PREDICTIVE HELICOPTER FLIGHT DATA MONITORING FOR FLIGHT SAFETY DESIGN
MetadataShow full item record
A method to augment existing Helicopter Flight Data Monitoring (HFDM) systems using physics-based models was proposed. It was suggested that physics-based models can be utilized to derive condition indicators (“events” or “exceedances”) with improved detection performance within the framework of typical HFDM systems. Data were collected from computer simulations and real-world helicopter training flights. Model-based analyses were performed using static model evaluations using flight data and dynamic simulations for predictive examination of potential hazards. In the case of dynamic simulations, neural networks were used to combine the results from simulated flight trajectories into a parametric monitoring metric. Results indicate that condition indicators defined using model-derived quantities such as performance metrics and dynamic responses generate a reduction in false alarms and missed detections relative to the existing HFDM events. Furthermore, pre-emptive simulation of potentially hazardous conditions was shown to yield condition indicators that are available sooner than typical HFDM events, allowing for timely detection of adverse flight states. The approach is general in that extension to other vehicles and flight states requires changes to model parameters and additional evaluations, which reduces the reliance on past experience when defining condition indicators for new operators. The results suggest that the application of the model-based approach can lead to improved accuracy and availability of HFDM events, with a corresponding potential for safety improvement in operations. Results for several flight conditions, including implementation considerations and directions for future investigation are discussed.