Using an irrelevance filter to facilitate efficient RUL analyses for electronic devices

1. A method for estimating a remaining useful life, RUL, of an electronic device, wherein during a surveillance mode, the method comprises:

receiving a set of time-series signals gathered from sensors in the electronic device while the electronic device is operating;

detecting statistical changes in the set of time-series signals that are deemed as anomalous signal patterns;

generating a set of anomaly alarms, wherein an anomaly alarm is generated for each of the anomalous signal patterns;

applying an irrelevance filter to the set of anomaly alarms to produce filtered anomaly alarms, wherein the irrelevance filter removes anomaly alarms associated with anomalous signal patterns that are not correlated with previous failures of similar electronic devices;

using a logistic-regression model to compute an RUL-based risk index for the electronic device based on the filtered anomaly alarms; and

when the risk index exceeds a risk-index threshold, generating a notification indicating that the electronic device has a limited remaining useful life.

2. The method of claim 1 , wherein the irrelevant filter removes an anomaly alarm associated with an anomalous signal pattern when the anomalous signal pattern matches a similar signal pattern that was previously observed from other similar electrical devices that have operated without incident.

3. The method of claim 1 , wherein detecting the statistical changes in the set of time-series signals includes:

performing a sequential probability ratio test, SPRT, on the set of time-series signals or on residual signals produced from the set of time-series signals, wherein the SPRT produces SPRT alarms for the anomalous signal patterns; and

wherein the SPRT alarms are the anomaly alarms.

4. The method of claim 1 , wherein detecting the statistical changes in the set of time-series signals is based at least in part on detecting the statistical changes in residual signals produced from the set of time series signals;

wherein the method further comprises, prior to the detecting:

using an inferential model to generate estimated values for the set of time-series signals; and

performing a pairwise differencing operation between actual values of the set of time-series signal and the estimated values for the set of time-series signals to produce the residual signals.

5. The method of claim 4 , wherein the inferential model comprises a Multivariate State Estimation Technique, MSET, model.

6. The method of claim 1 , wherein during an RUL-training mode, which precedes the surveillance mode, the method comprises:

receiving an RUL training set comprising time-series signals gathered from sensors in similar electronic devices while the similar electronic devices are run to failure;

receiving associated failure times for the similar electronic devices;

using an inferential model to generate estimated values for the RUL training set of time-series signals;

performing a pairwise differencing operation between actual values and the estimated values for the RUL training set of time-series signals to produce residuals;

performing a sequential probability ratio test, SPRT, on the residuals to produce SPRT alarms with associated tripping frequencies; and

training a logistic-regression model to predict an RUL for the electronic device based on correlations between the SPRT alarm tripping frequencies and the failure times for the similar electronic devices.

7. The method of claim 6 , wherein during the RUL-training mode, the method additionally configures the irrelevance filter by:

identifying relevant SPRT alarms that were generated during a time interval near the associated failure times of a similar electronic device; and

configuring the irrelevance filter to remove SPRT alarms that are not relevant.

8. The method of claim 7 , wherein while training the logistic-regression model to predict the RUL for the electronic device, the method considers SPRT alarm tripping frequencies associated with relevant SPRT alarms.

9. The method of claim 1 , wherein the time-series signals gathered from sensors in the electronic device include signals specifying one or more combinations of the following:

temperatures;

currents;

voltages;

resistances;

capacitances;

vibrations;

dissolved gas metrics;

cooling system parameters; and

control signals.

10. The method of claim 1 , wherein the electronic device is a utility system asset, a vehicle component, or a computing system device.

11. A non-transitory computer-readable storage medium storing instructions that when executed by a computer cause the computer to estimate a remaining useful life, RUL, of an electronic device, wherein the computer is caused to:

receive a set of time-series signals gathered from sensors in the electronic device while the electronic device is operating;

detect statistical changes in the set of time-series signals that are deemed as anomalous signal patterns;

generate a set of anomaly alarms, wherein an anomaly alarm is generated for each of the anomalous signal patterns;

apply an irrelevance filter to the set of anomaly alarms to produce filtered anomaly alarms, wherein the irrelevance filter removes anomaly alarms associated with anomalous signal patterns that are not correlated with previous failures of similar electronic devices;

use a logistic-regression model to compute an RUL-based risk index for the electronic device based on the filtered anomaly alarms; and

when the risk index exceeds a risk-index threshold, generate a notification indicating that the electronic device has a limited remaining useful life.

12. The non-transitory computer-readable storage medium of claim 11 ,

wherein the irrelevant filter is configured to remove an anomaly alarm associated with an anomalous signal pattern when the anomalous signal pattern matches a similar signal pattern that was previously observed from other similar electrical devices that have operated without failing.

13. The non-transitory computer-readable storage medium of claim 11 , wherein the instructions to detect the statistical changes in the set of time-series signals further include instructions that when executed cause the computer to:

perform a sequential probability ratio test, SPRT, on the set of time-series signals or on residual signals produced from the set of time-series signals, wherein the SPRT produces SPRT alarms for the anomalous signal patterns; and

wherein the SPRT alarms are the anomaly alarms.

14. The non-transitory computer-readable storage medium of claim 11 ,

wherein the instructions to detect the statistical changes in the set of time-series signals further include instructions that when executed cause the computer to:

detect the statistical changes in residual signals produced from the set of time series signals;

wherein the residual signals are produced by:

using an inferential model to generate estimated values for the set of time-series signals; and

performing a pairwise differencing operation between actual values of the set of time-series signal and the estimated values for the set of time-series signals to produce the residual signals.

15. The non-transitory computer-readable storage medium of claim 11 , further comprising instructions for causing the computer to perform an RUL-training mode comprising:

receiving an RUL training set comprising time-series signals gathered from sensors in similar electronic devices while the similar electronic devices are run to failure;

receiving associated failure times for the similar electronic devices;

using the inferential model to generate estimated values for the RUL training set of time-series signals;

performing a pairwise differencing operation between actual values and the estimated values for the RUL training set of time-series signals to produce residuals;

performing a sequential probability ratio test, SPRT, on the residuals to produce SPRT alarms with associated tripping frequencies; and

training a logistic-regression model to predict an RUL for the electronic device based on correlations between the SPRT alarm tripping frequencies and the failure times for the similar electronic devices.

16. The non-transitory computer-readable storage medium of claim 15 , wherein the irrelevance filter is configured to:

identify relevant anomaly alarms from the set of anomaly alarms that are generated during a time interval near a failure time of a similar electronic device; and

remove anomaly alarms from the set of anomaly alarms that are not relevant.

17. The non-transitory computer-readable storage medium of claim 11 , wherein while training the logistic-regression model to predict the RUL for the electronic device, the computing device is configured to consider SPRT alarm tripping frequencies associated with relevant SPRT alarms.

18. A system that estimates a remaining useful life, RUL, of an electronic device, the system comprising:

at least one processor and at least one associated memory; and

a notification mechanism configured to execute on the at least one processor, wherein during a surveillance mode, the notification mechanism is configured to iteratively:

receive a set of time-series signals gathered from sensors in the electronic device while the electronic device is operating;

detect statistical changes in the set of time-series signals that are deemed as anomalous signal patterns;

generate a set of anomaly alarms, wherein an anomaly alarm is generated for each of the anomalous signal patterns;

apply an irrelevance filter to the set of anomaly alarms to produce filtered anomaly alarms, wherein the irrelevance filter removes anomaly alarms associated with anomalous signal patterns that are not correlated with previous failures of similar electronic devices;

use a logistic-regression model to compute an RUL-based risk index for the electronic device based on the filtered anomaly alarms; and

when the risk index exceeds a risk-index threshold, generate a notification indicating that the electronic device is near a failure.

19. The system of claim 18 , wherein during an RUL-training mode, which precedes the surveillance mode, the notification mechanism is further configured to:

receive an RUL training set comprising time-series signals gathered from sensors in similar electronic devices while the similar electronic devices are run to failure;

receive associated failure times for the similar electronic devices;

use an inferential model to generate estimated values for the RUL training set of time-series signals;

perform a pairwise differencing operation between actual values and the estimated values for the RUL training set of time-series signals to produce residuals;

perform a sequential probability ratio test, SPRT, on the residuals to produce SPRT alarms with associated tripping frequencies; and

train a logistic-regression model to predict an RUL for the electronic device based on correlations between the SPRT alarm tripping frequencies and the failure times for the similar electronic devices.

20. The system of claim 19 , wherein the notification mechanism is configured to detect the statistical changes in the set of time-series signals by:

performing a sequential probability ratio test, SPRT, on the set of time-series signals or on residual signals produced from the set of time-series signals, wherein the SPRT produces SPRT alarms for the anomalous signal patterns; and

wherein the SPRT alarms are the anomaly alarms.