Detecting fault states of an aircraft

A method of detecting a fault state of an aircraft, comprising:

• accessing a training set of flight data for the aircraft, the training set including observations of the flight data, each observation of the flight data includes measurements of properties from sensors or avionic systems selected and transformed into a set of features;
• building a generative adversarial network (GAN) including a generative model and a discriminative model using the training set and the set of features, the generative model and the discriminative model being trained to respectively generate simulated flight data corresponding to the flight data, and evaluate the simulated flight data, until the simulated flight data has a corresponding distribution that matches a distribution of the flight data, the discriminative model further identifying a subset of features from the set of features that capture statistical properties of the distribution of the flight data;
• building an anomaly detection model to predict the fault state of the aircraft, the anomaly detection model being trained using the training set of flight data, the simulated flight data, and the subset of features;
• and
• deploying the anomaly detection model to predict the fault state of the aircraft using additional observations of the flight data.

A computer-readable storage medium for detecting a fault state of an aircraft, the computer-readable storage medium being non-transitory and having computer-readable program code stored therein that in response to execution by a processor, causes an apparatus to at least:

• access a training set of flight data for the aircraft, the training set including observations of the flight data, each observation of the flight data includes measurements of properties from sensors or avionic systems selected and transformed into a set of features;
• build a generative adversarial network (GAN) including a generative model and a discriminative model using the training set and the set of features, the generative model and the discriminative model being trained to respectively generate simulated flight data corresponding to the flight data, and evaluate the simulated flight data, until the simulated flight data has a corresponding distribution that matches a distribution of the flight data, the discriminative model being further configured to identify a subset of features from the set of features that capture statistical properties of the distribution of the flight data;
• build an anomaly detection model to predict the fault state of the aircraft, the anomaly detection model being trained using the training set of flight data, the simulated flight data, and the subset of features;
• and
• deploy the anomaly detection model to predict the fault state of the aircraft using additional observations of the flight data.