Fault prediction in valve systems through Bayesian framework

1. A method, comprising:

monitoring a plurality of parameters of a subsystem of a system related to opening and closing of a valve of the subsystem;

collecting a first set of data values for the plurality of parameters during a first operational phase of the system;

generating a first plurality of feature metrics based on first set of data values for the plurality of parameters;

detecting a fault in the subsystem in a second operational phase of the system, wherein the first operational phase and the second operational phase are a same phase of operation of the system collected at different legs of operation;

collecting a second set of data values for the plurality of parameters during the second operational phase of the subsystem;

generating a second plurality of feature metrics based on second set of data values for the plurality of parameters; and

defining a prefault state for the feature metrics according to a Bayesian framework based on the first plurality of feature metrics and the second plurality of feature metrics.

2. The method of claim 1 , wherein the plurality of parameters are selected from a group including:

temperature;

pressure;

state data for a component of the subsystem;

humidity; and

flowrate.

3. The method of claim 1 , wherein the plurality of feature metrics are selected from a group including:

a maximum value feature metric for each parameter of the plurality of parameters;

a minimum value feature metric for each parameter of the plurality of parameters; and

a mean value feature metric for each parameter of the plurality of parameters.

4. The method of claim 1 , wherein the plurality of feature metrics are selected from a group including:

an open/close imbalance feature metric derived from state data indicating a portion of time for which the valve is open relative to time for which the valve is closed;

a Maximum Information Coefficient feature metric for an entirety of a given operational phase for each parameter of the plurality of parameters; and

a Maximum Information Coefficient feature metric for a time window in the given operational phase after the valve is opened and closed for each parameter of the plurality of parameters.

5. The method of claim 1 , wherein monitoring the plurality of parameters includes normalizing operational data of the plurality of parameters.

6. A method, comprising:

collecting data from a plurality of sensors of a subsystem of a system related to opening and closing of a valve of the subsystem during a given operational phase of the system to identify a plurality of parameters;

generating a plurality of feature metrics based on the plurality of parameters, wherein the plurality of feature metric comprise at least one of:

a maximum value feature metric for each parameter of the plurality of parameters during the given operational phase;

a minimum value feature metric for each parameter of the plurality of parameters during the given operational phase;

a mean value feature metric for each parameter of the plurality of parameters during the given operational phase

an open/close imbalance feature metric derived from state data indicating a portion of time for which the valve is open relative to time for which the valve is closed during the given operational phase;

a Maximum Information Coefficient feature metric for an entirety of a given operational phase for each parameter of the plurality of parameters during the given operational phase; and

a Maximum Information Coefficient feature metric for a time window in the given operational phase after the valve is opened and closed for each parameter of the plurality of parameters during the given operational phase;

determining whether the feature metrics indicate a prefault state of the subsystem based on a Bayesian framework, wherein the Bayesian framework is based on feature metrics from at least two operational phases of the system; and

in response to determining that the feature metrics indicate the prefault state of the subsystem, generating a notification.

7. The method of claim 6 , wherein the plurality of parameters include at least one of:

temperature;

pressure;

state data for a component of the subsystem;

humidity; and

flowrate.

8. The method of claim 6 , wherein monitoring the plurality of parameters includes normalizing operational data of the plurality of parameters.

9. The method of claim 6 , wherein determining whether the feature metrics indicate the prefault state of the subsystem based on the Bayesian framework includes:

determining an aggregate probability of fault based on a deviation from a historic mean value for each feature metric of the feature metrics.

10. The method of claim 9 , wherein generating the notification includes generating a visualization of a data series plot for each feature metric indicating a current value of a given feature metric of the feature metrics relative to a confidence threshold that is based on a standard deviation from the historic mean value for the given feature metric.

11. The method of claim 6 , wherein generating the notification includes transmitting the notification to a maintenance scheduler, and wherein the notification includes an average time from detection of the prefault state to a fault in the subsystem.

12. A method, comprising:

generating a Bayesian framework for identifying prefault states in a subsystem of a system, comprising:

collecting a plurality of historical parameters related to opening and closing of a valve of the subsystem across a plurality of operational legs of the system;

generating a plurality of historical feature metrics based on the plurality of historical parameters and for the operational legs of the plurality of operational legs;

in response to detecting a fault in the subsystem, defining a prefault state corresponding to the historical feature metrics;

monitoring a plurality of operational parameters of the subsystem related to opening and closing of the valve during a given operational phase of an operational leg;

generating a plurality of operational feature metrics based on the plurality of operational parameters monitored during the given operational phase; and

in response to determining, using the generated Bayesian framework, that the operational feature metrics indicate the prefault state of the subsystem, generating a notification, wherein the notification includes an average time from detection of the prefault state to a fault in the subsystem.

13. The method of claim 12 , wherein the plurality of historical parameters and the plurality of operational parameters include:

temperature;

pressure;

state data for a component of the subsystem;

humidity; and

flowrate.

14. The method of claim 12 , wherein the plurality of historical feature metrics include:

a maximum value feature metric for each parameter of the plurality of historical parameters during a given operational leg of the plurality of operational legs;

a minimum value feature metric for each parameter of the plurality of historical parameters during the given operational leg of the plurality of operational legs;

a mean value feature metric for each parameter of the plurality of historical parameters during the given operational leg of the plurality of operational legs; and

an open/close imbalance feature metric derived from state data indicating a portion of time for which the valve is open relative to time for which the valve is closed during the given operational leg of the plurality of operational legs;

a Maximum Information Coefficient feature metric for an entirety of the given operational phase for each parameter of the plurality of parameters during the given operational leg of the plurality of operational legs; and

a Maximum Information Coefficient feature metric for a time window in the given operational leg after the valve is opened and closed for each parameter of the plurality of parameters.

15. The method of claim 12 , wherein the plurality of operational feature metrics include:

a maximum value feature metric for each parameter of the plurality of operational parameters during the given operational phase of the operational leg;

a minimum value feature metric for each parameter of the plurality of operational parameters during the given operational phase of the operational leg;

a mean value feature metric for each parameter of the plurality of operational parameters during the given operational phase of the operational leg;

an open/close imbalance feature metric derived from state data indicating a portion of time for which the valve is open relative to time for which the valve is closed;

a Maximum Information Coefficient feature metric for an entirety of the given operational phase for each parameter of the plurality of operational parameters; and

a Maximum Information Coefficient feature metric for a time window in the given operational phase after the valve is opened and closed for each parameter of the plurality of operational parameters.

16. The method of claim 12 , wherein monitoring the plurality of operational parameters includes normalizing operational data of the plurality of operational parameters.

17. The method of claim 12 , wherein generating the notification includes transmitting the notification to a maintenance scheduler.

18. A method, comprising:

collecting data from a plurality of sensors of a subsystem of a system related to opening and closing of a valve of the subsystem during a given operational phase of the system to identify a plurality of parameters;

generating a plurality of feature metrics based on the plurality of parameters;

determining whether the feature metrics indicate a prefault state of the subsystem based on a Bayesian framework by:

determining an aggregate probability of fault based on a deviation from a historic mean value for each feature metric of the feature metrics, wherein the Bayesian framework is based on feature metrics from at least two operational phases of the system; and

in response to determining that the feature metrics indicate the prefault state of the subsystem, generating a notification.