Aerospace Power, Propulsion, and Electronics
Deepak Chadha and Ben H. Jansen, Ph.D. Department of Electrical Engineering
Techniques are being developed for on-line monitoring of complex, nonlinear systems,
such as rocket engines. The primary aim is to detect impending failures in a timely manner
so that catastrophic failure can be avoided. The basic concept involves training
artificial neural networks (ANN) to accurately predict future values of sensor data from
past and present observations. Researchers hypothesize that a successfully trained network
captures the dynamics of the system from which sensor data are obtained. When the
predicted values start to deviate from actual sensor data, monitors have an early sign
that the underlying dynamics have changed and that the engine may be close to failing.
Two ANN architectures have been explored, the multi-layer perceptron (MLP) and the
cascade correlation network (CCN). Two criteria were examined as a means of detecting
change, the Total Sum of Squares (TSS) and the Weight Difference (WD). Research focused on
determining the sensitivity of the proposed method using artificially generated data.
Duffing's oscillator was used to produce chaotic time series sufficiently complicated that
they appeared to be "stochastic" time series. Findings concluded that an MLP
with ten hidden units operates better than a comparable CCN.
Return to Table of Contents
Ovidiu Crisan, Ph.D., Department of Electrical Engineering
Present and future developments of the U. S. space program which include three
important and difficult tasks-the Space Station, the return voyage to the moon, and a
journey to Mars-are heavily dependent on space research achievements. Along with other
objectives, advanced spacecraft require the implementation of large electrical power
systems. Thus, autonomous power system control and management technology are needed to
increase productivity and reliability and decrease human involvement and cost.
Substantial progress has been made in the last few years, with the NASA-Marshall Space
Center being a leading contributor in the field. In 1987, NASA-JSC initiated a program to
study Advanced Electrical Power Management Techniques for the Space Systems (ADEPTS). New
research directions, procedures, and methods are needed for JSC to advance the level of
technology. The solution calls for the utilization of multidisciplinary and high-level
theoretical and practical skills. The spacecraft electrical power system is a complex
subsystem which must operate in concert with other subsystems and payloads and for which
human monitoring and control are very difficult and costly.
Main tasks of the research are to define the spacecraft's power system configuration,
to define and model the operation, automation and control function for each of its
components, to define and model the monitoring, measurement, control and management
functions of the spacecraft main computer and, for an average daily cycle, by computer to
model and analyze the spacecraft's normal and possible faulty states. Proposals are drawn
to improve the structure and functions of the spacecraft.
The researchers are currently working on a new module for the space station electrical
power system (EPS), focusing on (1) EPS components modeling and EPS computer
implementation, (2) computers, computer network, and interface selection and
implementation , and (3) the definition of load and control management of the spacecraft
Electrical Power Distribution and Control Systems (EPDCS) and implementation. The Power
Control and Management System (PCMS) must have a monitoring system and a proper interface
with higher level operation management systems and must allow for easy expansion and the
inclusion of other expert system decision processes.
Designers of the module are considering problems of energy collection and conversion,
energy storage, power conversion, power distribution, and electrical power consuming
equipment.
Each of the main components of the spacecraft electrical power system has been modeled
and system operations tested in six areas: 1) energy collection and conversion, 2) energy
storage, 3) power conversion, 4) power distribution, 5) management and control, and 6)
grounding and protection subsystems.
Results show that for tasks of the main computer, mainly related to monitoring,
measurement, and economic dispatch, there is a need for a more complete definition. Also,
the time coordination of the protection system must be improved and possible multiple
faults modeled. To quantitatively define the overall performances of the system for a
given time-cycle, the states probability should be counted.
Return to Table of Contents
Robert D. Finch, Ph.D. Department of Mechanical Engineering
Acoustic signals can be used to monitor the integrity of construction facilities.
Research seeks to improve the resolution of defect detection by such systems. A system for
the detection of defects consists of an acoustic excitation, a transmission path, one or
more receivers, and hardware for signal processing and computation. Current research
involves three disciplines: (1) modeling of the monitored systems, (2) data acquisition
and signal processing, and (3) system identification.
The monitoring of structural composite beams and panels, as intended for aerospace
applications, may be considered a specific example. In modeling, the concern is to develop
computational techniques that are not only accurate but also efficient enough to permit
their being used as part of a software package for real-time recognition. In data
acquisition and signal processing, researchers seek the optimization of present techniques
and the development of new ones. For this purpose, this laboratory is developing a
database from experiments with beams and panels under a variety of conditions in a test
stand constructed specifically for this purpose. In system identification, the objective
is the classification of objects (as good, defective, etc.) from their received signals,
using techniques from pattern recognition, artificial neural nets, and inverse techniques
based upon ISSO modeling studies.
Return to Table of Contents
Sven G. Holmquist and Prahash Jayaram, graduate students; Ben H. Jansen, Ph.D.,
Department of Electrical Engineering
A control system is being developed for autonomous distribution and control of
electrical power systems using an object-oriented simulation model of the power system and
first-principle knowledge to detect, identify, and isolate faults. Each power system
component is represented as a separate object with knowledge of its normal behavior.
The reasoning process takes place at two different levels of abstraction: the Physical
Component Model (PCM) level and the Electrical Equivalent Model (EEM) level, with the PCM
being the lowest level of abstraction and the EEM the highest. At the EEM level, power
system components are reasoned in terms of their electrical equivalents (e.g., a resistive
load is thought of as a resistor). At the PCM level, however, detailed knowledge about the
component's specific characteristics is taken into account.
The control system operates in two simultaneous modes, a reactive and a proactive mode.
In the reactive mode, the control system receives measurement data from the power system
and compares these values with values determined through simulation to detect the
existence of a fault. The nature of the fault is then identified through a model-based
reasoning process mainly using the EEM. Compound component models are constructed at the
EEM level and are used in the fault identification process. In the proactive mode,
reasoning takes place at the PCM level. Individual components determine their future
health status utilizing a physical model and measured historical data. If changes in
health status seem imminent, the component warns the control system about its impending
failure.
During the past year, researchers completed the following projects:
The VIsolver was refined and tested for "larger" cases. The VIsolver is an
object that solves for currents and voltages of the power system using the modified nodal
formulation. This method can be used on networks containing voltage and current sources,
impedances, conductances, ideal two-ports, and switches.
The fault identification method-(the process of determining which component or
components caused the failure-has been modified. The concept of compound components has
been introduced to solve the problem of insufficient sensory information. Compound
components are models of several EEM's in series, parallel, or in a bridge configuration
combined into a single model component. The location of the available current-meters and
volt-meters guides the formation of the compound components so that in the (reduced)
environment the impedance of each compound component can be calculated based on sensor
data.
Models for some typical power systems components; e.g., generic solar panels and
batteries, based on first-principle knowledge, operating characteristics, and heuristics.
Development of the framework for concurrent hypothesis testing. The fault detection
method conceptualized is based on the generation and simultaneous pursuit of several
possible causes (hypotheses) for the misbehavior of the power system. Each of these
hypotheses will spawn a separate process, which will terminate when it has been
established that the hypothesis cannot be the cause of the problem. This approach leads to
a very efficient implementation on a multi-processor machine.
Return to Table of Contents
John Harris Miller, Jr., Ph.D., the Texas Center for Superconductivity, TCSUH and
Gemunu Gunaratne, Ph.D., Department of Physics
Interferometers
were fabricated consisting of arrays with several grain boundary junctions (GBJ's) in
parallel, by patterning 1400 Å thick YBa2Cu3O7-x (YBCO) films laser deposited onto SrTiO3
(STO) bi-crystal substrates with 24°ree; misorientation angles. The width of each
junction was 5 mm and the dimensions of each loop in the array were 10 mm x 50 mm. The
self-inductance parameters of the devices were estimated to be b = 2LIc1/I = 14 at 77 K
and even higher at lower temperatures. The magnetic field-dependence of the critical
current is shown in Figure 1 for a 6-junction interferometer at 77 K. Initially, after
cooling the sample in zero field, the critical current decreases with increasing field and
exhibits small oscillations, eventually saturating at a small residual value. When the
direction of the field sweep is reversed, a pronounced "satellite" peak is
observed as the field is decreased towards zero. If one continues to sweep the field in
the negative direction and then reverses the direction of the field sweep again, another
satellite peak, which is roughly the mirror image of the first is observed as the field is
swept in the positive direction towards the origin. Additional sweeps in the positive and
negative directions yield only the satellite peaks, whereas the central maximum at zero
field, Ic (H = 0), remains suppressed until the sample is warmed up above Tc and cooled
down again in zero field. This is somewhat similar to the hysteretic behavior observed in
the field-dependent critical current of granular Nb microbridges (Aumine et al., 1989).
Figure 1. Critical current vs. field fo a 6 JJ array at 77K, with a sweep
range of -2 gauss to +2 gauss (neglecting flux focusing effects). The magnetic field
begins at zero (at the center) and is slowly incresed to its maximum positive value (at
the right). The sweep direction is then reversed and the magnetic field is slowly
decreaded to its maximum negative value (at the left). Finally, the field is slowly swept
back to zero.
The separation between the satellite peaks is found to increase with increasing sweep
range, or maximum applied field Hmax. However, the difference DH* = |Hmax - Hpeak| between
each extreme of the field sweep and the position of each satellite peak is roughly
independent of Hmax for a given temperature, provided the maximum field is sufficiently
high that the condition |Hmax| > 2DH* is satisfied. The value of DH* thus appears to be
related to the field required for the system to reach the critical state, in which the
screening current flowing through each junction is about equal to the junction critical
current.
A one-dimensional (1D) parallel JJ array, or interferometer, with large B can be
represented by a Hamiltonian formally identical to the Frenkel-Kontorova (FK) model
(Frenkel and Kontorova, 1938, 1939) of spring-coupled balls in a washboard potential or,
equivalently, pendula in a gravitational potential coupled by torsion springs. The FK
model has been utilized to describe numerous physical systems, including metastable states
in pinned charge-density waves with internal degrees of freedom (Coppersmith, 1987). The
existence of irreversibility and critical state behavior in the field-dependent flux
profiles of 1D JJ arrays has been modeled by Parodi and Vaccarone (Parodi and Vaccarone,
1991). In the limit of large B, each loop in the array can trap several flux quanta and
the system is predicted to exhibit hysteresis in the magnetic behavior (Majhofer et al.,
1991), by analogy to the critical state model of type-II superconductors (Bean, 1962).
Initially, as the externally applied flux fx is increased from zero, screening currents
flowing through the junctions screen out most of the flux from the interior. When the
screening currents flowing through the two end junctions reach their corresponding
critical currents, flux starts to penetrate into the outer loops. As the applied field is
increased further, flux (= LIc1 per loop) penetrates into each loop one-by-one until a
"critical state" is attained, in which flux has penetrated all the way into the
interior loops of the array, and the (negative) magnetization M = <0 - 0x>/[µ0 x
loop area] has saturated to its maximum magnitude. In the FK model, this is equivalent to
stretching (or compressing) all of the springs to the point where the restoring force
acting on each ball in the washboard potential is maximum. If the field is then decreased,
flux "leaks out" of each loop one-by-one until opposite critical state is
attained, in which the trapped flux remaining in the loops is maximum (= NLIc1/2) in the
middle of the array and decreases towards the ends. A plot of M vs. field H thus forms a
hysteresis loop, just as in the critical state model.
The total magnetization of the array is proportional to the net circulating current.
One would therefore expect the critical current to be a maximum whenever the net
circulating current, and thus the magnetization, is equal to zero during an M-H hysteresis
loop. According to this argument, ΔH* would be constant and about equal to the
remnant magnetization, Mrem = NLIc1/ (2µ0Aloop), provided the maximum applied field was
sufficiently high that the critical state was maintained throughout the field sweep cycle,
i.e. Hmax > 2Mrem. This result would be consistent with the experiment, and
would account for the observed shift in satellite peak position with increasing sweep
range in plots of Ic vs. H.
References
Aumine J., E. Tanaka, S. Yamasaki, K. Tani, and A. Yonekura. "Hysteresis of
Critical Currents of Superconducting Bridges in Low Perpendicular Magnetic Fields."
J. Low Temp. Phys. 74 (1989): 263-275.
Bean, C.P. "Magnetization of Hard Superconductors." Phys. Rev. Lett.,8 (1962):
250-253.
Coppersmith, S.N. "Overdamped Frenkel -Kotorova Model with Randomness as a Dynamic
System: Mode Locking and Derivation of Discrete Maps." Phys. Rev. A.36 (1987):
3375-82.
Frenkel, J. and T. Kontorova, "On the Theory of Plastic Demortation and
Twinning," Zh. Eksp. Teor. Fiz 8 (1938):1340; J. Phys. [Moscow] (1939): 137.
Majhofer, A., T. Wolf, and W. Dietrich. "Irreversible Magnetization Effects in a
Network of Resistively Shunted Tunnel Junctions." Phys. Rev.B. 44 (1991): 9634-38.
Parodi, F. and R. Vaccarone. "Critical State and Flux Dynamics in Squid Arrays."
Physica C. 173(1991): 56-64.
Return to Table of Contents
Krishnaswamy Ravi-Chandar, Ph.D., and Shalini Gupta, graduate researcher, Department
of Mechanical Engineering
While the fracture mechanics of viscoelastic solids has been studied over the past
decades, problems remain in the application of the theory to solid propellants. This
program served to investigate the details of stress and deformation fields near the crack
tip using the method of Electronic Speckle Pattern Interferometry (ESPI), as a prelude to
the development of techniques that will prove useful in the characterization of the
fracture behavior of solid propellants.
The basic idea of the ESPI is similar to that of any two-beam interferometry. The two
light beams that interfere are, however, obtained from a diffusely reflecting surface.
This result implies that no special preparation of the specimen surface is necessary. The
image obtained is "speckled"; that is, it appears grainy. The speckle intensity
pattern initially generated before displacement of the reflecting surface, can be
correlated to the speckle pattern produced after displacement. If the initial speckle
pattern is stored as a digital image and then subtracted from speckle pattern after
displacement, an EPSI image is produced that can be correlated to the displacement. By
properly choosing the relative orientations of the two illuminating beams and the angle of
observation of the speckle patterns, in-plane and out-of-plane components of the
displacement can be obtained separately.
The sensitivity of the technique can also be altered easily by varying the same
parameters, thus making the speckle technique more flexible compared with other
interferometric techniques. The implementation of the method of ESPI to solid propellant
does not require any surface preparation since the solid propellant surface is already
"speckly." ESPI was implemented using a DATA Translation frame grabber on an
IBM/PC; the frame grabber permits real time implementation of the processing of ESPI. A
digitized 8-bit image is passed through an input look-up table and transformed into a
4-bit image and stored in the lower four bits of an 8-bit frame storage memory; this image
is the first speckle image.
The second digitized image, which is acquired continuously, is passed through another
look-up table that transforms the image into a 4-bit image stored in the upper four bits
of the same 8-bit frame storage memory. The two four-bit images in the buffer are
subtracted in real-time ESPI fringe patterns on the monitor, which can be correlated to
the displacements in the crack tip region. The technique of ESPI has been implemented on
solid propellant material. Work is currently in progress on the extraction of the
displacement components and interpretation of the fracture behavior in terms of classical
fracture mechanics of viscoelastic materials.
Return to Table of Contents
Jeffrey T. Williams, Ph.D., Donald R. Wilton, and Noel Purcell; NASA coordinator,
Dickie Arndt
Investigators initiated this study to determine the suitability of the Intrasystem
Electromagnetic Compatibility Analysis Program (IEMCAP) for predicting and analyzing
electromagnetic energy, coupling between devices within complex systems. Such tools as
EIMCAP, developed by the Air Force, are needed in the aerospace industry to predict
electromagnetic compatibility (EMC) and electromagnetic interference (EMI) between
electronic subsystems scheduled for installation in a close environment like a space
station. Coupling mechanisms chosen for study in this project included wire-to-wire and
case-to-case coupling.
Detailed calculations and experiments showed the IEMCAP is a good predictor of upper
bounds for wire-to-wire coupling at low frequencies. Results indicated that IEMCAP is not
a reliable predictor of wire-to-wire coupling at high frequencies, however, due,
apparently, to failure of its simple model at the frequencies.
The IEMCAP model for case-to-case modeling as found to be inadequate over a wide range
of frequencies. IEMCAP essentially assumes a dipole model for equipment case radiation,
with a constant determined by near field measurement at one field point. The research was
summarized in a report to NASA entitled "Space Systems EMC/EMI Analysis Using
IEMCAP," with Noel Purcell, Jeffrey T. Williams, and Donald R. Wilton credited as
authors of the work.
Return to Table of Contents
John C. Wolfe, Ph.D., Department of Electrical and Computer Engineering
ISSO investigators sought to develop a competitive proposal for a target-acquisition
radar (Seeker). A team was formed with Science Applications International Corp. (SAIC) and
Hughes Research Laboratory whereby design and testing was to be performed at SAIC, Hughes
would provide transmitter chips, and ISSO researchers would fabricate the actual radar.
Proof-of-concepts experiments were carried out and incorporated in the proposal, which was
submitted for review to the Army Strategic Defense Command.
The fabrication process comprised the following design goals: Lithographic resolution =
10-4d, where d is the depth of field Maximum depth-of-field = 1 cm Linewidth tolerance = 5
mm Edge slope of etched structures = < 5°ree; Etch depth = 0-100 mm Structural
material = Al or Si, TBD by University of Houston Electrical conductor = Al, Au, or Ag,
TBD by mutual agreement Dielectric = polymide
The monolithic seeker is being designed as an array of 5500 transmit/receive (T/R)
modules mounted vertically through an aluminum substrate (Fig. 1). These modules are
driven by an integrated waveguide feed structure on one side of the substrate. An
interconnect network containing dc-bias and transmit/receive control lines is defined on
the other. The network is covered by the antenna ground plane and a diamond window. The
T/R modules are embedded in the substrate. The waveguide is filled with polymide to reduce
its size and provide rigidity for the electric field probe. In addition, the bonding of
the T/R modules to the interconnect network is conformal, requiring no bond wires. The
result should be a robust package weighing less than one hundred grams, which is amenable
to high throughput packaging techniques. The design of the structure is expected to evolve
during the design phase, but the basic structure serves as a model for developing an
approach to manufacturing.
The University of Houston has experience in lithography and processing. The advanced
lithography program was established five years ago at the University to address problems
of 3-dimensional integration in automotive electronics.
Contents
ISSO -- Institute for Space Systems Operations
1992-1993 Annual Report
