Method and apparatus for machine learning based inlet debris monitoring

A method of inlet debris monitoring for a gas turbine engine of an aircraft, comprising:

• obtaining a data set of electrostatic charge data from an output signal of at least one electrostatic sensor of a gas turbine engine of an aircraft, the output signal indicative of an electrostatic charge in a nacelle of a particular gas turbine engine, the data set having initial dimensions corresponding to a window size and a sampling frequency of the output signal;
• utilizing a dimensional reduction technique to obtain a first set of basis vectors that represent the data set in a reduced dimensional space that is reduced with respect to the initial dimensions;
• utilizing the first set of basis vectors or a second set of reference basis vectors, which are based on historical electrostatic charge data for one or more reference gas turbine engines, to project the data set onto the reduced dimensional space and obtain a reduced dimensional representation of the data set;
• utilizing machine learning to determine whether the reduced dimensional representation of the data set indicates foreign object debris in the particular gas turbine engine;
• and
• based on the determination indicating detection of foreign object debris, providing a foreign object debris notification for the gas turbine engine of the aircraft;
• comparing model output of the reduced dimensional representation to that evaluated with the reference reduced dimensional representation;
• and
• performing the determination of whether foreign object debris has entered the inlet of the gas turbine engine based on the comparing step.

A method of inlet debris monitoring for a gas turbine engine of an aircraft, comprising:

• obtaining a data set of electrostatic charge data from an output signal of at least one electrostatic sensor of a gas turbine engine of an aircraft, the output signal indicative of an electrostatic charge in a nacelle of a particular gas turbine engine, the data set having initial dimensions corresponding to a window size and a sampling frequency of the output signal;
• utilizing a dimensional reduction technique to obtain a first set of basis vectors that represent the data set in a reduced dimensional space that is reduced with respect to the initial dimensions;
• utilizing the first set of basis vectors or a second set of reference basis vectors, which are based on historical electrostatic charge data for one or more reference gas turbine engines, to project the data set onto the reduced dimensional space and obtain a reduced dimensional representation of the data set;
• utilizing machine learning to determine whether the reduced dimensional representation of the data set indicates foreign object debris in the particular gas turbine engine;
• and
• based on the determination indicating detection of foreign object debris, providing a foreign object debris notification for the gas turbine engine of the aircraft;
• wherein the machine learning comprises non-neural network machine learning.