Search

Advances in sensors and avionics computation power suggest real-time structural load measurements could be used in flight control systems for improved safety and performance. A conventional transport flight control system determines the moments necessary to meet the pilot’s command while rejecting disturbances and maintaining stability of the aircraft. Control allocation is the problem of converting these desired moments into control effector commands. In this paper, a framework is proposed to incorporate real-time structural load feedback and structural load constraints in the control allocator. Constrained optimal control allocation can be used to achieve desired moments without exceeding specified limits on monitored load points. Furthermore, certain criteria can be minimized, such as loads on certain parts of the aircraft. Flight safety issues can be addressed by using system health monitoring information to change control allocation constraints during flight. The framework to incorporate structural l...

This paper presents a Takagi–Sugeno fuzzy model-based approach to modeling, controlling, and stabilizing an aircraft in both normal and damaged conditions. The major contributions of this paper are as follows. First, it introduces the Takagi–Sugeno modeling and the parallel distributed compensation control approach. Then, it demonstrates how the approach can be applied to aircraft in a systematic manner. Finally, the approach is applied to the modified NASA F-15 number 837 nonlinear model, and preliminary simulation results are presented. The benefits of using a Takagi–Sugeno model-based parallel distributed compensation control approach include a systematic controller design procedure, mathematical proof of stability for classes of failures, and intuitive insight into damage effects on an aircraft. This work is the first to present this control approach as applied to aircraft.

The performance of nonlinear flight-control algorithms, such as feedback linearization and dynamic inversion, is heavily dependent on the fidelity of the dynamic model being inverted. Incomplete or incorrect knowledge of the aircraft dynamics results in reduced performance and may lead to instability. A self-organizing parametrization structure is developed to augment the baseline dynamic inversion controller for a high-performance military aircraft. This algorithm is proven to be stable and can guarantee arbitrary tracking error performance. The training algorithm to grow the network and adapt the parameters is derived from Lyapunov theory. In addition to growing the network of basis functions, a pruning strategy is incorporated to keep the size of the network as small as possible. The controller is simulated for different situations, including control surface failures, modeling errors, and external disturbances. A performance measure of maximum tracking error is specified for the controllers a priori. Excellent tracking error minimization to a prespecified level using the adaptive component is achieved. The performance of the self-organizing radial-basis-function network-based controller is also compared with a fixed radial-basis-function network-based adaptive controller. While the fixed radial-basis-function network-based controller, which is tuned to compensate for control surface failures, fails to achieve the same performance under modeling uncertainty and disturbances, the self-organizing radial-basis-function network is able to achieve good tracking convergence under all specified error conditions.

Two methods for control system reconfiguration have been investigated. The first method is a robust servomechanism control approach (optimal tracking problem) that is a generalization of the classical proportional-plus-integral control to multiple input-multiple output systems. The second method is a control-allocation approach based on a quadratic programming formulation. A globally convergent fixed-point iteration algorithm has been developed to make onboard implementation of this method feasible. These methods have been applied to reconfigurable entry flight control design for the X-33 vehicle. Examples presented demonstrate simultaneous tracking of angle-of-attack and roll angle commands during failures of the fight body flap actuator. Although simulations demonstrate success of the first method in most cases, the control-allocation method appears to provide uniformly better performance in all cases.

A large, civilian, multiengine transport MD-11 airplane control system was recently modie ed to perform as an emergency backup controller using engine thrust only. The emergency backup system, referred to as the propulsion-controlled aircraft (PCA)system, would be used if a majorprimary e ight control system fails. To allow for longitudinal- and lateral-directional control, the PCA system requires at least two engines and is implemented through software modie cations. A e ight-test program was conducted to evaluate the PCA system high-altitude e ying characteristics and to demonstrate its capacity to perform safe landings. The cruise e ight conditions, several low approaches, and four landings without any aerodynamic e ight control surface movement were demonstrated; however, only one landing is presented. Results that show satisfactory performance of the PCA system in the longitudinal axis are presented. Test results indicate that the lateral-directional axis of the system performed well at high altitude but was sluggish and prone to thermal upsets during landing approaches. Flight-test experiences and test techniques are also discussed, with emphasis on the lateral-directional axis because of the dife culties encountered in e ight test.

When designing control laws for systems with constraints on the tracking performance, control allocation methods can be utilized. Control allocation methods are used when there are more command inputs than controlled variables. Control allocators can be used to address surface saturation limits, structural load limits, drag reduction constraints, or actuator failures. Most transport aircraft have many actuated surfaces compared to the three control variables (such as angle of attack, roll rate, and angle of sideslip). To distribute the control effort among the redundant set of actuators, a fixed mixer approach or online control allocation techniques can be utilized. The benefit of an online allocator is that complex constraints can be considered in the design; however, an online control allocator has the disadvantage of not guaranteeing a surface schedule, which can then produce unacceptable loads on the aircraft. The load uncertainty and complexity has prevented some controller designs from using advanced online allocation techniques. This paper considers actuator redundancy management for a class of over-actuated systems with real-time structural load limits using linear quadratic tracking applied to the generic transport model (a twin-engine heavy civil transport aircraft). With the inclusion of static load constraints in the allocator, the concern of overstressing the structures should be minimized or even eliminated. The results include three test cases. The first test case shows what happens when load constraints are applied over six left- and right-wing locations, with comparison to the same roll input run without load constraints. Test case two shows what happens when a large commanded roll is executed with the same load constraints as those used in test case one; this run is intended to stress the loads allocator. Test case three shows the robustness of the linear quadratic augmented allocator system to uncertainties; a 35-percent change in the control effectiveness plant model will be shown, in which the controller is kept the same as in test cases one and two with six load constraints.

In the presence of large uncertainty, a controller needs to be able to adapt rapidly to regain performance. Fast adaptation is referred to the implementation of adaptive control with a large adaptive gain to reduce the tracking error rapidly. However, a large adaptive gain can lead to high-frequency oscillations which can adversely affect robustness of an adaptive control law. As the adaptive gain increases, the time delay margin for a standard model-reference adaptive control decreases, hence loss of robustness. Optimal control modification is a new adaptive control method developed recently to achieve fast adaptation with robustness. Its formulation is based on the minimization of the L2 norm of the tracking error, posed as an optimal control problem. Computer simulations as well as pilot-in-the-loop high-fidelity simulations in a motion-based flight simulator demonstrate the effectiveness of the new adaptive law. In this study, we extend the optimal control modification to include a covariance-like adjustment mechanism of a time-varying adaptive gain to prevent persistent learning which can reduce robustness. The covariance update law can also include a forgetting factor in a similar context as a standard recursive least-squares estimation algorithm. The covariance adaptive gain adjustment allows an initial large adaptive gain to be set arbitrarily and provides the ability to drive the adaptive gain to a lower value as the adaptation has achieved sufficiently the desired tracking performance. Alternatively, a normalized adaptive gain may be used to reduce adaptation when the amplitude of an input basis function becomes large. Flight control simulation results demonstrate that both approaches can achieve significant robustness as measured by the time delay margin. Furthermore, a recent flight test program of the optimal control modification with normalization on a NASA F-18 aircraft demonstrates the effectiveness of the adaptive law.

National Aeronautics and Space Administration (NASA) researchers have conducted a series of flight experiments designed to study the effects of varying levels of adaptive controller complexity on the performance and handling qualities of an aircraft under various simulated failure or damage conditions. A baseline, nonlinear dynamic inversion controller was augmented with three variations of a model reference adaptive control design. The simplest design consisted of a single adaptive parameter in each of the pitch and roll axes computed using a basic gradient-based update law. A second design was built upon the first by increasing the complexity of the update law. The third and most complex design added an additional adaptive parameter to each axis. Flight tests were conducted using NASA s Full-scale Advanced Systems Testbed, a highly modified F-18 aircraft that contains a research flight control system capable of housing advanced flight controls experiments. Each controller was evaluated against a suite of simulated failures and damage ranging from destabilization of the pitch and roll axes to significant coupling between the axes. Two pilots evaluated the three adaptive controllers as well as the non-adaptive baseline controller in a variety of dynamic maneuvers and precision flying tasks designed to uncover potential deficiencies in the handling qualities of the aircraft, and adverse interactions between the pilot and the adaptive controllers. The work was completed as part of the Integrated Resilient Aircraft Control Project under NASA s Aviation Safety Program.

Advances in sensors and avionics computation power suggest real-time structural load measurements could be used in flight control systems for improved safety and performance. A convention transport flight control system determines the moments necessary to meet the pilot’s command, while rejecting disturbances and maintaining stability of the aircraft. Control allocation is the problem of converting these desired moments into control effector commands. In this paper, a framework is proposed to incorporate real-time structural load feedback and structural load constraints in the control allocator. Constrained optimal control allocation can be used to achieve desired moments without exceeding specified limits on monitored load points. Furthermore, certain criteria can be minimized, such as loads on certain parts of the aircraft. Flight safety issues could be addressed by using system health monitoring information to change control allocation constraints during flight. The framework to incorporate structural loads in the flight control system and an optimal control allocation algorithm will be described and then demonstrated on a nonlinear simulation of a generic transport aircraft with flight dynamics and static structural loads.

This paper presents a new adaptive control approach using Chebyshev orthogonal polynomials as basis functions in a least-squares functional approximation. The use of orthogonal basis functions improves the function approximation significantly and enables better convergence of parameter estimates. Flight control simulations demonstrate the effectiveness of the proposed adaptive control approach.

Singular value analysis can give conservative stability margin results. Applying structure to the uncertainty can reduce this conservatism. This paper describes flight-determined stability margins for the X-29A lateral-directional, multiloop control system. These margins are compared with the predicted unsealed structured singular values, scaled structured singular values, and conventional single-loop phase and gain margins. The algorithm was further evaluated with flight data by changing the roll-rate-to-aileron-command-feedback gain by ±20%. Minimum eigenvalues of the return difference matrix that bound the singular values are also presented. Extracting multiloop singular values from flight data and analyzing the feedback gain variations validates this technique as a measure of robustness. This analysis can be used for near-real-time flight monitoring and safety testing.

This paper presents an L(sub 1) adaptive control augmentation system design for multi-input multi-output nonlinear systems in the presence of unmatched uncertainties which may exhibit significant cross-coupling effects. A piecewise continuous adaptive law is adopted and extended for applicability to multi-input multi-output systems that explicitly compensates for dynamic cross-coupling. In addition, explicit use of high-fidelity actuator models are added to the L1 architecture to reduce uncertainties in the system. The L(sub 1) multi-input multi-output adaptive control architecture is applied to the X-29 lateral/directional dynamics and results are evaluated against a similar single-input single-output design approach.

Conventional aircraft generally employ mixing algorithms or lookup tables to determine control surface deflections needed to achieve moments commanded by the flight control system. Control allocation is the problem of converting desired moments into control effector commands. Next generation aircraft may have many multipurpose, redundant control surfaces, adding considerable complexity to the control allocation problem. These issues can be addressed with optimal control allocation. Most optimal control allocation algorithms have control surface position and rate constraints. However, these constraints are insufficient to ensure that the aircraft's structural load limits will not be exceeded by commanded surface deflections. In this paper, a framework is proposed to enable a flight control system with optimal control allocation to incorporate real-time structural load feedback and structural load constraints. A proof of concept simulation that demonstrates the framework in a simulation of a generic transport aircraft is presented.

In the presence of large uncertainties, a control system needs to be able to adapt rapidly to regain performance. Fast adaptation is referred to as the implementation of adaptive control with a large adaptive gain to reduce the tracking error rapidly; however, a large adaptive gain can lead to high-frequency oscillations which can adversely affect the robustness of an adaptive control law. A new adaptive control modification is presented that can achieve robust adaptation with a large adaptive gain without incurring high-frequency oscillations as with the standard model-reference adaptive control. The modification is based on the minimization of the Y2 norm of the tracking error, which is formulated as an optimal control problem. The optimality condition is used to derive the modification using the gradient method. The optimal control modification results in a stable adaptation and allows a large adaptive gain to be used for better tracking while providing sufficient robustness. A damping term (v) is added in the modification to increase damping as needed. Simulations were conducted on a damaged F-18 aircraft (McDonnell Douglas, now The Boeing Company, Chicago, Illinois) with both the standard baseline dynamic inversion controller and the adaptive optimal control modification technique. The results demonstrate the effectiveness of the proposed modification in tracking a reference model.

The National Aeronautics and Space Administration Dryden Flight Research Center (Edwards, California) is conducting ongoing flight research using adaptive controller algorithms. A highly modified McDonnell-Douglas NF-15B airplane called the F-15 Intelligent Flight Control System (IFCS) was used for these algorithms. This airplane has been modified by the addition of canards and by changing the flight control systems to interface a single-string research controller processor for neural network algorithms. Research goals included demonstration of revolutionary control approaches that can efficiently optimize aircraft performance for both normal and failure conditions, and to advance neural-network-based flight control technology for new aerospace systems designs. Before the NF-15B IFCS airplane was certified for flight test, however, certain processes needed to be completed. This paper presents an overview of these processes, including a description of the initial adaptive controller concepts followed by a discussion of modeling formulation and performance testing. Upon design finalization, the next steps are: integration with the system interfaces, verification of the software, validation of the hardware to the requirements, design of failure detection, development of safety limiters to minimize the effect of erroneous neural network commands, and creation of flight test control room displays to maximize human situational awareness.

This report describes the improvements and enhancements to a neural network based approach for directly adapting to aerodynamic changes resulting from damage or failures. This research is a follow-on effort to flight tests performed on the NASA F-15 aircraft as part of the Intelligent Flight Control System research effort. Previous flight test results demonstrated the potential for performance improvement under destabilizing damage conditions. Little or no improvement was provided under simulated control surface failures, however, and the adaptive system was prone to pilot-induced oscillations. An improved controller was designed to reduce the occurrence of pilot-induced oscillations and increase robustness to failures in general. This report presents an analysis of the neural networks used in the previous flight test, the improved adaptive controller, and the baseline case with no adaptation. Flight test results demonstrate significant improvement in performance by using the new adaptive controller compared with the previous adaptive system and the baseline system for control surface failures.

This paper presents a stable discrete-time adaptive law that targets modeling errors in a direct adaptive control framework. The update law was developed in our previous work for the adaptive disturbance rejection application. The approach is based on the philosophy that without modeling errors, the original control design has been tuned to achieve the desired performance. The adaptive control should, therefore, work towards getting this performance even in the face of modeling uncertainties/errors. In this work, the baseline controller uses dynamic inversion with proportional-integral augmentation. Dynamic inversion is carried out using the assumed system model. On-line adaptation of this control law is achieved by providing a parameterized augmentation signal to the dynamic inversion block. The parameters of this augmentation signal are updated to achieve the nominal desired error dynamics. Contrary to the typical Lyapunov-based adaptive approaches that guarantee only stability, the current approach investigates conditions for stability as well as performance. A high-fidelity F-15 model is used to illustrate the overall approach.

†This paper presents enhancements to a neural network based approach for directly adapting to aerodynamic changes resulting from damage or failures. This is a follow-on effort to flight tests performed on the NASA F-15 aircraft, as part of the Intelligent Flight Control System research effort. Previous results demonstrated the potential for improving performance under simulated damage conditions. However, little improvement was provided under simulated control surface failures, and the adaptive system tended to be prone to pilot induced oscillations. This paper presents an analysis of the previous flight tests and proposes an alternate input selection criterion, a technique for improving robustness through normalized learning rates, and a method for adaptively retrofitting a classical yaw damping controller. Simulation results demonstrate significant improvement in performance and robustness over the neural network implementation used in the previous flight tests.

A fault tolerant neurally augmented control scheme based on non-linear dynamic inversion is designed for the WVUL YF-22 aircraft model. The parameters of the model following adaptive flight controller are determined at a single flight condition and a neural network is used to compensate for inversion errors and changes in aircraft dynamics, including actuator failures. Three different neural networks are used: the Extended Minimal Resource Allocating Network, the Single Hidden Layer Network, and the Sigma Pi. Numerical simulations are performed at nominal flight conditions and failure conditions affecting the stabilator or the aileron. Performance assessment parameters are defined based on the angular rate tracking errors. The performance of the three neural networks is compared in terms of these parameters.

The objective of NASA?s Intelligent Flight Control System (IFCS) program is to develop and flight test control schemes that enhance control during a primary control surface failure or aerodynamic change due to a failure or modeling errors. The goal of the Generation 2 (Gen-2)] is to evaluate on-line neural flight control system that can provide adaptive control without explicit parameter identification. The Gen-2 approach does not require (1) information on the nature or the extent of the failure, (2) knowledge of the control surface positions, or (3) information on aerodynamic failures or un-modeled parameters. This tracking controller adds direct adaptive neural network signals to the control law1,2,3,4. These neural networks are used to generate command augmentation signals to compensate for errors due to un-modeled dynamics, including dynamics due to damage or failure. Flight demonstration started early in the year 2006, on the NASA F-15 tail number 837. The F-15 6-degree-offreedom (6-dof) simulator was used in the evaluation test, comparing stabilator and canard failure compensation with the neural network algorithm.

This paper proposes a methodology for the task of comparing the performance of flight control laws using flight data recorded under non-homogeneous atmospheric conditions and dissimilar pilot input. Small differences in atmospheric conditions can potentially impact significantly the evaluation of the controller performance and prevent consistent comparison especially in the case of reduced size aircraft (autonomous or remotely piloted). Consistent deterministic control inputs can only be guaranteed by some form of on-board excitation system. An algorithm is developed for flight data reduction to account for such dissimilarities. The method presented in this paper is developed for the specific purpose of comparing model following control laws based on non-linear dynamic inversion with neural network augmentation. Evaluation parameters based on angular rate tracking errors are defined and used for the comparison. The procedure is documented using flight data from the WVU YF22 research aircraft model - a platform for testing fault tolerant control laws.

This paper presents a neural network based adaptive controller with application to a high performance flight vehicle such as F-15 military aircraft. The baseline dynamic inversion controller is augmented with a Radial Basis Function Network (RBFN) that minimizes the inversion error which may occur due to imperfect modeling, approximate inversion, or sudden changes in aircraft dynamics such as failure of control surfaces. The RBFN is trained online using a simple growing and pruning algorithm in which hidden layer neurons are added or pruned based on the input to the network. The controller is simulated for 2 types of control surface failures with and without the adaptive neural network. The comparative results presented show the ability of the adaptive controller to compensate for errors arising due to the specified control surface failures.

Description of the performance of a simplified dynamic inversion controller with neural network augmentation follows. Simulation studies focus on the results with and without neural network adaptation through the use of an F-15 aircraft simulator that has been modified to include canards. Simulated control law performance with a surface failure, in addition to an aerodynamic failure, is presented. The aircraft, with adaptation, attempts to minimize the inertial cross-coupling effect of the failure (a control derivative anomaly associated with a jammed control surface). The dynamic inversion controller calculates necessary surface commands to achieve desired rates. The dynamic inversion controller uses approximate short period and roll axis dynamics. The yaw axis controller is a sideslip rate command system. Methods are described to reduce the cross-coupling effect and maintain adequate tracking errors for control surface failures. The aerodynamic failure destabilizes the pitching moment due to angle of attack. The results show that control of the aircraft with the neural networks is easier (more damped) than without the neural networks. Simulation results show neural network augmentation of the controller improves performance with aerodynamic and control surface failures in terms of tracking error and cross-coupling reduction.

A reconfigurable control law for the full X-33 flight envelope has been designed to accommodate a failed control surface and redistribute the control effort among the remaining working surfaces to retain satisfactory stability and performance. An offline nonlinear constrained optimization approach has been used for the X-33 reconfigurable control design method. Using a nonlinear, six-degree-of-freedom simulation, three example failures are evaluated: ascent with a left body flap jammed at maximum deflection; entry with a right inboard elevon jammed at maximum deflection; and landing with a left rudder jammed at maximum deflection. Failure detection and identification are accomplished in the actuator controller. Failure response comparisons between the nominal control mixer and the reconfigurable control subsystem (mixer) show the benefits of reconfiguration. Single aerosurface jamming failures are considered. The cases evaluated are representative of the study conducted to prove the adequate and safe performance of the reconfigurable control mixer throughout the full flight envelope. The X-33 flight control system incorporates reconfigurable flight control in the existing baseline system.

Two methods for control system reconfiguration have been investigated. The first method is a robust servomechanism control approach (optimal tracking problem) that is a generalization of the classical proportional-plus-integral control to multiple input-multiple output systems. The second method is a control-allocation approach based on a quadratic programming formulation. A globally convergent fixed-point iteration algorithm has been developed to make onboard implementation of this method feasible. These methods have been applied to reconfigurable entry flight control design for the X-33 vehicle. Examples presented demonstrate simultaneous tracking of angle-of-attack and roll angle commands during failures of the right body flap actuator. Although simulations demonstrate success of the first method in most cases, the control-allocation method appears to provide uniformly better performance in all cases.

Singular value analysis can give conservative stability margin results. Applying structure to the uncertainty can reduce this conservatism. This paper presents flight-determined stability margins for the X-29A lateral-directional, multiloop control system. These margins are compared with the predicted unscaled singular values and scaled structured singular values. The algorithm was further evaluated with flight data by changing the roll-rate-to-aileron command-feedback gain by +/- 20 percent. Minimum eigenvalues of the return difference matrix which bound the singular values are also presented. Extracting multiloop singular values from flight data and analyzing the feedback gain variations validates this technique as a measure of robustness. This analysis can be used for near-real-time flight monitoring and safety testing.

The application of the eigensystem synthesis technique to place the closed-loop eigenvalues and shape the closed-loop eigenvectors has not been practical for active flutter suppression, primarily because of the availability of only one control surface (aileron) for flutter suppression. The oblique-wing aircraft, because of its configuration, provides two independent surfaces (left and right ailerons), making the application of eigensystem synthesis practical. This paper presents the application of eigensystem synthesis using output feedback for the design of an active flutter suppression system for an oblique-wing aircraft. The results obtained are compared with those obtained by linear quadratic Gaussian techniques.

The U.S. Navy and NASA are currently involved in the design and development of an unsymmetric-skew-wing aircraft capable of 65° wing sweep and flight at Mach 1.6. A generic skew-wing aircraft model was developed for 45° wing skew at a flight condition of Mach 0.70 and 3048 m altitude. At this flight condition the aircraft has a wing flutter mode. An active implementable control law was developed using the linear quadratic Gaussian design technique. A method of modal residualization was used to reduce the order of the controller used for flutter suppression.