Method and system for determining a corrective action

• US 20080301499A1
• Filed: 05/31/2007
• Published: 12/04/2008
• Est. Priority Date: 05/31/2007
• Status: Active Grant

• Current Assignee: Solar Turbines Incorporated (Caterpillar Incorporated)
• Original Assignee: Solar Turbines Incorporated (Caterpillar Incorporated)
• Inventors: Grichnik, Anthony J., Seskin, Michael, Boggs, Brian Lee
• Granted Patent: US 8,015,134 B2
• Time in Patent Office: Days
• US Class Current: 714/25
• CPC Class Codes: G05B 23/0232 based on qualitative trend analysis, e.g. system evolution
• CPC Class Codes: G05B 23/0281 Quantitative, e.g. mathematical distance; Clustering; Neural networks; Statistical analysis

Abstract

A computer-implemented method is provided for determining a corrective action. The method may include obtaining diagnostic data and calculating, using the diagnostic data, a prognostic. The method may also include retrieving, when the prognostic is above an alarm level, historical records and calculating a Bayesian posterior probability using the diagnostic data and the historical records. Further, the method may include calculating a prognostic prescience metric using the diagnostic data and the historical records and determining, based on the Bayesian posterior probability and the prognostic prescience metric, the corrective action.

Claims

1. A computer-implemented method for determining a corrective action, comprising:

obtaining diagnostic data;

calculating, using the diagnostic data, a prognostic;

retrieving, when the prognostic is above an alarm level, historical records;

calculating a Bayesian posterior probability using the diagnostic data and the historical records;

calculating a prognostic prescience metric using the diagnostic data and the historical records; and

determining, based on the Bayesian posterior probability and the prognostic prescience metric, the corrective action.

2. The computer-implemented method of claim 1, wherein the corrective action comprises a repair to a machine.

3. The computer-implemented method of claim 1, wherein the corrective action comprises a medical intervention to a patient.

4. The computer-implemented method of claim 1, wherein the diagnostic data and the historical records include data indicating a status of a component of one or more machines.

5. The computer-implemented method of claim 1, wherein the diagnostic data and the historical records include data indicating a status of a biological function of one or more patients.

6. The computer-implemented method of claim 1, wherein the Bayesian posterior probability is calculated as

$p$

( A t | X ψ

) = p 

( X ψ

| A t ) * p 

( A t ) p 

( X ψ

| A t ) * p 

( A t ) + p 

( X ψ

| ∼

A t ) * p 

( ∼

A t )

provided that A_t represents a defined corrective action,

X represents the alarm being triggered,

p(A_t|X_ψ

) represents the normalized probability of needing the defined corrective action when the alarm occurred,

p(X_ψ

|A_t) represents the probability that the alarm is triggered when the defined corrective action is needed,

p(X_ψ

|˜

A_t) represents the probability that the alarm is triggered when the defined corrective action is not needed,

p(A_t) represents the probability of needing the defined corrective action, and

p(˜

A_t) represents the probability of not needing the defined corrective action

7. The computer-implemented method of claim 1, wherein the prognostic prescience metric is calculated as

$\mu$

= ∑

1 t 

p 

( A t | X ψ

, t ) * ( 

t / 

t 



max ) ∑

1 t 

( 

t / 

t 



max ) ,

provided that μ

represents the accuracy of the prognostic over time,

t represents a lower time measurement going back from the time of calculation,

tmax represents the upper time range to calculate the prognostic prescience metric,

p(A_t|X<sub>ψ

,t</sub>) is the probability of a failure at time t given that the alarm occurred at time t, and

e is the base of natural logarithms.

8. The computer-implemented method of claim 1, further including performing an economic calculation to determine the corrective action.

9. A computer-readable medium, tangibly embodied, including a tool for determining a corrective action, the computer-readable medium comprising instructions for:

obtaining diagnostic data;

calculating, using the diagnostic data, a prognostic;

retrieving, when the prognostic is above an alarm level, historical records;

calculating a Bayesian posterior probability using the diagnostic data and the historical records;

calculating a prognostic prescience metric using the diagnostic data and the historical records; and

determining, based on the Bayesian posterior probability and the prognostic prescience metric, the corrective action.

10. The computer-readable medium of claim 9, wherein the corrective action comprises a repair to a machine.

11. The computer-readable medium of claim 9, wherein the corrective action comprises a medical intervention to a patient.

12. The computer-readable medium of claim 9, wherein the diagnostic data and the historical records include data indicating a status of a component of one or more machines.

13. The computer-readable medium of claim 9, wherein the diagnostic data and the historical records include data indicating the status of a biological function of one or more patients.

14. The computer-readable medium of claim 9, wherein the Bayesian posterior probability is calculated as

$p$

( A t | X ψ

) = p 

( X ψ

| A t ) * p 

( A t ) p 

( X ψ

| A t ) * p 

( A t ) + p 

( X ψ

| ∼

A t ) * p 

( ∼

A t )

provided that A_t represents a defined corrective action,

X represents the alarm being triggered,

p(A_t|X_ψ

) represents the normalized probability of needing the defined corrective action when the alarm occurred,

p(X_ψ

|A_t) represents the probability that the alarm is triggered when the defined corrective action is needed,

p(X_ψ

|˜

A_t) represents the probability that the alarm is triggered when the defined corrective action is not needed,

p(A_t) represents the probability of needing the defined corrective action, and

p(˜

A_t) represents the probability of not needing the defined corrective action

15. The computer-readable medium of claim 9, wherein the prognostic prescience metric is calculated as

$\mu$

= ∑

1 t 

p 

( A t | X ψ

, t ) * ( 

t / 

t 



max ) ∑

1 t 

( 

t / 

t 



max ) ,

provided that μ

represents the accuracy of the prognostic over time,

t represents a lower time measurement going back from the time of calculation,

tmax represents the upper time range to calculate the prognostic prescience metric,

p(A_t|X<sub>ψ

,t</sub>) is the probability of a failure at time t given that the alarm occurred at time t, and

e is the base of natural logarithms.

16. The computer-readable medium of claim 9, wherein the instructions include performing an economic calculation to determine the corrective action.

17. A system comprising:

a platform;

at least one input device; and

at least one central processing unit in communication with the platform and the at least one input device, wherein the central processing unit;

obtains diagnostic data;

calculates, using the diagnostic data, a prognostic;

retrieves, when the prognostic is above an alarm level, historical records;

calculates a Bayesian posterior probability using the diagnostic data and the historical records;

calculates a prognostic prescience metric using the diagnostic data and the historical records; and

determines, based on the Bayesian posterior probability and the prognostic prescience metric, the corrective action.

18. The system of claim 17, wherein:

the corrective action comprises a repair to a machine, and

the diagnostic data and the historical records include data indicating a status of a component of one or more machines.

19. The system of claim 17, wherein:

the corrective action comprises a medical intervention to a patient, and

the diagnostic data and the historical records include data indicating a status of a biological function of one or more patients.

20. The system of claim 17, wherein the central processing unit calculates the Bayesian posterior probability as

$p$

( A t | X ψ

) = p 

( X ψ

| A t ) * p 

( A t ) p 

( X ψ

| A t ) * p 

( A t ) + p 

( X ψ

| ∼

A t ) * p 

( ∼

A t )

provided that A_t represents a defined corrective action,

X represents the alarm being triggered,

p(A_t|X_ψ

) represents the normalized probability of needing the defined corrective action when the alarm occurred,

p(X_ψ

|A_t) represents the probability that the alarm is triggered when the defined corrective action is needed,

p(X_ψ

|˜

A_t) represents the probability that the alarm is triggered when the defined corrective action is not needed,

p(A_t) represents the probability of needing the defined corrective action, and

p(˜

A_t) represents the probability of not needing the defined corrective action

21. The system of claim 17, wherein the central processing unit calculates the prognostic prescience metric as

$\mu$

= ∑

1 t 

p 

( A t | X ψ

, t ) * ( 

t / 

t 



max ) ∑

1 t 

( 

t / 

t 



max ) ,

provided that μ

represents the accuracy of the prognostic over time,

t represents a lower time measurement going back from the time of calculation,

tmax represents the upper time range to calculate the prognostic prescience metric,

p(A_t|X<sub>ψ

,t</sub>) is the probability of a failure at time t given that the alarm occurred at time t, and

e is the base of natural logarithms.

22. The system of claim 17, wherein the central processing unit performs an economic calculation to determine the corrective action.