A Virtual Pilot Algorithm for Synthetic HUMS Data Generation

Regime recognition is an important tool for creation of usage spectra, component lifetime prediction, and fatigue loads analysis. During development of regime recognition (RR) codes, it is often necessary to establish baseline performance and accuracy of RR algorithms using scripted flight test data. However, various problems arise when using scripted flight test data as "truth" measurements upon which RR code accuracy is based. During scripted flight tests, maneuvers may be imprecisely flown or data may be perturbed by environmental disturbances. An alternative to scripted flight test data lies in simulated, or "synthetic" flight data produced from a flight simulation model. In this case, a maneuver script is flown in simulation and the model outputs are used to build a simulated Health and Usage Monitoring System (HUMS) dataset. An algorithm is needed to generate control inputs in a similar manner to a human pilot so that a maneuver sequence can be flown correctly in simulation. This paper presents a so-called virtual pilot algorithm that generates simulated control inputs for a given maneuver sequence. The virtual pilot receives a maneuver script as input, and streams control inputs to the flight simulator to perform the required maneuvers in an accurate and reliable manner. The virtual pilot proposed here is formulated as a variable structure controller where each maneuver is mapped to a single feedback control law. By switching between feedback laws, different maneuver regimes are flown. A smoothing technique called setpoint tracking bumpless transfer is used to smooth control transitions. Example results show that the virtual pilot can generate realistic synthetic HUMS data for a variety of maneuver sequences. Virtual pilot fidelity is demonstrated through both conformance to the ADS-33 standards for selected Mission Task Elements and comparison to actual HUMS data.