SEARCH · Engineering Papers
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
This presentation highlights methods to classify aircraft from radar data and investigates how data from flight controllers can be used to train classifiers.
This document presents a static analysis tool used to evaluate the controllability of Lunar landers. This was created as part of the NASA Lunar Cargo Transportation and Landing by Soft Touchdown (Lunar CATALYST) program. This tool is capable of accepting typical design information such as location and direction of thrusters, maximum thruster forces, gravity vectors, and center of mass locations. The tool evaluates how far the center of gravity can move from its starting position while still maintaining control. This type of analysis is intended to support results produced by time domain simulations. The code created for this project was implemented in Python, and it was designed to be integrated into systems level optimization tools to yield first-cut results on optimal thruster placement.
Recent efforts by the National Aeronautics and Space Administration Langley Research Center have focused on expanding the flight envelope of the F/A-18 aircraft. Of particular concern has been the low speed, high angle-of-attack regime over which the conventional aerodynamic controls of the F/A-18 lose their effectiveness. In order to address this problem the F18 high alpha research vehicle was developed. This aircraft is essentially a modified F/A-18 which possesses thrust vectoring capabilities and hence increased maneuverability in this flight regime. The linear-quadratic-Gaussian with loop-transfer-recovery and H(infinity) design methodologies are used to design high performance controllers for the F18 at an operating point within the expanded envelope. In addition, how the control redundancy of the F18 can be used to maintain nominal performance, as well as nominal stability, in situations where failures occur, is shown.
The navigation and control of the space shuttle during atmospheric entry are discussed. A functional flow diagram presenting the basic approach to the deorbit targeting problem is presented. The major inputs to be considered are: (1) vehicle state vector, (2) landing site location, (3) entry interface parameters, (4) earliest desired time of landing, and (5) maximum cross range. Mathematical models of the navigational procedures based on controlled thrust times are developed.
Future solar sail missions, such as NASA's Solar Polar Imager Vision, will require sails with dimensions on the order of 50-500 m. We are examining a square sail design with moving mass (trim control mass, TCM) and quadrant rotation primary actuators plus pulsed plasma thrusters (PPTs) at the mast tips for backup attitude control. Quadrant rotation is achieved via roll stabilizer bars (RSB) at the mast tips. At these sizes, given the gossamer nature of the sail supporting structures, flexible modes may be low enough to interact with the control system, especially as these actuators are located on the flexible structure itself and not on the rigid core. This paper develops a practical analysis of the flexible interactions using state-space systems and modal data from finite element models of the system. Torsion and bending of the masts during maneuvers could significantly affect the function of the actuators while activation of the membrane modes could adversely affect the thrust vector direction and magnitude. Analysis of the RSB and TCM dynamics for developing high-fidelity simulations is included. For control analysis of the flexible system, standard finite-element models of the flexible sail body are loaded and the modal data is used to create a modal coordinate state-space system. Key parameters include which modes to include, which nodes are of interest for force inputs and displacement outputs, connecting nodes through which external forces and torques are applied from the flex body to the core, any nominal momentum in the system, and any steady rates. The system is linearized about the nominal attitude and rate. The state-space plant can then be analyzed with a state-space controller, and Bode, Nyquist, step and impulse responses generated. The approach is general for any rigid core with a flexible appendage. This paper develops a compensator for a simple two-mass flex system and extrapolates the results to the solar sail. A finite element model of the 20 m solar sail by ATK Space Systems, recently validated in ground tests, is used to demonstrate the sail analysis approach.
Background: The number of applications of deep learning algorithms in bioinformatics is increasing as they usually achieve superior performance over classical approaches, especially, when bigger training datasets are available. In deep learning applications, discrete data, e.g. words or n-grams in language, or amino acids or nucleotides in bioinformatics, are generally represented as a continuous vector through an embedding matrix. Recently, learning this embedding matrix directly from the data as part of the continuous iteration of the model to optimize the target prediction – a process called ‘end-to-end learning’ – has led to state-ofthe-art results in many fields. Although usage of embeddings is well described in the bioinformatics literature, the potential of end-to-end learning for single amino acids, as compared to more classical manually-curated encoding strategies, has not been systematically addressed. To this end, we compared classical encoding matrices, namely one-hot, VHSE8 and BLOSUM62, to end-to-end learning of amino acid embeddings for two different prediction tasks using three widely used architectures, namely recurrent neural networks (RNN), convolutional neural networks (CNN), and the hybrid CNN-RNN. Results: By using different deep learning architectures, we show that end-to-end learning is on par with classical encodings for embeddings of the same dimension even when limited training data is available, and might allow for a reduction in the embedding dimension without performance loss, which is critical when deploying the models to devices with limited computational capacities. We found that the embedding dimension is a major factor in controlling the model performance. Surprisingly, we observed that deep learning models are capable of learning from random vectors of appropriate dimension. Conclusion: Our study shows that end-to-end learning is a flexible and powerful method for amino acid encoding. Further, due to the flexibility of deep learning systems, amino acid encoding schemes should be benchmarked against random vectors of the same dimension to disentangle the information content provided by the encoding scheme from the distinguishability effect provided by the scheme.
Background: The number of applications of deep learning algorithms in bioinformatics is increasing as they usually achieve superior performance over classical approaches, especially, when bigger training datasets are available. In deep learning applications, discrete data, e.g. words or n-grams in language, or amino acids or nucleotides in bioinformatics, are generally represented as a continuous vector through an embedding matrix. Recently, learning this embedding matrix directly from the data as part of the continuous iteration of the model to optimize the target prediction – a process called ‘end-to-end learning’ – has led to state-of-the-art results in many fields. Although usage of embeddings is well described in the bioinformatics literature, the potential of end-to-end learning for single amino acids, as compared to more classical manually-curated encoding strategies, has not been systematically addressed. To this end, we compared classical encoding matrices, namely one-hot, VHSE8 and BLOSUM62, to end-to-end learning of amino acid embeddings for two different prediction tasks using three widely used architectures, namely recurrent neural networks (RNN), convolutional neural networks (CNN), and the hybrid CNN-RNN. Results: By using different deep learning architectures, we show that end-to-end learning is on par with classical encodings for embeddings of the same dimension even when limited training data is available, and might allow for a reduction in the embedding dimension without performance loss, which is critical when deploying the models to devices with limited computational capacities. We found that the embedding dimension is a major factor in controlling the model performance. Surprisingly, we observed that deep learning models are capable of learning from random vectors of appropriate dimension. Conclusion: Our study shows that end-to-end learning is a flexible and powerful method for amino acid encoding. Further, due to the flexibility of deep learning systems, amino acid encoding schemes should be benchmarked against random vectors of the same dimension to disentangle the information content provided by the encoding scheme from the distinguishability effect provided by the scheme.
The Juno Magnetic Field investigation (MAG) characterizes Jupiter's planetary magnetic field and magnetosphere, providing the first globally distributed and proximate measurements of the magnetic field of Jupiter. The magnetic field instrumentation consists of two independent magnetometer sensor suites, each consisting of a tri-axial Fluxgate Magnetometer (FGM) sensor and a pair of co-located imaging sensors mounted on an ultra-stable optical bench. The imaging system sensors are part of a subsystem that provides accurate attitude information (to approx. 20 arcsec on a spinning spacecraft) near the point of measurement of the magnetic field. The two sensor suites are accommodated at 10 and 12 m from the body of the spacecraft on a 4 m long magnetometer boom affixed to the outer end of one of 's three solar array assemblies. The magnetometer sensors are controlled by independent and functionally identical electronics boards within the magnetometer electronics package mounted inside Juno's massive radiation shielded vault. The imaging sensors are controlled by a fully hardware redundant electronics package also mounted within the radiation vault. Each magnetometer sensor measures the vector magnetic field with 100 ppm absolute vector accuracy over a wide dynamic range (to 16 Gauss = 1.6 x 10(exp. 6) nT per axis) with a resolution of approx. 0.05 nT in the most sensitive dynamic range (+/-1600 nT per axis). Both magnetometers sample the magnetic field simultaneously at an intrinsic sample rate of 64 vector samples per second. The magnetic field instrumentation may be reconfigured in flight to meet unanticipated needs and is fully hardware redundant. The attitude determination system compares images with an on-board star catalog to provide attitude solutions (quaternions) at a rate of up to 4 solutions per second, and may be configured to acquire images of selected targets for science and engineering analysis. The system tracks and catalogs objects that pass through the imager field of view and also provides a continuous record of radiation exposure. A spacecraft magnetic control program was implemented to provide a magnetically clean environment for the magnetic sensors, and residual spacecraft fields andor sensor offsets are monitored in flight taking advantage of Juno's spin (nominally 2 rpm) to separate environmental fields from those that rotate with the spacecraft.
Remote sensing offers the vantage of monitoring a vast area of the Earth continuously. Once developed and launched, a satellite gives years of service in collecting data from the land, the oceans, and the atmosphere. Since the 1980s, attempts have been made to relate disease occurrence with remotely sensed environmental and geophysical parameters, using data from Landsat, SPOT, AVHRR, and other satellites. With higher spatial resolution, the recent satellite sensors provide a new outlook for disease control. At sub-meter to I 10m resolution, surface types associated with disease carriers can be identified more accurately. The Ikonos panchromatic sensor with I m resolution, and the Advanced Land Imager with 1 Om resolution on the newly launched Earth Observing-1, both have displayed remarkable mapping capabilities. In addition, an entire array of geophysical parameters can now be measured or inferred from various satellites. Airborne remote sensing, with less concerns on instrument weight, size, and power consumption, also offers a low-cost alternative for regional applications. NASA/GSFC began to collaborate with the Mahidol University on malaria and filariasis control using remote sensing in late 2000. The objectives are: (1) To map the breeding sites for the major vector species; (2) To identify the potential sites for larvicide and insecticide applications; (3) To explore the linkage of vector population and transmission intensity to environmental variables; (4) To monitor the impact of climate change and human activities on vector population and transmission; and (5) To develop a predictive model for disease distribution. Field studies are being conducted in several provinces in Thailand. Data analyses will soon begin. Malaria data in South Korea are being used as surrogates for developing classification techniques. GIS has been shown to be invaluable in making the voluminous remote sensing data more readily understandable. It will be used throughout this study to clearly demonstrate the spatial relationship between the disease intensities, geophysical variables, and socioeconomic parameters. Asides from malaria and filariasis, application of remote sensing to the control of other diseases have been vigorously pursued by NASA's Environment and Health Initiative. The current program includes projects on Rift Valley fever, St. Louis encephalitis, dengue fever, ebola, African dust and diseases, meningitis, asthma, bartonellosis, cholera, and urban health concerns. Results from these projects indicate that remote sensing will play an increasingly important role in disease control in the future.
Two algorithms are disclosed that, with the use of a 3-axis accelerometer, will be able to determine the angles of attack, sideslip and roll of a capsule-type spacecraft prior to entry (at very high altitudes, where the atmospheric density is still very low) and during entry. The invention relates to emergency situations in which no reliable attitude and attitude rate are available. Provided that the spacecraft would not attempt a guided entry without reliable attitude information, the objective of the entry system in such case would be to attempt a safe ballistic entry. A ballistic entry requires three controlled phases to be executed in sequence: First, cancel initial rates in case the spacecraft is tumbling; second, maneuver the capsule to a heat-shield-forward attitude, preferably to the trim attitude, to counteract the heat rate and heat load build up; and third, impart a ballistic bank or roll rate to null the average lift vector in order to prevent prolonged lift down situations. Being able to know the attitude, hence the attitude rate, will allow the control system (nominal or backup, automatic or manual) to cancel any initial angular rates. Also, since a heat-shield forward attitude and the trim attitude can be specified in terms of the angles of attack and sideslip, being able to determine the current attitude in terms of these angles will allow the control system to maneuver the vehicle to the desired attitude. Finally, being able to determine the roll angle will allow for the control of the roll ballistic rate during entry.
Modal regulators and deformation trackers are designed for an open-loop fluttering wing model. The regulators are designed with modal coordinate and accelerometer inputs respectively. The modal coordinates are estimated with simulated fiber optics. The robust stability of the closed-loop systems is compared in a structured singular-value vector analysis. Performance is evaluated and compared in a gust alleviation and flutter suppression simulation. For the same wing and flight condition two wing-shape-tracking control architectures are presented, which achieve deformation control at any point on the wing.
This paper establishes the connective stability concept in the framework of the comparison principle and vector Liapunov functions, as a natural setting for resolving the complexity vs reliability problem in the control of large-scale dynamic systems. The central result is the following: a stable complex system when composed as a competitive structure of the interconnected stable subsystems remains stable despite the on-off participation of the subsystems. This result is important in that it can be used efficiently to synthesize reliable complex systems by multilevel feedback control.
Stability analyses and simulation data and results are presented for an initial Control Moment Gyroscope system proposed for the Apollo Telescope Mount cluster (later named Skylab) using momentum vector feedback. A compensation filtering technique is presented which significantly improved analytical and simulation performance of the system. This technique is quite similar to the complementary filtering technique and represents an early NASA application.
The Augmented Deflector Exhaust Nozzle (ADEN) was tested in PSL-3 at NASA-Lewis Research Center using an F404 engine. The ADEN is a flight weight Single Expansion Ramp Nozzle with thrust vectoring, an internal cooling system utilizing the available engine fan flow, and a variable area throat controlled by the engine control system. Test conditions included dry and max A/B operation at nozzle pressure ratios from 2.0 to 15.0. High nozzle pressure loading was simulated to verify structural integrity at near maximum design pressure. Nozzle settings covered the full range in throat area and + or - 15 deg deflection angle. Test results demonstrated expected aerodynamic performance, cooling system effectiveness, control system stability, and mechanical integrity.
The following problem is examined: given a nonlinear control system dot-x(t) = f(x(t)) + the sum to m terms(i=1) u sub i (t)g sub i (x(t)) on R(n) and a point x(0) in R(n), approximate the system near x(0) by a linear system. One approach is to use the usual Taylor series linearization. However, the controllability properties of both the nonlinear and linear systems depend on certain Lie brackets of the vector field under consideration. This suggests that a linear approximation based on Lie bracket matching should be constructed at x(0). In general, the linearizations based on the Taylor method and the Lie bracket approach are different. However, under certain mild assumptions, it is shown that there is a coordinate system for R(n) near x(0) in which these two types of linearizations agree. The importance of this agreement is indicated by examining the time responses of the nonlinear system and its linear approximation and comparing the lower order kernels in Volterra expansions of each.
The following problem is examined: given a nonlinear control system dot-x(t) = f(x/t/) + the sum to m terms (i = 1) u sub i (t)g sub i (x/t/) on R(n) and a point x(0) in R(n), approximate the system near x(0) by a linear system. One approach is to use the usual Taylor series linearization. However, the controllability properties of both the nonlinear and linear systems depend on certain Lie brackets of the vector field under consideration. This suggests that a linear approximation based on Lie bracket matching should be constructed at x(0). In general, the linearizations based on the Taylor method and the Lie bracket approach are different. However, under certain mild assumptions, it is shown that there is a coordinate system for R(n) near x(0) in which these two types of linearizations agree. The importance of this agreement is indicated by examining the time responses of the nonlinear system and its linear approximation and comparing the lower order kernels in Volterra expansions of each.
The creation of a computer module used to calculate the size of the horizontal control surfaces of a conceptual aircraft design is discussed. The control surface size is determined by first calculating the size needed to rotate the aircraft during takeoff, and, second, by determining if the calculated size is large enough to maintain stability of the aircraft throughout any specified mission. The tail size needed to rotate during takeoff is calculated from a summation of forces about the main landing gear of the aircraft. The stability of the aircraft is determined from a summation of forces about the center of gravity during different phases of the aircraft's flight. Included in the horizontal control surface analysis are: downwash effects on an aft tail, upwash effects on a forward canard, and effects due to flight in close proximity to the ground. Comparisons of production aircraft with numerical models show good accuracy for control surface sizing. A modified canard design verified the accuracy of the module for canard configurations. Added to this stability and control module is a subroutine that determines one of the three design variables, for a stable vectored thrust aircraft. These include forward thrust nozzle position, aft thrust nozzle angle, and forward thrust split.
Considerable experience was acquired with Ada at the NASA Dryden Flight Research Facility during the on-going High Alpha Technology Program. In this program, an F-18 aircraft was highly modified by the addition of thrust-vectoring vanes to the airframe. In addition, substantial alteration was made in the original quadruplex flight control system. The result is the High Alpha Research Vehicle. An additional research flight control computer was incorporated in each of the four channels. Software for the research flight control computer was written in Ada. To date, six releases of this software have been flown. This paper provides a detailed description of the modifications to the research flight control system. Efficient ground-testing of the software was accomplished by using simulations that used the Ada for portions of their software. These simulations are also described. Modifying and transferring the Ada for flight software to the software simulation configuration has allowed evaluation of this language. This paper also discusses such significant issues in using Ada as portability, modifiability, and testability as well as documentation requirements.