System and method for early detection of system component failure

• US 20060265625A1
• Filed: 05/19/2005
• Published: 11/23/2006
• Est. Priority Date: 05/19/2005
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: International Business Machines Corporation
• Inventors: Dubois, Andrew J. Jr., Yashchin, Emmanuel, Jensen, David L., Ross, Christopher D., Khabibrakhmanov, Ildar, Restivo, Stephen, Evans, Vaughn Robert
• Granted Patent: US 7,401,263 B2
• Time in Patent Office: Days
• US Class Current: 714/6
• CPC Class Codes: G07C 3/00 Registering or indicating the condition or the working of machines or other apparatus, other than vehicles

Abstract

A system and method for detecting trends in time-managed lifetime data for products shipped in distinct vintages within a time window. Data is consolidated from several sources and represented in a form amenable to detection of trends using a criterion for measuring failure. A weight is applied to the failure measures, the weight increasing over the time the products are in the time window. A function of weighted failures is used to define a severity index for proximity to an unacceptable level of failures, and an alarm signal is triggered at a threshold level that allows the level of false alarms to be pre-set.

Claims

1. A method for detecting trends in time-managed lifetime data, comprising the steps of:

storing in a database time-managed lifetime data for a product;

establishing a criterion from said data for measuring failure of said product;

comparing measured failures of the product within a time window against expected failures of the product within said time window; and

triggering an alarm signal when a value of said comparison exceeds a threshold, said threshold being chosen to limit false alarms to a pre-specified rate.

2. A method as in claim 1, wherein said product is comprised of components and is shipped in a sequence of discrete vintages within said time window, said time-managed lifetime data for each said vintage being updated periodically with new information as each said vintage progresses through said time window.

3. A method as in claim 2, wherein said comparison is a computation or simulation analysis determining a probability that a hypothetical sequence of vintages having said expected failures will produce a failure statistic less than or equal to said failure statistic for said observed failures, said probability being an index of severity for said criterion.

4. A method as in claim 3, wherein said failure statistic is produced by

establishing a weight to be applied to a value of said criterion, said weight being proportional to a volume of said product within a vintage and increasing over time within said time window;

defining and computing for each said vintage in said sequence a cumulative function based on said weight applied to a value of said criterion, said value of said criterion being reduced by a reference value before application of said weight; and

defining a maximum value of said function over said vintages.

5. A method as in claim 1, wherein the database is derived from multiple sources.

6. A method as in claim 4, wherein said function is the function

S₀=0, S_i=max[0, S_i-1+W_i (X_i−

k)]

for vintages i=1 to N, where X_i is the value of said criterion for vintage i, W_i is said weight to be applied to said criterion, and k is said reference value.

7. A method as in claim 6, wherein said criterion is a rate of replacement of said product and said weight is a measure of service time of said product within a vintage.

8. A method as in claim 4, wherein said threshold is a trigger value, slightly less than one, of said severity index, the probability of a false alarm being the difference between one and said threshold.

9. A method as in claim 1, wherein said criterion for measuring failure of said product measures failure of a component of said product.

10. A method as in claim 6, further comprising the steps of:

determining whether the product is active;

if the product is active, triggering a supplemental alarm signal when said failure statistic is defined as the value S_N.

11. A method as in claim 1, further comprising the steps of:

determining whether the product is active;

if the product is active, triggering a tertiary alarm signal when said comparison is a computation or simulation analysis determining a probability that a hypothetical sequence of vintages having said expected failures will produce within an active period a cumulative total of said expected failures greater than or equal to the cumulative total of said observed failures.

12. A method as in claim 10, further comprising the step of triggering a tertiary alarm signal, if the product is active, when said comparison is a computation or simulation analysis determining a probability that a hypothetical sequence of vintages having said expected failures will produce within an active period a cumulative total of said expected failures greater than or equal to the cumulative total of said observed failures.

13. A method as in claim 12, further comprising the steps of:

combining said severity index for said criterion with a severity index corresponding to said secondary alarm signal and a severity index corresponding to said tertiary alarm signal into a function; and

triggering an alarm signal when said combined function exceeds a threshold.

14. A system for detecting trends in time-managed lifetime data, comprising:

means for storing in a database time-managed lifetime data for a product;

means for establishing a criterion from said data for measuring failure of said product;

means for comparing measured failures of the product within a time window against expected failures of the product within said time window; and

means for triggering an alarm signal when a value of said comparison exceeds a threshold, said threshold being chosen to limit false alarms to a pre-specified rate.

15. A system as in claim 14, wherein said product is comprised of components and is shipped in a sequence of discrete vintages within said time window, said time-managed lifetime data for each said vintage being updated periodically with new information as each said vintage progresses through said time window.

16. A system as in claim 15, wherein said comparison is a simulation analysis determining a probability that a hypothetical sequence of vintages having said expected failures will produce a failure statistic less than or equal to said failure statistic for said observed failures, said probability being an index of severity for said criterion.

17. A system as in claim 16, wherein said failure statistic is produced by a statistical analysis module, which module:

establishes a weight to be applied to a value of said criterion, said weight being proportional to a volume of said product within a vintage and increasing over time within said time window;

defines and computes for each said vintage in said sequence a cumulative function based on said weight applied to a value of said criterion, said value of said criterion being reduced by a reference value before application of said weight; and

defines a maximum value of said function over said vintages.

18. A system as in claim 14, wherein the database is derived from multiple sources by a data integration module.

19. A system as in claim 17, wherein said function is the function

S₀=0, S_i=max[0, S_i-1+W_i (X_i−

k)],

for vintages i=1 to N, where X_i is the value of said criterion for vintage i, W_i is said weight to be applied to said criterion, and k is said reference value.

20. A system as in claim 19, wherein said criterion is a rate of replacement of said product and said weight is a measure of service time of said product within a vintage.

21. A system as in claim 17, wherein said threshold is a trigger value, slightly less than one, of said severity index, the probability of a false alarm being the difference between one and said threshold.

22. A system as in claim 14, wherein said criterion for measuring failure of said product measures failure of a component of said product.

23. A system as in claim 19, further comprising:

means for determining whether the product is active;

means for triggering a supplemental alarm signal when said failure statistic is defined as the value S_N, said triggering means being operable if the product is active.

24. A system as in claim 14, further comprising:

means for determining whether the product is active;

means for triggering a tertiary alarm signal when said comparison is a computation or simulation analysis determining a probability that a hypothetical sequence of vintages having said expected failures will produce within an active period a cumulative total of said expected failures greater than or equal to the cumulative total of said observed failures, wherein said triggering means is operable if the product is active.

25. A system as in claim 23, further comprising means for triggering a tertiary alarm signal when said comparison is a computation or simulation analysis determining a probability that a hypothetical sequence of vintages having said expected failures will produce within an active period a cumulative total of said expected failures greater than or equal to the cumulative total of said observed failures, wherein said triggering means is operable if the product is active.

26. A system as in claim 25, further comprising:

means for combining said severity index for said criterion with a severity index corresponding to said secondary alarm signal and a severity index corresponding to said tertiary alarm signal into a function; and

means for triggering an alarm signal when said combined function exceeds a threshold.

27. A computer implemented software tool for detecting trends in time-managed lifetime data, comprising:

first computer code for storing in a database time-managed lifetime data for a component of a product;

second computer code for establishing a criterion from said data for measuring failure of said component;

third computer code for comparing measured failures of the component within a time window against expected failures of the component within said time window; and

fourth computer code for triggering an alarm signal when a value of said comparison exceeds a threshold, said threshold being chosen to limit false alarms to a pre-specified rate.