You're invited to attend
(Advisor: Prof. Dimitri Mavris)
"Retrospective and Exploratory Analyses for Enhancing the Safety of Rotorcraft Operations"
Wednesday, November 17
Collaborative Visualization Environment (CoVE)
Weber Space Science and Technology Building (SST II)
Helicopters are versatile aerial vehicles and are ideal for a variety of operations. However, recent studies have shown that the number and rate of accidents associated with helicopters is increasing. In its the 2017 most wanted list, the National Transportation Safety Board (NTSB) recommended expanding the use of flight data recorders in helicopters. This means that the amount of helicopter flight data for routine operations is expected to grow rapidly in the coming years, and it is important to properly utilize the information for knowledge discovery or anomaly detection. Data mining techniques are widely used in the commercial fixed-wing aviation domain to retrospectively improve operational safety. On the other hand, studies on anomaly detection for flight data in the rotorcraft domain are not as prevalent, potentially due to the previous lack of installed flight data recorders and vague definitions of phases of flight. In this research, one of the objectives is to develop a framework for improving flight safety specific to rotorcraft operation through retrospectively discovering potential anomalies in flight data records. To pave the way for the task of anomaly detection, we first focus on phases of flight identification and several methods are proposed to find the homogeneous flight segments. To compare these techniques, a few flight samples with pre-labeled flight phases were used and results show that a regression-based method along with a filtering approach is capable of identifying flight phases in different altitude regions.
Additionally, exceedance analyses are typically used in flight data monitoring (FDM) for anomaly detection. However, they usually rely on pre-defined thresholds which might vary depending on the type of operations or the vehicles considered. To detect anomalies without defining thresholds in advance, a sequential approach was proposed and it contains three modules for detecting different levels of anomalies. To ensure the effectiveness of the methods selected, synthetic and simulated data are used to test the proposed approach before applying candidate methods to the real flight segments. A specific group of initial climb and the approach segments is used to demonstrate the validity of the methods chosen in this study. Our tests show that functional principal component analysis (FPCA) and a convolutional variational autoencoder (CVAE) along with DBSCAN are capable of identifying shape anomalies in flight parameters. Although the detected anomalies might not directly be associated with hazardous events, it may be useful to assist helicopter operators in discovering patterns not conforming to their standard operating procedures or that do not follow normal operations.
Another objective relevant to the exploratory analysis is to develop an efficient methodology to explore the safety envelope of some flight maneuvers, and to acquire recovery trajectories for a corresponding hazardous event. The autorotation maneuver is selected as a use case due to its time criticality and low occurrences during routine helicopter operations. Surrogate modeling is used to facilitate the process and with this implementation, the response can be predicted without going through an optimization. Among all the types of surrogates tested in this study, the Gaussian process regression with MaxPro design is an adequate method because it can tackle functions in both the low and the high-dimensional spaces. To predict the required controls for an unobserved location in the operational space, a surrogate which contains two Gaussian processes is proposed for handling functional responses in unequal lengths. Finally, a sensitivity analysis is conducted to identify the key parameters that affect the shape of the safety envelope.
- Prof. Dimitri Mavris – School of Aerospace Engineering (advisor)
- Prof. Jonnalagadda V R Prasad – School of Aerospace Engineering
- Prof. Daniel P Schrage – School of Aerospace Engineering
- Dr. Alexia P Payan – School of Aerospace Engineering
- Mr. Charles Cliff Johnson – Federal Aviation Administration