Integrating design and field management of gas turbine engine components with a probabilistic model

• US 10,330,018 B2
• Filed: 02/25/2015
• Issued: 06/25/2019
• Est. Priority Date: 03/24/2014
• Status: Active Grant

• Current Assignee: Rolls-Royce Corporation (Rolls-Royce Holdings Plc)
• Original Assignee: Rolls-Royce Corporation (Rolls-Royce Holdings Plc)
• Inventors: Stillinger, Jeffrey D., Grey, Zachary J., Fitzpatrick, Rhoderik J.
• Primary Examiner(s): Cassity, Robert A
• Assistant Examiner(s): Zidanic, Michael
• Application Number:
  US14/631,101
• Publication Number:
  US 20150269490A1
• Time in Patent Office: 1,581 Days
• Field of Search: None
• CPC Class Codes: F02C 9/00 Controlling gas-turbine plants; Controlling fuel supply in air- breathing jet-propulsion plants controlling air intakes F02C7/057; controlling turbines F01D; controlling compressors F04D27/00; controlling in general G05
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

Embodiments of a gas turbine engine lifecycle decision assistant apply a probabilistic-based process founded on a Bayesian mathematical framework to intelligently combine analytical models, expert judgment, and data during the development and field management of gas turbine engines. The process integrates physics-based and high-fidelity models with data and expert judgment that evolves over the course of the gas turbine engine lifecycle. Among other things, embodiments of the gas turbine engine lifecycle decision assistant can improve future predictive models and understanding while at the same time reducing risk and uncertainty in the service management of existing products.

Claims

1. A gas turbine engine lifecycle decision assistance system for quantifying uncertainties during the lifecycle of one or more gas turbine engine components, the system comprising:

a processor; and

one or more machine-accessible storage media coupled to the processor, the one or more machine-accessible storage media comprising;

a bidirectional probabilistic analysis subsystem comprising a probabilistic model of conditional dependencies between a plurality of random variables associated with a plurality of different sources of uncertainty in the gas turbine engine component lifecycle, wherein each source of uncertainty is associated with a random variable of the plurality of random variables, and wherein the probabilistic model includes an acyclic directed graph that includes a plurality of nodes and one or more edges, wherein each node corresponds to a random variable of the plurality of variables, and wherein each edge corresponds to a conditional dependency between a pair of random variables, the probabilistic model arranged to connect at least two of the plurality of different sources of uncertainty by a common random variable of the plurality of random variables, the bidirectional probabilistic analysis subsystem to, when executed by the processor;

compute a joint probability distribution for the probabilistic model;

periodically receive new evidence from one or more of the different sources of uncertainty over the course of the component lifecycle;

in response to the new evidence, re-compute the joint probability distribution;

determine a current phase of the gas turbine engine component lifecycle, wherein the current phase comprises a pre-production certification phase or a post-production certification phase;

communicate the re-computed joint probability distribution to a component design subsystem in response to determining the current phase is the pre-production certification phase; and

communicate the re-computed joint probability distribution to a field management subsystem in response to determining the current phase is the post-production certification phase, wherein the one or more machine-accessible storage media comprises the component design subsystem and/or the field management subsystem.

2. The system of claim 1, wherein the bidirectional probabilistic analysis subsystem, when executed by the processor, computes the joint probability distribution in response to a request for a quantification of uncertainty relating to an aspect of the gas turbine engine component lifecycle.

3. The system of claim 2, wherein the bidirectional probabilistic analysis subsystem, when executed by the processor, receives the request from the component design subsystem and/or the field management subsystem.

4. The system of claim 1, wherein the bidirectional probabilistic analysis subsystem, when executed by the processor, connects the plurality of different sources of uncertainty in the gas turbine engine component lifecycle to the probabilistic model, and wherein at least one of the different sources of uncertainty relates to the pre-production certification phase of the component lifecycle and at least one of the different sources of uncertainty relates to the post-production certification phase of the component lifecycle.

5. The system of claim 1, wherein the bidirectional probabilistic analysis subsystem, when executed by the processor, connects the plurality of different sources of uncertainty in the gas turbine engine component lifecycle to the probabilistic model, and wherein at least one of the different sources of uncertainty relates to the pre-production certification phase of the component lifecycle including one or more of a design phase, a manufacture phase, and a test phase.

6. The system of claim 1, wherein the bidirectional probabilistic analysis subsystem, when executed by the processor, connects the plurality of different sources of uncertainty in the gas turbine engine component lifecycle to the probabilistic model, and wherein at least one of the different sources of uncertainty relates to the post-production certification phase of the component lifecycle including one or more of a use phase, and a service phase.

7. The system of claim 1, wherein the bidirectional probabilistic analysis subsystem, when executed by the processor, connects the plurality of different sources of uncertainty in the gas turbine engine component lifecycle to the probabilistic model, and wherein the plurality of different sources of uncertainty include at least one analytical model, at least one source of empirical data, and at least one source of expert knowledge.

8. A method for quantifying uncertainty during different phases of the lifecycle of a manufactured component, the method comprising, with at least one computing device:

identifying at least two sources of uncertainty that are associated with different phases of the manufactured component lifecycle, wherein each source of uncertainty is associated with a random variable of a plurality of random variables;

connecting the at least two sources of uncertainty by a common random variable of the plurality of random variables in a Bayesian network, wherein the Bayesian network includes an acyclic directed graph that includes a plurality of nodes and one or more edges, wherein each node corresponds to a random variable of the plurality of variables, and wherein each edge corresponds to a conditional dependency between a pair of random variables;

computing a joint probability distribution for the Bayesian network using the common random variable and at least one random variable associated with each of the at least two sources of uncertainty;

determining a current phase of the manufactured component lifecycle, wherein the current phase comprises a pre-production certification phase or a post-production certification phase;

communicating the re-computed joint probability distribution to a component design subsystem in response to determining the current phase is the pre-production certification phase; and

communicating the re-computed joint probability distribution to a field management subsystem in response to determining the current phase is the post-production certification phase.

9. The method of claim 8, comprising receiving new evidence relating to at least one of the random variables and propagating the new evidence through the Bayesian network.

10. The method of claim 9, comprising forward-propagating the new evidence through the Bayesian network if the new evidence relates to the pre-production certification phase of the gas turbine engine lifecycle.

11. The method of claim 10, comprising back-propagating the new evidence through the probabilistic model if the new evidence relates to the post-production certification phase of the gas turbine engine lifecycle.

12. The method of claim 8, comprising computing the joint probability distribution in response to a request for a quantification of uncertainty relating to an aspect of the manufactured component lifecycle.

13. The method of claim 12, comprising receiving the request from the component design subsystem and/or the field management subsystem.

14. The method of claim 8, comprising connecting a source of uncertainty relating to the pre-production certification phase of the component lifecycle to the Bayesian network.

15. The method of claim 8, comprising connecting a source of uncertainty relating to the post-production certification phase of the component lifecycle to the Bayesian network.

16. The method of claim 8, comprising connecting at least one analytical model, at least one source of empirical data, and at least one source of expert knowledge to the Bayesian network.

17. The method of claim 16, comprising propagating output of the at least one analytical model, output of the at least one source of empirical data, and output of the at least one source of expert knowledge through the Bayesian network.

18. A gas turbine engine lifecycle decision assistant for quantifying uncertainties during the lifecycle of a gas turbine engine component, the gas turbine engine lifecycle decision assistant comprising:

computer program instructions embodied in one or more machine-accessible storage media and executable by at least one processor to;

create a Bayesian network of conditional dependencies between a plurality of random variables associated with a plurality of different sources of uncertainty in the gas turbine engine component lifecycle, wherein each source of uncertainty is associated with a random variable of the plurality of random variables, and wherein the Bayesian network includes an acyclic directed graph that includes a plurality of nodes and one or more edges, wherein each node corresponds to a random variable of the plurality of variables, and wherein each edge corresponds to a conditional dependency between a pair of random variables, the Bayesian network arranged to connect at least two of the plurality of different sources of uncertainty by a common random variable of the plurality of random variables;

compute a joint probability distribution for the Bayesian network;

receive new evidence from one or more of the different sources of uncertainty over the course of the component lifecycle;

in response to the new evidence, re-compute the joint probability distribution;

determine a current phase of the gas turbine engine component lifecycle, wherein the current phase comprises a pre-production certification phase or a post-production certification phase;

communicate the re-computed joint probability distribution to a component design subsystem in response to determining the current phase is the pre-production certification phase; and

communicate the re-computed joint probability distribution to a field management subsystem in response to determining the current phase is the post-production certification phase.

19. The gas turbine engine lifecycle decision assistant of claim 18, wherein the computer program instructions are to compute the joint probability distribution in response to a request for a quantification of uncertainty relating to an aspect of the gas turbine engine component lifecycle, and wherein the computer program instructions are to receive the request from the component design subsystem and/or the field management subsystem.

20. The gas turbine engine lifecycle decision assistant of claim 18, wherein the computer program instructions are to connect the plurality of different sources of uncertainty in the gas turbine engine component lifecycle to the Bayesian network, and wherein at least one of the different sources of uncertainty relates to the pre- production certification phase of the component lifecycle and at least one of the different sources of uncertainty relates to the post-production certification phase of the component lifecycle, and wherein the plurality of different sources of uncertainty include at least one analytical model, at least one source of empirical data, and at least one source of expert knowledge.