JP2008072047A - Model preparing device, analyzing device for process abnormality, and those method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a model preparing device by which high-precision quality-model can be prepared even when there are small number of samples of the same recipe. <P>SOLUTION: The device includes: a process data editing part 22 for extracting process feature value from process data acquired in time series during the time that a process is carried out; an inspection data editing part 27 for finding difference between a measurement value in inspection data which is information for inspection result of an object processed in the process and a design value; and an analysis part 32 which, when analysis by data mining and multivariate analysis, etc. are carried out by using the process feature value and the inspection data, a process-quality model is prepared with the difference between the measurement value obtained as a result of the process found by the inspection data editing part and the design value as target variables, and with the extracted process feature value data as explaining variables. By using the difference as the target variable, the data obtained by different recipes can be used as samples, so that required number of samples can be easily acquired, thereby the high-precision model can be prepared. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a model creation device for creating a model for analyzing an abnormality of an object to be processed and a manufacturing apparatus related to the state of a process, a process abnormality analysis device for performing analysis using the model created thereby, and It is related with those methods and programs.

  The manufacturing process of various products including a semiconductor / liquid crystal panel must be appropriately managed in order to improve the manufacturing yield of the product or maintain a good yield.

  A semiconductor device is manufactured through a semiconductor process having 100 steps or more, and is manufactured using a plurality of complicated semiconductor manufacturing apparatuses. For this reason, there are many cases where a relationship between a parameter indicating the state of each manufacturing apparatus (process apparatus) and characteristics of a semiconductor device manufactured using each manufacturing apparatus is not clearly required. On the other hand, in the semiconductor process, there is also a demand that each process must always be strictly managed so that the yield of manufactured semiconductor devices is improved.

In order to solve such a problem, in the invention disclosed in Patent Document 1, a wide variety of process data and processing results generated during process execution are acquired, analyzed by data mining, and a model of correlation between the process data and the processing results Create Using this model, the processing result can be predicted at the time of process execution.
JP 2004-186445 A

  For example, correlation between process processing results such as pattern wiring line width in an array manufacturing process of a liquid crystal display and process data and process feature data in each process such as a resist coating process and an exposure process in the array manufacturing process (= model) Is obtained, process mining and multivariate analysis are performed using process processing results such as line width as objective variables and process (feature) data as explanatory variables. Since the correlation between the process result and the process (feature amount) data is expected to change when the normal recipe, that is, the manufacturing condition changes, the analysis is performed for each recipe, and the correlation (= model) is obtained for each.

  In order to obtain a highly accurate model in data mining and multivariate analysis used when obtaining the correlation (= model) between process results and process (features) data, process result data and process (features) data must be A considerably large number of data samples are required. Also, the number of cases in which the process results are abnormal must be included in the sample.

  However, the inspection process for obtaining the process result is often a sampling inspection rather than a complete inspection, and process result data of all executed processes cannot be obtained. Further, when process processing is performed for a plurality of recipes using the same process apparatus, the number of data that can be used for data mining and multivariate analysis for obtaining a model for each recipe is further reduced. Therefore, it becomes difficult to obtain a highly accurate model, and it takes a long time to accumulate data, and the start of model creation is delayed.

  The present invention is capable of creating a model with high accuracy even when the number of samples of the same recipe is small, and a model creation device, a process abnormality analysis device, a method thereof, and a program with high accuracy of abnormality analysis using the model The purpose is to provide.

  The model creation method according to the present invention is a method for creating a model representing a correlation between a process processing result and process data and a process feature amount in the process from which the process processing result is obtained. A step of obtaining an inspection result of the target object, a step of obtaining a design value of the process, a step of extracting process feature amount data from process data obtained at the time of executing the process, a measurement value in the inspection result, and the design A step of obtaining a difference between the value and a step of creating the model using the obtained difference as an objective variable and the extracted process feature data as an explanatory variable.

  In addition, a model creation apparatus suitable for carrying out such a method includes a process data editing means for extracting a process feature amount from process data acquired in time series during a period in which the process is being executed, and processing by the process. Data mining using inspection data editing means for obtaining a difference between the measured value in the inspection data, which is information on the inspection result of the target product, and the design value, the process feature amount, and the inspection data When the analysis by multivariate analysis is executed, the difference between the measured value obtained as a result of the process processing obtained by the inspection data editing means and the design value is used as an objective variable, and the extracted process feature data is explained. And an analysis means for creating a process-quality model as a variable.

  The process abnormality analysis method according to the present invention is an analysis method in a process abnormality analysis apparatus that detects a process abnormality for each unit target product based on process data obtained at the time of execution of a process in a manufacturing system including one or a plurality of manufacturing apparatuses. A step of acquiring time series process data and storing it in the process data storage means, a step of extracting a process feature quantity from the process data stored in the process data storage means, a process feature quantity, and a process feature A process is performed by obtaining a difference between a design value and an actual measurement value from abnormality analysis rule data for performing abnormality detection of a product manufactured by the manufacturing system from a quantity, and adding the design value of the unit target product to the obtained difference. Abnormality determination that calculates the processing result and determines the presence or absence of abnormality of the product from the calculation result It was to run a step.

  In addition, a process abnormality analysis apparatus suitable for carrying out such a method detects a process abnormality for each unit target product based on process data obtained at the time of process execution in a manufacturing system including one or a plurality of manufacturing apparatuses. A process abnormality analysis apparatus, a process data storage means for storing the time-series process data, a process data editing means for extracting a process feature quantity from the process data stored in the process data storage means, and a process feature quantity An abnormality analysis rule data storage means for storing an abnormality analysis rule for detecting an abnormality of a product manufactured by the manufacturing system, and a difference between a design value and an actual measurement value is obtained from the process feature value and the abnormality analysis rule data. Process processing by adding the design value of the unit target product to the obtained difference Results were calculated, and configured to include an abnormality judging means for judging the presence or absence of the product abnormality from the calculation result.

  Further, the program according to the present invention is a process data editing means for extracting a process feature amount from process data acquired in time series during a period in which the process is being executed, and a target product processed in the process. Analysis by data mining or multivariate analysis using the inspection data editing means for obtaining the difference between the measured value in the inspection data which is information about the inspection result and the design value, the process feature amount and the inspection data When executing the process-quality model, the difference between the measured value obtained as a result of the process processing obtained by the inspection data editing means and the design value is used as an objective variable, and the extracted process feature data is used as an explanatory variable. It was made to function as an analysis means for creating.

  Further, the process data editing means for extracting a process feature quantity from time-series process data stored in the process data storage means, the process feature quantity, and an abnormality of a product manufactured by the manufacturing system from the process feature quantity Obtain the difference between the design value and the actual measurement value from the abnormality analysis rule stored in the abnormality analysis rule data storage means for performing detection, and calculate the process processing result by adding the design value of the product to the obtained difference, It can be made to function as an abnormality determination means for determining the presence or absence of an abnormality of the product from the calculation result.

  Here, the target products manufactured by the manufacturing process include semiconductors and FPDs (flat panel displays: displays using liquid crystal, PDP, EL, FED, etc.). “Unit target product” may be a target product that is grasped in a normal counting unit such as one semiconductor wafer, one glass substrate, or grasped in units of products such as one lot of these products. The target product may be a target product that is a unit of a product such as a region set on a large glass substrate. The output of the abnormality notification information includes various processes such as output to a display device, notification by mail transmission or the like, and saving in a storage device. The “product abnormality” corresponds to an abnormality in the predicted value state in the embodiment, and the “abnormality in the manufacturing apparatus” corresponds to an abnormality in the process state in the embodiment. An abnormality of a manufacturing apparatus (process apparatus) causes an abnormality in process data obtained from the manufacturing apparatus, such as a malfunction of an apparatus for executing the process of the apparatus or an abnormality of a sensor incorporated in the apparatus. including.

  In the present invention, for example, when a mask used in an exposure process is different but most other process conditions are the same, a single model can cope with different recipes representing mask types. It is to make. As a result, a large number of samples can be taken when creating a model, the accuracy of analysis is improved, and a highly accurate model can be created. And the abnormality analysis result determined based on the highly accurate model also becomes highly accurate.

  In the present invention, since the model is created using the difference between the design value and the measured value as the objective variable, the model can be created using samples of different recipes. As a result, a highly accurate model can be created even if the number of samples of the same recipe is small, and the accuracy of abnormality analysis using the model is also increased.

  FIG. 1 shows a manufacturing system including a model creating apparatus according to a preferred embodiment of the present invention. This system includes a process device 2, an inspection device 3, and a model creation device 10. These apparatuses are connected to each other by an EES (Equipment Engineering System) network 7 which is an apparatus network for exchanging process-related information more detailed than production management information at high speed. Although not shown in the figure, the EES network 7 is also connected to other process devices and inspection devices used in earlier stages of the manufacturing process and later stages. The system further includes a production management system 9 including a MES (Manufacturing Execution System) and a MES network 8 that transmits production management information connected to the production management system 9. The EES network 7 and the MES network 8 are connected via a router 12. The production management system 9 existing on the MES network 8 can access each device on the EES network 7 via the router 12.

  This manufacturing system manufactures a semiconductor and a liquid crystal panel, for example, and the process apparatus 2 executes a process for manufacturing a semiconductor, a liquid crystal panel, and the like. In the liquid crystal panel manufacturing system, a predetermined number of glass substrates to be processed are set in the cassette 1, and are used between the process device 2 and the inspection device 3 in units of the cassette, as well as devices used in the previous process with these devices. While moving between devices used in later processes, predetermined processing is performed in each device. A predetermined number of glass substrates mounted on the cassette 1 form the same lot. An RF-ID (radio frequency identification) tag 1 a is attached to the cassette 1. The tag 1a is electromagnetically coupled to the RF-ID read / write head 6 and reads / writes arbitrary data without contact, and is also called a data carrier. The tag 1a stores information such as the lot ID and the shipping time of the preceding apparatus.

  In the liquid crystal panel manufacturing system of this embodiment, since it is necessary to manage each glass substrate (target product: product in this embodiment), an ID (product ID) is assigned to each product. This product ID can be set, for example, by combining a lot ID and an identification number in the lot. That is, if the lot ID is “01201” and the number of sheets that can be set in the lot is one digit, the product ID of the first glass substrate in the lot (the identification number in the lot is “1”) is It is possible to set “012011” with the identification number in the lot added to the digit. The product ID can be set by the process data collecting device 4 built in the process device 2.

  Of course, the product IDs of all glass substrates stored in place of the lot ID or together with the lot ID are recorded in the tag 1a, and the process device 2 (process data collection device 4) is stored in the tag 1a. All product IDs may be acquired. When there is one glass substrate set in the cassette 1, the ID recorded on the tag 1a can be used as it is as the product ID.

  The process apparatus 2 is an apparatus that executes a predetermined process on the glass substrate. The process device 2 is, for example, a resist solution coating device that applies a resist solution to the surface of a glass substrate, or a device that performs process processing for exposing or developing the applied resist layer.

  The process device 2 acquires the recipe ID sent from the production management system 9 via the router 12 from the MES network 8. The process device 2 has a correspondence table between recipe IDs and processes to be actually performed, and executes processes according to the acquired recipe IDs.

  The process device 2 includes a process data collection device (process information collection device) 4. The process data collection device 4 is connected to the EES network 7. The process data collection device 4 collects process state data (process state information), which is information related to the state of the manufacturing process, in time series during a period in which the process is being executed in the process device 2. Further, a device ID is set in the process device 2, and the device ID is also passed to the model creation device 10 via the process data collection device 4.

  The process data includes, for example, the voltage and current during operation of the process device 2 and the residence time from when the process device that executes a certain process is delivered to when the process device is input to the next process. Further, when the process apparatus 2 is an exposure apparatus or a development apparatus for producing a liquid crystal panel, there are a soft bake hot plate temperature, lamp power and illuminance in exposure, and the like. The process device 2 includes a detection device for detecting these process data, and the output of the detection device is given to the process data collection device 4.

  The process data collection device 4 collects the delivery time of the previous stage device read from the tag 1a via the RF-ID read / write head 6 and the entry time to the process device 2 on which the glass substrate is set. By taking the difference between the exit time and the entry time, the residence time from the preceding apparatus can be calculated. Further, the RF-ID read / write head 6 writes the delivery time and the like to the tag 1a when the glass substrate is delivered from the process apparatus 2 as necessary.

  The process data collection device 4 has a communication function. The process data collection device 4 collects all the process data generated in the process device 2 and associates the collected process data with the lot ID and outputs it to the EES network 7. The type of data to be collected is not limited to the above, and it does not prevent obtaining more information.

  The inspection device 3 inspects the product processed by the process device 2. The inspection result data obtained by the inspection device 3 is output to the EES network 7 via the inspection data collection device 5. In addition, although the example in which the inspection data collection device 5 is built in the inspection device 3 is shown, the inspection data collection device 5 may be installed outside the inspection device 3, or conversely, the both are integrated and the inspection device is integrated. It may have a function of outputting inspection data to the EES network 7 itself. The inspection result data here includes, for example, inspection result data on the film thickness and film quality of a film layer formed on a glass substrate, the width of a pattern generated by exposure and development, and the like.

  An RF-ID read / write head 6 is connected to the inspection device 3. The RF-ID read / write head 6 reads / writes data from / to the tag 1a of the cassette 1 in which a product (glass substrate) set in the inspection apparatus 2 is stored. The read data includes a lot ID. The inspection data collection device 5 built in the inspection device 3 has a communication function, collects the inspection result data and the lot ID (product ID), associates the inspection result data with the lot ID (product ID), and EES. Output to network 7.

  The production management system 9 uses recipe No. which is information for specifying the type of process as production instruction information. (Process identification information) is sent to the process device 2. The process apparatus 2 has its recipe no. A predetermined process corresponding to is executed.

  The model creation device 10 acquires process data and inspection result data output from the data collection devices 4 and 5 via the EES network 7 and associates each data with the product ID as a key and stores them in the database 11.

  The model creation apparatus 10 is a general personal computer from the viewpoint of hardware, and each function of the apparatus is realized by an application program that runs on an operating system such as Windows (registered trademark). . The model creation device 10 also uses the database 11. The database 11 may be provided in a storage device such as an internal hard disk device or an external hard disk device included in the computer constituting the model creation device 10 or may be provided in another computer that communicates with the model creation device 10.

  The model creation device 10 includes an input device 13 such as a keyboard and an output device 14 such as a display. An operator (operator) can input operator data, maintenance data, failure data, and the like by operating the input device 13. Input information input by such an operation is also registered in the database 11. Further, the model creation device 10 has a function of creating a process-quality model based on the process data and the inspection result data.

  FIG. 2 shows an internal configuration of the model creation apparatus 10. In the model creation apparatus 10, the following processing function units are realized by application programs that run on the operating system. That is, the processing function unit of the model creation apparatus 10 includes a process data editing unit 22, a data combination unit 24, an inspection data editing unit 27, and an analysis unit 32. Each of these units can be realized by dedicated hardware (circuit).

  Furthermore, as a storage unit for storing data to be accessed by each processing function unit, a process data storage unit 21, a process feature data storage unit 23, an analysis data storage unit 31, an inspection data storage unit 28, a design value data storage A unit 29 and a design value measurement value difference data storage unit 30 are provided. Each of these storage units is provided in the database 11. However, it is not prohibited to provide each storage unit in a storage device such as a memory or a hard disk of the model creation device 10 other than the database 11 or in a storage device of another computer that communicates with the model creation device 10.

  The model creation device 10 can also be configured as follows. That is, the computer connected to the EES network 7 is a client computer that is in charge of communication with the process device 2 and the inspection device 3 and processing of the human machine interface, and a server computer that communicates with the client computer. Provided to realize each processing function unit in the server computer. In addition, the model creation apparatus 10 can be communicated with a process apparatus or the like at a production site via a communication line such as the Internet in a remote place. In addition, various modifications may be made to the configuration of the computer that implements the model creation apparatus 10 and the data transfer method.

  The process data stored in the process data storage unit 21 is associated with the product ID and the device ID (see FIG. 3A). The process data includes date and time information (date + time) when the process data is collected in addition to various process data collected by the process data collection device 4.

  The process data includes process control data and process detection data. The process control data is a state of control data and a control signal for the process device 2 to operate. The control data or control signal includes a soft bake hot plate temperature setting value, a lamp power setting value, an illuminance setting value, a resist solution flow rate setting value, a resist solution temperature setting value, a resist solution opening / closing valve ON / OFF, and the like. The process detection data is data acquired by various detectors of the process apparatus 2 and includes soft bake hot plate temperature, lamp power, illumination, resist film thickness, timing signal, ambient temperature, and the like.

  The process data storage unit 21 stores process data in chronological order according to date and time information for each lot ID or product ID. The process data storage unit 21 includes temporary storage means such as a ring buffer, and deletes process data (overwrites new process data) at a predetermined timing after the end of the process.

  The process data editing unit 22 calls time-series process data stored in the process data storage unit 21 and calculates a process feature amount in units of lots. The process feature amount is not limited to, for example, a value calculated from process data values such as the peak value, sum total, and average value of the process data for the same product ID, and the process data value exceeds a set threshold. There are various things such as time.

  The process data editing unit 22 acquires the recipe ID output from the production management system 9 together with the product ID. A recipe is a set of commands, settings, and parameters for a predetermined process apparatus, and there are a plurality of recipes depending on processing objects, processes, and apparatuses, and are managed by the production management system 9. Each recipe is given a recipe ID. The recipe for the glass substrate processed by the process apparatus 2 is specified by the product ID, the apparatus ID, and the recipe ID.

  The process data editing unit 22 acquires a set of product ID, device ID, and recipe ID possessed by the production management system 9 shown in FIG. First, the process data editing unit 22 accesses the production management system (MES) 9 and searches for the corresponding recipe ID using the product ID of the glass substrate to be analyzed and the device ID that identifies the process device 2 as keys. Next, the process data editing unit 22 acquires the retrieved recipe ID from the production management system 9 directly or via the process data collection device 4. When acquiring via the process data collection device 4, the process data collection device 4 acquires the recipe ID of the process in progress from the production management system (MES) 9, and combines the device ID of the process device 2 and the process data. You may make it pass to the model production apparatus 10. FIG.

  The process data editing unit 22 combines the calculated process feature data and the acquired recipe ID using the product ID and the device ID as keys, and the combined data is the process feature data for the corresponding device ID. Store in the storage unit 23. Therefore, the data structure of the process feature data storage unit 23 is as shown in FIG.

  On the other hand, the inspection data given from the inspection data collection device 5 is stored in the inspection data storage unit 28 in association with the product ID and the device ID (see FIG. 4A). The inspection result data stored in the inspection data storage unit 28 is raw data of detection results obtained by the inspection device 3.

  The inspection data editing unit 27 accesses the inspection data storage unit 28 and acquires the measurement value. The measurement value acquired at this time is associated with the product ID and the device ID. Next, the inspection data editing unit 27 searches the information (FIG. 4B) output from the product management system (MES) 9 with the product ID and device ID of the measurement value, and acquires the recipe ID. Then, the inspection data editing unit 27 searches the design value data storage unit 29 with the device ID and recipe ID of the acquired measurement value, and acquires corresponding design value data. The data (device ID + recipe ID + design value data: FIG. 4C) stored in the design value data storage unit 29 is registered in advance via the input device 13 or the network. The inspection data editing unit 27 obtains the difference between the acquired measurement value and the design value, and stores the difference data in the design value measurement value difference data storage unit 30 together with the product ID, device ID, and recipe ID (FIG. 4D). reference).

  In addition, the acquisition of the recipe ID in the inspection data editing unit 27 is performed using the recipe ID acquired by the inspection device 3 from the production management system 9 as one of the inspection raw data via the inspection data collection device 5, as in the process data editing unit 22. You may make it acquire.

  The data combining unit 24 acquires each data stored in the process feature value data storage unit 23 and the design value measurement value difference data storage unit 30, and obtains difference data, device ID, recipe between the process feature value data and the design value measurement value IDs are combined in units of product IDs (FIGS. 5A and 5B). Then, the data combining unit 24 stores the combined data in the analysis data storage unit 31. The data structure stored in the analysis data storage unit 31 is a table structure in which product ID, device ID, recipe ID, process feature value, and difference data of design value measurement values are associated (FIG. 5C).

  In this embodiment, all the data generated by combining by the data combining unit 24 is stored in the analysis data storage unit 31 and is analyzed. However, the analysis data storage unit 31 or separately prepared The data obtained by storing in the combined data storage unit, reading the stored combined data, and eliminating the abnormal data of the process feature amount may be used as analysis data. Abnormal data means, for example, data including numerical values that are impossible in practice. Deletion of such data can be realized by a method of preprocessing analysis data that is generally performed.

  The analysis unit 32 reads out the analysis data stored in the analysis data storage unit 31, performs analysis according to a predetermined procedure, and creates a process-quality model that is a set of process state rules for generating non-defective products or defective products. To do. The process-quality model obtained by the analysis unit 32 is stored and retained in the database 11 or the like as analysis result data and used for subsequent evaluation.

  This process-quality model is a model that represents a correlation between data mining or multivariable analysis using design value measurement value difference data as an objective variable and process feature data as an explanatory variable. Furthermore, as data mining and multivariate analysis, normalization test processing, PLS linear regression analysis processing, correlation analysis processing, principal component analysis processing, cluster analysis processing, moving principal component analysis processing, stepwise processing, Q statistic analysis processing, Techniques such as linear regression analysis processing, T2 statistic analysis processing, decision tree analysis processing, correlation diagram processing, Mahalanobis distance analysis processing, and the like can be combined.

  As a basic algorithm for analysis processing for creating a model, for example, the technology of the invention disclosed in Japanese Patent Application Laid-Open No. 2004-186445 and Japanese Patent Application Laid-Open No. 2005-197323 can be used.

  Here, in the present embodiment, the objective variable used when performing the analysis is “difference between the measured value obtained as a result of the process and its design value”. That is, for example, as shown in FIG. 6, in the process of forming a pattern, there are some processes that have the same most other process conditions except that the mask used in the exposure process is different. In other words, if the difference in mask pattern is used as a recipe, the number of recipes is equal to “the number of size patterns × n layers”. On the other hand, the manufacturing conditions such as the resist film thickness, soft bake hot plate temperature, and development processing time in development are not different depending on the mask pattern.

  In such a case, since the masks to be used are different, at least the recipe information for the masks to be used is different. Conventionally, a model is created for each mask. On the other hand, in the present embodiment, a single model can cope with different recipes representing mask types.

  In other words, using the fact that the difference between the measured value design value is not related to the recipe, first, a model is created using the difference d between the measured value and the design value as the objective variable and the process data and process feature data as the explanatory variables. Then, one model is created regardless of the recipe. The predicted value of the actual measurement value can be obtained by adding the difference value between the measurement value obtained from the model and the design value to the design value in the recipe.

  Specifically, if the design value (definition for each recipe) is D (r), the difference between the design value and the measurement value is d, and the recipe type is r, the relationship between the measurement value and the design value is expressed by the following equation. It is done. d is a provisional objective variable (= prediction target).

y = D (r) + d (1)
The explanatory variable is process (feature amount) data, and examples thereof include resist film thickness, exposure lamp power, and development time. If each of these process feature data is Xi (i = 1, 2,..., P), d is
d = b + a1 · X1 + a2 · X2 + ... + ap · Xp (2)

It can be defined by the multiple regression equation.

  Since one generic model may be created by modeling the difference between the measured value and the design value, the model analysis may be performed once. Then, the actual prediction target (for example, line width) y is calculated by substituting d obtained in Expression (2) and the design value into Expression (1).

  FIG. 8 is a flowchart illustrating functions of the analysis unit. First, the analysis unit 32 acquires a design value and a measurement value from the design value data and the inspection result data using the recipe ID as a key, and calculates the difference (S1). Next, the analysis unit 32 concatenates the “difference between the design value and the measured value” and the “process data, process feature data” with the product ID as a key to be analysis data (S2). The analysis unit 32 extracts analysis data using the recipe ID that is the target of the generic model (S3). Then, the analysis unit 32 uses the difference between the design value and the measured value as an objective variable, process data, and process feature data as explanatory variables, and performs data mining and multivariate analysis to obtain a generic model (S4).

  FIG. 9 shows a manufacturing system including the process abnormality analyzer 40. As apparent from comparison with FIG. 1, a process abnormality analysis device 40 is arranged instead of the model creation device 10. In addition, an abnormality display device 50 is connected to the EES network 7. In this embodiment, the presence or absence of abnormality is determined in real time without waiting for the inspection result in the inspection apparatus. The process abnormality analysis apparatus 40 has an abnormality detection and classification function that uses the process-quality model generated by using the model creation apparatus 10 and predicts the quality of the product that is being processed and identifies the cause of the abnormality. is doing. The process abnormality analysis apparatus 40 is a general personal computer from the viewpoint of hardware, and each function of the apparatus is realized by an application program that runs on an operating system such as Windows (registered trademark). Yes. Components similar to those of the embodiment of FIG. 1 are denoted by the same reference numerals, and detailed description thereof is omitted.

  FIG. 10 shows the internal configuration of the process abnormality analyzer 40. The process abnormality analysis device 40 includes a process data storage unit 41 that stores the process data of the process device 2 sent from the process data collection device 4, and a process feature amount from various process data stored in the process data storage unit 41. Based on the process feature value data stored in the process feature value data storage unit 43 and the process feature value data storage unit 43 for storing the process feature value calculated by the process data edit unit 42 An abnormality determination unit 44 that determines whether there is an abnormality, a determination result data storage unit 48 that stores a determination result of the abnormality determination unit 44, and an abnormality detection factor analysis rule that is used when the abnormality determination unit 44 performs a determination process. An abnormality detection factor analysis rule data storage unit 46 to be stored and an abnormality detection factor analysis rule data storage thereof 46, an abnormality detection factor analysis rule editing unit 45 for adding / changing an abnormality detection factor analysis rule, and a design value data storage unit 49 for storing design value data used for abnormality determination. Yes. Each storage unit may be set in a storage device (database) external to the process abnormality analysis device 40 or may be provided in an internal storage device.

  The data structures of the process data storage unit 41 and the process feature amount data storage unit 43 are the same as those of the process data storage unit 21 and the process feature amount data storage unit 23 in the model creation apparatus 10. The function of the process data editing unit 42 is the same as that of the process data editing unit 22 in the model creation device 10.

  The abnormality detection factor analysis rule editing unit 45 acquires a model obtained by the above-described model creation device 10 or manual analysis, defines an abnormality analysis rule, and stores it in the abnormality analysis rule data storage unit 46.

  Furthermore, the abnormality analysis rule editing unit 45 also registers abnormality notification information corresponding to the abnormality analysis rule. Thereby, as shown in FIG. 11, the data structure of the abnormality analysis rule data storage unit 46 has a table structure in which the product ID, the recipe ID of the process device, the abnormality analysis rule, and the abnormality notification information are associated with each other. .

The abnormality notification information includes information for displaying the result determined based on the abnormality analysis rule on the abnormality display device 50, information for specifying an output destination such as a notification destination for notifying the determination result, and specific notification contents There is. The notification destination is, for example, a mail address of a person in charge. Both the abnormality display device 50 and the notification destination may be registered, or only one of them may be registered. When a plurality of output destinations are provided, for example, it is possible to classify according to the degree of abnormality or the abnormality location obtained by the determination, and distribute according to the classification. A plurality of abnormality display devices, notification destinations, and notification contents can be designated for one classification. The anomaly analysis rule can be used in combination with techniques such as multiple linear regression, PLS linear regression, decision tree, Mahalanobis distance, principal component analysis, moving principal component analysis, DISSIM, Q statistic, and T ² statistic. .

  For example, classify the results obtained by the abnormality detection rules (abnormality judgment formulas) according to the degree of abnormality, and associate the notification destination and notification content with each. Specify multiple notification destinations and notification contents per category. Furthermore, a device ID and a recipe ID may be added as a key for searching for a rule. In the case of a rule including process feature values of a plurality of devices, the device ID and recipe ID of each device can be included.

  The abnormality determination unit 44 includes an abnormality analysis unit 44a, an abnormal process data storage unit 44b, an abnormality notification unit 44c, and a determination result storage unit 44d. The abnormality analysis unit 44 a uses the abnormality detection factor analysis rule stored in the abnormality detection factor analysis rule data storage unit 46 to perform abnormality determination according to the process feature amount read from the process feature amount data storage unit 43. More specifically, the abnormality analysis unit 44a has a function of executing the flowchart shown in FIG.

  The abnormality analysis unit 44a accesses the process feature value data storage unit 43, extracts process feature value data for the product to be processed, and acquires recipe information (ID) (S11). Next, the abnormality analysis unit 44a accesses the design value data storage unit 49 using the acquired recipe ID as a key, and acquires the design value, upper limit, and lower limit value of the recipe (S12). The data structure of the design value data storage unit 49 is as shown in FIG. Further, the abnormality analysis unit 44a acquires an abnormality detection factor analysis rule using the generic model from the abnormality detection factor analysis rule data storage unit 46 (S13).

  Then, the abnormality analysis unit 44a applies the process feature amount to the acquired abnormality detection factor analysis rule, predicts the difference between the design value and the measured value, and uses the predicted value as the design value of the processing target product (specifying the recipe ID). ) Are added to calculate a predicted value of the process processing result (S14). The generic model created by the model creation device 10 of the present embodiment predicts “difference between design value and measurement value” corresponding to abnormality analysis of products having a plurality of recipe IDs. Therefore, various analysis / analysis algorithms similar to the conventional ones are executed using such a model to obtain “difference between design value and measurement value”. Since what is finally desired is the predicted value of the process processing result, the abnormality analysis unit 44a obtains the predicted value by adding the design value to the obtained difference.

  Then, it is determined whether or not the evaluation (predicted value calculation) of all the specified determination formulas has been completed (S15). If it has not been completed, an abnormality determination is performed (S16). This abnormality determination is performed for each product to be processed (in this example, one glass substrate / liquid crystal panel).

When an abnormality is detected, display is performed on the abnormality display device, or other designated notification destination is notified (S17). When an abnormality is detected, the corresponding process data is stored (permanently) as abnormal process data and can be reanalyzed. Further, the result of the abnormality determination is stored together with the PLS predicted value, the predicted value contribution rate, the Q, T ² statistic, and the like (S18).

  When an abnormality is detected by the abnormality analysis unit 44a, the abnormal process data storage unit 44b reads out process data for a product determined to be abnormal from the process data storage unit 41 and stores abnormal process data as abnormal process data. Stored in the unit 47. At this time, the abnormality determination result (value of y) may be associated and registered.

  The abnormality output unit 44c outputs an abnormality message to the specified abnormality display device when an abnormality is detected by the abnormality analysis unit 44a. The abnormality message to be output is stored in the abnormality analysis rule data storage unit 46. In addition, when abnormal factor analysis is performed, detailed data such as contribution rate is also output. Furthermore, the abnormality output unit 44c also has a function of outputting an abnormality message by a specified method for a specified abnormality notification destination when an abnormality is detected by the abnormality analysis unit 44a. As an example, the abnormality output unit 44c transmits mail to a designated address. The abnormality message to be output is stored in the abnormality analysis rule data storage unit 46. In addition, when abnormal factor analysis is performed, detailed data such as a contribution rate may be output together.

The abnormality result storage unit 44d stores the result of abnormality determination in the abnormality analysis unit 44a in the determination result data storage unit 48 as determination result data. That is, the result of the abnormality determination is stored together with the PLS predicted value, the predicted value contribution rate, the Q, T ² statistic, and the like, and can be searched from the abnormality display device or the like. This determination result data may be stored for all determination results, or may be stored only when determined to be abnormal.

It is a figure which shows an example of the manufacturing system containing the model production apparatus which is suitable embodiment of this invention. It is a figure which shows an example of the internal structure of a model production apparatus. It is a figure which shows an example of the data structure of the various data which a model production apparatus processes. It is a figure which shows an example of the data structure of the various data which a model production apparatus processes. It is a figure which shows an example of the data structure of the various data which a model production apparatus processes. It is a figure explaining the example from which a recipe is different. It is a figure which shows an example of the data structure of a design value data storage part. It is a flowchart which shows the function of an analysis part. It is a figure which shows an example of the manufacturing system containing the process abnormality analyzer which is suitable embodiment of this invention. It is a figure which shows an example of the internal structure of a process abnormality analyzer. It is a figure which shows an example of the data structure of the various data which a process abnormality analysis apparatus processes. It is a flowchart which shows the function of a determination part.

Explanation of symbols

1 cassette 1a tag 2 process device 3 inspection device 4 data collection device 5 inspection data collection device 6 RF-ID read / write head 7 EES network 8 MES network 9 production management system 10 model creation device 11 database 12 router 13 input device 14 display Device 21 Process data storage unit 22 Process data editing unit 23 Process feature data storage unit 24 Data combining unit 27 Inspection data editing unit 28 Inspection data storage unit 29 Design data storage unit 31 Analysis data storage unit 32 Analysis unit 40 Process abnormality analysis Device 41 Process data storage unit 42 Process data editing unit 43 Process feature data storage unit 44 Abnormality determination unit 44a Abnormality analysis unit 44b Abnormal process data storage unit 44c Abnormality notification unit 44d Determination result storage unit 45 Abnormality detection factor analysis loop Edit unit 46 abnormality detection factor analysis rule data storage unit 47 abnormal process data storage unit 48 determination result data storage unit 49 design value data storage unit

Claims (6)

A method of creating a model representing a correlation between a process processing result and process data and a process feature amount in the process from which the process processing result is obtained,
Obtaining an inspection result of the target product obtained as a result of executing the process;
Obtaining design values for the process,
Extracting process feature data from process data obtained during the process execution;
Obtaining a difference between the measured value in the inspection result and the design value;
Creating the model using the obtained difference as an objective variable and the extracted process feature data as an explanatory variable;
The model creation method characterized by performing. Process data editing means for extracting process feature values from process data acquired in time series during a period in which the process is being executed;
An inspection data editing means for obtaining a difference between the measured value in the inspection data, which is information about the inspection result of the target product processed in the process, and the design value;
Using the process feature value and the inspection data, when performing analysis by data mining or multivariate analysis, the measured value obtained as a result of the process processing obtained by the inspection data editing means and its design value Analysis means for creating a process-quality model using the difference as an objective variable and the extracted process feature data as an explanatory variable;
A model creation device comprising: In a manufacturing system comprising one or a plurality of manufacturing apparatuses, an analysis method in a process abnormality analysis apparatus that detects a process abnormality for each unit target product based on process data obtained at the time of process execution,
Obtaining the time-series process data and storing it in the process data storage means;
Extracting a process feature amount from the process data stored in the process data storage means;
A difference between a design value and an actual measurement value is obtained from the process feature amount and an abnormality analysis rule data for detecting an abnormality of the product manufactured by the manufacturing system from the process feature amount, and the difference between the unit target product is calculated to the obtained difference. A process abnormality analysis method, comprising: calculating a process processing result by adding a design value; and executing an abnormality determination step for determining whether or not the product has an abnormality from the calculation result. In a manufacturing system composed of one or a plurality of manufacturing apparatuses, a process abnormality analysis apparatus that detects a process abnormality for each unit target product based on process data obtained during process execution,
Process data storage means for storing the time-series process data;
Process data editing means for extracting process feature values from the process data stored in the process data storage means;
Anomaly analysis rule data storage means for storing an anomaly analysis rule for detecting an anomaly of a product manufactured by the manufacturing system from a process feature amount;
A difference between a design value and an actual measurement value is obtained from the process feature value and the abnormality analysis rule data, a process processing result is calculated by adding a design value of the unit target product to the obtained difference, and the product is calculated from the calculation result. An abnormality determining means for determining the presence or absence of an abnormality,
A process abnormality analysis apparatus comprising: Computer
Process data editing means for extracting process feature values from process data acquired in time series during the period in which the process is being executed;
Inspection data editing means for obtaining a difference between the measured value in the inspection data, which is information about the inspection result of the target product processed in the process, and the design value;
Using the process feature value and the inspection data, when performing analysis by data mining or multivariate analysis, the measured value obtained as a result of the process processing obtained by the inspection data editing means and its design value Analysis means for creating a process-quality model using the difference as an objective variable and the extracted process feature data as an explanatory variable,
Program to function as. Computer
Process data editing means for extracting process feature values from time-series process data stored in the process data storage means;
The difference between the design value and the actual measurement value is obtained from the process feature value and the abnormality analysis rule stored in the abnormality analysis rule data storage means for detecting the abnormality of the product manufactured by the manufacturing system from the process feature value. An abnormality determination means for calculating a process processing result by adding the design value of the product to the difference and determining the presence or absence of an abnormality of the product from the calculation result;
Program to function as.

Images