Web-based system of product performance assessment and quality control using adaptive PDF fitting

• US 20070192060A1
• Filed: 02/14/2006
• Published: 08/16/2007
• Est. Priority Date: 02/14/2006
• Status: Active Grant

• Current Assignee: Seagate Technology LLC (Seagate Technology Holdings Plc)
• Original Assignee: Seagate Technology LLC (Seagate Technology Holdings Plc)
• Inventors: Ng, Hwaliang, Yoap, Timothy, Yam, Hongsee, Lim, Teck
• Granted Patent: US 7,565,266 B2
• Time in Patent Office: Days
• US Class Current: 702/181
• CPC Class Codes: G05B 2219/1112 Bit addressing, handling
• CPC Class Codes: G05B 2223/02 Indirect monitoring, e.g. monitoring production to detect faults of a system
• CPC Class Codes: G05B 23/0221 Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods
• CPC Class Codes: G06F 18/2321 using statistics or function optimisation, e.g. modelling of probability density functions
• CPC Class Codes: G06F 18/2433 Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection

Abstract

A data analysis system analyzes data sets that are characterized by a wide variety of probability density functions (PDFs), and also analyzes mixed populations, i.e., a single data set containing subsets of data that each fit a different distribution. The data analysis system receives quality control data from a product manufacturing or product testing floor, determines whether the data is a mixed population, and fits the data with a variety of PDFs. The data analysis system selects the best fitting PDF based on statistical characteristics calculated for each fitted PDF. The data analysis system generates performance capability parameters based on the best fitting PDF, and may generate reports illustrating the performance capability parameters. The data analysis system also adjusts statistical control limits to provide a more reliable process trigger and control plan. In this manner, the data analysis system may more accurately report product performance.

Claims

1. A method comprising:

receiving data for a product;

fitting the data with a plurality of probability density functions (PDFs);

selecting a PDF that best fits the data;

generating one or more performance capability parameters for the product based on the best fitting PDF; and

generating a report that presents at least one of the performance capability parameters.

2. The method of claim 1, further comprising:

providing a user interface for allowing a user to separate the data into two or more data sets, wherein one of the data sets is a main distribution data set;

fitting the main distribution data set with the plurality of PDFs; and

selecting a PDF that best fits the main distribution data set.

3. The method of claim 1, wherein receiving data for a product further comprises receiving real-time data from a product testing floor.

4. The method of claim 1, further comprising adjusting statistical control limits for the product based on the data.

5. The method of claim 4, wherein adjusting statistical control limits comprises generating a first set of data samples using the selected PDF, calculating new control limits from the first set of data samples, generating a second set of data samples, determining an average run length (ARL) value from the second set of data samples and the new control limits, and comparing the determined ARL value to an acceptable ARL value.

6. The method of claim 1, wherein selecting the PDF comprises comparing an actual defect rate with an estimated defect rate from each of the PDFs;

and computing a validity rating for the data and each of the PDFs.

7. The method of claim 6, wherein comparing the actual defect rate with the estimated defect rate comprises computing a χ

² indicator for each of the PDFs and selecting the smallest χ

² indicator.

8. The method of claim 6, wherein computing the validity rating comprises computing based on a p-value, kurtosis, and skewness of the data.

9. The method of claim 6, further comprising adjusting a specification limit when the actual defect rate is zero.

10. The method of claim 1, wherein fitting the data with a plurality of PDFs comprises fitting the data with at least two of the group consisting of a normal distribution, an exponential distribution, a Weibull distribution, a 3-Parameter Weibull distribution, a uniform distribution, a Laplace distribution, a power transformation, a logistic distribution, or a 3-parameter LogLogistic distribution.

29. The method of claim 1, further comprising:

automatically separating the data into two or more data sets when the data comprises two or more data distributions, wherein one of the data sets is a main distribution data set;

fitting the main distribution data set with the plurality of PDFs; and

selecting a PDF that best fits the main distribution data set.

30. The method of claim 29, wherein automatically separating the data into two or more data sets comprises automatically separating the data into three data sets when the data comprises three data distributions, wherein the data sets comprise a main distribution, an upper tail section, and a lower tail section.

31. The method of claim 29, further comprising:

determining a proportion of each of the lower tail section, the main distribution, and the upper tail section with respect to the data; and

computing defects per opportunity (DPO) rates for each of the lower tail section and the upper tail section.

11. A system comprising:

a data acquisition module that acquires data for a product;

a probability density function (PDF) fitting module that fits the data with a plurality of PDFs, selects a PDF that best fits the data, and generates one or more performance capability parameters for the product based on the best fitting PDF; and

a report generator that generates a report that presents at least one of the performance capability parameters.

12. The system of claim 11, further comprising:

a database that stores the data; and

a PDF repository that stores the plurality of PDFs.

13. The system of claim 11, further comprising a user interface for allowing a user to separate the data into two or more data sets, wherein one of the data sets is a main distribution data set, wherein the PDF fitting module fits the main distribution data set with the plurality of PDFs and selects a PDF that best fits the main data set.

14. The system of claim 11, wherein the data acquisition module acquires real-time data from a product testing floor.

15. The system of claim 11, further comprising a control limit determination module that adjusts statistical control limits for the product based on the data.

16. The system of claim 11, wherein the plurality of PDFs include at least two of the group consisting of a normal distribution, an exponential distribution, a Weibull distribution, a 3-Parameter Weibull distribution, a uniform distribution, a Laplace distribution, a power;

transformation, a logistic distribution, or a 3-parameter LogLogistic distribution.

27. The system of claim 11, further comprising a control limit determination module that generates a first set of data samples using the selected PDF, calculates new control limits from the first set of data samples, generates a second set of data samples, determines an average ran length (ARL) value from the second set of data samples and the new control limits, and compares the determined ARL value to an acceptable ARL value.

32. The system of claim 11, further comprising a trimming module that automatically separates the data into two or more data sets when the data comprises two or more data distributions, wherein one of the data sets is a main distribution data set, and wherein the PDF fitting module fits the main distribution data set with the plurality of PDFs and selects a PDF that best fits the main data set.

17. A computer-readable medium comprising instructions for causing a processor to:

receive data for a product;

fit the data with a plurality of probability density functions (PDFs);

select the PDF that best fits the data;

generate one or more performance capability parameters for the product based on the best fitting PDF; and

generate a report that presents at least one of the performance capability parameters.

18. The computer-readable medium of claim 17, further comprising instructions to cause the processor to:

provide a user interface for allowing a user to separate the data into two or more data sets, wherein one of the data sets is a main distribution data set;

fit the main distribution data set with the plurality of PDFs; and

select a PDF that best fits the main distribution data set.

19. The computer-readable medium of claim 17, further comprising instructions to cause the processor to receive real-time data from a product testing floor.

20. The computer-readable medium of claim 17, further comprising instructions to cause the processor to adjust statistical control limits for the product based on the data.

21. The computer-readable medium, of claim 17, further comprising instructions to cause the processor to fit the data with at least two of the group consisting of a normal distribution, an exponential distribution, a Weibull distribution, a 3-Parameter Weibull distribution, a uniform distribution, a Laplace distribution, a power transformation, a logistic distribution, or a 3-parameter LogLogistic distribution.

28. The computer-readable medium of claim 17, further comprising instructions to cause the processor to:

generate a first set of data samples using the selected PDF;

calculate new control limits from the first set of data samples;

generate a second set of data samples;

determine an average run length (ARL) value from the second set of data samples and the new control limits; and

compare the determined ARL value to an acceptable ARL value.

22. A method comprising:

obtaining parameters associated with a probability density function (PDF) that best fits a set of data for a product;

adjusting statistical control limits based on the parameters;

receiving data associated with the product;

triggering a user when a predetermined amount of the received data is outside of the adjusted statistical control limits;

using a one-sided pre-control chart when one of the parameters is at or above a threshold; and

using a confidence interval control chart when the parameter is below the threshold.

24. The method claim 22, wherein the parameter is Z_ST and the threshold is 10.

25. The method of claim 22, wherein using a one-sided pre-control chart comprises:

obtaining a specification limit;

determining a lower pre-control limit using the specification limit; and

generating a report that shows the received data relative to the specification limit and the lower pre-control limit.

26. The method of claim 22, wherein using a confidence interval control chart comprises:

determining an upper confidence level and a lower confidence level; and

generating a report that shows the received data relative to the upper confidence level and the lower confidence level.

23. (canceled)