Machine event duration analytics and aggregation for machine health measurement and visualization
The unexpected failure of a turbomachine can be costly and dangerous. Processes may collect data from a turbomachine, calculate durations of machine events from the collected data, and apply a model to those machine-event durations to predict future machine-event durations and/or detect trends in the machine-event durations. This predictive output may be utilized to inform downstream functions regarding the health of the turbomachine. For example, a downstream function may utilize the predictive output to detect degradation or a potential future failure in the turbomachine and trigger remedial functions, such as alerts and/or controls, to prevent or mitigate the degradation or failure of the turbomachine.
1 . A method of controlling an engine based on predicting an abnormal state in the engine, the method comprising using at least one hardware processor to:
receive operating data from an electronic control unit of an engine, wherein the operating data includes values for a plurality of parameters of the engine, and wherein the values for the plurality of parameters include values for one or more machine events;
calculate a duration of each of the one or more machine events from the values for the one or more machine events;
generate a dataset that correlates the calculated durations with values for one or more of the plurality of parameters;
receive, from a trained machine-learning model, a predictive output that includes a predicted future duration of each of the one or more machine events;
determine whether or not the predictive output is indicative of a future abnormal operating state of the engine; and,
when the predictive output is indicative of a future abnormal operating state of the engine, execute at least one remedial function to control the engine, the at least one remedial function including transmitting a control command to the electronic control unit of the engine to initiate a transition of the engine from a first state to a second state that is different than the first state, so as to prevent or mitigate the future abnormal operating state.
2 . The method of claim 1 , wherein each value for each of the one or more machine events is a binary value that indicates either a presence of the machine event or an absence of the machine event.
3 . The method of claim 1 , wherein the dataset comprises, for each calculated duration of a machine event, a feature vector comprising a value of that calculated duration and values for the one or more parameters during the calculated duration of that machine event.
4 . The method of claim 1 , wherein the one or more parameters comprise operating parameters of the engine.
5 . The method of claim 1 , wherein the one or more parameters represent an event history of the engine.
6 . The method of claim 1 , wherein the one or more parameters represent a repair history of the engine.
7 . The method of claim 1 , wherein the machine-learning model comprises a regression.
8 . The method of claim 1 , wherein the predictive output comprises a confidence value for each predicted future duration.
9 . The method of claim 1 , wherein the at least one remedial function further includes sending an alert to at least one recipient.
10 . The method of claim 1 , wherein the one or more machine events comprise a machine event that is required for startup or shutdown of the engine.
11 . The method of claim 1 , wherein the one or more machine events comprise a process control mode for operation of the engine.
12 . The method of claim 1 , wherein the one or more machine events comprise an emissions mode of the engine.
13 . The method of claim 1 , wherein the one or more machine events comprise a state of a valve in the engine.
14 . The method of claim 1 , wherein the one or more machine events comprise an alert state of the engine.
15 . The method of claim 1 , wherein determining whether or not the predictive output is indicative of a future abnormal operating state of the engine comprises determining whether or not the predicted future duration of at least one of the one or more machine events satisfies one or more criteria.
16 . The method of claim 1 , further comprising using the at least one hardware processor to apply a statistical model to at least the calculated durations to detect a trend, wherein the predictive output includes an indication of the trend.
17 . The method of claim 1 , wherein the engine is a gas turbine engine or gas compressor.
18 . A system comprising:
at least one hardware processor; and
a memory storing software configured to, when executed by the at least one hardware processor,
receive operating data from an electronic control unit of an engine, wherein the operating data includes values for a plurality of parameters of the engine, and wherein the values for the plurality of parameters include values for one or more machine events,
calculate a duration of each of the one or more machine events from the values for the one or more machine events,
generate a dataset that correlates the calculated durations with values for one or more of the plurality of parameters,
receive, from a trained machine-learning model, a predictive output that includes a predicted future duration of each of the one or more machine events,
determine whether or not the predictive output is indicative of a future abnormal operating state of the engine, and,
when the predictive output is indicative of a future abnormal operating state of the engine, execute at least one remedial function to control the engine, the at least one remedial function including transmitting a control command to the electronic control unit of the engine to initiate a transition of the engine from a first state to a second state that is different than the first state, so as to prevent or mitigate the future abnormal operating state.
19 . A non-transitory computer-readable medium having instructions stored thereon, wherein the instructions, when executed by a processor, cause the processor to:
receive operating data from an electronic control unit of an engine, wherein the operating data includes values for a plurality of parameters of the engine, and wherein the values for the plurality of parameters include values for one or more machine events;
calculate a duration of each of the one or more machine events from the values for the one or more machine events;
generate a dataset that correlates the calculated durations with values for one or more of the plurality of parameters;
receive, from a trained machine-learning model, a predictive output that includes a predicted future duration of each of the one or more machine events;
determine whether or not the predictive output is indicative of a future abnormal operating state of the engine; and,
when the predictive output is indicative of a future abnormal operating state of the engine, execute at least one remedial function to control the engine, the at least one remedial function including transmitting a control command to the electronic control unit of the engine to initiate a transition of the engine from a first state to a second state that is different than the first state, so as to prevent or mitigate the future abnormal operating state.
20 . A non-transitory computer-readable medium of claim 19 , wherein the instructions further cause the at least one processor to receive, from a statistical model, a trend, wherein the predictive output includes an indication of the trend.