WO1996026539A1

WO1996026539A1 - Method and device for analyzing abnormality of production line and method and device for controlling production line

Info

Publication number: WO1996026539A1
Application number: PCT/JP1995/000288
Authority: WO
Inventors: Isao Miyazaki; Yoshihiko Kiyokuni; Ken Uchikawa; Tetsuo Noguchi; Jun Nakazato; Masao Sakata; Seiji Ishikawa; Kazuko Ishihara; Naoki Go; Masaki Koma; Koichi Kubouchi; Tsutomu Okabe; Kazunori Nemoto
Original assignee: Hitachi, Ltd.
Priority date: 1995-02-24
Filing date: 1995-02-24
Publication date: 1996-08-29

Abstract

An abnormality analyzing technique of automatically detecting abnormality of the production line of an industrial product, investigating the cause of the detected abnormality, and predicting the influence of the detected abnormality on the succeeding processes. For detecting abnormality, manufacturing data and inspection data are collected and recorded in a process data base. The manufacturing data and inspection data pertinent to an abnormality detecting parameter designated in an abnormality detecting file recorded in a file data base are read from the process data base. An abnormality detecting method recorded in a prerecorded arithmetic method file is designated and applied to the data. When an abnormality is detected, a cause investigating parameter of a cause investigating file recorded in the file data base is designated and the manufacturing data and inspection data pertinent to the parameter are read from the process data base. The cause of abnormality is investigated by applying a cause investigating method read from a prerecorded arithmetic method file to the data. When the abnormality is detected, in addition, a variation predicting parameter of a prerecorded variation predicting file is designated and the manufacturing data and inspection data pertinent to the parameter are read from the process data base. The variation is predicted by applying a variation predicting method read from the arithmetic method file to the data.

Description

Description Damage Analysis method and equipment for production line and control method and equipment for production line

The present invention provides an abnormality analysis technology that automatically detects an abnormality that has occurred in a manufacturing line of an industrial product, tracks the cause of the detected abnormality, and predicts an influence of the detected abnormality on a subsequent process. The present invention relates to a control technique for utilizing the results obtained in a control method of the production line. Background technology

2. Description of the Related Art A semiconductor integrated circuit device (hereinafter, referred to as an IC), which is an example of a semiconductor product among industrial products, is a semiconductor wafer (hereinafter, referred to as a wafer) in a so-called pre-process (hereinafter, referred to as an IC manufacturing line) of an IC manufacturing line. In the meantime, integrated circuit devices are manufactured through a wide variety of manufacturing processes such as photolithography, oxide film formation, thin film formation, and impurity implantation. In other words, IC manufacturing lines have many different manufacturing processes. In addition, since each manufacturing process of the IC manufacturing line is an ultra-fine processing, a dimensional variation of a wafer, a foreign substance attached to a wafer, a defect of a transferred pattern, and the like cause a product defect. Therefore, in the IC manufacturing line, the film thickness of the thin film formed on the wafer, the dimension of the pattern, the foreign matter adhering to the wafer and the defect in the appearance of the pattern are quantitatively inspected. In addition, defects in each manufacturing process are quickly found through inspections in each manufacturing process, and prevention of the flow of defective wafers in subsequent manufacturing processes is being implemented. In addition, manufacturing equipment for each manufacturing process and its surrounding ring Strict environmental controls are in place. With this strict management, it is implemented to prevent defects from being created in the wafer in each 3R fabrication process.

By the way, in the final process of the IC manufacturing line, a final inspection is performed to determine whether many products (pellets) formed on a wafer are good or defective. That is, a so-called probe inspection is performed in which a probe needle is brought into contact with the wafer to perform an electrical characteristic inspection of the product. In other words, product defects are ultimately found by probe inspection. Therefore, a failure analysis method for analyzing the cause of product failure by investigating the correlation between probe inspection data and inspection data for each manufacturing process has been proposed.

For example, there are the following two methods for analyzing product defects in conventional IC production lines.

(1) The first product failure analysis method is a method in which foreign matter inspection data and appearance defect inspection data are collated by a pattern matching method, and a product failure is predicted based on the overlap of both data.

(2) In the second product failure analysis method, geometric foreign matter inspection data and appearance defect inspection data, and logical one-dimensional probe inspection data that is compressible characteristic data are converted into a data array. After being transformed by a data transformation processing method such as data coordinate transformation, it is collated by a pattern matching method, and a product defect is analyzed by overlapping of both data.

Examples of this type of failure analysis method are described in the Japanese Patent Office Publication Nos. 2-2,872,241,4,15,18,45,5. No. 16 0 2 39 and No. 6—2 7 5 6 8 8.

However, the above-mentioned product failure analysis method for IC manufacturing lines There are the following problems.

(1) Since abnormalities that occur in each manufacturing process in the IC production line are not detected, abnormality analysis of the IC production line linked to the occurrence of the abnormality has not been performed.

(2) In addition, the effect of the abnormality on each manufacturing process after the manufacturing process in which the abnormality occurred is not predicted.

(3) When analyzing the causes of product defects (follow-up investigation), the history of manufacturing equipment and inspection equipment is not used.

A first object of the present invention is to provide a production line abnormality analysis technique capable of detecting an abnormality occurring in a production line by using production data and inspection data.

A second object of the present invention is to provide a production line abnormality analysis technology that can automatically determine the cause of the detected abnormality.

A third object of the present invention is to provide a production line abnormality analysis technique capable of predicting the influence of a detected abnormality on a subsequent production process.

Today, a wide variety of ICs are being developed. For this reason, in recent IC manufacturing lines, a so-called mixed line in which various types of wafers flow simultaneously on the same line is used. Therefore, in an IC manufacturing line, many kinds of wafers are flowing simultaneously in various and many manufacturing processes. Moreover, in the IC manufacturing line, the same manufacturing process may use a plurality of the same type of manufacturing equipment, and the same type of manufacturing equipment may pass wafers that have undergone different manufacturing steps many times. . In other words, IC production lines are in an extremely complex mixed production mode, and the impact on the production process after an abnormality that has occurred in complex mixed production becomes extremely complicated and serious. Therefore, human control of IC manufacturing lines is limited. Has reached. Therefore, there is a demand for the development of control technology using computers.

A fourth object of the present invention is to provide a computer-based production line control technique that improves control accuracy by utilizing abnormality analysis data. Disclosure of the invention

The present invention has an abnormality detection method including the following steps. That is, manufacturing data and inspection data necessary for abnormality detection processing for a production line to be subjected to abnormality analysis are selected as an abnormality detection parameter, and an edited abnormality detection file is recorded in a file database in advance. An operation method required for the abnormality detection processing is selected for the production line to be prepared, and an edited operation method file is prepared in advance. A step in which data and inspection data are recorded in a process database; and a step in which the abnormality detection file is searched and an abnormality detection parameter to be subjected to abnormality detection processing is specified by a human operation or automatically. The manufacturing data and / or inspection data corresponding to the specified abnormality detection The step read from the process database and the calculation method to be used from the calculation method file are specified manually or automatically by a human operation, and the calculation method is applied to the read data, causing an error. Are detected, and a step in which the calculation result is output.

With this abnormality detection method, the production line is constantly monitored dynamically, and any abnormalities that occur should be detected dynamically and quickly and accurately. For this reason, not only can rigorous and monotonous monitoring of the production line be performed, which can relieve humans from the work of detecting abnormalities, but also monitoring of the production line. Can greatly improve the reliability of the abnormality detection work.

Further, the present invention includes an abnormality factor tracking method including the following steps in addition to the abnormality detection method. In other words, the manufacturing data and inspection data required for the abnormality factor tracking process for the manufacturing line to be analyzed are selected as the factor tracking parameters, and the edited factor tracking file is recorded in the file database in advance. A step of preparing an edited calculation method file for an error factor tracking process for a manufacturing line to be analyzed for anomalies and editing the file in advance; and searching for the factor tracking file and tracking the factor from now on. A step in which the factor tracking parameter to be processed is specified by a human operation or automatically, and the manufacturing data and Z or inspection data corresponding to the specified factor tracking parameter are based on the process data. And the operation method to be used from the operation method file is determined by a human operation. Te or automatically specified, the steps of the calculation method to read de one data is applied has been factor is additionally & a step, the tracking result is output.

By this abnormal factor tracking method, if an abnormal event is detected, the cause of the abnormal event is automatically and quickly and accurately tracked and specified. By identifying the cause of this anomaly, it is possible to quickly take countermeasures for anomalies in the production line, thereby shortening the period during which an anomaly occurs and ensuring that the same product defect continues in large quantities. Accidents can be avoided and ultimately product yield can be increased. Also, when launching a new product, unexpected abnormalities tend to occur. It can be raised to the standard and new products can be launched early.

Furthermore, the present invention provides a fluctuation prediction including the following steps in addition to the abnormality detection method. It has a measurement method. That is, the manufacturing data and inspection data required for the fluctuation prediction process for the production line to be analyzed for abnormality are selected as the fluctuation prediction parameters, and the edited fluctuation prediction file is recorded in the file database in advance. A calculation method file selected and edited in advance for a fluctuation prediction process relating to the production line is recorded in a file database in advance; and the fluctuation prediction parameter is searched for the fluctuation prediction file and processed. Is specified, a step in which manufacturing data and / or inspection data corresponding to the specified fluctuation prediction parameter is read out from the process database, and a step in which the calculation method file is searched and used from now on. Operation method is specified, and the read data Calculation method is applied, and scan Tetsubu abnormal fluctuations in the process of manufacturing the manufacturing process or later after being detected is predicted, a step, the prediction result is output.

With this fluctuation prediction method, if an abnormality is detected in the production line, the effect of the abnormality on subsequent processes can be dynamically and quickly predicted accurately. By predicting fluctuations in the post-process due to this abnormality, it is possible to take prompt measures for abnormality in the production line, so that the period during which the abnormality occurs can be shortened, and the same product defect continues in large quantities. Accidents can be avoided, and ultimately the product yield can be improved. Also, when launching a new product, unexpected abnormalities are likely to occur. A brief description of the drawing

FIG. 1 shows an embodiment of the present invention equipped with an abnormality analysis device for an IC manufacturing line. FIG. 2 is a block diagram showing an IC IS construction line control device. Fig. 2 shows an example of a file recorded on a file basis. (A) is an error detection file, (b) is a factor addition file, and (c) is a fluctuation prediction file. . FIG. 3 is a block diagram showing an abnormality detection system of an abnormality analysis device for an IC manufacturing line according to one embodiment of the present invention. FIG. 4 is an explanatory diagram showing an abnormality detection method using a control value deviation test method. FIG. 5 is an explanatory diagram showing an abnormality detection method using the repetition test method. FIG. 6 is an explanatory diagram showing an abnormality detection method using another repetition test method. FIG. 7 is an explanatory diagram showing an abnormality detection method using the histogram outlier test method. FIG. 8 is an explanatory diagram showing an abnormality detection method using the histogram mean difference test method. FIG. 9 is a block diagram showing a factor tracking system of an abnormality analyzer for an IC manufacturing line according to an embodiment of the present invention. FIG. 10 is an explanatory diagram showing a factor tracking method for identifying an abnormal factor by comparing inspection data with probe inspection data. FIG. 11 is an explanatory diagram showing a factor tracking method for identifying an abnormal factor manufacturing apparatus by comparing probe inspection data for each lot of the manufacturing apparatus. FIG. 12 is an explanatory diagram showing a factor tracking method for identifying an abnormal cause charge position by comparing probe inspection data for each lot of a charge position. FIG. 13 is an explanatory diagram showing a factor tracking method for tracking an abnormality caused by maintenance of a manufacturing apparatus. FIG. 14 is an explanatory diagram showing a factor tracking method for tracking an abnormality caused by a change in manufacturing conditions of a manufacturing apparatus. FIG. 15 is a flowchart showing an embodiment of a factor tracking method for automatically identifying the cause of an abnormality by comparing probe inspection data with inspection data in a manufacturing process. FIG. 16 is a flowchart showing an embodiment of a factor tracking method for dynamically identifying an abnormal factor manufacturing apparatus by comparing probe inspection data for each lot of the manufacturing apparatus. FIG. 17 is a flowchart showing an embodiment of a factor tracking method for automatically specifying an abnormal cause charger position by comparing probe inspection data for each mouth of a charging position of a manufacturing apparatus. Fig. 18 5 is a flowchart showing an embodiment of a factor tracking method for automatically tracking an abnormality caused by maintenance of a manufacturing apparatus. FIG. 19 is a flowchart showing an embodiment of a factor tracking method for automatically tracking an abnormality caused by a change in manufacturing conditions of a manufacturing apparatus. FIG. 20 is a block diagram showing a fluctuation prediction system of an abnormality analyzer for an IC manufacturing line according to an embodiment of the present invention. FIG. 21 is an explanatory diagram showing a fluctuation prediction method for predicting an influence after the manufacturing process on the basis of inspection data of the manufacturing process in which an abnormality is detected. FIG. 22 is an explanatory diagram showing a fluctuation prediction method for predicting the influence of a maintenance operation on a manufacturing apparatus. FIG. 23 is an explanatory diagram showing a fluctuation prediction method for predicting the influence of a change in manufacturing conditions on a manufacturing apparatus. BEST MODE FOR CARRYING OUT THE INVENTION

The present invention will be described below with reference to the accompanying drawings in order to explain the present invention in more detail.

In the present embodiment, the control device of the manufacturing line according to the present invention is constructed as a control device for automatically controlling a so-called front-end process in a manufacturing line for manufacturing an IC which is an example of a semiconductor product by a computer. An abnormality analysis device for an IC production line (hereinafter, referred to as an abnormality analysis device), which is an embodiment of the present invention, for analyzing an abnormality occurring in an IC production line is incorporated.

As shown in Fig. 1, the so-called pre-process (hereinafter referred to as the "manufacturing line") in the IC manufacturing line 1 has a wide variety of processes such as a photolithography process, an oxide film forming process, a thin film forming process, and an impurity implantation process. Each of the manufacturing steps 2 includes various semiconductor manufacturing apparatuses (hereinafter referred to as manufacturing apparatuses) 3 and various inspection apparatuses 4. The final process of the production line 1 has a probe inspection process 5 In this probe inspection step 5, a probe inspection device 6 is provided. Then, while the wafer 7 as the work of the production line 1 goes through each of the pre-specified production steps 2, the integrated circuit elements are built into the wafer 7, and the wafer 7 is produced by the product of the production line 1. A pellet 8 is manufactured. Note that, in FIG. 1, the arrow 9 indicates the flow of the wafer 7.

The control device of the production line comprehensively monitors the entire production line based on the production measurement surface measured in advance, and controls the production line 1 by issuing commands as needed to the supervisor of each production process 2 The host computer 10 includes an abnormality analyzer 12 via a communication network 11 and a manufacturing data collection terminal device (not shown) attached to each manufacturing device 3. ) And an inspection data collection terminal device (not shown) attached to each inspection device 4 as appropriate.

The anomaly analyzer 12 is constructed by a personal computer or the like, and includes a central processing unit (hereinafter, referred to as a CPU) 13 and an input device for executing data input composed of a mouse, a keyboard, and the like. 1 and 4, an output device 15 composed of a display device for displaying data and a printer for outputting data, etc., and an external memory 16 are provided. Access database 17 and file database 18 are defined respectively. Note that the abnormality analysis device 12 can also be constructed by a part of the host computer 10.

The manufacturing apparatus 3 installed in each manufacturing process 2 performs predetermined processing on the wafer 7 in the manufacturing process 2, and then stores product information, manufacturing conditions, manufacturing history, maintenance history, and failure history (hereinafter, manufacturing data). The information is transmitted in real time to the abnormality analyzer 12 via the network 11. The abnormality analyzer 12 records each manufacturing data transmitted from each manufacturing apparatus 3 in the process database 17. The inspection equipment 4 installed in the manufacturing process 2 is configured to quantitatively inspect the processing state of the wafer 7 in the manufacturing process 2. For example, the inspection device 4 installed in the photolithography process performs pattern dimension inspection, film thickness inspection, foreign matter inspection, and appearance inspection. Each inspection device 4 transmits the inspection data to the abnormality analysis device 12 via the network 11 in real time. The abnormality analyzer 12 records each inspection data sent from each inspection device 4 in the process database 17. In addition, the same inspection device may be shared by the plurality of manufacturing processes 2 for the inspection device 4.

The probe inspection device 6 in the probe inspection process 5 measures the electrical characteristics of the product pellets 8 by bringing the probe needles into contact with the wafer 7 and checks whether each pellet 8 is good or defective based on the measurement results. I do. The probe inspection device 6 transmits the inspection data (hereinafter referred to as probe inspection data) of the thermal characteristics inspection for each pellet 8 to the abnormality analysis device 12 via the network 11 in real time. The abnormality analysis device 12 records the probe inspection data transmitted from the probe inspection device 6 in the process database 17.

On the other hand, there are two types of data in the file database 18. The first data is data taught in advance by humans based on past results, and the second data is analyzed abnormally based on the analysis result data output from the abnormality analyzer 12 every moment. Device 1 Data that the body automatically learns and records. Fig. 2 shows an embodiment of the files recorded in the file database 18. (a) is an abnormality detection file, (b) is a factor tracking file, and (c) is a fluctuation prediction file. is there.

The CPU 13 is configured to execute the abnormality analysis method based on the data of the process database 17 and the data of the file database 18. In other words, the CPU 13 is an abnormality detection system for the production line It is called a normal detection system. 20), a production line abnormality tracking system (hereinafter referred to as a factor tracking system) 30 and a production line fluctuation prediction system (hereinafter referred to as a fluctuation prediction system) 40. By the way, the abnormality analysis method includes abnormality detection processing, factor tracking processing, fluctuation prediction processing, processing that links abnormality detection processing and factor tracking processing, and also links abnormality detection processing and fluctuation prediction processing. There is a processing that was done. The output contents of the abnormality detection system 20 include the type of the abnormality, the manufacturing process of the abnormality, the time of the abnormality, the content of the abnormality, the manufacturing data, and the inspection data. The output contents of the factor tracking system 30 include the type of the variable factor, the manufacturing process, the manufacturing device, the inspection device, the changed content, and the timing. The output contents of the fluctuation prediction system 40 include the type, the manufacturing process, the time, the content, and the manufacturing data whose fluctuation is predicted.

Hereinafter, an abnormality analysis method according to an embodiment of the present invention using each system of the abnormality analysis device 12 according to the above configuration will be described in order.

First, an abnormality detection method by the abnormality detection system 20 will be described.

As shown in FIG. 3, the abnormality detection system 20 includes a start unit 21, an abnormality detection parameter specification unit 22, a data reading unit 23, a sorting unit 24, an abnormality detection calculation unit 25, and a detection result. An output unit 26 is provided. In addition, the starting unit 21 includes a manual mode unit 21a and an automatic mode unit 21b. The manual mode is activated when the manual mode section 21a is selected in the activation section 21 and all processing steps proceed in an interactive manner between a person and a computer. The automatic mode is activated when the automatic mode unit 21b is selected by the activation unit 21 and all processing steps automatically proceed. Hereinafter, an embodiment of the automatic mode will be described.

When the automatic mode part 2 1 b is selected and started in the start part 21, the abnormality detection parameter specification part 22 detects an abnormality in the file database 18. Access the file 27 and automatically specify the abnormality detection record 28 of the abnormality detection processing to be executed in the abnormality detection file 27, and the item to be processed first in the abnormality detection record 28 28 a is transferred to the data reading section 23. The data reading section 23 reads the manufacturing data and the inspection data from the process database 17 as parameters (characters) relating to the designated item 28a. Next, the sorting unit 24 sorts the read data into data necessary in the abnormality detection calculation unit 25. The anomaly detection calculation unit 25 reads out a predetermined test method (calculation method) that matches the footed data from the calculation method file 29 containing the test method as described later, and executes the calculation based on the read. Then, based on the sorted data, a singular point / special point is extracted and detected as abnormal. The detected abnormality is output by the detection result output unit 26 to an output device 15 such as a printer / display device, and is also output to the factor tracking system 30 and the fluctuation prediction system 40, respectively. Further, the result output unit 26 transmits an update signal to the abnormality detection parameter designating unit 22 to start the next abnormality detection process. The calculation method of abnormality detection by each test method of the abnormality detection calculation unit 25 will be specifically described with reference to FIGS.

First, the control value outlier test method shown in Fig. 4 will be described.

This control value deviation test method is an example of a test method for detecting sudden fluctuations. In the embodiment shown in FIG. 4, the parameters related to the item specified by the abnormality detection parameter specifying unit 22 are each lot in the source-gate photo process (hereinafter referred to as SG photo process) for the type 4 MSRAM. This is development dimension inspection data for each development. Now, when the previous abnormality detection processing is completed, the abnormality detection parameter designating section 22 reads the file data in order to specify the abnormality detection record 28 to be subjected to the next abnormality detection processing. The error detection record 28 specified at the end of the error detection file 27 in Fig. 2 stored in the base 18 is searched, and the next error detection item 28 a is specified to read the data. Transfer to The data reading unit 23 reads necessary data from the process database 17 according to the specified item, and transfers the data to the sorting unit 24. The sorting section 24 sorts the read data to virtually create the graph of FIG. 4 and transfers it to the abnormality detection calculation section 25. In FIG. 4, the abscissa indicates the lot numbers, and the ordinate indicates the development dimension inspection values for each lot. The abnormality detection calculation unit 25 reads the control value deviation test method from the calculation method file 29 and calculates whether or not the control value deviation has occurred in the graph of FIG. 4, and the control value deviation has occurred. In this case, it is detected as abnormal. In the embodiment shown in FIG. 4, an abnormality of the lot number A23 is detected in the development dimension inspection in the SG photo process for the product type 4 MSRAM. The detection result output unit 26 outputs the detected abnormality to an output device 15 such as a printer or a display device, and transmits an abnormality detection signal to the factor tracking system 30 and the fluctuation prediction system 40. The detection result output unit 26 transmits an update signal to the abnormality detection file designating unit 22.

Next, the run test shown in FIG. 5 will be described.

This repetition test method is an example of a test method for detecting a trend change. In the embodiment shown in FIG. 5, the parameters of the item specified by the abnormality detection parameter specifying section 22 are the same as in FIG. 4 for each unit in the SG photo process for the type 4 MSRAM. This is the developed dimension inspection data. Anomaly detection parameter specification unit 2 2 searches anomaly detection record 28 of anomaly detection file 27 of file database 18, specifies anomaly detection item to be processed from now on, and transfers it to data read unit 23 Data read unit 2 3 process data according to this designated item The necessary data is read from the overnight base 17 and transferred to the sorting section 24. The sorting section 2 sorts the read data and virtually creates the graph of FIG. 5 in which the horizontal axis indicates each lot number and the vertical axis indicates the development dimension inspection value for each lot. The data is transferred to the abnormality detection calculation section 25. The abnormality detection calculation unit 25 reads the repetition test method from the calculation method file 29 and determines whether the developed dimension value is continuously lower than the average value or continuously higher than the average value in the graph of FIG. Calculation is performed, and if a linked part occurs, it is detected as abnormal. In the embodiment shown in FIG. 5, abnormalities of the lot numbers A12 to A20 are detected in the development dimension inspection in the SG photo process. Note that the processing of the detection result output unit 26 is the same as that of the control value outlier test method, and a description thereof will be omitted.

FIG. 6 shows another embodiment of the repeated test method. In the embodiment shown in FIG. 6, the parameter of the item specified by the abnormality detection parameter specifying unit 22 is the same as in FIGS. 4 and 5 for each lot in the SG photo process for the type 4 MSR AM. This is development dimension inspection data. The abnormality detection parameter specification unit 22 searches the abnormality detection record 28 of the abnormality detection file 27 of the file database 18, specifies the next abnormality detection item, and transfers it to the data reading unit 23. The data reading unit 23 reads necessary data from the process database 17 according to the specified item, and transfers the data to the sorting unit 24. The sorting unit 24 sorts the read data, and displays the graph of Fig. 6 in which the horizontal axis indicates each lot number and the vertical axis indicates the inspection value of the development dimension for each slot. It is created virtually and transferred to the abnormality detection calculation unit 25. In the case of executing the second run test method, the abnormality detection calculation unit 25 reads out the second run test method from the calculation method file 29, and the developed dimension values are continuously displayed in the graph of FIG. Calculates whether it has risen or has fallen continuously, and a continuous point occurs If so, it is detected as abnormal. In the embodiment shown in FIG. 6, abnormalities of lot numbers A12 to A20 are detected. Note that the abnormality detection calculation unit 25 carries out the combination of the above-described control value deviation test method and the two consecutive test methods, and generates an abnormality detection signal regardless of which test method detects an abnormality. .

Next, the histogram outlier test method shown in FIG. 7 will be described. This histogram outlier test method is an example of a test method for detecting sudden fluctuation. In the embodiment shown in FIG. 7, the parameter specified by the abnormality detection parameter setting unit 22 is a deviation value related to the yield of the probe inspection process at the present time for the type 4 MECL. Incidentally, the yield of the probe inspection process is the percentage of the number of non-defective products to the number of berets on one wafer. The abnormality detection parameter specification unit 22 searches the abnormality detection record 28 of the abnormality detection file 27 of the file database 18, specifies the next abnormality detection item, and transfers the parameters to the data readout unit 23. . The data reading section 23 reads necessary data from the access database 17 in accordance with the designated parameters, and transfers it to the sorting section 24. The sorting unit 24 sorts the read data and virtually creates the graph in Fig. 7 with the horizontal axis indicating each lot number and the vertical axis indicating the inspection value of the development dimension for each lot. Then, the data is transferred to the abnormality detection calculation unit 25. In FIG. 7, the yield value is plotted on the horizontal axis, and the frequency is plotted on the vertical axis. The abnormality detection calculation unit 25 reads the histogram test method from the error test method file 29 and calculates whether or not there is a deviation value outside the reference range in FIG. In the case of the embodiment shown in FIG. 7, since there is a deviation value, an abnormality detection signal is generated.

FIG. 8 shows a mean difference test method which is another embodiment of the histogram test method. ing. In the embodiment shown in FIG. 8, the parameters of the item specified by the abnormality detection parameter overnight specification unit 22 are two types of periods corresponding to the yield of the probe inspection process for the type 4 ME. It is the deviation value of. The abnormality detection parameter specification section 22 searches the abnormality detection record 28 of the error detection file 27 of the file database 18, specifies the next abnormality detection item, and transfers it to the data reading section 23. The data reading unit 23 reads necessary data from the process database 17 according to the designated item, and transfers the data to the sorting unit 24. The sorting unit 24 sorts the read data and virtually creates the graph in Fig. 8 with the horizontal axis indicating each lot number and the vertical axis indicating the inspection value of the development dimension for each lot. Then, it is transferred to the abnormality detection calculation section 25. In FIG. 8, the horizontal axis represents the yield value, and the vertical axis represents the frequency. The upper graph shows the deviation from the current time to one week before, and the lower graph shows the deviation from one week before to one week before. The anomaly detection calculation unit 25 reads the second histogram test method from the calculation method file 29, obtains the average value of the deviation values of the upper and lower graphs, and determines whether the difference between the average values is within the reference range. Is calculated, and if out of range, it is detected as abnormal. In the case of the embodiment shown in FIG. 8, no abnormality detection signal is generated because it is within the range. The abnormality detection calculation unit 25 performs the above-described two types of histogram test methods in combination, and generates an abnormal signal regardless of which test method detects an abnormality.

As described above, the IC manufacturing line 1 is automatically and constantly monitored by the abnormality detection system 20 of the abnormality analysis device 12, and any abnormality that occurs is detected automatically and quickly and accurately. This not only relieves humans from the rigorous and monotonous monitoring or abnormality detection of the IC manufacturing line 1, but also monitors or monitors the IC manufacturing line 1. The reliability of the normal detection work can be greatly improved.

Next, a factor tracking method using the factor tracking system 30 will be described.

The factor tracking system 30 is a system that tracks and investigates the cause of the abnormality detected by the abnormality detection system 20.As shown in Fig. 9, the activation unit 31 and the factor tracking parameter It has a designation unit 32, a data readout unit 33, a sorting unit 34, a factor tracking calculation unit 35, and a tracking result output unit 36. Further, the activation unit 31 includes a manual mode unit 31 a and an automatic mode unit 31 b as in the abnormality detection system 20. First, the factor tracking method in the case of the manual mode will be described.

When the abnormality detection system 20 alerts the user of the abnormality detection, the monitor of the production line activates the activation unit 31 manually. In the automatic mode, when an abnormality detection signal is transmitted from the abnormality detection system 20 to the activation unit 31, the factor tracking system S 0 is automatically activated. With this activation, the factor tracking parameter designation unit 32 accesses the factor tracking file 37 of the file database 18 and causes the factor tracking record 38 to be displayed on the display device of the output device 15. The monitor designates the item 38 a of the factor tracking process to be executed from among the displayed factor tracking records 38 with the input device 14 such as a mouse, and transfers the item 38 a to the data reading unit 33. The data reading unit 33 reads necessary data from the process database 17 according to the item 38a designated in this way and transfers it to the sorting unit 34. The sorting section 34 sorts the read data into necessary data in the factor tracking operation section 35, creates the graphs shown in FIGS. 10 to 14 and displays them on the output device 15 Let it. On the other hand, the factor tracking calculation unit 35 reads a factor tracking method described later from a calculation method file 39 in which a factor tracking method described later is recorded as a calculation method. The observer observes this graph while confirming the factor tracking method of the computation file 39, and observes the factor abnormality. Identify the cause. The purging result output unit 36 outputs the specified abnormal cause to an output device 15 such as a printer or a display device, and updates the next factor purging process.

Hereinafter, the factor tracking methods shown in FIGS. 10 to 14 will be specifically described.

FIG. 10 shows an embodiment of a factor tracking method for identifying the cause of an abnormality by comparing the inspection data of a predetermined manufacturing process with the probe inspection data. In this embodiment, the parameters corresponding to the items designated by the observer in the factor tracking parameter designation section 32 are the developed dimension inspection data and the completed dimension inspection data for each lot in the SG photo, and the lot dimension. This is probe inspection data for each case. The data read unit 33 reads predetermined data from the process database 17 according to the designated parameters, and transfers the read data to the sorting unit 34. The sorting unit 3 creates the graph shown in FIG. 10 based on the transferred data and displays the graph on the monitor of the output device 15. In FIG. 10, the horizontal axis indicates each lot number, and the vertical axis indicates the inspection value for each lot. The upper graph shows the development dimensions, the middle graph shows the completed dimensions, and the lower graph shows the deviation in the probe inspection yield. The observer reads the calculation method file 39 of the factor tracking calculation unit 35, and checks the inspection data against the probe inspection data to confirm the factor tracking method for identifying the cause of the abnormality. . Then, the observer tracks the common points of these graph changes as abnormal factors. For example, in FIG. 10, since the development dimension, the completed dimension, and the yield deviation value decrease at the lot number A22, immediately before the development dimension inspection is performed on the lot number A22, the SG photo process is performed. It is concluded that there is an anomaly factor in the work before the development dimension inspection. The tracking result output unit 36 sends this result to the output device 15 Output.

Figure 11 shows that when the same processing in the same manufacturing process was performed by different manufacturing equipment of the same model, the probe inspection data of each lot passed through each manufacturing equipment was compared with each other. An embodiment of a factor tracking method for identifying a manufacturing apparatus that has caused an abnormality is shown. In the present embodiment, the parameter specified by the monitor in the factor tracking parameter specifying unit 32 is a probe related to the yield of each of the first, second, and third exposure apparatuses in the SG photo process. This is inspection data. The data reading unit 33 reads out predetermined data from the process database 17 and transfers it to the sorting unit 34. The sorting unit 34 creates the graph shown in FIG. 11 based on the transferred data, and displays the graph on the monitor of the output device 15. By the way, the process database 17 contains the history of manufacturing equipment that records which unit has passed which machine and which manufacturing machine, and also contains probe inspection data for each lot. The sorting unit 34 can create the graph shown in FIG. 11 by matching the manufacturing device history and the probe inspection data for each lot. In FIG. 11, the horizontal axis represents the deviation value of the yield, and the vertical axis represents the frequency. The upper graph is for Unit 1, the middle graph is for Unit 2, and the lower graph is for Unit 3. The observer reads the calculation method file 39 of the factor tracking calculation unit 35, and checks the probe inspection data for each port after passing through each manufacturing device. Check the factor tracking method to identify the manufacturing equipment that was used. Then, the observer tracks the differences between these graph changes as abnormal factors. For example, in FIG. 11, since the deviation value of the second unit is reduced, it is concluded that there is a cause of abnormality in the second unit of the exposure apparatus in the SG photo process. The tracking result output unit 36 outputs this result. Force device 15 to output.

Fig. 12 shows that when the same processing in the same manufacturing process is charged to different positions in the same manufacturing equipment and executed, the probe inspection data for each port passing through each charge position is compared with each other. This shows an embodiment of a factor tracking method for specifying the position of a charger of a manufacturing apparatus that has caused an abnormality. In the present embodiment, the parameters specified by the observer in the factor tracking parameter specification unit 32 are the upper charge position, the middle charge position, and the lower charge position of the high-temperature low-pressure CVD device (hereinafter referred to as HLDCVD device) in the SG photo process. Probe inspection data for each yield for each position. The data reading unit 33 reads out predetermined data from the process database 17 and transfers it to the sorting unit 34. The sorting section 34 creates the graph shown in FIG. 12 based on the transferred data, and displays the graph on the monitor of the output device 15. By the way, the process database 17 contains the history of manufacturing equipment that records which lot was charged to which manufacturing equipment and at which location, and the probe inspection data for each lot. Therefore, the factor tracking calculation unit 35 can create the graph shown in FIG. 12 by matching the history of the manufacturing apparatus with the probe inspection data by mouth. In FIG. 12, the horizontal axis represents the deviation value of the yield, and the vertical axis represents the frequency. The upper graph shows the upper charge position, the middle graph shows the upper charge position, and the lower graph shows the upper charge position. The observer read the calculation method file 39 of the factor tracking calculation unit 35 and collated the probe test data for each mouth after passing through each change position, which caused the abnormality. Check the factor tracking method to identify the position of the manufacturing equipment charge. The observer then tracks the differences in these graphs as anomalies. For example, in Fig. 12 Since the deviation value is decreasing, it is concluded that there is an abnormal factor in the middle charger position of the HLDCVD equipment in the SG photo process. The tracking result output unit 36 causes the output device 15 to output this result.

FIG. 13 shows an embodiment of a factor tracking method for tracking an abnormality caused by maintenance of a manufacturing apparatus. In the present embodiment, the parameters specified by the monitoring person in the factor tracking parameter specifying unit 32 are manufacturing apparatus history data, film thickness inspection data, and yield of probe inspection, which indicate the cleaning time for the HLDCVD apparatus in the SG photo process. Data. The data reading unit 33 reads predetermined data from the process database 17 and transfers it to the sorting unit 34. The sorting section 34 creates the graph shown in FIG. 13 based on the transferred data and displays it on the monitor of the output device 15. By the way, the process database 17 records the maintenance history that records when cleaning work etc. were performed for each manufacturing device, and the manufacturing time that recorded when each lot passed through each manufacturing device. Since the equipment history is recorded and the probe inspection data for each lot is recorded, the reading unit 34 matches the maintenance history with the manufacturing equipment history in chronological order, and the manufacturing equipment S history and probe. By matching the inspection data with each mouth, the graph shown in Figure 13 can be created. In Fig. 13, the upper part shows the time series of the cleaning work for the HLDCVD equipment in the SG photo process, the middle part shows the graph on the film thickness inspection value, and the lower part shows the graph on the deviation value of the yield in the probe inspection. In the graph relating to the film thickness inspection value, the horizontal axis represents the lot number, and the vertical axis represents the film thickness value. In the graph related to yield, the horizontal axis shows the lot number, and the vertical axis shows the deviation value. Then, the time series of the cleaning operation and each lot number are matched with the passage time of the HLDCVD equipment that has been cleaned. You. The K-viewer reads the calculation method file 39 of the factor tracking calculation unit 35 and checks the factor tracking method for pursuing an abnormality caused by the maintenance of the manufacturing apparatus. Then, the observer tracks the displacement point as an abnormal factor in the graph of the film thickness inspection value before and after the cleaning time and the graph of the yield deviation value. For example, in FIG. 13, since there is a displacement point after cleaning, it can be concluded that there is an abnormal factor in the cleaning work of the HLDCVD apparatus in the SG photo process. The additional output unit 36 causes the output device 15 to output this result.

FIG. 14 shows an embodiment of a factor tracking method for tracking an abnormality caused by a change in manufacturing conditions in a manufacturing apparatus. In the present embodiment, the parameters specified by the observer in the factor tracking parameter specification section 32 include the manufacturing apparatus history data, the development dimension inspection data, and the probe inspection regarding the change of the exposure condition of the exposure apparatus in the SG photo process. This is yield data. The data read section 33 reads predetermined data from the process database 17 and transfers the read data to the sorting section 34. The sorting section 34 creates the graph shown in FIG. 14 based on the transferred data, and displays it on the monitor of the output device 15. Incidentally, the manufacturing equipment history in the process database 17 records when and how the manufacturing conditions were changed for each manufacturing equipment, and the time when each lot passed through each manufacturing equipment. Since the process database 17 contains development dimension inspection data and probe inspection data for each port, the sorting section 34 has the manufacturing equipment history and development dimension inspection data and data. By matching the probe test data with the time series and each lot number, the graph shown in FIG. 14 can be created. In Fig. 14, the upper part shows the time series of changes in the exposure conditions of the exposure equipment in the SG photo process, the middle part shows the graph related to the development dimension inspection values, and the lower part shows the professional 6 is a graph relating to a deviation value of the yield of the probe inspection. In the graph relating to the developed dimension inspection values, the horizontal axis represents the lot number, and the vertical axis represents the developed dimension value. In the graph related to yield, the horizontal axis represents the lot number, and the vertical axis represents the deviation value. Then, the time series relating to the exposure condition changing work and each lot number are matched with the passage time of the exposure apparatus whose condition has been changed. The observer reads the calculation method file 39 of the factor tracking calculation unit 35, and confirms the factor tracking method for tracking the abnormality caused by the change of the manufacturing condition in the manufacturing apparatus. Then, the observer tracks the displacement point as an abnormal factor in the graph of the developed dimension inspection value and the graph of the yield deviation value before and after the change time. For example, in FIG. 14, since the displacement point exists after the change, it is concluded that there is an abnormal factor in the change of the condition of the exposure apparatus in the SG photo process. The tracking result output unit 36 causes the output device 15 to output this result.

Next, the factor tracking method in the automatic mode of the factor tracking system 30 will be described with reference to FIGS.

FIG. 15 is a flowchart showing an embodiment of a factor tracking method for automatically specifying a cause of abnormality by comparing probe inspection data with inspection data of a predetermined manufacturing process. In the present embodiment, the parameters specified by the additional factor parameter setting unit 32 are first the probe inspection data for each lot and the development dimension inspection data for each lot in the SG photo process. That is, when an abnormality detection signal based on the probe inspection data is transmitted from the abnormality detection system 20, the factor tracking parameter setting unit 32 initially specifies a parameter to be subjected to factor tracking processing. Then, the factor tracking file 37 in FIG. 3 stored in the file database 18 is searched to specify the factor tracking record 38, and the developed dimension inspection data for each mouth in the SG photo process is firstly obtained. Abnormal factor candidate items And transfer it to the data reading unit 33. The data readout unit 33 reads out the development dimension inspection inspection data in the SG photo process, which is the abnormal factor candidate parameter, from the process database 17 together with the probe inspection data, and transfers it to the factor tracking operation unit 35. The factor additional calculation unit 35 calculates a correlation coefficient between the probe inspection data and the developed dimension inspection data. The distribution diagram shown in FIG. 15 schematically illustrates the calculation of the correlation coefficient. In FIG. 15, the horizontal axis represents the yield value, and the vertical axis represents the development dimension. Each plot shows the yield value (horizontal axis) and development size (vertical axis) of each lot. The straight line is the calculated correlation coefficient straight line. Subsequently, the factor tracking calculation unit 35 compares whether the calculated correlation coefficient is equal to or greater than a preset reference value. If the value is equal to or higher than the reference value, it is concluded that the exposure process, which is a process prior to the development dimension inspection in the SG photo process, is one of the abnormal factor candidates. The follow-up result output unit 36 outputs this conclusion to the output device 15 and registers it as a factor candidate in the factor tracking file 37.

Thereafter, the item relating to the probe inspection yield abnormality retrieved earlier by the factor tracking parameter specification unit 32 is updated, and the correlation coefficient between the next inspection data and the probe inspection data is calculated by the factor tracking calculation unit 35. At the same time, the routine in which the calculated correlation coefficient is compared with the reference value is repeated. This routine is repeated for all the abnormal cause candidate parameters corresponding to the items related to the probe inspection yield abnormality. Then, when tracking of all the specified abnormal cause candidate parameters is completed, the parameters concluded as abnormal cause candidates are determined and output in order of ascending and descending correlation coefficients. Further, as the output of the tracking result, the correlation coefficient may be output according to the order of ascending and descending for all the parameters examined. Figure 16 shows that when the same process in the same manufacturing process was performed by different manufacturing equipment of the same model, probe inspection data for each lot passed through each manufacturing equipment was compared with each other to detect abnormalities. 5 is a flowchart illustrating an embodiment of a factor tracking method for automatically specifying a manufacturing apparatus that has caused a cause. In the present embodiment, the parameters of the item specified first by the factor tracking parameter specifying section 32 are the first, second, and third units of the exposure device in the SG photo process. Probe inspection data on yield. That is, when an abnormality detection signal based on the probe test data is transmitted from the abnormality detection system 20, the factor tracking parameter specification unit 32 searches the factor tracking file 37 and probes as a factor tracking record 38. In addition to reading out the inspection yield abnormality record, probe inspection data for each of the first, second, and third exposure equipment in the SG Hote process is used as the parameter of the first abnormality factor candidate item. And transfer it to the data readout unit 33. The data reading unit 33 reads from the process database 17 probe inspection data for each yield of each of the first, second, and third exposure equipment in the SG photo process, which is one of the parameters for this abnormal factor candidate. To the factor tracking calculation unit 35. The factor tracking calculator 35 calculates the deviation value of the yield for each unit, calculates the average value of all the deviation values, and obtains the difference between the average value and the deviation value of each unit. The graph shown in FIG. 16 is a schematic diagram of this arithmetic processing. In Fig. 16, the horizontal axis shows the yield deviation and the vertical axis shows the frequency. (A) is the graph for the first unit, (b) is the graph for the second unit, (c) is the graph for the third unit, and (d) is the graph showing the difference between the average value and the second unit. It is. Subsequently, the factor tracking calculation unit 35 compares whether the calculated difference is equal to or greater than a preset reference value. For example, as shown in Figure 16, the difference value of Unit 2 If is greater than or equal to the reference value, it is concluded that the second exposure apparatus in the SG photo process is one of the candidate abnormal factors. The tracking result output unit 36 outputs this conclusion to the output device 15 and registers it in the factor tracking file 37 as a factor candidate.

After that, from the items of the probe inspection yield abnormality factor tracking record retrieved earlier by the factor additional parameter specification unit 32, the probe inspection data on the yield of each unit of the manufacturing equipment in the next process is the next abnormality factor candidate. The difference between the average value of all designated units and the deviation value of each unit is specified by the factor tracking calculation unit 35, and the routine of comparing the calculated difference value with the reference value is repeated. It is. This routine is repeated for the manufacturing equipment of all the processes specified in the items of the cause tracking record of the probe inspection yield abnormality. When the tracking of all the specified manufacturing equipment is completed, the manufacturing equipments concluded as abnormal cause candidates are ranked and output in the order of the difference value. Further, as the output of the arterial vein result, the difference value may be output according to the order of the magnitude with respect to all the inspected manufacturing apparatuses.

Figure 17 compares probe inspection data for each port after each charge position when the same process in the same manufacturing process is charged to different positions in the same manufacturing equipment. 6 is a flowchart showing an embodiment of a factor tracking method for automatically specifying a charge position of a manufacturing apparatus that has caused an abnormality. In the present embodiment, the parameter of the item specified first by the factor tracking parameter specifying unit 32 is the upper charging position, the middle charging position, and the lower charging position of the HLDCVD apparatus in the SG photo process. This is a probe inspection day for yield. That is, when the abnormality detection signal based on the probe inspection data is transmitted from the abnormality detection system 20, the factor tracking parameter specification unit 32 Search the factor tracking file 37 and read out the record of the probe inspection yield abnormality as the factor tracking record 38.Also, regarding the yield at each of the upper charge position, middle charge position, and lower charge position of the HLDCVD equipment in the SG photo process The probe inspection data is designated as a parameter of the first abnormality factor candidate item, and is transferred to the data reading unit 33. The data reading unit 33 processes the probe data for each of the yield factors at the upper, middle, and lower stages of the HLDCVD system in the SG photo process, which are the candidate parameters for the abnormal factors, in the process database. Read from 17 and transfer to factor addition calculation unit 35. The factor addition calculation unit 35 calculates the deviation value of the yield for each charge position, calculates the average value of all the deviation values, and calculates the difference between the average value and the deviation value of each unit. Ask. The graph shown in FIG. 17 is a schematic diagram of this arithmetic processing. In Fig. 17, the deviation value of the yield is plotted on the horizontal axis, and the frequency is plotted on the vertical axis. Curve A is the deviation value of the middle charging position, and curve B is the average value. Accordingly, the factor tracking calculation unit 35 compares whether the calculated difference value is equal to or larger than a preset reference value. For example, when the difference value at the middle charge position is equal to or larger than the reference value as shown in the graph of FIG. 17, the middle charge position of the HLDCVD apparatus in the SG photo process is one of the abnormal cause candidates. It is concluded. The tracking result output unit 36 outputs this conclusion to the output device 15 and registers it as a factor candidate in the factor tracking file 37.

Thereafter, the probe inspection data on the yield at each charge position of the manufacturing equipment in the next process is obtained from the items of the factor tracking record of the probe inspection yield abnormality previously searched by the factor tracking parameter specifying unit 32. The difference between the average value of all the designated charge positions and the deviation value of each charge position is designated as an abnormal cause candidate parameter overnight. Is calculated, and the routine in which the calculated difference value is compared with the reference value is returned. This routine is repeated for all manufacturing equipment specified in the items of the tracking record for the cause of the probe inspection yield abnormality. Then, when the tracking of all the specified manufacturing apparatuses is completed, the charging positions of the manufacturing apparatuses concluded as the abnormal cause candidates are ranked and output in the order of the difference value. In addition, as the output of the tracking result, the difference value may be output according to the order of the magnitude for each charging position for all the inspected manufacturing apparatuses.

FIG. 18 is a flowchart showing an embodiment of a factor tracking method for automatically tracking an abnormality caused by maintenance of a manufacturing apparatus. In the present embodiment, the parameters specified first by the factor additional parameter specifying unit 32 are manufacturing apparatus history data, film thickness inspection data, and processing data relating to the cleaning timing of the HLDCVD apparatus in the SG photo process. This is the yield data for one inspection. That is, when an abnormality detection signal based on the probe inspection data is transmitted from the abnormality detection system 20, the factor tracking parameter specification unit 32 searches the factor tracking file 37 and obtains the probe inspection yield as the factor tracking record 38. While reading out the record of the abnormality, specify the manufacturing equipment history data, the film thickness inspection data and the yield data of the probe inspection regarding the cleaning timing of the HLDCVD equipment in the SG photo process as the first abnormality factor candidate parameter, and read the data. Transfer to The data readout unit 33 processes the abnormality parameter candidate parameters, such as manufacturing equipment history data on the cleaning timing of the HLDCVD equipment in the SG photo process, film thickness inspection data, and yield data of the probe inspection, on a process data basis. From the source 17 and transferred to the factor tracking calculation unit 35. Factor tracking calculation unit 35 is the average of the deviations of the lots processed before cleaning by the HLDCVD equipment and the average of the deviations of the lots processed after cleaning. And calculate the difference between the two averages. The graph shown in FIG. 18 is a schematic diagram of this arithmetic processing. In Figure 18, the horizontal axis shows the yield deviation, and the vertical axis shows the frequency. Curve A is the average of the deviations of the lots treated before washing, and curve B is the average of the deviations of the lots treated after washing. Then, the factor tracking calculation unit 35 compares whether the calculated difference value is equal to or larger than a preset reference value. For example, as shown in the graph of Fig. 18, when the average value after cleaning is larger than the average value before cleaning and the difference value is equal to or more than the reference value, cleaning work of the HLDCVD apparatus in the SG photo process is performed. It is concluded that this is one of the possible causes of the anomaly. The tracking result output unit 36 outputs this conclusion to the output device 15 and registers it in the factor tracking file 37 as a factor candidate.

Thereafter, from the items of the factor tracking record of the probe inspection yield abnormality previously searched by the factor tracking parameter overnight specification unit 32, the data related to the maintenance of the manufacturing equipment in the next process is specified as the next candidate parameter of the abnormality factor and specified. The difference between the average value before and after the maintenance work of all the manufacturing apparatuses thus calculated is calculated by the factor tracking calculation unit 35, and the routine of comparing the calculated difference value with the reference value is repeated. This routine is returned for all equipment maintenance operations for all processes specified in the items in the probe tracking yield factor anomaly tracking record. Then, when the tracking of the maintenance work of all the specified manufacturing equipment is completed, the maintenance work of the manufacturing equipment concluded as a candidate for an abnormal factor is ranked and output in the order of the difference value. You. Further, as the output of the tracking result, the difference value may be output in accordance with the order of the magnitude for each maintenance operation for all the inspected manufacturing apparatuses. Figure 19 shows dynamic changes in manufacturing equipment caused by changes in manufacturing conditions. 5 is a flowchart showing an embodiment of a factor tracking method to be followed. In the present embodiment, the parameters initially specified by the factor additional parameter specification section 32 are manufacturing apparatus history data, development dimension inspection data, and probe data relating to changes in the exposure conditions of the SS optical apparatus in the SG photo process. This is the inspection yield data. That is, when an abnormality detection signal based on the probe inspection data is transmitted from the abnormality detection system 20, the factor tracking parameter specification unit 32 searches the factor tracking file 37 and performs the probe inspection as the factor tracking record 38. In addition to reading out the record of the yield abnormality, the manufacturing equipment history data, the current dimension inspection data, and the yield data of the probe inspection regarding the change of the exposure condition of the exposure equipment in the SG Photo process were specified as the first abnormality cause candidate parameter. Then, the data is transferred to the data reading unit 33. The data readout unit 33 processes the process parameters based on the manufacturing equipment history data, development dimension inspection data, and probe inspection yield data related to changes in the exposure conditions of the exposure equipment in the SG photo process, which are the abnormal parameter candidate parameters. From the source 17 and transferred to the factor tracking calculation unit 35. The factor tracking calculation unit 35 calculates the average value of the deviation values of the lot groups processed before the change of the exposure condition of the exposure apparatus and the average value of the deviation values of the lot groups processed after the change. At the same time, find the difference between the two averages. The graph shown in FIG. 19 is a schematic diagram of this arithmetic processing. In Fig. 19, the horizontal axis shows the deviation value of the yield, and the vertical axis shows the frequency. Curve A is the average of the deviations of the lots processed before the change, and curve B is the average of the deviations of the lots processed after the change. Subsequently, the factor tracking calculator 35 compares whether the calculated difference value is equal to or larger than a preset reference value. For example, as shown in the graph of Fig. 19, when the average value after the change is larger than the average value before the change and the difference value is equal to or larger than the reference value, the exposure condition of the exposure apparatus in the SG photo process is used. Change is an abnormal factor It is concluded that it is one of the supplements. The tracking result output unit 36 outputs this conclusion to the output device 15 and registers it in the factor tracking file 37 as a factor candidate.

Thereafter, from the probe inspection yield abnormality record retrieved earlier by the factor tracking parameter designating unit 32, the data relating to the change in the condition of the manufacturing equipment in another process is designated as the next abnormal factor candidate parameter, and all of the specified manufacturing The difference between the average values before and after the device condition change is calculated by the factor tracking calculation unit 35, and the routine for comparing the calculated difference value with the reference value is repeated. This routine is repeated for changes in the manufacturing equipment conditions for all processes specified in the items of the probe inspection yield abnormality record. When the tracking of the change in conditions of all the specified manufacturing equipment is completed, the change of the conditions of the manufacturing equipment that is concluded as a candidate for the cause of abnormality is determined and output in the order of the difference value. You. Further, as the output of the additional result, the difference value may be output in accordance with the order of the magnitude every time the condition is changed for all the manufacturing apparatuses examined.

As described above, the IC manufacturing line is automatically and constantly monitored by the abnormality analyzer, and if an abnormality is detected, the cause of the abnormality is automatically and quickly and accurately tracked and identified. By identifying the cause of this abnormality, it is possible to quickly take countermeasures against the abnormality on the IC manufacturing line, so that the period during which the abnormality occurs can be shortened and the same product defect continues in large quantities Accidents can be avoided and ultimately product yield can be improved. Also, when launching a new product, unexpected abnormalities are more likely to occur, but since abnormalities can be detected quickly and the cause can be quickly identified, product yields can be taken while taking countermeasures against abnormalities. Can be raised to a predetermined level, and new products can be launched early. Next, a fluctuation prediction method using the fluctuation prediction system 40 will be described with reference to FIG.

The fluctuation prediction system 40 predicts, for the abnormality detected by the abnormality detection system 20, how the abnormality will affect the processes after the abnormality occurrence process. The fluctuation prediction system 40 includes a start unit 41, a fluctuation prediction parameter overnight setting unit 42, a data reading unit 43, a reading unit 44, a fluctuation prediction calculation unit 45, and a prediction result output unit 46. I have. The starter 41 has only an automatic mode, and all steps automatically proceed.

When the activation unit 41 is automatically activated by the abnormality detection signal from the abnormality detection system 20, the fluctuation prediction parameter specification unit 42 stores the fluctuation prediction file of FIG. 20 stored in the file database 18. 4 Search for 7. As a result of this search, a fluctuation prediction record 48 used in the fluctuation prediction processing to be executed in the fluctuation prediction file 47 is searched, and the first item to be processed is automatically read out. Transferred to 3. The data reading section 43 reads necessary data from the process database 17 according to the parameters of the transferred items and transfers the data to the sorting section 44. The sorting section 44 sorts the read data, virtually creates the graphs shown in FIGS. 21 to 23 necessary for the fluctuation prediction operation section 45, and appropriately creates the fluctuation prediction operation section 45. Transfer to The fluctuation prediction calculation unit 45 reads the fluctuation prediction methods shown in FIGS. 21 to 23 described later from the calculation method file 49, calculates the fluctuation amount of the created graph, and calculates the calculation result as the prediction result. Output as The prediction result output unit 46 outputs the prediction result to the output device 15 and updates the next fluctuation prediction process. Also, in order to automatically verify whether the predicted parameters in the subsequent process change as predicted, the prediction result output unit 46 stores the prediction results in the abnormality detection record 28 of the abnormality detection file 27. register. This error detection file By registering the file 27 in the abnormality detection record 28, the fluctuation prediction system 40 is linked to the abnormality detection system 20.

Hereinafter, the fluctuation prediction methods shown in FIGS. 21 to 23 will be specifically described.

Incidentally, when the abnormality detection system 20 finds the prediction result of the fluctuation prediction system 40 as an abnormality detection item when searching the abnormality detection file 27 described above, it executes the abnormality detection processing described above for the abnormality detection item. .

FIG. 21 shows an embodiment of a fluctuation prediction method for predicting an influence after a manufacturing process based on inspection data of a manufacturing process in which an abnormality is detected. In the present embodiment, the inspection data in which the abnormality is detected is the actual dimension inspection data in the SG photo process. When an abnormality detection signal regarding the development dimension inspection data in the SG photo process is transmitted from the abnormality detection system 20, the fluctuation prediction parameter designating section 42 searches the fluctuation prediction file 47 to execute the SG photo process. The parameter of the first item is read from the fluctuation prediction record 48 that causes the development dimension to be abnormal, and is transferred to the data reading unit 43. The data reading section 43 reads necessary data from the process database 17 in accordance with the transferred parameters and transfers it to the sorting section 44. The sorting unit 44 sorts the read data to virtually create the graph shown in FIG. 21 and transfers it to the fluctuation prediction calculation unit 45. The fluctuation prediction calculation unit 45 reads from the calculation method file 49 a fluctuation prediction method for predicting the influence of the manufacturing process after the manufacturing process, based on the inspection data of the manufacturing process in which the abnormality is detected. Is calculated using the coefficient of variation specified in the calculation method file 49. That is, the variation of the parameter to be predicted for variation = abnormal parameter X variation coefficient is calculated. In this example, a product of type 4 MECL corresponding to the product of lot A22 was used. The yield in the lobe inspection process is specified as a parameter for fluctuation prediction. That is, the yield fluctuation amount = developed dimension deterioration amount X1.2 is calculated. Figure 21 illustrates this fluctuation prediction method. In FIG. 21, the abscissa indicates each lot number, and the ordinate indicates the inspection value for each lot. The upper part is a graph relating to the deterioration of the development dimension, and the lower part is a graph relating to the fluctuation of the deviation value of the yield of the probe inspection target for fluctuation prediction. The amount of variation in the probe inspection yield for which variation is to be predicted is expressed by the difference from the preceding and following lots. The prediction result output unit 46 outputs the prediction result to the output device 15 and updates the next fluctuation prediction process. Also, in order to automatically verify whether the predicted parameters of the post-process that have been predicted change as predicted, the prediction result output unit 46 stores the prediction results in the abnormality detection record 28 of the abnormality detection specification file 27. register.

FIG. 22 shows an embodiment of a fluctuation prediction method for predicting the influence of a maintenance operation on a manufacturing apparatus. In the present embodiment, the inspection data in which an abnormality is detected is the film thickness inspection data after the processing by the HLDCVD apparatus in the SG photo process. When an abnormality detection signal is transmitted from the abnormality detection system 20 to the film thickness inspection data after processing by the HLDCVD device in the SG photo process, the fluctuation prediction parameter designation unit 4 2 changes the fluctuation prediction file 4. 7 is retrieved, the parameters of the first item are read from the fluctuation prediction record 48 relating to the film thickness abnormality after processing by the HLDCVD apparatus in the SG photo process, and transferred to the data reading unit 43. The data reading section 43 reads necessary data from the process database 17 according to the transferred parameters and transfers the data to the sorting section 44. The sorting unit 44 sorts the read data to virtually create the graph shown in FIG. 22 and transfers it to the fluctuation prediction calculation unit 45. The fluctuation prediction calculation unit 45 receives the method for the manufacturing equipment from the calculation method file 49. The fluctuation prediction method for predicting the effect of the maintenance work is read, and the fluctuation amount of the parameter is calculated using the fluctuation coefficient specified in the calculation method file 49. According to the embodiment illustrated in FIG. 22, the process of the fluctuation prediction specified by the fluctuation prediction operation unit 45 is a probe inspection process, and the parameter of the fluctuation prediction corresponds to the product of the lot A22. Yield in the probe inspection process for Type 4 MECL. Then, the fluctuation prediction calculation unit 45 calculates the yield fluctuation amount = the film thickness deterioration amount X1.5. In FIG. 22, the upper part is a time series of the cleaning operation for the HLDCVD apparatus in the SG photo process, the middle part is a graph relating to the film thickness inspection value, and the lower part is a graph showing the deviation value of the yield in the probe inspection. In the graph relating to the film thickness inspection value, the horizontal axis represents the lot number, and the vertical axis represents the film thickness value. In the graph related to yield, the horizontal axis represents the lot number, and the vertical axis represents the deviation value. The time series of the cleaning operation and each lot number are matched with the passage time of the HLDCVD equipment that has been cleaned. The fluctuation amount of the probe inspection yield for fluctuation prediction is expressed by the difference from the preceding and following lots. The prediction result output unit 46 outputs the prediction result to the output device 15 and updates the next fluctuation prediction process. The prediction result output unit 46 registers the prediction result in the abnormality detection record 28 of the abnormality detection specification file 27 in order to automatically verify whether the predicted parameter in the post-process changes as predicted. I do.

FIG. 23 shows an embodiment of a fluctuation prediction method for predicting the influence of a change in manufacturing conditions on a manufacturing apparatus. In this embodiment, the inspection data in which an abnormality is detected is development dimension inspection data in the SG photo process. When an abnormality detection signal is transmitted from the abnormality detection system 20 to the developed dimension inspection data in the SG photo process, the fluctuation prediction parameter specifying unit 42 searches the fluctuation prediction file 47 to execute the SG photo processing. Abnormal development dimension in process The first parameter is read from the fluctuation prediction record 48 and the data is transferred to the data reading unit 43. The data reading section 43 reads necessary data from the process database 17 according to the transferred parameters and transfers the data to the sorting section 44. The sorting section 44 sorts the read data, virtually creates the graph shown in FIG. 23, and transfers it to the fluctuation prediction calculation section 45. The fluctuation prediction calculation unit 45 reads the fluctuation prediction method for predicting the effect of the change in the manufacturing conditions on the manufacturing equipment from the calculation method file 49, and calculates the fluctuation amount of the parameter by using the fluctuation coefficient specified in the calculation method file 49. Use to calculate. According to the embodiment illustrated in FIG. 23, the process of the fluctuation prediction designated by the fluctuation prediction operation unit 45 is a probe inspection process, and the parameter of the fluctuation prediction is a product of the cutout A22. This is the yield in the probe inspection process of the corresponding product type 4 MECL. Then, the fluctuation predicting operation unit 45 calculates the yield fluctuation amount-deterioration amount XI. In FIG. 23, the upper part is a time series relating to the change of the exposure condition of the exposure apparatus in the SG photo process, the middle part is a graph relating to the development dimension inspection value, and the lower part is a graph relating to the deviation value of the yield of the probe inspection. In the graph related to the developed dimension inspection values, the horizontal axis represents the lot number, and the vertical axis represents the developed dimension value. In the graph related to yield, the horizontal axis represents the lot number, and the vertical axis represents the deviation value. Then, the time series relating to the exposure condition change work and each lot number are matched with the passage time of the exposure apparatus whose condition has been changed. The fluctuation amount of the inspection yield for the fluctuation prediction target probe is expressed by the difference between the front and rear mouths. The prediction result output unit 46 outputs the prediction result to the output device 15 and updates the next fluctuation prediction process. In addition, in order to automatically verify whether the predicted parameters in the post-process that have been predicted fluctuate as predicted, the prediction result output unit 46 stores the prediction in the abnormality detection record 28 of the abnormality detection specification file 27. Register the measurement results.

As described above, the IC manufacturing line is automatically and constantly monitored by the abnormality analyzer, and if an abnormality is detected, the effect of the abnormality on the subsequent process is automatically and quickly and accurately predicted. . Prediction of post-process fluctuations caused by these abnormalities makes it possible to take prompt measures for abnormalities in the IC manufacturing line, thereby shortening the period during which abnormalities occur and increasing the number of identical product defects. It is possible to avoid accidents that cause a spliced gun, and ultimately increase the product yield. In addition, unexpected abnormalities tend to occur when new products are launched. Can be launched early. Industrial applicability

As described above, the method and apparatus for analyzing a production line abnormality and the method and apparatus for controlling a production line according to the present invention include an IC, other semiconductor products, a liquid crystal display device, a printed wiring board, a photomask, a magnetic disk, and a compact. It is useful as a method and apparatus for analyzing abnormalities in production lines of industrial products such as disks and automobiles, and is particularly suitable for use in production lines with a large number of production lines and a wide variety of production types.

Claims

The scope of the claims

1. The manufacturing data and inspection data required for the abnormality detection processing for the production line to be analyzed for anomalies are selected as the anomaly detection parameters and the collected anomaly detection file is recorded in the file database in advance. Steps in which an arithmetic method necessary for the abnormality detection process is selected for the production line to be analyzed for abnormality and an edited arithmetic method file is prepared in advance, and manufacturing data and inspection data in each manufacturing process of the production line to be analyzed in the future And a step in which the abnormality detection file is searched and abnormality detection parameters to be subjected to abnormality detection processing are specified by a human operation or automatically, and

A step in which manufacturing data and Z or inspection data corresponding to the specified abnormality detection parameter are read from the process database; and a calculation method to be used from the calculation method file is manually or automatically specified and read. Applying the calculation method to the output data to detect an abnormality; and

An abnormality detection method including a step of outputting a calculation result,

An abnormality analysis method for a production line, comprising:

2. Anomaly detection parameters include the time when the anomaly detection method is executed, execution frequency, product information, manufacturing conditions for each manufacturing device, manufacturing history, maintenance history, failure history, inspection result data for each manufacturing process, and final data. Inspection of product inspection process 2. The abnormality analysis method for a production line according to claim 1, wherein the data includes result data.

3. The manufacturing line according to claim 1 or 2, wherein the calculation method required for the abnormality detection processing includes a control value out-of-value test method, a continuous test method, a moving average method, and a cumulative sum method. Abnormal analysis method.

4. The calculation method file is stored in the file database in advance, the calculation method file stored in this file database is searched, the calculation method is automatically specified, and the specified calculation method is read. 3. The method for analyzing an abnormality of a production line according to claim 1 or claim 2, wherein the abnormality is dynamically applied and an abnormality is detected as soon as possible.

5. The abnormality detection method according to claim 1, and

The production data required for tracking the cause of the

The data and the inspection data are selected as the factor tracking parameters, and the edited factor tracking file is recorded in the file database in advance; and

A step in which an operation method necessary for an abnormal factor tracking process is selected for a production line to be subjected to an abnormal analysis and an edited operation method file is prepared in advance. The factor tracking file is searched and the factor to be subjected to a factor additional process is tracked. A step in which the parameters are specified either manually or automatically,

Reading the manufacturing data and / or inspection data corresponding to the specified factor tracking parameter from the process database; and specifying and automatically specifying a calculation method to be used from the calculation method file by human operation or automatically. Calculation method for the converted data Is applied to track the factors;

An abnormal factor tracking method including: a step of outputting a tracking result.

6. Factor tracking parameters shall include product information, manufacturing conditions of each manufacturing equipment, manufacturing history, maintenance history, failure history, inspection result data for each manufacturing process, and inspection result data for the final product inspection process. The method for analyzing an abnormality of a production line according to claim 5, wherein the method is characterized in that:

7. The calculation method required for the failure factor tracking process is a method in which the factor is added based on the correlation between the inspection data of the manufacturing process in which the abnormality is detected and the inspection data of the manufacturing process upstream of the manufacturing process. 7. The production line abnormality analysis method according to claim 5, wherein the method includes:

8. The calculation method required for the process of tracking the causes of abnormalities is as follows. For a lot group in which the same machining was performed by different manufacturing equipment of the same model in the manufacturing process upstream of the manufacturing process where the abnormality was detected, 7. The abnormality analysis of a manufacturing line according to claim 5, wherein the method includes a method of tracking a factor by a correlation between inspection data of each lot passing through a manufacturing apparatus. Method.

9. The calculation method required for the process of tracking the cause of the abnormality is a lot where the same processing is charged to different positions of the same manufacturing equipment at different positions in the manufacturing process upstream of the manufacturing process where the error was detected. Claims 5 or 6 characterized in that the method includes a method in which a factor is tracked by a correlation between test data for each mouth after passing through each charge position for each group. Abnormality analysis method of the production line described in section.

10. In the calculation method required for the cause tracking process, the same processing was performed before and after the maintenance time of the same manufacturing equipment in the manufacturing process upstream of the manufacturing process where the abnormality was detected. Maintenance of lots Claims 5 or 6 which include a method in which a factor is tracked by a correlation between the inspection data of the lot group before the nonce and the inspection data of the lot group after the maintenance. Abnormality analysis method of the production line described in the paragraph.

1 1. The calculation method required for the process of tracking the cause of the abnormality is that the same processing is performed before and after the timing of changing the manufacturing conditions of the same manufacturing equipment g in the manufacturing process upstream of the manufacturing process where the abnormality was detected. For the implemented lot group, it includes a method to track the factors by the correlation between the inspection data of the lot group before the manufacturing condition change and the inspection data of the lot group after the manufacturing condition change. 7. The abnormality analysis method for a production line according to claim 5 or claim 6.

1 2. The calculation method file is recorded in the file database in advance, the calculation method file stored in this file database is searched, the calculation method is automatically specified, and the specified calculation method is read. 7. The manufacturing line abnormality analysis method according to claim 5, wherein the abnormality factor is tracked by being automatically applied to data.

1 3. The abnormality detection method according to claim 1, and

A step in which manufacturing data and inspection data necessary for the fluctuation prediction processing for the production line to be analyzed for abnormality are selected as fluctuation prediction parameters and a fluctuation prediction file edited and recorded in a file database in advance; A calculation method file required for the fluctuation prediction process for the production line is selected and edited, and the edited calculation method file is recorded in a file database in advance;

A step in which the fluctuation prediction file is searched and a fluctuation prediction parameter to be processed is specified; and

The manufacturing data and Z or Inspection data is read from the process database, the calculation method file is searched, and a calculation method to be used is specified, and the calculation method is applied to the read data, and an abnormality is detected. A step of predicting a change in a manufacturing process after a later manufacturing process, and a step of outputting a prediction result.

A method for analyzing abnormalities in a production line, comprising:

14. The production line according to claim 13, wherein the prediction result is recorded in an abnormality detection file of the abnormality detection method, and whether or not an abnormality actually occurs is verified by the abnormality detection method. Abnormal analysis method.

15 5. Variation prediction parameters include product information, manufacturing conditions for each manufacturing equipment, manufacturing history, maintenance history, failure history, inspection result data for each manufacturing process, inspection result data for the final product inspection process, and variation coefficient. The method for analyzing an abnormality of a production line according to claim 13 or 14, characterized by including

16. A production line control method characterized by being provided with the method for analyzing a production line abnormality according to claim 1, claim 5, or claim 13.

1 7. An abnormality detection file edited and selected for the production data and inspection data required for the abnormality detection processing for the production line to be analyzed for abnormalities, and abnormalities for the production line to be analyzed for abnormalities. A file database that contains the calculation method files that have been selected and edited for the calculation method required for the detection process;

A process database that stores manufacturing data and inspection data in each manufacturing process of the manufacturing line to be analyzed in the future, An abnormality detection parameter designating unit in which the abnormality detection file is searched and an abnormality detection parameter to be processed is specified manually or automatically;

A data reading unit for reading manufacturing data and Z or inspection data corresponding to the specified abnormality detection parameter from the process database;

An abnormality detection operation unit for searching for the operation method file and manually or automatically specifying an operation method to be used, and applying the operation method to the read data to detect an abnormality; ,

An abnormality analysis device for a production line, comprising: an abnormality detection system including an abnormality detection result output unit that outputs a calculation result.

18. Anomaly detection system according to claim 17, and

The production data and inspection data required for the abnormal factor tracking process for the production line to be analyzed for abnormalities were selected and edited as the factor tracking parameters, and the factor tracking file was edited. A file database that contains the calculation method files in which the necessary calculation methods have been selected and edited,

A factor tracing parameter designating unit in which the factor tracing file is searched and processed from now on; a parameter tracing parameter designating unit in which parameters are designated manually or automatically;

A data reading unit for reading manufacturing data and / or inspection data corresponding to the designated factor tracking parameter from the process database;

The calculation method file is searched, and the calculation method to be used from now on is specified manually or automatically, and the read data is given the corresponding operation method. A factor tracking unit for applying a calculation method to track the factor,

An abnormality analysis device for a manufacturing line, comprising: an abnormality factor tracking system including a tracking result output unit for outputting a tracking result.

1 9. Anomaly detection system according to claim 17, and

A fluctuation prediction file in which the manufacturing data and inspection data required for the fluctuation prediction processing for the production line to be analyzed for abnormalities are selected and edited as fluctuation prediction parameters, and the fluctuation prediction file required for the fluctuation prediction processing for the production line to be analyzed for abnormality. A file database that contains calculation method files whose calculation methods have been selected and edited;

A variation prediction parameter specifying unit in which the variation prediction file is searched and the variation prediction parameter to be processed is specified by a manual operation or automatically;

A data reading unit for reading manufacturing data and Z or inspection data corresponding to the designated fluctuation prediction parameter from the process database;

The manufacturing method after the calculation method file is searched and the calculation method to be used is specified manually or automatically, and the calculation method is applied to the read data, and an abnormality is detected. A fluctuation prediction operation unit for predicting fluctuations in the subsequent manufacturing process;

A fluctuation prediction system that includes a prediction result output unit that outputs a prediction result.

20. A control device for a production line, characterized by comprising the abnormality analysis device for a production line according to claim 17, claim 18, or 19.