Real Flight Test Data Research Articles

Recurrent neural networks for aerodynamic parameter estimation with Lyapunov stability analysis

ABSTRACT Aircraft parameter estimation is crucial for designing accurate and reliable flight dynamic models, essential for integration into flight simulators and developing effective control laws. These parameters, encompassing aerodynamic characteristics, play a pivotal role in enhancing aircraft performance, safety and overall operational efficiency. This paper explores aircraft parameter estimation using two Recurrent Neural Network (RNN) approaches: (i) the weight-bias method (RNN-WB) and (ii) the gradient-based method (RNN-Gradient). RNN-WB is based on the Hopfield neural network and relies on pre-computation of weight and bias information, applying non-linearity to the states of the network. RNN-Gradient computes direct gradients and applies non-linearity to the error. Simulated flight data is employed in MATLAB implementations to assess estimation results, considering both performance metrics and computation time. The estimated parameters were also compared with estimates obtained from Unscented Kalman Filter (UKF). A novel methodology is introduced to establish the asymptotic stability of RNN-WB and RNN-Gradient methods, utilizing the normalized Lyapunov energy functional. The asymptotic stability ensures the reliability and robustness of these techniques for parameter estimation. The RNN-Gradient shows a 20% performance improvement over RNN-WB. Satisfactory simulation results from both algorithms provide a promising foundation for their potential applications to real flight test data.

Balancing predictability and flexibility through operation volume-constrained visual flight rule operations in low altitude airspaces

The advancement in uncrewed aircraft systems such as small drones and advanced air mobility vehicles such as Electric Vertical Take Off and Landing aircraft (eVTOL) has called for airspace integration at low altitudes of both traditional aircraft, such as helicopters and new entrants, such as drones and eVTOL. Currently, the trajectories and necessary buffers around them of flights operating under visual flight rules are not possible to predict. This research proposes the use of flight mission characteristics to model the trajectory and evaluates temporal, lateral and vertical deviations to define the safety buffers which can be used to generate operation volumes, geo-fencing such low-altitude flights and separating them from other traffic in a safe and efficient manner. Real flight test data obtained for the purposes of this study and pilots’ interviews assure high fidelity and practicability of the proposal.

Aircraft Parameter Estimation using Feedforward Neural Networks With Lyapunov Stability Analysis

Aerodynamic parameter estimation is critical in the aviation sector, especially in design and development programs of defense-military aircraft. In this paper, new results of the application of Artificial Neural Networks (ANN) to the field of aircraft parameter estimation are presented. The performances of Feedforward Neural Network (FFNN) with Backpropagation and FFNN with Backpropagation using Recursive Least Square (RLS) are investigated for aerodynamic parameter estimation. The methods are validated on flight data simulated using MATLAB implementations. The normalized Lyapunov energy functional has been used to derive the convergence conditions for both the ANN-based estimation algorithms. The estimation results are compared on the basis of performance metrics and computation time. The performance of FFNN-RLS has been observed to be approximately 10% better than FFNN-BPN. Simulation results from both algorithms have been found to be highly satisfactory and pave the way for further applications to real flight test data.

Prediction of aircraft engine parameters based on ARMA model

Abstract The high-pressure rotor speeds, low-pressure rotor speeds, and exhaust temperature of the aircraft engine are the key parameters reflecting the performance of aircraft engines. To realize the trend monitoring during the flight test and the processing of data outliers in flight data recorder, the time series analysis and modeling method is used to establish a suitable ARMA model through data processing, series property analysis, model identification, order determination, modeling, model diagnosis and other steps. Fit the real flight test data of an engine. The results show that the prediction interval within 3 steps of the ARMA model has high accuracy, and has good engineering practicability in real-time flight monitoring and data processing.

Optimized Energy Management Strategy for Hybrid Fuel Cell Powered Drones in Persistent Missions using Real Flight Test Data

AbstractThis paper proposes an optimized energy management strategy (EMS) for hybrid electric fuel cell/battery-based drones focusing on fuel economy while extending sources lifespans in persistent missions. An off-the-shelf drone fed by a 650W fuel cell is selected as a case study, where the power splitting is conventionally managed by a simple rule-based method. Then, a multi-objective genetic algorithm is used to optimize the proposed EMS parameters considering three scenarios regarding battery state of charge, namely charge sustaining, depleting, and increasing. Therefore, advantages of rule-based strategy and genetic algorithm are combined in an online EMS to fit on drone applications. Extensive simulation results demonstrate that the proposed approach allows power sources to operate within their rated area, prolonging their service life, and leading to 5.1% of fuel consumption reduction. Thus, the autonomy will be increased depending on the carried hydrogen quantity, and the world record can be extended by about 37min. It may also lead to benefit in the operating cost achieving 1450C during one fuel cell stack lifecycle. In fleet tasks, the benefit can be further multiplied.

Hybrid fuel cell powered drones energy management strategy improvement and hydrogen saving using real flight test data

This paper deals with hybrid electric fuel cell-powered drones energy management while targeting hydrogen saving and power supply system efficiency improvement. In this context, a commercially available quadcopter powered by the Intelligent Energy 650 W power module is adopted as a case study. Its power supply system is based on fuel cell and battery, and the power is conventionally managed using a basic rule-based strategy. To improve power management, a frequency separation rule-based approach is first proposed, and then an equivalent consumption minimization strategy is implemented for fuel economy seeking. An experimental flight test is carried out using a battery-powered hexacopter to extract a real power profile for load requirement modeling. The obtained load profile is repeated several times replicating the hovering phase to obtain a larger mission lifetime. Extensive simulation results clearly show that the proposed power management strategies enables power sources operating in their nominal area, extending their lifetimes, and inducing 3% minimization in hydrogen consumption. This optimization extends the drone endurance as much as the carried fuel amount, and it can increase the world endurance record by 21.81min. It has also an economical benefit, which consists in the operating cost gain reaching 853.2€ per fuel cell module lifecycle. In fleet missions, this gain may further be increased.

Navigator Notes

Navigator Notes

An improved online system identification method for tiltrotor aircraft

An online system identification method for tiltrotor aircraft flight dynamics modeling is developed in this paper. A general parametric flight dynamics model which can be used for online identification purpose in the three flight modes of tiltrotor aircraft (i.e., helicopter mode, conversion mode, and fixed-wing mode) is developed, and an unideal noise model is also introduced in order to minimize the parameter identification error caused by measurement noise. An adaptive model structure identification algorithm is established by introducing local accuracy criteria, so that a time variant optimal model structure is obtained in real time. A weighted recursive least squares algorithm is established for parameter identification, and a time variant weighting matrix is designed to provide best numerical performance in the identification procedure at any time. A comprehensive nonlinear flight dynamics model of a notional tiltrotor aircraft is used to generate simulation flight test data in all flight modes, and the identification results of the notional tiltrotor aircraft in fixed wing mode, helicopter model as well as conversion mode are obtained and verified. Finally, the flight dynamics model of an unmanned tiltrotor aircraft in helicopter mode is identified by using the real flight test data. The identification results show that the online system identification method for tiltrotor aircraft developed in this paper is effective, efficient and robust.

Air Data Sensor Fault Detection and Diagnosis in the Presence of Atmospheric Turbulence: Theory and Experimental Validation With Real Flight Data

Managing air data sensor fault detection and diagnosis (FDD) in the presence of atmospheric turbulence is challenging since the effects of faults and turbulence are coupled. Existing FDD approaches cannot decouple the faults from the turbulence. To address this challenge, this brief first proposes a novel kinematic model that incorporates the effects of the turbulence. This model is valid inside the entire flight envelope, and there is no need to design a linear parameter varying system. Then, the double-model adaptive estimation algorithm is extended to achieve unbiased state estimation even in the presence of unknown disturbances. The proposed approach is validated using generated turbulence data with various scale lengths and intensities. More importantly, the proposed approach is successfully validated using the real flight test data of a business jet when it is experiencing atmospheric turbulence.

Identification of Physical Helicopter Models Using Subspace Identification

Subspace identification is a powerful tool due to its well-understood techniques based on linear algebra (orthogonal projections and intersections of subspaces) and numerical methods like singular value decomposition. However, the state space model matrices, which are obtained from conventional subspace identification algorithms, are not necessarily associated with the physical states. This can be an important deficiency when physical parameter estimation is essential. This holds for the area of helicopter flight dynamics, where physical parameter estimation is mainly conducted for mathematical model improvement, aerodynamic parameter validation, and flight controller tuning. The main objective of this study is to obtain helicopter physical parameters from subspace identification results. To achieve this objective, the subspace identification algorithm is implemented for a multirole combat helicopter using both FLIGHTLAB simulation and real flight-test data. After obtaining state space matrices via subspace identification, constrained nonlinear optimization methodologies are utilized for extracting the physical parameters. The state space matrices are transformed into equivalent physical forms via the "sequential quadratic programming" nonlinear optimization algorithm. The required objective function is generated by summing the square of similarity transformation equations. The constraints are selected with physical insight. Many runs are conducted for randomly selected initial conditions. It can be concluded that all of the significant parameters can be obtained with a high level of accuracy for the data obtained from the linear model. This strongly supports the idea behind this study. Results for the data obtained from the nonlinear model are also evaluated to be satisfactory in the light of statistical error analysis. Results for the real flight-test data are also evaluated to be good for the helicopter modes that are properly excited in the flight tests.

MagSLAM: Aerial simultaneous localization and mapping using Earth's magnetic anomaly field

Instances of spoofing and jamming of global navigation satellite systems (GNSSs) have emphasized the need for alternative navigation methods. Aerial navigation by magnetic map matching has been demonstrated as a viable GNSS-alternative navigation technique. Flight test demonstrations have achieved accuracies of tens of meters over hour-long flights, but these flights required accurate magnetic maps which are not always available.Magneticmap availability and resolution vary widely around the globe. Removing the dependency on prior survey maps extends the benefits of aerial magnetic navigation methods to small unmanned aerial systems (sUAS) at lower altitudes where magnetic maps are especially undersampled or unavailable. In this paper, a simultaneous localization andmapping (SLAM) algorithm known as FastSLAM was modified to use scalar magnetic measurements to constrain a drifting inertial navigation system (INS). The algorithm was then demonstrated on real magnetic navigation flight test data. Similar in performance to the map-based approach, MagSLAM achieved tens of meters accuracy in a 100-minute flight without the use of a prior magnetic map. Aerial SLAM using Earth's magnetic anomaly field provides a GNSS-alternative navigation method that is globally persistent, impervious to jamming or spoofing, stealthy, and locally accurate to tens of meters without the need for a magnetic map.

INS/GNSS integration using recurrent fuzzy wavelet neural networks

In recent years, aided navigation systems through combining inertial navigation system (INS) with global navigation satellite system (GNSS) have been widely applied to enhance the position, velocity, and attitude information of autonomous vehicles. In order to gain the accuracy of the aided INS/GNSS in GNSS gap intervals, a heuristic neural network structure based on the recurrent fuzzy wavelet neural network (RFWNN) is applicable for INS velocity and position error compensation purpose. During frequent access to GNSS data, the RFWNN should be trained as a highly precise prediction model equipped with the Kalman filter algorithm. Therefore, the INS velocity and position error data are obtainable along with the lost intervals of GNSS signals. For performance assessment of the proposed RFWNN-aided INS/GNSS, real flight test data of a small commercial unmanned aerial vehicle (UAV) were conducted. A comparison of test results shows that the proposed NN algorithm could efficiently provide high-accuracy corrections on the INS velocity and position information during GNSS outages.

Learning-Based Anomaly Detection and Monitoring for Swarm Drone Flights

This paper addresses anomaly detection and monitoring for swarm drone flights. While the current practice of swarm flight typically relies on the operator’s naked eyes to monitor health of the multiple vehicles, this work proposes a machine learning-based framework to enable detection of abnormal behavior of a large number of flying drones on the fly. The method works in two steps: a sequence of two unsupervised learning procedures reduces the dimensionality of the real flight test data and labels them as normal and abnormal cases; then, a deep neural network classifier with one-dimensional convolution layers followed by fully connected multi-layer perceptron extracts the associated features and distinguishes the anomaly from normal conditions. The proposed anomaly detection scheme is validated on the real flight test data, highlighting its capability of online implementation.

Preliminary Design of a Model-Free Synthetic Sensor for Aerodynamic Angle Estimation for Commercial Aviation.

Heterogeneity of the small aircraft category (e.g., small air transport (SAT), urban air mobility (UAM), unmanned aircraft system (UAS)), modern avionic solution (e.g., fly-by-wire (FBW)) and reduced aircraft (A/C) size require more compact, integrated, digital and modular air data system (ADS) able to measure data from the external environment. The MIDAS project, funded in the frame of the Clean Sky 2 program, aims to satisfy those recent requirements with an ADS certified for commercial applications. The main pillar lays on a smart fusion between COTS solutions and analytical sensors (patented technology) for the identification of the aerodynamic angles. The identification involves both flight dynamic relationships and data-driven state observer(s) based on neural techniques, which are deterministic once the training is completed. As this project will bring analytical sensors on board of civil aircraft as part of a redundant system for the very first time, design activities documented in this work have a particular focus on airworthiness certification aspects. At this maturity level, simulated data are used, real flight test data will be used in the next stages. Data collection is described both for the training and test aspects. Training maneuvers are defined aiming to excite all dynamic modes, whereas test maneuvers are collected aiming to validate results independently from the training set and all autopilot configurations. Results demonstrate that an alternate solution is possible enabling significant savings in terms of computational effort and lines of codes but they show, at the same time, that a better training strategy may be beneficial to cope with the new neural network architecture.

Pattern recognition of UAV flight data based on semi-supervised clustering

UAV is an unmanned aerial vehicle controlled by a remote radio signal or a trajectory planning software carried itself. It is widely used in military, civil and scientific research fields. However, due to the lack of real-time decision-making ability, the UAV has high fault rate. The flight quality assessment of UAV and the construction of fault prediction model can be used for debugging and fault-removing to customer’s aircraft, and also to increase the added value of the civilian UAV products. Before building a fault prediction model, a very important step is to identify the pattern of sampled data. For each group of flight data, the efficiency and accuracy rate of manual quality evaluation and fault identification are not acceptable. Based on the UAV flight data accumulated in the big data platform of an UAV production company in Shenyang, Liaoning Province of China, this paper proposes a semi-supervised clustering technique to do automatic pattern recognition for the sampling points. According to the characteristics of UAV flight data, two different methods are designed to choose initial centroids. Meanwhile, we use the existing normal flight data to train distance thresholds to combine some clusters to eliminate the resulting error clustering. Real flight data or flight test data with manually added labels are used to run the proposed algorithms to verify the recognition results. The experimental results show that the proposed methods greatly improve the efficiency and accuracy of adding precise labels to the historical flight data and play a role in assisting the manual recognition of sampling points while strengthening the management and statistics.

Toward delay-tolerant multiple-unmanned aerial vehicle scheduling system using Multi-strategy Coevolution algorithm

Autonomous bridge inspection operations using unmanned aerial vehicles take multiple task assignments and constraints into account. To efficiently execute the operations, a schedule is required. Generating a cost optimum schedule of multiple-unmanned aerial vehicle operations is known to be Non-deterministic Polynomial-time (NP)-hard. This study approaches such a problem with heuristic-based algorithms to get a high-quality feasible solution in a short computation time. A constructive heuristic called Retractable Chain Task Assignment algorithm is presented to build an evaluable schedule from a task sequence. The task sequence representation is used during the search to perform seamless operations. Retractable Chain Task Assignment algorithm calculates and incorporates slack time to the schedule according to the properties of the task. The slack time acts as a cushion which makes the schedule delay-tolerant. This algorithm is incorporated with a metaheuristic algorithm called Multi-strategy Coevolution to search the solution space. The proposed algorithm is verified through numerical simulations, which take inputs from real flight test data. The obtained solutions are evaluated based on the makespan, battery consumption, computation time, and the robustness level of the schedules. The performance of Multi-strategy Coevolution is compared to Differential Evolution, Particle Swarm Optimization, and Differential Evolution–Fused Particle Swarm Optimization. The simulation results show that Multi-strategy Coevolution gives better objective values than the other algorithms.

Development of virtual sensors to estimate critical aircraft flight parameters

This paper addresses the estimation of flight parameters in nominal and degraded flight conditions when one or several specific sensors become faulty. Anticipating more stringent regulations for future aircraft, reconstructing through time the missing signals is a challenging objective for Flight Control Systems, especially because some of them are essential for control purposes. To this end, this paper proposes a solution based on virtual sensors which makes use of an Adaptive Extended Kalman Filter. This allows the longitudinal flight parameters of a civil aircraft to be estimated on-line before and after the occurrence of one or several failures. This solution offers interesting monitoring capabilities including the detection of common mode failures (i.e., simultaneous failures occurring on redundant sensors), and also provides the capability of correcting a wrong aircraft weight in addition to the sensors monitoring. Typical results using real flight test data are shown to evaluate the performances of the approach in terms of detection and estimation capabilities.

A heuristic reference recursive recipe for adaptively tuning the Kalman filter statistics part-1: formulation and simulation studies

Since the innovation of the ubiquitous Kalman filter more than five decades back it is well known that to obtain the best possible estimates the tuning of its statistics \({\mathbf{X}}_{\mathbf{0}}\), \({\mathbf{P}}_{\mathbf{0}}\), \(\Theta \), R and Q namely initial state and covariance, unknown parameters, and the measurement and state noise covariances is very crucial. The manual and other approaches have not matured to a routine approach applicable for any general problem. The present reference recursive recipe (RRR) utilizes the prior, posterior, and smoothed state estimates as well as their covariances to balance the state and measurement equations and thus form generalized cost functions. The filter covariance at the end of each pass is heuristically scaled up by the number of data points and further trimmed to provide the \({\mathbf{P}}_{\mathbf{0}}\) for subsequent passes. The importance of \({\mathbf{P}}_{\mathbf{0}}\) as the probability matching prior between the frequentist approach via optimization and the Bayesian approach of the Kalman filter is stressed. A simultaneous and proper choice for Q and R based on the filter sample statistics and other covariances leads to a stable filter operation after a few iterations. A typical simulation study of a spring, mass and damper system with a weak nonlinear spring constant by RRR shows it to be better than earlier techniques. Part-2 of the paper further consolidates the present approach based on an analysis of real flight test data.

Multiplicative Error State Kalman Filter vs Nonlinear Complimentary Filter for a High Performance Aircraft Attitude Estimation

<p>Modern control law designs increasingly use aircraft attitude information to improve aircraft manoeuverability. Attitude information allows for gravity term compensations in the longitudinal as well as lateral directional control laws of a typical fighter aircraft. Methodologies and comparisons of multiplicative error state Kalman filter (MEKF) and nonlinear complimentary filter for estimation of attitudes of a high performance aircraft using its onboard autonomous sensors is presented. Shows a problem in pitch angle estimation beyond ± 80 deg in the MEKF and a solution is proposed for the same for the first time. Also presents novel aiding sensor modelling for the implementation of attitude heading reference system for this class of aircraft for the first time. The filter formulations are evaluated using full range manuoevering real flight test data.</p>

Recalibration of a stereoscopic camera system for in-flight wing deformation measurements

A decalibration of a stereoscopic camera system caused by slight movements of the cameras can influence the accuracy of the measured 3D positions significantly. Especially for large scale in-flight applications this is difficult to avoid, e.g. due to the high loads and the vibration level occurring during dynamic flight manoeuvres. Thus a practicable approach for a correction of the results by a recalibration of the camera system is necessary. The image pattern correlation technique (IPCT) delivers large area surface results which enables the assessment of its triangulation error in detail as a measure for the quality of the results. The objective of the presented recalibration is a minimisation of the overall triangulation error by a correction of the external camera parameters. The criteria to assess the reliability of the 3D-surface results and the deformation results derived from are described as well as the limitations of the method. A wing deformation measurement on a VUT100 Cobra aeroplane by means of stereoscopic IPCT was used as a test case to demonstrate the applicability of the recalibration method on real flight test data.