Test data analyzing system and test data analyzing program

• US 7,424,336 B2
• Filed: 07/11/2006
• Issued: 09/09/2008
• Est. Priority Date: 07/11/2005
• Status: Expired due to Fees

• Current Assignee: Hitachi High-Technologies Corporation (Hitachi, Ltd.)
• Original Assignee: Hitachi High-Technologies Corporation (Hitachi, Ltd.)
• Inventors: Ono, Makoto, Konishi, Junko
• Primary Examiner(s): Kosowski; Alexander J
• Application Number:
  US11/483,740
• Publication Number:
  US 20070021855A1
• Time in Patent Office: 791 Days
• Field of Search: 700108-110, 700/121, 324/500, 324/765, 438/14, 438/17
• US Class Current: 700/108
• CPC Class Codes: G05B 19/41875 characterised by quality surveillance of production
• CPC Class Codes: G05B 2219/32196 Store audit, history of inspection, control and workpiece data into database
• CPC Class Codes: G05B 2219/32206 Selection from a lot of workpieces to be inspected
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
• CPC Class Codes: Y02P 90/02 Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
• CPC Class Codes: Y02P 90/80 Management or planning

Abstract

Disclosed is a test data analyzing method and system for use in estimation of a defect cause of a product, such as, an integrated circuit, a liquid crystal display, an optical transceiver, a thin film magnetic head, etc., which is fabricated through plural processes. The estimation of a defect cause is achieved by selecting a wafer number to be analyzed, reading test data, reading fabrication line data, counting frequency of machine codes by wafers, grouping test data by machine codes or frequencies, comparing test data distributions between groups by machine codes, and comparing results between machine codes.

Claims

1. A test data analyzing system for estimating a fabrication machine that makes a defect cause in a product that is fabricated through plural processes, including two or more repetitions, the system comprising:

a fabrication line data management unit, which records information of machine code numbers specifying fabrication machines that process the product in a time series, as fabrication line data for the respective products;

a test data management unit, which stores test data obtained from an inspection done on the product by an inspection machine; and

a data analysis unit, which (1) presents a user interface that displays the list of the products and the test data to a user, accents the input of the selection of the product and the inspection items by the user, (2) reads a fabrication line data of each of the selected products, groups the products per each machine code number per usage frequency of the fabrication machine, (3) reads the test data of the selected inspection item corresponding to each product to prepare a test data distribution for each group, (4) calculates a significance probability P value from the ratio between each variance of the test data distribution per usage frequency of the fabrication machine by machine code numbers, (5) displays a box-and-whisker plot of the usage frequency of the fabrication machine and the test data, on a basis of an order sorted by the significance probability P value as designated by a user.

2. The system of claim 1, wherein the data analysis unit selects, as a candidate of a fabrication machine with a defect cause, only the machine code number that has a minus coefficient of correlation in a scatter diagram with the horizontal axis representing usage frequency of the fabrication machine and the vertical axis representing test data.

3. A test data analyzing method for supporting a user to estimate a fabrication machine creating a defect cause in a product that is fabricated through plural processes, including two or more repetitions, the method comprising:

presenting a user interface that displays a list of products and test data to the user and accepting input of a selection of a product and inspection items by the user;

reading fabrication line data that includes information of machine code numbers specifying fabrication machines that process the product in a time series;

reading test data obtained from an inspection on the product by an inspection machine;

grouping products, according to the fabrication line data, per each machine code number per usage frequency of the fabrication machine, and preparing a test data distribution of the respective groups;

calculating a significance probability P value from the ratio between each variance of the test data distribution per usage frequency of the fabrication machine by the machine code numbers; and

displaying box-and-whisker plot of the usage frequency of the fabrication machine and the test data, on a basis of an order sorted by the significance probability P value as designated by the user.

4. The test data analyzing method of claim 3, wherein the significance probability P value is calculated by selecting, as a candidate of a fabrication machine with a defect cause, only the machine code number that has a minus coefficient of correlation in a scatter diagram with the horizontal axis representing usage frequency of the fabrication machine and the vertical axis representing test data.

5. A computer-readable medium containing a test data analyzing program for supporting a user to determine a fabrication machine creating a defect cause in a product that is fabricated through plural processes, including two or more repetitions, the program, when implemented, causing operations comprising:

presenting a user interface that displays a list of products and test data to the user, and accepting input of a selection of a product and inspection items by the user;

reading fabrication line data that includes information of machine code numbers specifying fabrication machines that process the product in a time series;

reading test data obtained from an inspection on the product by an inspection machine;

grouping products, according to the fabrication line data, per each machine code number per usage frequency of the fabrication machine, and preparing a test data distribution of the respective groups;

calculating a significance probability P value from the ratio between each variance of the test data distribution per usage frequency of the fabrication machine by the machine code numbers; and

displaying a box-and-whisker plot of the usage frequency of the fabrication machine and the test data, on a basis of an order sorted by the significance probability P value as designated by the user.

6. The computer-readable medium of claim 5, wherein the significance probability P value is calculated by selecting, as a candidate of a fabrication machine with a defect cause, only the machine code number that has a minus coefficient of correlation in a scatter diagram with the horizontal axis representing usage frequency of the fabrication machine and the vertical axis representing test data.