the first environmental data within a time period before the monitoring time of the first environmental data to obtain a first group of quality data and a first group of second environmental data corresponding to the first environmental data; extracting the first quantity of quality data and the second quantity of second environmental data that are closest to the monitoring time of the first environmental data within a time period after the monitoring time of the first environmental data to obtain a second group of quality data and a second group of second environmental data corresponding to the first environmental data; taking the respective means of the first group of quality data and the first group of second environmental data corresponding to the first environmental data and the first environmental data as a set of matching environmental data and quality data, and taking the respective means of the second group of quality data and the second group of second environmental data corresponding to the first environmental data and the first environmental data as another set of matching environmental data and quality data.

[0086] It should be noted that in order to better extract the above-mentioned first quantity of quality data in the time periods before and after the monitoring time of the first environmental data, the extraction of quality data can be carried out in the first time period before and after the monitoring time of the first environmental data according to the above-mentioned relevant description; similarly, in order to better extract the above-mentioned second quantity of second environmental data in the time periods before and after the monitoring time of the first environmental data, the extraction of second environmental data can be carried out in the second time period before and after the monitoring time of the first environmental data according to the above-mentioned relevant description.

[0087] Figure 6The figure shows an exemplary schematic diagram for determining the quality data and the second environmental data that match the first environmental data, wherein the horizontal axis represents time (Time), the vertical axis represents the specific value (Value) of each data, environmental data 1 is marked by the black dots in the figure, environmental data 2 is marked by the black triangles in the figure, quality data is marked by the black diamonds in the figure, and environmental data 1 serves as the first environmental data. In this example, the first quantity is 5, the second quantity is 3, the total amount of environmental data 1 is 3, and the sample quantity threshold is 5, for example, so that 5 quality data are extracted in the first time period before and after the monitoring time of each environmental data 1, and 3 environmental data 2 are extracted in the second time period before and after the monitoring time of each environmental data 1. Here, the first time period threshold and the second time period threshold are both 12 hours, for example, and the time span of the dotted box in the figure is 12 hours. As shown Figure 6 As shown, each environmental data 1 is adjacent to two dotted boxes and thus corresponds to two dotted boxes, and there are solid boxes and dotted boxes in each dotted box, wherein the quality data marked by black diamonds in the solid box are the first quantity of quality data extracted for the environmental data 1 corresponding to the dotted box where the solid box is located, and the quality data extracted for the environmental data 1 include the first group of quality data located in the solid box before the monitoring time of the environmental data 1 and the second group of quality data located in the solid box after the monitoring time of the environmental data 1; the environmental data 2 marked by triangles in the dotted box are the second quantity of environmental data 2 extracted for the environmental data 1 corresponding to the dotted box where the dotted box is located, and the environmental data 2 extracted for the environmental data 1 include the first group of environmental data 2 located in the dotted box before the monitoring time of the environmental data 1 and the second group of environmental data 2 located in the dotted box after the monitoring time of the environmental data 1.

[0088] Reference Figure 6 , the respective means of the first group of quality data and the first group of environmental data 2 corresponding to an environmental data 1 and the environmental data 1 are used as a group of matching environmental data and quality data, thereby forming a sample for performing the partial correlation analysis of step S120 and the fitting regression equation of step S130; the respective means of the second group of quality data and the second group of environmental data 2 corresponding to the same environmental data 1 and the environmental data 1 are used as another group of matching environmental data and quality data, thereby forming a sample for the partial correlation analysis of step S120 and the fitting regression equation of step S130. It can be seen that the same environmental data 1 can form two samples, realizing the expansion of the number of samples. It should be noted that the two samples formed by the same environmental data 1 are of great significance for analyzing the correlation between the second environmental parameter and the quality parameter when the value of the first environmental parameter remains unchanged.

[0089] It should be noted that the first quantity and the second quantity, the first initial duration and the first duration threshold, and the second initial duration and the second duration threshold determined according to the embodiment of the present application can ensure as much as possible that during the execution of the product quality monitoring method, the first quantity of quality data and the second quantity of second environmental data that meet the requirements are extracted. However, in the process of expanding the number of samples, it may be impossible to extract twice the first quantity of quality data or twice the second quantity of second environmental data, such as Figure 6 As shown, only two environmental data 2 are extracted for the latest monitored environmental data 1 within 12 hours after its monitoring time. In the embodiment of the present application, in the case where twice the first amount of quality data cannot be extracted or twice the second amount of second environmental data cannot be extracted, as long as the quality data and the second environmental data are extracted, it is allowed to continue to execute, that is, as long as the quality data that can be extracted is averaged as the matching quality data of the corresponding first environmental data, and similarly, as long as the second environmental data that can be extracted is averaged as the matching second environmental data of the corresponding first environmental data.

[0090] The product quality monitoring method provided in the embodiment of the present application can also be used to analyze semiconductor manufacturing plants in which only a single environmental parameter fluctuates. Specifically, the functional relationship between the single environmental parameter and the quality parameter is fit by using the environmental data obtained by monitoring the single environmental parameter and the quality data obtained by monitoring the product quality parameter, and the quality data that matches the monitored environmental data can be determined through the above related content.

[0091] Corresponding to the product quality monitoring method for a semiconductor manufacturing plant provided in the above embodiment, another embodiment of the present application provides a product quality monitoring system for a semiconductor manufacturing plant, such as Figure 7 As shown, the product quality monitoring system 1300 includes: a memory 1310 and a processor 1320, and a program stored in the memory 1310 and executable on the processor 1320. When the program is executed by the processor 1320, each process of each embodiment of the above-mentioned product quality monitoring method can be implemented and the same technical effect can be achieved. In some examples, the product quality monitoring system 1300 may also include auxiliary sub-devices such as a power supply component 1330, a network interface 1340, and an input / output interface 1350.

[0092] The product quality monitoring system provided in the embodiment of the present application is an intelligent monitoring system that can perform correlation analysis between multiple environmental parameters and quality parameters, thereby exploring the true relationship between the environment of a semiconductor manufacturing plant and the quality of manufactured products through multi-angle analysis; in addition, it can support single-factor analysis and multi-factor interactive analysis, thereby allowing users to flexibly input relevant data of a single environmental parameter or multiple environmental parameters for analysis according to actual needs to achieve corresponding single-factor monitoring or multi-factor monitoring, which is of great significance for accurately and efficiently monitoring the quality of products manufactured by semiconductor manufacturing plants.

[0093] Those skilled in the art can understand that all or part of the steps of the above method embodiments can be completed by executing a computer program. To this end, another embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, each process of each embodiment of the above product quality monitoring method can be implemented and the same technical effect can be achieved. Among them, computer-readable storage media, such as mobile hard disks, read-only memories (Read-Only Memory, referred to as ROM), random access memories (Random Access Memory, referred to as RAM), disks or optical disks, and other media that can store program codes.

[0094] Finally, it should be noted that the terms "first", "second" and "third" are used for descriptive purposes only and should not be understood as indicating or implying relative importance. According to the embodiments of the present application as described above, these embodiments do not describe all the details in detail, nor do they limit the application to the specific embodiments described. Obviously, many modifications and changes can be made based on the above description. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that technicians in the relevant technical field can make good use of the present application and the modified use based on the present application. The present application is only limited by the claims and their full scope and equivalents.

Claims

1. A product quality monitoring method for a semiconductor manufacturing plant, comprising: Acquiring target data, wherein the target data includes environmental data obtained by monitoring various environmental parameters of a semiconductor manufacturing plant and quality data obtained by monitoring quality parameters of products manufactured by the semiconductor manufacturing plant; Conduct partial correlation analysis based on environmental data and quality data to identify key environmental parameters related to quality parameters from a variety of environmental parameters; Fitting a regression equation between the key environmental parameters and the quality parameters based on the environmental data and the quality data; The quality parameters are predicted according to the regression equation and / or the quality parameters are controlled by adjusting the key environmental parameters.

2. The product quality monitoring method according to claim 1, wherein: The monitoring frequency of environmental parameters is lower than that of quality parameters, and the matched environmental data and quality data are used to perform the partial correlation analysis and fit the regression equation; The product quality monitoring method further includes: extracting a first quantity of quality data having a closest interval to the monitoring time of the environmental data, and determining a mean value of the extracted first quantity of quality data as the quality data matching the environmental data.

3. The product quality monitoring method according to claim 2, wherein: The monitoring frequencies of the multiple environmental parameters are not completely the same and the monitoring frequency of the first environmental parameter is the lowest, and the first environmental parameter is monitored to obtain the first environmental data; Extracting a first quantity of quality data that is closest to the monitoring time of the environmental data includes: extracting the first quantity of quality data that is closest to the monitoring time of the first environmental data.

4. The product quality monitoring method according to claim 3, wherein: The multiple environmental parameters include the first environmental parameter and a second environmental parameter whose monitoring frequency is higher than that of the first environmental parameter, and the second environmental parameter is monitored to obtain second environmental data; The product quality monitoring method further includes: extracting a second quantity of second environmental data having a closest monitoring time interval to the first environmental data, and determining the average of the second quantity of second environmental data extracted as the second environmental data matching the first environmental data.

5. The product quality monitoring method according to claim 4, wherein: Extracting the first quantity of quality data that is closest to the monitoring time of the first environment data includes: extracting the quality data that is not more than a first time period apart from the monitoring time of the first environment data, the first time period gradually increasing from a first initial time period until the first quantity of quality data is extracted or increasing to a first time period threshold if the first quantity of quality data is not extracted; Extracting a second quantity of second environment data that is closest to the monitoring time of the first environment data, including: extracting second environment data that is no more than a second time interval away from the monitoring time of the first environment data, the second time interval gradually increasing from a second initial time interval until the second quantity of second environment data is extracted or increasing to a second time interval threshold if the second quantity of second environment data is not extracted; The first initial duration and the first duration threshold are determined according to the monitoring frequency of the quality parameter and are each negatively correlated with the quality parameter monitoring frequency. The second initial duration and the second duration threshold are determined according to the monitoring frequency of the second environmental parameter and are each negatively correlated with the second environmental parameter monitoring frequency.

6. The product quality monitoring method according to claim 5, wherein: The first quantity is determined according to the total amount of quality data obtained by monitoring the quality parameters, and there is a positive correlation between the first quantity and the quality data; The second quantity is determined according to the total amount of second environmental data obtained by monitoring the second environmental parameter, and there is a positive correlation between the two.

7. The product quality monitoring method according to claim 4, further comprising: Determining whether the total amount of first environmental data obtained by monitoring the first environmental parameter is not less than a sample size threshold; When the total amount of the first environmental data obtained by monitoring is not less than the sample size threshold, the first quantity of quality data closest to the monitoring time of the first environmental data is extracted, and the second quantity of second environmental data closest to the monitoring time of the first environmental data is extracted.

8. The product quality monitoring method according to claim 7, further comprising executing the following steps when the total amount of the first environmental data obtained by monitoring is less than the sample size threshold: Extracting the first quantity of quality data and the second quantity of second environment data that are closest to the monitoring time of the first environment data within a period before the monitoring time of the first environment data, so as to obtain a first group of quality data and a first group of second environment data corresponding to the first environment data; Extracting the first quantity of quality data and the second quantity of second environment data that are closest to the monitoring time of the first environment data within a period after the monitoring time of the first environment data, so as to obtain a second group of quality data and a second group of second environment data corresponding to the first environment data; The respective averages of the first set of quality data and the first set of second environmental data corresponding to the first environmental data and the first environmental data are taken as a set of matching environmental data and quality data, and the respective averages of the second set of quality data and the second set of second environmental data corresponding to the first environmental data and the first environmental data are taken as another set of matching environmental data and quality data.

9. The product quality monitoring method according to claim 1, wherein: The various environmental parameters include parameters that characterize the environmental characteristics and the waiting time between different process steps in the product manufacturing process.

10. A product quality monitoring system for a semiconductor manufacturing plant, comprising: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the product quality monitoring method as described in any one of claims 1 to 9.

Images