Method, system, and program storage device for automating prognostics for physical assets

1. A method of calculating end-of-life (EOL) predictions for a physical asset, the method comprising:

accessing a state-space model for the physical asset, the state-space model being a physics-based model describing a state of the physical asset at a particular time given measurements or observations for the physical asset;

inferring a current state of the physical asset;

deriving a long-term prediction for the physical asset based on the inferred current state of the physical asset and the state-space model for the physical asset; and

generating an EOL probability distribution function for the physical asset based on the long-term prediction, the EOL probability distribution function describing a range of estimates of EOL for the physical asset and their corresponding confidence intervals.

2. The method of claim 1 , wherein the obtaining comprises deriving the state-space model using training data.

3. The method of claim 2 , wherein the training data is incomplete and the obtaining further comprises using Gaussian Process Regression to derive the state-space model from the incomplete training data.

4. The method of claim 2 , wherein the training data is incomplete and the obtaining further comprises using a Radial basis functions (RBF) model to derive the state-space model from the incomplete training data.

5. The method of claim 2 , wherein the training data is complete and the obtaining further comprises using symbolic regression to derive the state-space model from the complete training data.

6. The method of claim 2 , wherein the training data is complete and the obtaining further comprises using a Lagrange polynomial to derive the state-space model from the complete training data.

7. The method of claim 2 , wherein the training data is complete and the obtaining further comprises using Gaussian Process Regression to derive the state-space model from the complete training data.

8. The method of claim 2 , wherein the training data is complete and the obtaining further comprises using a Radial basis functions (RBF) model to derive the state-space model from the complete training data.

9. The method of claim 1 , wherein the inferring comprises using calculating a posterior probability distribution function (PDF).

10. The method of claim 1 , wherein the inferring comprises using a Kalman filter.

11. The method of claim 1 , wherein the inferring comprises using a particle filter.

12. A system comprising:

a server machine communicatively coupled via a computer network to a device, the server machine comprising one or more processors and configured to:

access a state-space model for a physical asset, the state-space model being a physics-based model describing a state of the physical asset at a particular time given measurements or observations for the physical asset;

infer a current state of the physical asset;

derive a long-term prediction for the physical asset based on the inferred current state of the physical asset and the state-space model for the physical asset; and

generate an end-of-life (EOL) probability distribution function for the physical asset based on the long-term prediction, the EOL probability distribution function describing a range of estimates of EOL for the physical asset and their corresponding confidence intervals.

13. The system of claim 12 , wherein the device is an industrial asset.

14. The system of claim 13 , wherein the industrial asset is a gas turbine.

15. A non-transitory computer-readable medium embodying instructions that, when executed by a processor, perform operations comprising:

accessing a state-space model for a physical asset, the state-space model being a physics-based model describing a state of the physical asset at a particular time given measurements or observations for the physical asset;

inferring a current state of the physical asset;

deriving a long-term prediction for the physical asset based on the inferred current state of the physical asset and the state-space model for the physical asset; and

generating an end-of-life (EOL) probability distribution function for the physical asset based on the long-term prediction, the EOL probability distribution function describing a range of estimates of EOL for the physical asset and their corresponding confidence intervals.

16. The non-transitory computer-readable medium of claim 15 , wherein the obtaining the state-space model comprises deriving a state-space model using training data.

17. The non-transitory computer-readable medium of claim 16 , wherein the training data is incomplete and the obtaining further comprises using Gaussian Process Regression to derive the state-space model from the incomplete training data.

18. The non-transitory computer-readable medium of claim 16 , wherein the training data is incomplete and the obtaining further comprises using a Radial basis functions (RBF) model to derive the state-space model from the incomplete training data.

19. The non-transitory computer-readable medium of claim 16 , wherein the training data is complete and the obtaining further comprises using symbolic regression to derive the state-space model from the complete training data.

20. The non-transitory computer-readable medium of claim 16 , wherein the training data is complete and the obtaining further comprises using a Lagrange polynomial to derive the state-space model from the complete training data.