JP2002110493A - Method and apparatus for extracting failure in machining process - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable anyone to quickly extract failure factors causing the product yield to drop in machining processes such as diffusion steps to meet the actual situation in a semiconductor manufacturing factory. SOLUTION: Using manufacturing apparatus history information I2 including troubles and maintenance information about manufacturing apparatus and inspecting apparatus, and quality result information I1 expressing the yields and electric characteristic values of products, a multi-step multivariate analyzing means 4 analyzes the relations between both information to limit, and extracts failure factors in the manufacturing apparatus history information I2.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for extracting abnormalities in a processing process such as a diffusion process in a semiconductor manufacturing plant.

[0002]

2. Description of the Related Art Japanese Unexamined Patent Publication No. Hei 10-135091 discloses a conventional method and apparatus for analyzing and specifying factors of quality deterioration of a product processed in a processing step such as a diffusion step in semiconductor manufacturing. Some have been disclosed.

In this conventional method and apparatus, factors affecting quality include manufacturing apparatus history, manufacturing conditions such as gas pressure and flow rate set in the manufacturing apparatus, and inspections performed to evaluate processing results of the apparatus. The main results of the survey are the inline measured values that are the results.

When analyzing the cause of quality deterioration, quality result information such as the processing yield of a product and electrical characteristic values are used as objective variables (variables representing results to be finally obtained), and the history information of the manufacturing apparatus described above is used. Information such as manufacturing conditions, in-line measured values, etc. are explanatory variables (variables for explaining objective variables)
The factors are extracted by analyzing the causal relationship between them using a multi-stage multivariate analysis means. The multi-stage multivariate analysis means performs a plurality of analyzes in which the number of explanatory variables used in one analysis is fixed, and abnormal items (explanatory variables) are automatically narrowed down using a known variable increase / decrease method. After that, the final analysis is performed using only the items narrowed down in each analysis.

[0005] Further, when creating analysis information, if the number of data cannot be collected and analysis cannot be performed with a single product type, products under the same manufacturing conditions are collectively analyzed as a "product type group". With this, it is possible to cope with high-mix low-volume production. Furthermore, when the history information of the manufacturing apparatus does not remain, a temporary manufacturing apparatus number is newly assigned, so that it is possible to automatically perform the extraction processing without prior data processing and human judgment for missing data. I have to.

[0006]

By the way, in the above-mentioned conventional analysis method, the history information of the manufacturing apparatus is also used as an explanatory variable.
This is also caused by a human error in troubleshooting or maintenance. However, such human errors are currently analyzed only based on human intuition and experience, and there is a problem that it is not easy for everyone to analyze the factors of quality deterioration. .

An object of the present invention is to solve the above-mentioned problems, and an object of the present invention is to provide a method and an apparatus for extracting abnormalities in a machining process step, which can promptly extract the cause of quality deterioration in accordance with the actual situation. .

[0008]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention according to the first aspect of the present invention is directed to an abnormality for extracting an abnormal factor causing a reduction in product yield in a processing process such as a diffusion process in a semiconductor manufacturing plant. An extraction method, in which a causal relationship between two pieces of information is multi-staged using apparatus history information including information on troubles and maintenance related to a manufacturing apparatus and an inspection apparatus and quality result information indicating a product yield and an electrical characteristic value. The analysis is performed by a multivariate analysis method, thereby narrowing down and extracting an abnormal factor included in the apparatus history information.

That is, quality result information indicating product yield and electrical characteristic values is used as an objective variable, and device history information such as device troubles and maintenance is used as an explanatory variable, and an abnormal candidate is extracted by multiple regression analysis to assist analysis. However, since an effective analysis result cannot be obtained if the number of explanatory variables is too large, a multi-stage multivariate analysis method is used. As described above, the multi-stage multivariate analysis method is a method in which the number of explanatory variables used in one analysis is fixed, and abnormal items (explanatory variables) are automatically narrowed down using a known variable increase / decrease method. Is performed a plurality of times, and the final analysis is performed using only the items narrowed down in each analysis. As a result, it is possible to automatically and rapidly perform an abnormality extraction with high accuracy in accordance with the actual situation in response to the infinite number of explanatory variables.

According to a second aspect of the present invention, there is provided an abnormality extracting apparatus for extracting an abnormal factor which causes a reduction in product yield in a processing process such as a diffusion process in a semiconductor manufacturing plant. A storage unit for storing apparatus history information including maintenance information and quality result information indicating a product yield or an electrical characteristic value; and an apparatus using the quality result information and apparatus history information stored in the storage device. A multi-stage multivariate analysis means for analyzing a causal relationship between the history and the quality result by a multi-stage multivariate analysis method, and narrowing down and extracting an abnormal factor included in the device history information.

The multi-stage multivariate analysis means includes: a data search unit for searching a storage device for quality result information as an objective variable of multiple regression analysis and device history information as an explanatory variable based on input parameters; Analysis data creation unit that creates analysis data from the extracted data, automatic extraction processing unit that performs multivariate analysis of the analysis data in multiple stages and extracts abnormalities, and work data that stores the work data being processed A configuration including a storage unit and a central control unit that controls these processes can be provided.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A method and an apparatus for extracting an abnormality in a machining process according to an embodiment of the present invention will be described below with reference to FIGS.

In FIG. 1, reference numeral 1 denotes an input device for inputting parameters and manufacturing conditions necessary for extracting abnormalities, and 2 includes quality result information I1 and information on troubles and maintenance of manufacturing devices and inspection machines. Device history information I2
Storage device for storing the manufacturing history information I3 and the analysis result information I4, 3 is a central processing unit for performing abnormality extraction, 4 is a multi-stage multivariate analysis means belonging to the central processing unit 3 and performing analysis, and 5 is It is a display device / printing device that displays or prints the extraction result and the like.

The multi-stage multivariate analysis means 4 includes a data search unit A for searching necessary information from the storage device 2, an analysis data generation unit B for generating data for analysis from the searched data, , A central control unit D for controlling these processes, and a work data storage unit E for storing and transferring work data of each process.

As shown in FIG. 2, in the diffusion step, lots L1 to L6 are sequentially sent to the devices DA, DB, and DC, respectively, and manufactured (processed) at the times indicated by black circles. Maintenance MA, MB, respectively
It is assumed that maintenance is performed at the time indicated by MC. Therefore, the manufacturing history information of the lots L1 to L6 and the device D
A maintenance record of A, DB, and DC, and device history information that is a record of trouble occurrence / cancellation when a trouble TC occurs in the device DC are accumulated. In the inspection process, the yields Y1 to Y6 of lots L1 to L6
And electrical characteristics are measured and recorded as quality result information.

In such a diffusion process, when the average yield of a certain product falls below a target value as shown in FIG. 3, a process for extracting a factor that has affected the yield has been described with reference to FIGS.
This will be described with reference to FIG.

As shown in FIG. 4, a target product type as a parameter for analysis and an analysis designation target period P1-P2 as shown in FIG. Is set from the input device 1.
Here, P1−P2 is determined by the operator from the change in the average yield. Next, according to the set parameters, the analysis data search unit A2 obtains the manufacturing history I21 and the device history information I31 in the diffusion process from the manufacturing device history information I2, and the yield information I11 in the inspection process from the quality result information I1, Search for. The searched data is stored in the manufacturing history information storage unit E1, the yield information storage unit E2, and the device history information storage unit E3.

Next, as shown in FIG. 5, the analysis failure data processing section B1 of the analysis data creating section B performs each processing based on the information of the manufacturing history information storage section E1 and the apparatus history information storage section E3. A mark indicating whether or not the lot is after the apparatus history such as apparatus maintenance is added and registered in the processing information storage unit E4.

For example, when the yield is lowered and the cause of the factor extraction is performed in the lots L1 to L3 of the same product type, the maintenance MA, MB, MC
Before the maintenance MB, a mark indicating that it is after the trouble TC is given to the lot L2, and before the maintenance MB, a mark indicating that it is the maintenance MA, MC, and after the trouble TC is given to the lot L3.
A mark indicating that it is before the MB and after the maintenance MC and the trouble TC is added.

Then, the explanatory variable data creating section B2 reads data from the processing information storage section E4, extracts only a certain number of apparatus histories and manufacturing history information from the beginning of the chronologically read apparatus histories, and outputs the data from the yield information storage section E2. The data is registered in the analysis data storage unit E5 in combination with the above data.
All the remaining device history information is also divided and extracted at a fixed number (manufacturing conditions and manufacturing ancillary conditions that are fractions are set to a certain number retroactively from the last condition),
It is registered in the analysis data storage unit E5 in combination with the data from the yield information storage unit E2.

Next, as shown in FIG. 6, in the primary multivariate analysis processing section C1 of the automatic extraction processing section C, reading is performed for each of the fixed number of data from the analysis data storage section E5, and the yield data is read. A multiple regression analysis (quantification class I) is performed using the apparatus history information as an explanatory variable as an objective variable, and manufacturing conditions and manufacturing incidental conditions in which the dispersion ratio F value is higher than a reference value set in advance by an operator by a variable increase / decrease method. Is extracted. At this time, if a multicollinearity error occurs, the reliability of the data is improved by removing the device history information from the analysis range. The result is registered in the temporary analysis result storage unit E6. The processing of the first-order multivariate analysis processing unit C1 is performed for all the division registration data. Next, the secondary multivariate analysis processing unit C2 reads the data analysis result from the temporary analysis result storage unit E6, and the primary multivariate analysis processing unit C1 performs a primary multivariate analysis process on the manufacturing conditions and manufacturing incidental conditions extracted. The same processing as that of the section C1 is performed, and the apparatus history information having a higher dispersion ratio F value than the reference value set in advance by the operator, that is, the factor that has affected the yield reduction is extracted and registered in the analysis result storage section E7.

In the analysis result display processing section C3, based on the data in the analysis result storage section E7, a graph showing a result of the dispersion ratio F value for each apparatus history information, an average yield, a yield distribution graph for each manufacturing condition and manufacturing incidental condition, etc. Display (device) and printing device 5
View through Further, the data storage processing unit C4 registers the analysis result and the parameter in the storage device 2, thereby enabling the result display and the re-analysis with the parameter changed as necessary.

As the parameter, information on electric characteristics may be used instead of the above-mentioned yield information. In addition, as in the case of the conventional abnormality extraction method described above, when the number of data cannot be analyzed for a single product type and analysis is not possible, the product types with the same manufacturing conditions can be analyzed as a whole, and the history information of the manufacturing equipment remains. If not, a temporary manufacturing apparatus may be registered.

[0024]

As described above, according to the present invention, even in a process having a complicated manufacturing process such as a diffusion process in a semiconductor manufacturing factory, when a quality deterioration occurs, the causal relationship between the quality result and the device history is determined. The cause can be extracted by automatically and quickly analyzing, and the yield can be improved by taking prompt action based on the extraction result.

[Brief description of the drawings]

FIG. 1 is an overall configuration diagram showing an abnormality extraction apparatus for a processing process step according to an embodiment of the present invention.

FIG. 2 is a conceptual diagram of a diffusion process to be subjected to abnormality extraction by the abnormality extraction device of FIG. 1;

FIG. 3 is an explanatory diagram showing an average yield of a certain product in the diffusion step of FIG. 2;

FIG. 4 is a block diagram illustrating a data search process when an abnormality is extracted by the abnormality extraction device of FIG. 1;

FIG. 5 is a block diagram illustrating an analysis data creation process when an abnormality is extracted by the abnormality extraction device of FIG. 1;

FIG. 6 is a block diagram illustrating an automatic extraction process when an abnormality is extracted by the abnormality extraction device of FIG. 1;

[Explanation of symbols]

 2 storage device 4 multi-stage multivariate analysis means A data retrieval section B analysis data creation section C automatic extraction processing section

Claims (2)

[Claims] 1. An abnormality extracting method for extracting an abnormal factor causing a decrease in product yield in a processing process such as a diffusion process in a semiconductor manufacturing plant, the device including information on trouble and maintenance related to a manufacturing apparatus and an inspection apparatus. By analyzing the causal relationship between the two pieces of information using the history information and the quality result information indicating the product yield and the electrical characteristic value by a multi-stage multivariate analysis method, the abnormal factors included in the apparatus history information are narrowed down. A method for extracting abnormalities in a machining process step, characterized by extracting. 2. An abnormality extracting apparatus for extracting an abnormal factor causing a decrease in product yield in a processing process such as a diffusion process in a semiconductor manufacturing factory, the apparatus including information on trouble and maintenance related to a manufacturing apparatus and an inspection apparatus. A storage unit for storing history information and quality result information indicating a product yield or an electrical characteristic value; and a causal relationship between the device history and the quality result using the quality result information and the device history information stored in the storage device. An abnormality extraction device for a machining process step, comprising: a multi-stage multivariate analysis means for analyzing a relationship by a multi-stage multivariate analysis method and narrowing down and extracting an abnormal factor included in the device history information.