Automatic generation of baysian diagnostics from fault trees

• US 20050160324A1
• Filed: 12/24/2003
• Published: 07/21/2005
• Est. Priority Date: 12/24/2003
• Status: Active Grant

• Current Assignee: The Boeing Co.
• Original Assignee: The Boeing Co.
• Inventors: Milford, Richard I., Przytula, Krzysztof W.
• Granted Patent: US 7,158,958 B2
• Time in Patent Office: Days
• US Class Current: 714/43
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

Fault trees are automatically converted to Bayesian networks for assisting in system reliability, failure analysis and diagnostics by using information from the fault tree structure to create the Bayesian network structure, creating parameters of the Bayesian network using information from the fault tree, obtaining information about observation nodes for the Bayesian network from a list of observations that augments information contained in the fault tree, and inserting the observation nodes into the Bayesian network. The fault tree is pre-processed into an intermediate format prior to conversion that may include adding reliability values from a separate text document when the fault tree is in such format that requires it.

Claims

1. A method of conversion of fault trees to Bayesian networks for assisting in system reliability, failure analysis and diagnostics comprising the steps of:

creating a structure of a Bayesian network using information from a fault tree;

creating parameters of the Bayesian network using information from the fault tree;

obtaining information about observation nodes for the Bayesian network from a list of observations that augments information contained in the fault tree; and

inserting the observation nodes into the Bayesian network.

2. The method of claim 1 wherein the step of creating a structure of a Bayesian network comprises the steps of:

creating a node in the Bayesian network for each node in the fault tree;

naming each node in the Bayesian network the same as a corresponding node in the fault tree;

labeling each node in the Bayesian network with a unique identifier from the fault tree if available, and creating the unique identifiers if they are not available in the fault tree;

assigning to each Bayesian network node two states, one being a failure state and one being an OK state;

connecting the nodes in the Bayesian network with links as corresponding nodes in the fault tree are connected and wherein the direction of connections between the nodes is equivalent to the direction up the fault tree.

3. The method of claim 1 wherein the step of creating parameters of the Bayesian network comprises the steps of:

defining prior probability of failure for each root node of the Bayesian network;

building conditional probability tables based upon the type of logic operations performed in the fault tree; and

defining conditional probabilities for each of the nodes in the Bayesian network that are not root nodes using corresponding conditional probability tables.

4. The method of claim 3, further comprising the step of pre-processing the fault tree into an intermediate format prior to conversion.

5. The method of claim 4 wherein pre-processing step comprises syntax checking of the fault tree.

6. The method of claim 4 wherein the pre-processing step comprises including reliability values from a separate text document when the fault tree is in such format that requires it.

7. The method of claim 1 further comprising the step of performing a system diagnostic procedure using the Bayesian network.

8. The method of claim 7 further comprising the step of taking repair actions based upon diagnostic recommendations resulting from the diagnostic procedure.

9. The method of claim 1 further comprising the step of performing system failure analysis using the Bayesian network.

10. The method of claim 1 further comprising the step of performing system reliability analysis using the Bayesian network.

11. The method of claim 1 wherein the information about observation nodes is based on spacecraft telemetry information.

12. The method of claim 1 wherein the steps are performed automatically.

13. A system for performing diagnostic procedures comprising:

means for creating a structure of a Bayesian network using information from a fault tree;

means for creating parameters of the Bayesian network using information from the fault tree operable for working in conjunction with the means for creating a structure of a Bayesian network;

means for obtaining information about observation nodes for the Bayesian network from a list of observations that augments information contained in the fault tree operable for working in conjunction with the means for creating a structure of a Bayesian network; and

means for inserting the observation nodes into the Bayesian network operable for working in conjunction with the means for creating a structure of a Bayesian network.

14. The system of claim 13 wherein the means for creating a structure of a Bayesian network comprises:

means for creating a node in the Bayesian network for each node in the fault tree;

means for naming each node in the Bayesian network the same as a corresponding node in the fault tree;

means for labeling each node in the Bayesian network with a unique identifier from the fault tree if available, and creating the unique identifiers if they are not available in the fault tree;

means for assigning to each Bayesian network node two states, one being a failure state and one being an OK state;

means for connecting the nodes in the Bayesian network with links as corresponding nodes in the fault tree are connected and wherein the direction of connections between the nodes is equivalent to the direction up the fault tree wherein all said means are in operable communication with a means for controlling operation of the system.