Granted Patent B2

US 8,219,437 · App. 12/170,577 · Granted Jul 10, 2012

Abstract

Claims (46)

1. A system for constructing plans for operating a production plant to achieve one or more production goals, the system comprising:

at least one processor;

a diagnosis engine implemented using the at least one processor and operative to:

determine a current plant condition based at least partially on a previously executed plan, at least one corresponding observation from the plant, and a plant model of the plant,

formulate a heuristic based on at least one diagnostic objective related to identifying resources of the plant causing intermittent or persistent faults, and to

update a belief model representing the current state of the plant based at least partially on the previously executed plan, the at least one corresponding observation from the plant, and the plant model; and

a planner implemented using the at least one processor and operatively coupled with the diagnosis engine to perform a partial-plan search using the heuristic to construct a plan for execution in the plant that will concurrently achieve at least one production goal and facilitate the at least one diagnostic objective;

wherein the diagnosis engine is operative to calculate plan failure probabilities based on action failure probabilities and the current plant condition, to formulate the heuristic based at least partially on the plan failure probabilities, and to provide information gain criterion to the planner for one or more possible plans based at least partially on the current plant condition and the plant model; and

wherein the planner uses the information gain criterion to construct the plan.

2. The system of claim 1 , wherein the diagnosis engine is operative to derive failure probabilities for a plurality of fault hypothesis in the plant in formulating the heuristic based at least partially on the current plant condition.

3. The system of claim 2 , wherein the diagnosis engine is operative to formulate the heuristic as a lower and upper plan failure probability bounding heuristic based at least partially on the current plant condition.

4. The system of claim 3 , wherein the planner is operative to perform search space pruning to speed up the partial-plan search in constructing the plans.

5. The system of claim 3 , wherein the planner is operative to evaluate partial plans according to the heuristic to preferentially construct informative plans based on the at least one diagnostic objective.

6. The system of claim 3 , wherein the planner is operative to selectively construct plans with a plan failure probability closest to a target value T.

7. The system of claim 1 , wherein the diagnosis engine is operative to formulate the heuristic as a lower and upper plan failure probability bounding heuristic based at least partially on the current plant condition.

8. The system of claim 7 , wherein the planner is operative to perform search space pruning to speed up the partial-plan search in constructing the plans.

9. The system of claim 7 , wherein the planner is operative to evaluate partial plans according to the heuristic to preferentially construct informative plans based on the at least one diagnostic objective.

10. The system of claim 7 , wherein the planner is operative to selectively construct plans with a plan failure probability closest to a target value T.

11. The system of claim 1 , wherein the planner is operative to evaluate partial plans according to the heuristic to preferentially construct informative plans based on the at least one diagnostic objective.

12. The system of claim 1 , wherein the planner performs an A* partial-plan search using the heuristic to construct the plans.

13. A method of constructing plans for operating a production plant to achieve one or more production goals, the method comprising:

using at least one processor, determining a current plant condition based at least partially on a previously executed plan, at least one corresponding observation from the plant, and a plant model of the plant;

using the at least one processor, formulating a heuristic based on at least one diagnostic objective related to identifying resources of the plant causing intermittent or persistent faults;

using the at least one processor, updating a belief model representing the current state of the plant based at least partially on the previously executed plan, the at least one corresponding observation from the plant, and the plant model;

using the at least one processor, performing a partial-plan search using the heuristic;

using the at least one processor, constructing a plan for execution in the plant that will concurrently achieve at least one production goal and facilitate the at least one diagnostic objective based on the partial-plan search;

using the at least one processor, calculating plan failure probabilities based on action failure probabilities and the current plant condition;

using the at least one processor, generating information gain criterion for one or more possible plans based at least partially on the current plant condition and the plant model;

wherein the at least one processor formulates the heuristic based at least partially on the plan failure probabilities; and

wherein the at least one processor constructs the plan using the information gain criterion.

14. The method of claim 13 , wherein formulating the heuristic comprises deriving failure probabilities for a plurality of fault hypothesis in the plant based at least partially on the current plant condition.

15. The method of claim 13 , wherein the heuristic is formulated as a lower and upper plan failure probability bounding heuristic.

16. The method of claim 13 , further comprising performing search space pruning to speed up the partial-plan search in constructing the plans.

17. The method of claim 13 , wherein performing the partial-plan search comprises evaluating partial plans according to the heuristic to preferentially construct informative plans based on the at least one diagnostic objective.

18. A non-transitory computer readable medium having computer executable instructions for performing the steps of:

determining a current plant condition based at least partially on a previously executed plan, at least one corresponding observation from the plant, and a plant model of the plant;

formulating a heuristic based on at least one diagnostic objective related to identifying resources of the plant causing intermittent or persistent faults;

performing a partial-plan search using the heuristic;

updating a belief model representing the current state of the plant based at least partially on the previously executed plan, the at least one corresponding observation from the plant, and the plant model;

constructing a plan for execution in the plant that will concurrently achieve at least one production goal and facilitate the at least one diagnostic objective based on the partial-plan search;

calculating plan failure probabilities based on action failure probabilities and the current plant condition; and

generating information gain criterion for one or more possible plans based at least partially on the current plant condition and the plant model;

wherein the heuristic is formulated based at least partially on the plan failure probabilities; and

wherein the plan is constructed using the information gain criterion.

19. The system of claim 1 , wherein the diagnosis engine performs the partial-plan search using the at least one diagnostic objective and a current state of a partial plan.

20. The system of claim 1 , wherein the diagnosis engine formulates the heuristic at least partially based on the current plant condition.

Assignments (11)