JP2006319220A - Unit, method, and program for fault equipment estimation, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accurately estimate fault equipment causing quality deterioration via multiple pieces of manufacturing equipment and that of products manufactured through multiple routes in the manufacturing process. <P>SOLUTION: A fault equipment estimation unit 200 comprises a history DB 102 for storing product history information in a manufacturing system, a quality DB 103 for storing quality information, an analysis data preparation 106 for searching the history DB 102 and quality DB 103 for target data and then preparing analysis data, and a data analyzer 107 for estimating the fault equipment by using this analysis data. The data analyzer 107 has a fault event classification 108 for classifying analysis data by manufacturing equipment and a fault equipment evaluator 109 for estimating the manufacturing equipment causing the quality deterioration of products. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention provides an abnormal facility for estimating an abnormal facility that causes a failure when a product is manufactured in a manufacturing system having many processes and facilities, such as a semiconductor device manufacturing process. The present invention relates to an estimation device, an abnormal equipment estimation method, and the like.

  A semiconductor device is manufactured by repeatedly performing a series of processes such as cleaning, film formation, resist coating, exposure, and etching on a wafer as many as a predetermined number of layers to form wirings and circuit elements. Each of these processes includes elements that can cause malfunction of the semiconductor device, such as contamination of foreign matter and defective formation of a layer film. Therefore, when quality characteristics are deteriorated by various inspections performed during the manufacturing process, there are many possible manufacturing equipments, and identify which manufacturing equipment caused the abnormality. It was difficult.

  Therefore, as a method for analyzing the quality information and manufacturing history information on the quality characteristics of the manufactured semiconductor device and estimating the manufacturing equipment that causes the abnormality, pay attention to the variation in quality characteristics between the equipment arranged in each process. Thus, a method for estimating an abnormal facility has been proposed. For example, in the semiconductor device manufacturing apparatus (or manufacturing method) disclosed in Patent Document 1, first, as shown in FIG. 20, the quality characteristics of the wafers processed in each manufacturing facility arranged for each process are tabulated. To do. Subsequently, a value obtained by comparing the characteristic value variation 2001 between the manufacturing facilities arranged in the process with the characteristic value variation 2002 in the manufacturing facility is quantified, and the process having the largest value may be most abnormal. Estimated high.

As another method, a method has been proposed in which attention is paid to a difference between an average value of characteristic values of the entire wafer and an average value of characteristic values of wafers processed in individual manufacturing facilities. For example, the abnormality cause detection apparatus (or abnormality cause detection method) disclosed in Patent Document 2 takes into account the variation 2101 in the characteristic values of the entire wafer and the number of processes 2102 for each manufacturing facility, as shown in FIG. After that, the difference 2103 between the average value of the characteristic values of the entire wafer and the average value of the characteristic values of the wafers processed in each manufacturing facility is obtained. Then, based on the difference 2103, the amount of abnormality in each manufacturing facility is quantified, and the facility having the largest value is estimated to be an abnormal facility.
JP2000-12640 (release date: January 24, 2000) JP 2002-43200 (Publication date: February 8, 2002)

  However, in the technologies described in Patent Documents 1 and 2, there are a plurality of facilities that are in charge of the same manufacturing process, and an abnormality caused by different types of facilities that perform different manufacturing processes affects each other. There is a problem that correct estimation cannot be performed.

  For example, in the example shown in FIG. 11, the defect rate is high only in the wafer group 1101 processed in the manufacturing facility A1 in the process A and processed in the manufacturing facility B1 in the process B. As described above, the deterioration of quality characteristics is not always caused only by a specific facility in charge of a single process, but often occurs by a combination of specific facilities. However, in the techniques described in Patent Documents 1 and 2, abnormal facilities are estimated on the assumption that only a specific facility in charge of one process causes an abnormality. For this reason, when a deterioration in quality characteristics occurs in a specific combination of facilities, it is impossible to correctly estimate the abnormal facility that is the cause.

  In addition, in the estimation of abnormal facilities in Patent Documents 1 and 2, when quantifying an abnormality, a statistical test is performed on the assumption that a normal population is uniformly distributed according to a normal distribution. Yes. Therefore, the distribution of the quality characteristic of the wafer assumed in these documents is a distribution centered on a certain average value as shown in FIG. 22, for example. If all the manufacturing processes in the same manufacturing facility arranged in each process are uniform, the distribution of quality characteristics is expected to be close to the distribution as shown in FIG. For example, when the deterioration of quality characteristics is due to a single abnormal event (that is, the product is manufactured with a single manufacturing path that identifies the manufacturing equipment responsible for each process, and the deterioration of quality characteristics is placed in the manufacturing process) 22), the difference between the quality characteristic acquired from the manufacturing process and the distribution as shown in FIG. 22 is statistically tested to estimate an abnormal manufacturing facility. be able to.

  However, in an actual manufacturing process, a wafer is processed using a number of manufacturing facilities in various processes, and even if the manufacturing equipment of the same type is arranged in the same manufacturing process, it is processed by each manufacturing facility. A small amount of individual difference occurs. For this reason, even if no abnormality has occurred, the distribution of quality characteristics of the wafer is not uniform as shown in FIG. 22, and abnormal manufacturing facilities are found due to the difference between the distribution of normal quality characteristics and abnormal quality characteristics. Is difficult to estimate. In many cases, abnormalities that occur simultaneously in different processes cause different abnormalities to deteriorate the same quality characteristics, and wafers whose quality characteristics are deteriorated due to different causes are mixed in the entire manufactured wafer. As shown in FIG. 11, a wafer group 1101 processed by the manufacturing facility A1 and the manufacturing facility B1, a wafer group 1102 processed by the manufacturing facility A1 and the manufacturing facility B2, and a wafer group 1103 processed by the manufacturing facility Am. When deterioration of different quality characteristics occurs, a plurality of distributions are mixed in the inspection result as shown in FIG. In this case, the quality characteristic has a more complicated distribution.

  Here, reference numerals 1201, 1202, and 1203 shown in FIG. 12 correspond to 1101, 1102, and 1103 in FIG. 11, respectively. In this specification, compared to a distribution with a low defect rate as shown at 1202 in FIG. 12, the distribution of wafers with reduced quality characteristics as shown at 1201 and 1203 (that is, due to the combination of the same abnormal equipment). Each of these is called an abnormal event. As shown in FIG. 12, when the quality information of the wafer to be analyzed includes a plurality of abnormal events, that is, the product is manufactured by various manufacturing paths, and each manufacturing path is caused by a different manufacturing facility. In the case where quality degradation occurs, the techniques disclosed in Patent Documents 1 and 2 cannot correctly estimate abnormal equipment.

  The present invention has been made in view of the above-described problems, and the object thereof is a product manufactured through a plurality of manufacturing paths existing in the manufacturing process due to a decrease in quality caused by a plurality of manufacturing facilities. It is an object of the present invention to provide an abnormal equipment estimation device and an abnormal equipment estimation method capable of accurately estimating the abnormal equipment that causes the deterioration of quality.

  In order to solve the above-mentioned problem, the abnormal equipment estimation apparatus according to the present invention includes a plurality of processing facilities and a plurality of manufacturing facilities for performing the same processing, and a plurality of manufacturing facilities for each processing step. An abnormal equipment estimation device for estimating a manufacturing equipment that causes a reduction in the quality of the product in a manufacturing system that manufactures a product by performing processing using any of the above, wherein the product manufactured in the manufacturing system Quality information acquisition means for acquiring the quality information in which the identification information and the quality inspection result of the product are associated from the manufacturing system, the product identification information, and the identification information of the manufacturing facility in charge of each process at the time of manufacturing the product The history information acquisition means for acquiring the history information associated with and from the manufacturing system, and the inspection result and the manufacturing equipment identification information for each product identification information Data analysis means for generating analysis data, and when the above analysis data is classified for each manufacturing facility, select the analysis data in order of decreasing variation in quality results in the classified analysis data and manufacture the analysis data Classifying means for sequentially classifying into the analysis data group for each facility, and calculating the average value of each quality inspection result included in the analysis data group classified by the classification means, and the average in each classified analysis data group It is characterized by comprising estimation means for calculating the degree of influence of each manufacturing facility on the quality degradation by comparing the values with each other and estimating the manufacturing equipment that causes the product quality degradation.

  The abnormal equipment estimation apparatus of the present invention comprises a plurality of processing steps, and a plurality of manufacturing equipments for performing the same processing are provided, and processing is performed using any of the plurality of manufacturing equipments for each processing step. This is for estimating an abnormal facility that causes a deterioration in the quality of a product manufactured by a manufacturing system such as a semiconductor device that manufactures the product. In this abnormal equipment estimation device, for example, quality information acquisition means for acquiring the quality information such as the quality inspection result of each product performed by the inspection process in the manufacturing system from the manufacturing system, and at the time of manufacturing the product History information acquisition means for acquiring history information indicating which manufacturing facility is in charge of each process from the manufacturing system is provided.

  Furthermore, the abnormal equipment estimation device includes data for creating analysis data indicating the manufacturing history and quality information (quality inspection result) of each product manufactured in the manufacturing system in association with the quality information and history information. Classifying the analysis data by manufacturing equipment, in order from the one with the smallest variation in quality results in the classified analysis data, when creating means and analysis data are classified by manufacturing equipment Means and estimation means for calculating the degree of influence of each classified manufacturing facility on quality deterioration and estimating manufacturing equipment causing quality deterioration are provided.

  In the abnormal equipment estimation apparatus of the present invention, when analysis data is classified for each manufacturing equipment in a manufacturing system composed of a plurality of processes and manufacturing equipment, variations in quality results included in the classified analysis data group are reduced. The analysis data is selected in order from the product, and the analysis data is sequentially classified for each series of manufacturing equipment that has processed each product, and the impact on the quality deterioration is calculated for each classified manufacturing equipment. Estimated production facilities. Therefore, it is possible to accurately estimate the abnormal equipment that causes the deterioration of quality caused by multiple manufacturing facilities and the quality of products manufactured through multiple manufacturing paths existing in the manufacturing process. be able to. And an exact measure can be quickly implemented with respect to the abnormality which has generate | occur | produced in the manufacturing system.

  In addition, according to the above configuration, in a series of manufacturing equipment paths related to product manufacturing, how much each manufacturing equipment has a relatively low impact on quality compared to other manufacturing equipment. Can be confirmed. Therefore, abnormal equipment can be estimated more accurately.

  In the abnormal equipment estimation apparatus according to the present invention, the classification means classifies the analysis data using a decision tree analysis technique using the quality information as an objective variable and the history information as an explanatory variable, and the estimation means includes: By calculating the influence of the classification by the lower node of the decision tree created by the classification means on the classification by the higher node, the influence of the manufacturing facility assigned to each node on the quality degradation is calculated, It is preferable to estimate a production facility that causes a reduction in product quality.

  The decision tree analysis method is a method of classifying the relationship between each item (namely, quality information and history information) in analysis data into a tree shape, and is a method used for classification of a large amount of data.

  According to the above configuration, a series of products processed in the manufacturing process by classifying the analysis data using the decision tree analysis technique with the quality information as an objective variable and the history information as an explanatory variable. A decision tree is created in which each manufacturing facility (that is, the route of the manufacturing facility) is represented in a tree shape. Then, the estimating means calculates the degree of influence of the manufacturing facility assigned to each node on the quality degradation by calculating the degree of influence of the classification by the lower node of the decision tree on the classification by the upper node.

  As a result, in the series of manufacturing equipment related to product manufacturing, the degree of influence that each manufacturing equipment has on the quality degradation relative to other manufacturing equipment is accurately calculated. be able to. Therefore, the abnormal equipment can be estimated more accurately.

  In the abnormal equipment estimation apparatus according to the present invention, the quality information includes evaluation value information relating to arbitrary quality, and the estimation means is classified according to the number of data classified by the lower nodes and the lower nodes. The degree of influence of the lower node on the upper node may be calculated based on the average value of the evaluation value in the data and the average value and standard deviation of the evaluation value of the data classified by the upper node.

  In the abnormal equipment estimation apparatus of the present invention, it is preferable that the data creation means evaluates the quality information of the product step by step using cluster analysis.

  Here, “product quality information is evaluated step by step” means that each product manufactured with respect to its quality is ranked based on the quality inspection results obtained in the quality inspection process performed in the manufacturing system. Means to do.

  According to the above configuration, the quality information is divided into a predetermined number (usually about several ranks) of ranks, so that data classification processing and influence degree calculation performed thereafter can be easily performed. In addition, this clustering can be performed automatically by using cluster analysis.

  The abnormal equipment estimation apparatus of the present invention preferably classifies data into two according to whether or not the condition is met in the decision tree analysis method.

  According to said structure, the process in a classification means can be simplified and processing time can be shortened.

  In addition, the abnormal equipment estimation method according to the present invention includes a plurality of processing steps, and a plurality of manufacturing facilities for performing the same processing are provided, and each processing step is performed using any one of the plurality of manufacturing facilities. An abnormal equipment estimation method performed in an abnormal equipment estimation device for estimating a manufacturing equipment that causes a reduction in quality of the product in a manufacturing system that performs processing to manufacture a product, wherein the quality information acquisition means includes the manufacturing A quality information acquisition step for acquiring from the manufacturing system the quality information in which the identification information of the product manufactured in the system and the quality inspection result of the product are associated with each other, and the history information acquisition means includes the product identification information and the product A history information acquisition step for acquiring history information associated with identification information of a manufacturing facility in charge of each process at the time of manufacturing from the manufacturing system, A data creation step for creating analysis data in which inspection results and manufacturing equipment identification information are associated with each other for each product identification information; and a classification means for classifying the above analysis data for each manufacturing equipment. In this case, the classification step of sequentially classifying the analysis data into the analysis data group for each manufacturing facility, and the estimation means, selecting in order from the smallest variation of the quality results in the classified analysis data, The average value of each quality inspection result included in the analysis data group classified by the classification step is calculated, and the average value in each classified analysis data group is compared with each other, so that each manufacturing facility gives quality degradation. And an estimation step for estimating a production facility that causes a reduction in product quality.

  According to the abnormal equipment estimation method of the present invention, when analysis data is classified for each manufacturing equipment in a manufacturing system composed of a plurality of processes and manufacturing equipment, variations in quality results included in the classified analysis data group are reduced. The analysis data is selected in order from the product, and the analysis data is sequentially classified for each series of manufacturing equipment that has processed each product, and the impact on the quality deterioration is calculated for each classified manufacturing equipment. Estimated production facilities. Therefore, it is possible to accurately estimate the abnormal equipment that causes the deterioration of quality caused by multiple manufacturing facilities and the quality of products manufactured through multiple manufacturing paths existing in the manufacturing process. be able to. And an exact measure can be quickly implemented with respect to the abnormality which has generate | occur | produced in the manufacturing system.

  The abnormal equipment estimation device may be realized by a computer. In this case, the abnormal equipment estimation device can be realized by a computer by operating the computer as each of the means. That is, the abnormal equipment estimation program of the abnormal equipment estimation apparatus and the computer-readable recording medium on which the abnormal equipment estimation program is recorded also fall within the scope of the present invention.

  That is, the abnormal equipment estimation program according to the present invention is an abnormal equipment estimation program for operating any one of the above abnormal equipment estimation apparatuses, and is for causing a computer to function as each of the above-described means. The recording medium according to the present invention is a computer-readable recording medium in which the abnormal facility estimation program is recorded.

  As described above, the abnormal equipment estimation device according to the present invention acquires quality information in which the identification information of the product manufactured in the manufacturing system is associated with the quality inspection result of the product from the manufacturing system. Means, history information acquisition means for acquiring from the manufacturing system history information in which product identification information is associated with the identification information of the manufacturing facility in charge of each process at the time of manufacturing the product, and inspection for each product identification information Data creation means for creating analysis data in which results and identification information of manufacturing equipment are associated with each other, and when the analysis data is classified for each manufacturing equipment, variation in quality results in the classified analysis data is reduced. A classification means for selecting analysis data in order, and sequentially classifying analysis data into analysis data groups for each manufacturing facility, and an analysis data group classified by the classification means Calculate the average value of each quality inspection result included, and compare the average value in each classified analysis data group with each other to calculate the degree of influence each manufacturing facility has on quality degradation, resulting in product quality degradation And an estimation means for estimating the manufacturing equipment that causes the above.

  Therefore, according to the present invention, it is a cause for the deterioration of quality caused by a plurality of manufacturing facilities and the quality of products manufactured through a plurality of manufacturing paths existing in the manufacturing process. Abnormal equipment can be estimated accurately. And there exists an effect that an exact countermeasure can be rapidly implemented with respect to the abnormality which has generate | occur | produced in the manufacturing system.

  An embodiment of the present invention will be described below with reference to FIGS. In the present embodiment, as an example of the present invention, an abnormal facility estimation apparatus that estimates a facility that causes a deterioration in quality characteristics in a semiconductor device manufacturing process will be described as an example.

  FIG. 2 shows a schematic configuration of a semiconductor device manufacturing system 201 including an abnormal equipment estimation apparatus 200 according to the present embodiment. As shown in FIG. 2, a manufacturing system 201 represents a manufacturing line for processing an input wafer, and a plurality of processing steps (steps) 202a, 202b, 202c... (These are collectively referred to as 202x). As a result, the semiconductor device is completed. Here, the process 202x is a process divided for each manufacturing process unit for the wafer, such as a cleaning process, an exposure process, and an etching process. And, for example, at least one manufacturing facility 203 is assigned in accordance with each process, such as a cleaning apparatus for the cleaning process and a film forming apparatus for the film forming process. In FIG. 2, facilities A1, A2,... Am that perform the process A (for example, the cleaning process) in the processing step 202a are set as facilities A1, A2,... Am, and facilities B1, B2 that perform the process B (for example, the exposure process) in the processing step 202b. ,... Bn.

  Further, the manufacturing system 201 is provided with an inspection process 204 as necessary, such as a film thickness measurement inspection process for measuring the thickness of the formed thin film and a line width measurement inspection process for measuring the line width of the formed circuit. ing. Also in this inspection process 204, for example, at least one inspection facility 206 according to each inspection, such as a film thickness measurement apparatus for a film thickness measurement inspection process and a line width measurement apparatus for a line width measurement inspection process. Is assigned. In each of these inspection steps 204, it is possible to grasp whether or not the wafer has been processed normally in each processing step 200x.

  Various manufacturing equipments 203 arranged in the manufacturing system 201 are connected to the analysis server 101 in the abnormal equipment estimation apparatus 200 via a network 205. Thereby, the processing history of the wafer is transmitted from each processing step 202x to the analysis server 101 via the network 205, and the processing history of the wafer and the quality inspection result of the wafer (also referred to as quality characteristics) are transmitted from each inspection step 204. Sent. The analysis server 101 collects the transmitted processing history and quality characteristics, stores them in a database (DB: Data Base), and performs data analysis. The network 205 is connected to a display terminal 110 arranged in the manufacturing process or in the analyst's room. The display terminal 110 can display the result of analysis performed by the analysis server 101 by an analyst's operation on a monitor (not shown) provided in the terminal. As described above, the abnormal equipment estimation apparatus 200 according to the present embodiment is configured to include the analysis server 101 and the display terminal 110, and based on the wafer processing information transmitted through the network 205, the manufacturing system 201. , Estimation of abnormal equipment that causes deterioration of quality characteristics of the wafer is performed.

  In FIG. 1, the structure of the abnormal equipment estimation apparatus 200 with which the said manufacturing system 201 was equipped is shown. As shown in FIG. 1, the analysis server 101 includes a history DB 102 (history information acquisition unit) that stores a processing history (history information) of a wafer in the manufacturing system 201, a quality DB 103 (quality information acquisition unit) that stores quality information, A condition DB 104 that stores various analysis conditions, a data registration unit 105 (quality information acquisition means and history information acquisition means) that registers each data transmitted from the manufacturing equipment 203 in the DB, and a history DB 102 according to the conditions stored in the condition DB 104 The data is retrieved from the quality DB 103 and processed, and an analysis data creation unit 106 (data creation means) that creates new analysis data. The analysis data created by the analysis data creation unit 106 is used to estimate abnormal equipment. It comprises a data analysis unit 107 (classification means / estimation means) for performing.

  The data analysis unit 107 includes an abnormal event classification unit (classification unit) 108 that classifies wafer information included in the analysis data for each path (ie, abnormal event) of each manufacturing facility that has processed the wafer. An abnormal facility that analyzes the manufacturing equipment in each manufacturing path classified by the abnormal event classification unit 108, quantitatively evaluates the abnormal equipment, and estimates the manufacturing equipment that causes the quality degradation of the manufactured semiconductor device And an evaluation unit (estimating means) 109.

  In other words, the abnormal equipment evaluation unit 109 reduces the quality of the semiconductor device based on the quality characteristic value (quality information) included in the analysis data regarding the manufacturing equipment existing in each manufacturing path classified by the abnormal event classification unit 108. By calculating the degree of influence on the manufacturing facility, a manufacturing facility that causes a deterioration in the quality of the semiconductor device is estimated.

  Further, the “abnormal event” means an abnormality of the equipment that causes a certain quality abnormality. For example, an abnormality P related to quality occurs in a wafer that is a material of a semiconductor device, and this abnormality P occurs when the wafer is processed by the equipment A1 in the manufacturing process A and further processed by the equipment B1 in the manufacturing process B. Suppose you want to. Thus, when the abnormality P is caused by a combination of the equipment A1 and the equipment B1, this abnormality becomes one abnormal event. Moreover, when the same abnormality P also occurs by processing in the equipment C1 in charge of another manufacturing process C, the abnormality P is an abnormal event different from the above-described abnormal event. That is, the same abnormality generated by processing with the same equipment (this equipment is not limited to one) is called an abnormal event.

  In a manufacturing process of a semiconductor device or the like, as described above, a certain quality abnormality may occur due to a plurality of causes (abnormal events). In the abnormal equipment estimation apparatus 200 according to the present embodiment, a plurality of simultaneously occurring abnormal events are estimated from the quality information and the history information, such as the above-described abnormality P, and wafers in which the quality abnormality has occurred are respectively determined. After classifying each abnormal event, a process of estimating a device causing each abnormal event is performed.

  The result analyzed by the data analysis unit 107 is displayed on a monitor (not shown) provided in the display terminal 110. The display terminal 110 also has an input unit (not shown), and the analyst changes the analysis conditions stored in the condition DB 104 using the input unit or does not use the condition DB 104. An analysis condition can be designated for the analysis server 101. Note that the condition DB 104 stores information on analysis conditions such as, for example, how far back in the past the wafer data is to be analyzed, whether to limit the model and the type of failure, and the like.

  Next, an abnormal facility estimation method in the manufacturing system 201 will be described. First, a history information and quality information registration processing method will be described.

  In the manufacturing system 201, the manufacturing equipment 203, when processing a wafer, assigns a wafer ID (product identification information) assigned to the processed wafer, a lot ID of a lot to which the wafer belongs, a process number assigned to the process, and a process The name and ID (production equipment identification information) of the manufacturing facility that performed the processing, the processing end date and time when the processing on the wafer is completed, and other processing information are transmitted to the analysis server 101 as processing history information.

  Here, the wafer ID is a number assigned to each wafer at the time of introduction into the manufacturing system 201 so as not to overlap all the wafers. Further, the lot ID is a number assigned to each lot when it is put into the manufacturing system 201 so as not to be duplicated in all lots. The process number is a number assigned to each process so as not to overlap in all processes of the manufacturing system 201.

  In the analysis server 101, the processing history information sent from the manufacturing facility 203 is registered in the history DB 102 by the data registration unit 105 according to the format as shown in FIG. FIG. 3 is a schematic diagram illustrating an example of a data table stored in the history DB 102. Each time a wafer is processed by the manufacturing system 201, the data registration unit 105 adds data related to the processing to the data table (see FIG. 3) in the history DB. The history DB 102 may store other processing information not shown in FIG. 3 as additional information.

  In addition, for the discrimination between the wafer ID and the lot ID in each manufacturing facility 203, a conventionally known discrimination method generally used in a manufacturing system such as a semiconductor device can be used. Usually, the manufacturing system is collectively managed by a computer system called CIM. This system manages information on which equipment the lot is transported to, information on what kind of processing is performed in each equipment, and the like.

  Therefore, for example, when the wafer ID and the lot ID are determined by the CIM, the lot ID and the wafer ID are transmitted from the CIM 208 (see FIG. 2) through the network 205 when the lot is transported to the manufacturing facility 203. Is done. Thereby, each manufacturing equipment 203 can easily obtain the ID of the lot or wafer that it is processing.

  As another discrimination method, a method of reading the wafer ID and the lot ID from the barcode or IC tag attached to the lot case or the wafer itself without using the CIM may be adopted.

  Further, since the name of the manufacturing facility 203 is determined in advance for each facility, each facility stores the name, and the information is stored in the analysis server 101 together with information such as a wafer ID and a lot ID each time processing is performed. Just send it. As with the wafer ID and lot ID, the process number can be transmitted using the above-mentioned CIM, or can be read from a barcode or an IC tag and transmitted.

  On the other hand, when the inspection of the wafer is completed, the inspection facility 206 arranged in the inspection process 204 transmits the inspection result of the wafer together with the above processing history information to the analysis server 101 as quality information. The transmitted quality information is aggregated in wafer units by the data registration unit 105. For example, if it is quality information of electrical characteristic inspection, it is registered at any time in the quality DB 103 according to a format as shown in FIG. FIG. 4 is a schematic diagram illustrating an example of a data table stored in the quality DB 103. The “wafer ID”, “lot ID”, “process number” and the like shown in FIG. 4 are the same as the history information. The contents of the inspection vary depending on the type of inspection performed in the inspection process 204. Therefore, in the quality DB 103, “lot ID”, “wafer ID”, “model name”, “process number”, and “inspection date” are displayed. For items other than, different formats are prepared for each examination type. Further, the quality information may be aggregated in units of wafers inside the manufacturing facility 203 before being transmitted to the analysis server 101. In this case, the data registration unit 105 only registers quality data.

  The “model name” shown in the data tables shown in FIGS. 3 and 4 indicates the type of the semiconductor device manufactured by the manufacturing system 201, such as an IC for XX. Since the manufacturing conditions differ if the model name is different, it is desirable to record it in each data table.

  Next, the flow of analysis processing performed by the analysis server 101 will be described using the flowchart shown in FIG.

  In the analysis processing in the analysis server 101, first, the analysis data creation unit 106 creates analysis data (step S501), and then the data analysis unit 107 estimates abnormal equipment based on the analysis data (step S502).

  In the analysis data creation step S501, first, quality data to be analyzed is retrieved from the quality DB 103 in accordance with the analysis conditions stored in the condition DB 104 of the analysis server 101 (step S5011). Next, the quality data is classified by rank (step S5012). Furthermore, the history data of the wafer included in the quality data is retrieved from the history DB 102 to create history data (step S5013). Then, the analysis data is created by combining the quality data and the history data (step S5014).

  In the abnormal equipment estimation step S502, the abnormal event classification unit 108 in the data analysis unit 107 performs a decision tree analysis on the analysis data created in the analysis data creation step S501, and information on each wafer is obtained for each abnormal event. Classify. After that, the abnormal facility evaluation unit 109 evaluates the influence degree of each manufacturing facility 203 from the decision tree created by the abnormal event classification unit 108, and estimates the abnormal facility (step S5022).

  Next, each process in the analysis data creation step S501 will be described in more detail. In step S5011, the analysis data creation unit 106 determines the model name according to the model name of the analysis target wafer stored in the condition DB 104, the inspection date and time, and the quality characteristic name (for example, open defect, short defect, etc.) to be analyzed. Further, using the inspection date and time as a search key, data relating to the wafer ID and quality characteristics to be analyzed is extracted from the quality DB 103.

  The search key is not limited to the model name and the inspection date and time described above, and other information may be used. Further, the analysis data creation unit 106 may perform a search according to an instruction of an analyst using the display terminal 110 without using the condition DB 104.

  In FIG. 6A, “model name” = “AA”, “inspection date / time” = “2004/6/5 00:00:00 to 2004/6/5 23:59:59. FIG. 4 shows the result of extracting corresponding data from the data table shown in FIG. 4 with “thing” as a search key and the quality characteristic to be analyzed as “open failure rate”.

  In step S5012, the standard value stored in the condition DB 104 is compared with the quality characteristics of each data extracted in step S5011, and the characteristics of which range the quality characteristic values are included in are labeled, so that the rank of the data The division is performed (see FIG. 6B). In FIG. 7, an example of the table which shows the standard value memorize | stored in condition DB104 is shown. The table shown in FIG. 7 shows the relationship between the standard value of the open defect rate y, which is one of the quality characteristics, and the rank, and in order from the wafer having the highest quality according to the “open defect rate” value of the wafer. , “OK”, “A”, “B”, and “C”, ranks are defined. FIG. 6B shows a ranking of FIG. 6A based on this information.

  In step S5013, the history DB 103 is searched using the wafer ID included in the quality data extracted from the quality DB 103 in step S5011 as a search key, and the name of the manufacturing facility that processed the wafer in each process is extracted from the wafer history information. Line up in the direction. FIG. 8 shows an example in which history information is searched using “wafer ID” = “AAA-1” included in the quality data shown in FIG. FIG. 8A is a table showing the result of searching the history DB 103 using “wafer ID” = “AAA-1” as a search key, and FIG. 8B shows each of the tables shown in FIG. It is a table | surface which shows the result of having taken out the name of the manufacturing equipment which processed the wafer at the process, and arranging it in the horizontal direction. FIG. 8 shows a search result and a rearrangement result by using “wafer ID” = “AAA-1” as a search key as an example. In the present embodiment, such processing is performed for all wafer IDs included in the quality data.

  In step S5014, the quality data created in step S5012 and the history data created in step S5013 are combined using the wafer ID to create analysis data. FIG. 9 shows a table as an example of the analysis data. This figure shows an example of analysis data created from FIG. 6 (b) and FIG. 8 (b).

  The above is the flow of the analysis data creation process (step S501) by the analysis data creation unit 106.

Next, details of the abnormal equipment estimation step S502 performed by the data analysis unit 107 will be described.
In step S5021, when the data analysis unit 107 receives the analysis data created in step S501 from the analysis data creation unit 106, the abnormal event classification unit 108 in the data analysis unit 107 performs decision tree analysis on the analysis data. Classify by abnormal event. Then, a tree-shaped decision tree rule describing each abnormal event is created. In decision tree analysis, there are two methods of branching whether or not a single condition is met, and a method of branching into plural depending on possible conditions. In this embodiment, the latter method is used.

  FIG. 10 is an example of a decision tree created by the abnormal event classification unit, and shows a decision tree created when a decision tree analysis is performed on the analysis data shown in FIG. In FIG. 10, the wafers included in the analysis data shown in FIG. 9 are first classified according to which manufacturing equipment is processed with “process number” = “3300”. Here, the wafer processed by “equipment name” = “CA-01” is classified as 1001 in the first branch of FIG. 10, and further classified by the manufacturing facility processed by “process number” = “1320”. Is done. In decision tree analysis, a point representing a branch condition such as 1001 is called a node.

  The node creation method will be described later, but the nodes are created so that the same ranks are collected in the classified wafer group. That is, in FIG. 10, the wafer group shown in FIG. 9 is first a “CA-01” wafer group, a “CA-02” wafer group, and a “CA-03” facility with “process number” = “3300”. "", Which indicates that the data included in each data group includes many of the same ranks as compared with the case of classification using other processes.

  In addition, the wafer classified by the node 1002 in the wafer group classified by the node 1001, that is, the wafer processed by “equipment name” = “CT-02” in “process number” = “1320” is “rank”. = “OK” is gathered. On the other hand, wafers processed with “equipment name” = “CT-01” are further classified according to manufacturing equipment processed with “process number” = “0820”. By repeating the above branching for all the wafers until reaching the lowest node having no branching, the original data group is classified into a plurality of wafer groups in which the same rank is concentrated. In other words, by tracing the branch from the first node to the last node, it is possible to determine which rank the quality data of the wafer processed by the combination of the process number assigned to that node and the manufacturing equipment is most frequently ranked. Can be estimated.

Therefore, the decision tree created by the analysis can be rewritten into a simple If-Then rule using history information. For example, the node 1002 in FIG. 10 can be rewritten to the following rule.
“If“ Step (3300) ”=“ CA-01 ”AND“ Step (1320) ”=“ CT-02 ”Then“ Rank ”=“ OK ””
This rule indicates that wafer groups processed by a combination of specific manufacturing facilities are classified into the same group, and the wafers classified into the group have characteristics of similar quality characteristics. In decision tree analysis, all data reaches one of the conclusions according to the rules expressed in the decision tree. Therefore, analysis data is subjected to decision tree analysis to find and divide a group of analysis data common to both processing equipment and quality characteristics. That is, the process of step S5021 is to separate individual abnormal events from quality information in which a plurality of abnormal events are mixed as shown in FIG. 12, using the conditions resulting from one or more manufacturing facilities shown in FIG. It is.

  The decision tree analysis used in the present invention is a general-purpose technique used for classification and prediction of a large amount of data, and several methods have already been proposed for a branch creation method and a useless branch avoidance method. Therefore, detailed description is omitted here, and only rough processing common to each method will be described.

  FIG. 13 shows a rough processing flow of decision tree analysis. In the decision tree analysis, first, an item to be classified is set as an objective variable, an item used as a classification condition is set as an explanatory variable, and a root node is created (step S1301). In the example of analysis data shown in FIG. 9, in step S1301, the objective variable is “rank” of quality information, and the explanatory variables are “process (0001)”, “process (0002)”,. . The root node is the highest node that is the starting point of analysis, and all data belongs to this node. Note that “wafer ID” and “open failure rate” are not used in step S5021 shown in FIG.

  Next, one of the explanatory variables set in step S1301 is selected to classify the data (step S1302), and the result is evaluated according to a predetermined classification accuracy index (step S1303). As the index of classification accuracy, “Gini diversity index”, “information gain”, and the like have been proposed. This operation is repeated for each explanatory variable (step S1304), and when all the explanatory variables have been evaluated (Yes in S1304), the data is classified using the explanatory variable having the best classification accuracy. Then, a new node is created (step S1305).

  In the above S1302 to S1304, the data is divided from the processing history (history information) in each process on the condition of the processing equipment, for example, how the data is neatly divided under each condition using the variation in quality characteristics as an index. Is evaluated. The fact that each data group is divided neatly indicates that the quality of data belonging to each divided data group is similar and that there is a difference in the quality of the data groups. Therefore, adopting the explanatory variable having the best classification accuracy in S1305 to create a new node means selecting a facility having the most difference in quality characteristics and creating a new node. In other words, the process of sequentially creating nodes in this series of steps means performing an operation of successively extracting equipment that has a difference in quality characteristics.

  Then, the processes from step S1302 to S1305 are repeated until a predetermined end condition is satisfied (step S1306). The “predetermined end condition” referred to here is, for example, until the number of data divided by the node is equal to or less than a predetermined number, or until the variation is reduced to some extent as compared with the original data. And the like.

  If the termination condition is not satisfied (No in S1306), a new classification condition (node) is added (step S1307), and data is classified (step S1302). A decision tree is created by performing such processing.

  As described above, in the decision tree analysis according to the present embodiment, the result of dividing the data by all combinations that can be divided by the node is compared, and the manufacturing that focuses on the node that can be divided most “cleanly” is focused on that node. Adopted as equipment. Here, the “most beautiful” means, for example, using a value such as a variance value or an information amount (entropy) to calculate the variation of each classified data group of quality results, and each is the most organized ( That is, the variation is small.

  That is, when the analysis data is classified for each manufacturing facility, the abnormal event classification unit 108 selects the analysis data in order of decreasing variation in the quality results in the classified analysis data, and processes the analysis data for each product. The decision tree is created by performing a process of sequentially classifying each of the series of manufacturing facilities that have been performed.

  Note that the above-described root node is a node created first. That is, the root node created in S1301 is a simple branch point for first dividing the analysis data, and the branch condition has not yet been determined at this point. Then, all branch conditions conceivable in S1302 to S1304 are tried. As a result, the branch condition (here, the manufacturing equipment) that can be classified most beautifully in S1305 is adopted, and at this point, the branch is determined and a new node is created at the branch destination. Further, the same processing is performed on the new node based on the data classified according to the branch condition determined earlier, and the next branch condition is determined. A decision tree is created by repeating this operation until the node satisfies the end condition.

  In step S5021, combinations of individual abnormal events and the manufacturing equipment that causes them can be extracted from the created decision tree. However, these rules are an enumeration of manufacturing equipment to categorize events most efficiently, and it is clear which manufacturing equipment actually has the most impact on quality characteristics and which manufacturing equipment should take measures. hard. Therefore, in step S5022, the influence degree of the manufacturing equipment assigned to the node is evaluated by quantitatively evaluating the influence degree that each node constituting the decision tree created in step S5021 has on the classification result. Ranking abnormal equipment.

  Next, a more detailed processing flow in step S5022 will be described based on the flowchart shown in FIG. First, the number N of data classified by the branch from the root node to the node to be evaluated and the average value M of the quality characteristics are calculated (step S1401). Next, the average value μ and the standard deviation σ of the quality characteristics of the data classified by the branch up to the node positioned higher than the node to be evaluated are calculated (step S1402), and the classification at the upper node is performed from these values. On the other hand, the degree of influence f of the classification at the evaluation target node is calculated (step S1403). This evaluation is repeated until all the nodes other than the root node are evaluated (Yes in step S1404). After the above evaluation is performed on all nodes (Yes in step S1404), the influence F of the manufacturing facility is calculated from the influence f of the node (step S1405). Finally, the evaluation value W of the manufacturing facility is calculated from the influence F of the manufacturing facility (step S1406), and it is estimated that the degree of abnormality is larger as the manufacturing facility has a larger evaluation value W.

FIG. 15 shows an example in which the process shown in FIG. 14 is performed using the analysis data shown in FIG. 9 and a part of the decision tree shown in FIG. In FIG. 10 and FIG. 15, the notation method of the decision tree is different, but both are the same decision tree. Processing when the node 1501 is evaluated in step S1401 will be described. The node 1501 classifies the analysis data of FIG. 9 according to the following rules.
“If“ Step (3300) ”=“ CA-01 ”AND“ Step (1320) ”=“ CT-01 ”AND“ Step (0820) ”=“ CT-01 ”AND“ Step (1800) ”=“ SU -02 "Then node 1501"
At this time, it is assumed that the number of data of a wafer that meets this rule is N, and the average value of the “open failure rate” of the wafer is M.

The node 1502 above the node 1501 classifies the analysis data of FIG. 9 according to the following rules.
“If“ Step (3300) ”=“ CA-01 ”AND“ Step (1320) ”=“ CT-01 ”AND“ Step (0820) ”=“ CT-01 ”Then node 1502”
At this time, the average value of the “open failure rate” of wafers that match this rule is μ, and the standard deviation is σ.

  In step S1403, the node influence degree f is calculated from the statistical values calculated in steps S1401 and S1402 by the following formula.

  The degree of influence f stochastically represents how much the average quality characteristic value of the classified wafer changes when the condition of the lower node is added to the classification up to the upper node. In other words, the influence degree f is an index representing how much the quality characteristic changes when the condition of the lower node is added to the data group divided under the condition up to the upper node. Since the standard value of the “open failure rate” exists only on the upper side, Equation (3) is used as the value of z in the example of FIG.

  In step S1404, the degree of influence between these upper node and lower node is calculated for all nodes other than the root node.

  Next, in step S1405, the influence F of the manufacturing facility is calculated from the influence f of the node calculated in step S1403. The influence F of the manufacturing facility can be obtained by averaging the influences f of the nodes to which the same facility name is assigned in the decision tree. Here, the case where the process number is different is determined as a different manufacturing facility. For example, the same process “process (1800)” = “SU-02” is assigned to the node 1501 and the node 1003 shown in FIG. 10, but it is determined that the same manufacturing equipment is assigned to them. Average the impact. The same equipment name is assigned to the node 1502 and the node 1503, but since the process numbers are different, it is determined that they are different manufacturing equipments, and the influence is not averaged. In this way, the influence F for each manufacturing facility is obtained. The influence degree F is an influence degree that the processing performed by the manufacturing equipment on the wafer with the process number on the entire manufacturing line has an effect on the deterioration of the quality characteristics.

  Finally, in step S1406, the evaluation value W of the manufacturing facility is calculated from the influence F for each manufacturing facility using the following formula.

  The above equation (5) is an inverse function of the above equation (1) with respect to the influence F of the manufacturing facility, and the evaluation value W is an integral value of the standard normal distribution with respect to the value Z. Therefore, the above formula (4) stochastically represents the deviation of the characteristic value when a certain wafer is processed by the manufacturing facility. That is, the above formula (4) quantifies the influence of the manufacturing facility on the characteristic value.

  FIG. 16 shows a table in which the manufacturing facilities assigned to the decision tree of FIG. 10 are evaluated by the above processing and are arranged in the order of evaluation values. FIG. 17 shows a decision tree in which the top 10 manufacturing facilities are marked in the table of FIG. The average failure rate in the table is an average of the open failure rate of data classified into each node at the node to which the corresponding facility is assigned.

  In the table shown in FIG. 16, when only the average defect rate of the wafers classified into each manufacturing facility is compared, “0880 RS-01” is the highest, but when the evaluation values are compared, “3300 CA-03” “3000 AG-01” is high. When the relationship between the nodes to which these manufacturing facilities are assigned is examined in FIG. 17, “0880 RS-01” is assigned to the node 1701, but “3000 CA-03” having a high evaluation value is assigned to the upper node. There are an assigned node 1702 and a node 1703 to which “3000 AG-01” is assigned.

  From these results, the defect rate of “0880 RS-01” in the table of FIG. 16 is high because “3300 CA-03” of 1702 which is the upper node and “3000 AG-01” of node 1703. It can be seen that the influence of the defect rate reduction at “is large. As compared with these upper nodes, it can be seen that the influence of the lower node “0880 RS-01” alone is small.

  On the other hand, “1800 SU-01” ranked sixth in the table of FIG. 16 has a low average defect rate of 0.862 but a high evaluation value of 0.944. . In FIG. 17, this manufacturing facility is assigned to node 1704 and node 1705, and the rank of the classification result is A. However, these nodes have a large influence on the failure rate as compared with the failure rate value of the upper node. By improving the state of “1800 SU-01”, the failure of the wafer classified as the node 1704 and the node 1705 is improved. It shows that there is room to improve the rate.

  In this way, in the decision tree analysis method, when tracing from the highest node of the decision tree (tree shape) to a lower node i, the wafer to be analyzed is classified under the conditions represented by each node. The rate of deterioration of quality characteristics (for example, the wafer defect rate) is a. For example, when the assigned equipment A1 for the process A,..., The assigned equipment E3 for the process E are allocated in order from the highest node, the process A = equipment A1,. The defective rate when narrowed down is a. Furthermore, if the failure rate when tracing from the lower node i to the next lower node (i + 1) is b, the above equation (5) is used to narrow down the conditions to the node i, not the failure rate b itself. By adding node (i + 1) as a condition to the wafer defect rate a, it is possible to calculate how much the defect rate has increased. By this calculation, it can be seen whether the failure rate is increased due to the node (i + 1) or whether the failure rate is deteriorated by the cumulative number of higher nodes.

  The decision tree analysis does not restrict a certain condition (in this case, a manufacturing facility) from appearing at a plurality of nodes. Therefore, equipment that appears on the route from the highest node to a certain lower node may appear on the completely different route from the highest node. Therefore, the average rate of increase in the relative defect rate as described above is averaged for the same equipment on each route to calculate how much the equipment has an adverse effect on the entire wafer. FIG. 16 shows the ranking of the possibility of abnormality of each facility based on the ranking.

  In other words, if data is analyzed using the decision tree analysis technique, the influence on the quality deterioration of a certain manufacturing facility in the path of the manufacturing facility classified by the abnormal event classification unit 108 is different from that of other manufacturing facilities in the path. The relative value of how large (or small) it is compared can be known. And the influence degree (for example, evaluation value shown in FIG. 16) given to the quality fall of the said manufacturing equipment is computable by averaging the said relative value in each path | route about one manufacturing equipment.

  As described above, the abnormal equipment evaluation unit 109 evaluates the manufacturing equipment that is estimated to be the cause of the quality characteristic deterioration, and estimates the abnormal equipment. The created decision tree, the evaluation value of each manufacturing facility, and other information are transmitted to the display terminal 110 through the network 205, and can be browsed freely by the analyst.

  As described above, according to the abnormal equipment estimation apparatus according to the present embodiment, the abnormal event classification unit 108 performs multi-branch decision tree analysis on the analysis data created from the history DB 102 and the quality DB 103, and the abnormal equipment The evaluation unit 109 evaluates the manufacturing equipment assigned to each node of the decision tree. Therefore, according to this abnormal equipment estimation apparatus, it is possible to accurately estimate manufacturing equipment having a large influence on quality characteristics by comprehensively judging individual events classified by the decision tree. Effective and quick countermeasures can be implemented for manufacturing equipment.

  In the decision tree analysis used in the present embodiment, as described above, a method of branching into a plurality according to possible conditions is adopted, but the present invention is not limited to this.

  FIG. 18 shows an example of a decision tree created when the abnormal event classification unit 108 performs a decision tree analysis with two branches. As shown in FIG. 18, the node 1801 and the node 1802 classify wafers according to whether or not the condition “process (3300) = CA-03” is met. If it does not meet the above conditions, it is classified as node 1802. In such a decision tree, the degree of influence of a node is evaluated only on the side that matches the condition. For example, in the decision tree shown in FIG. 18, the node 1801 is evaluated, but the node 1802 is not evaluated. By using such an evaluation method, the processing time can be shortened.

  Moreover, in the abnormal equipment estimation apparatus 200 according to the present embodiment, the analysis data creation unit 106 classifies quality data using the standard values stored in the condition DB 104. It is not limited.

  As another method, the analysis data creation unit 106 can automatically perform quality data using a technique such as cluster analysis. Various methods have been proposed for cluster analysis. As an example, FIG. 19 shows the flow of ranking processing using the K-means method.

When dividing the data by using a K-means clustering method into K rank, it sets a first appropriate data to the central _{V i} of rank i (1 ≦ i ≦ K) ( step S1901). Next, the distance between each data and V _i is calculated, the data is assigned to the rank i having the closest center V _i (step S1902), and the average value C _i of the data assigned to each rank is calculated for each rank. (Step S1903). Finally, the center V _{i of} rank i is compared with the average C _i (step S1904). If V _i and C _i are the same value, the ranking is determined. If the values are different, the average C _i is newly set. The processing from step S1902 onward is repeated for the center V _i .

  In this way, by performing cluster analysis on the analysis data, the data can be classified according to the distribution of quality characteristics without providing standard values in the condition DB 104, and quality characteristics for which standard values are not defined. Also, the abnormal equipment estimation according to the present invention can be applied. In addition, according to this method, the rank threshold value is automatically changed according to the distribution of quality characteristics in the analysis data, so that the characteristics of quality characteristic values are displayed more clearly than when standard values are provided. Classification can be performed. Therefore, the accuracy of data analysis in the data analysis unit 107 can be further improved.

  The cluster analysis method is not limited to the K-means method, and various methods such as a method using a neural network such as a Fuzzy-c-means method and an LVQ method can be used.

  The present invention is not limited to the above-described embodiments, and various modifications can be made within the scope shown in the claims. That is, embodiments obtained by combining technical means appropriately modified within the scope of the claims are also included in the technical scope of the present invention.

  Finally, each block of the abnormal equipment estimation apparatus 200, particularly the analysis data creation unit 106 and the data analysis unit 107, may be configured by hardware logic, or may be realized by software using a CPU as follows. Good.

  That is, the abnormal equipment estimation apparatus 200 includes a CPU (central processing unit) that executes instructions of a control program that realizes each function, a ROM (read only memory) that stores the program, and a RAM (random access memory) that expands the program. ), A storage device (recording medium) such as a memory for storing the program and various data. The object of the present invention is to record the program code (execution format program, intermediate code program, source program) of the abnormal facility estimation program of the abnormal facility estimation apparatus 200, which is software that realizes the above-described functions, in a computer-readable manner. This can also be achieved by supplying the recording medium to the abnormal equipment estimation apparatus 200 and reading and executing the program code recorded on the recording medium by the computer (or CPU or MPU).

  Examples of the recording medium include tapes such as magnetic tapes and cassette tapes, magnetic disks such as floppy (registered trademark) disks / hard disks, and disks including optical disks such as CD-ROM / MO / MD / DVD / CD-R. Card system such as IC card, IC card (including memory card) / optical card, or semiconductor memory system such as mask ROM / EPROM / EEPROM / flash ROM.

  Moreover, the abnormal equipment estimation apparatus 200 may be configured to be connectable to a communication network, and the program code may be supplied via the communication network. The communication network is not particularly limited. For example, the Internet, intranet, extranet, LAN, ISDN, VAN, CATV communication network, virtual private network, telephone line network, mobile communication network, satellite communication. A net or the like is available. Also, the transmission medium constituting the communication network is not particularly limited. For example, even in the case of wired such as IEEE 1394, USB, power line carrier, cable TV line, telephone line, ADSL line, etc., infrared rays such as IrDA and remote control, Bluetooth ( (Registered trademark), 802.11 wireless, HDR, mobile phone network, satellite line, terrestrial digital network, and the like can also be used. The present invention can also be realized in the form of a computer data signal embedded in a carrier wave in which the program code is embodied by electronic transmission.

  According to the present invention, since efficient production management can be performed in various industrial product manufacturing systems, it can be effectively used in the field of production technology.

It is a block diagram which shows the structure of the abnormal equipment estimation apparatus concerning this Embodiment. It is a schematic diagram which shows schematic structure of the manufacturing system of the semiconductor device containing the abnormal equipment estimation apparatus concerning this Embodiment. It is a schematic diagram which shows an example of the data table stored in log | history DB of the abnormal equipment estimation apparatus shown in FIG. It is a figure which shows an example of the data table stored in quality DB of the abnormal equipment estimation apparatus shown in FIG. It is a flowchart which shows the flow of the analysis process performed with the analysis server in the abnormal equipment estimation apparatus shown in FIG. (A) is a figure which shows the result which took out the data regarding the quality characteristic which performs wafer ID and analysis from the data table shown in FIG. 4, (b) ranks the quality data of the table shown in (a). It is a figure which shows a result. It is a figure which shows an example of the table which shows the standard value of the quality characteristic stored in condition DB of the abnormal equipment estimation apparatus shown in FIG. (A) is a figure which shows an example of the log | history data taken out from log | history DB by using wafer ID as a search key, (b) takes out the name of the manufacturing equipment which processed the wafer in each process from (a). It is a figure which shows the result arranged in the horizontal direction. It is a figure which shows an example of the list of the analysis data produced from FIG.6 (b) and FIG.8 (b). FIG. 10 is an example of a decision tree created by an abnormal event classification unit, and is a schematic diagram showing a decision tree created when a decision tree analysis is performed on the analysis data shown in FIG. 9. It is a schematic diagram which shows the example in which a quality characteristic falls resulting from composite equipment. It is a graph which shows distribution of the occurrence frequency with respect to the defect rate when the deterioration of quality characteristics occurs due to a plurality of abnormal events. It is a flowchart which shows the flow of the decision tree creation process performed by the abnormal event classification | category part. It is a flowchart which shows the flow of the evaluation process of the abnormal equipment performed by the abnormal equipment evaluation part. It is a schematic diagram which shows notionally the evaluation method of the manufacturing equipment performed by the abnormal equipment evaluation part. It is a figure which shows the estimation result of abnormal equipment with a table. It is a figure which shows the relationship between the estimation result of an abnormal installation, and a decision tree. It is a schematic diagram which shows other embodiment of this invention, Comprising: It is a schematic diagram which shows the decision tree produced when using the bifurcation method. It is a flowchart which shows other embodiment of this invention, and shows the flow of the process of ranking of the quality data using the K-means method. It is a figure explaining the method of estimation of the abnormal installation in patent documents 1. It is a figure explaining the method of estimation of the abnormal installation in the abnormality cause detection apparatus of patent document 2. It is a graph which shows the relationship between the quality characteristic and the occurrence frequency when the deterioration of the quality characteristic occurs due to a single abnormal event.

Explanation of symbols

101 analysis server 102 history DB (history information acquisition means)
103 Quality DB (quality information acquisition means)
104 Condition DB
105 Data registration unit (history information acquisition means / quality information acquisition means)
106 Analysis data creation unit (data creation means)
107 Data analysis unit (classification / estimation)
108 Abnormal event classification part (classification means)
109 Abnormal equipment evaluation section (estimating means)
DESCRIPTION OF SYMBOLS 110 Display terminal 200 Abnormal equipment estimation apparatus 201 Manufacturing system 203 Manufacturing equipment

Claims (8)

In a manufacturing system that includes a plurality of processing steps and that includes a plurality of manufacturing facilities that perform the same processing, and that uses each of the plurality of manufacturing facilities to manufacture a product for each processing step. An abnormal equipment estimation device for estimating a production equipment that causes a reduction in product quality,
Quality information acquisition means for acquiring, from the manufacturing system, quality information in which identification information of a product manufactured in the manufacturing system is associated with a quality inspection result of the product;
History information acquisition means for acquiring history information associated with identification information of a product and identification information of a manufacturing facility in charge of each process at the time of manufacturing the product from the manufacturing system;
Data creation means for creating analysis data in which the quality inspection result and the identification information of the manufacturing equipment are associated with each other for the identification information of the product;
When the above analysis data is classified for each manufacturing facility, the analysis data is sequentially sorted into analysis data groups for each manufacturing facility by selecting in descending order of variation in the quality inspection results in the classified analysis data. Classification means to go,
The average value of each quality inspection result included in the analysis data group classified by the classification means is calculated, and the respective average values in each classified analysis data group are compared with each other, so that each manufacturing facility gives quality degradation. An abnormal equipment estimation device comprising: an estimation unit that calculates an influence degree and estimates a manufacturing equipment that causes a reduction in product quality. The classification means classifies the analysis data using a decision tree analysis technique, using the quality information as an objective variable and the history information as an explanatory variable,
The estimation means calculates the influence of the classification by the lower node of the decision tree created by the classification means on the classification by the higher node, thereby determining the influence of the manufacturing equipment assigned to each node on the quality degradation. The abnormal equipment estimation apparatus according to claim 1, wherein a manufacturing equipment that causes a reduction in product quality is estimated. The quality inspection result is an evaluation value for arbitrary quality,
The estimation means includes the number of data classified by the lower node, the average value of the evaluation values in the data classified by the lower node, and the average value and standard deviation of the evaluation values of the data classified by the upper node The abnormal equipment estimation apparatus according to claim 2, wherein the degree of influence of the lower node on the upper node is calculated based on the above.   The abnormal equipment estimation apparatus according to any one of claims 1 to 3, wherein the data creation means evaluates the quality information of the product step by step using cluster analysis.   The abnormal equipment estimation apparatus according to claim 2, wherein, in the decision tree analysis method, the data is classified into two according to whether or not a condition is met. In a manufacturing system that includes a plurality of processing steps and that includes a plurality of manufacturing facilities that perform the same processing, and that uses each of the plurality of manufacturing facilities to manufacture a product for each processing step. An abnormal facility estimation method in an abnormal facility estimation device for estimating a production facility that causes a reduction in product quality,
A quality information acquisition step in which quality information acquisition means acquires, from the manufacturing system, quality information in which identification information of a product manufactured in the manufacturing system and a quality inspection result of the product are associated with each other;
A history information acquisition step in which history information acquisition means acquires from the manufacturing system history information in which identification information of a product and identification information of a manufacturing facility in charge of each process at the time of manufacturing the product are associated;
A data creation step for creating analysis data in which the data creation means associates the inspection result with the identification information of the manufacturing equipment for each identification information of the product;
When the classifying means classifies the above analysis data for each manufacturing facility, the analysis data is selected in descending order of variation in quality results in the classified analysis data, and the analysis data for each manufacturing facility is selected. A classification step that sequentially classifies
The estimation means calculates the average value of each quality inspection result included in the analysis data group classified by the classification step, and compares the average value in each classified analysis data group with each other, so that each manufacturing facility An abnormal facility estimation method comprising: an estimation step of calculating an influence on quality degradation and estimating a production facility that causes a product quality degradation.   An abnormal equipment estimation program for operating the abnormal equipment estimation device according to any one of claims 1 to 5, wherein the abnormal equipment estimation program causes a computer to function as each of the above means.   A computer-readable recording medium in which the abnormal equipment estimation program according to claim 7 is recorded.

Images