|
University of Houston • University of Houston-Clear Lake • ISSO Annual Report Y2004 • 145-149
High-Performance Martian Space Radiation Mapping
Abstract--To ensure the health of space personnel and the cost-efficiency and safe-operation of space materiel and equipment, we performed a preliminary investigation to find ways to apply high-performance and parallel computer techniques to enable more speedy, accurate, efficient, and user-friendly space radiation studies--modeling, analysis, simulation, and visualization--for Mars missions and current and future space missions. The need to modernize NASA radiation codes is understood. We are currently obtaining the code and seeking support to restructure the computation to match newly developed computer resources of parallel, multithread, network cluster and reconfigurable parallel FPGA platforms. The success of applying parallel techniques to enhance deterministic HZETRN and stochastic FLUKA Monte Carlo radiation transport analysis/ simulation codes used by NASA scientists will greatly enhance space radiation understanding. This research is also vital to safer nuclear energy and anti-aging, cancer treatment with applications in the public sector.
To protect the health of space personnel and to ensure the safe-operation of space materiel and equipment, we carried out a preliminary investigation to find ways to apply high-performance and parallel computer techniques to enable more speedy, efficient, and accurate space radiation simulations for the analysis of Mars missions and and other current or future space missions. While studying the current execution mode and the platform of NASA HZETRN1 space radiation transport analysis methods, our contact at NASA Langley Research Center verified the "huge" needs to modernize the code to take advantage of the newly developed computer technologies that may run programs faster, more accurately, and with more user-friendly technology. We are currently obtaining the actual code and seeking support to (1) analyze parallel computation threads, (2) schedule computation threads to parallel processor resources, (3) update the core interpolation and integration routines of the code, and (4) port to MPP, cluster, or reconfigurable parallel FPGA platforms. While deterministic space radiation transport code such as HZETRN can produce faster, but approximated, results, it is imperative to also study the complementing, much more accurate, stochastic Monte Carlo space radiation transport simulation FLUKA code for improvement.
Mars Radiation
With any practical amount of shielding during Mars missions, every body cell in an
astronaut would still be traversed by at least one heavy ion and several hundred protons.
Such space radiation, primarily caused by Galactic Cosmic Rays (GCR) and Solar Energetic
Particles (SEP), is very distinct from terrestrial radiation (x-ray or gamma-rays) with
unknown risk to human health, as shown in Cucinotta.2 Unlike Earth, thin
atmosphere and lack of global magnetic field make Mars extremely vulnerable to harmful
radiation. Therefore, direct measurements of radiation levels (including the relative
contributions of protons, neutrons, and heavy ions) and of Martian atmospheric
characteristics are a pre-requisite for any human mission.
Space Radiation Risk
For over 35 years, NASA's radiation simulation and analysis group has long been burdened
with great responsibility to assess and certify mission duration for all astronauts and
equipment. Research to analyze, model, simulate, and visualize space radiation is critical
to predict human risk. Possible health risks from space radiation for long-term space
travel include cancer, cataracts, acute radiation sickness, hereditary effects, and damage
to the central nervous system.3 Understanding the space radiation environment
is essential for risk assessment of orbit and selection of the crew and provides a
scientific basis of countermeasures comprised of shielding materials, radio-protectants,
and pharmaceuticals.
Complex Radiation Computation
Research is being performed to mathematically model high-energy particles as they pass
through the walls of a spacecraft as well as parts of the human body. Data have been
collected over the years on the effects of space radiation on humans4 but to
understand the effects for many different cases, a mathematical model needs to be created
to simulate different conditions in space. As a complex computation written in FORTRAN,
NASA's current space radiation code is facing greater challenge than ever, comprising a
large amount of data gathered from Martian Rover daily in addition to the data collected
since 2001 by Mars Odyssey Orbiter.4 Comparable to scientific weather forecasting and
ocean simulation, the task of visualizing the space radiation map of the entire Martian
atmosphere plus the interplanetary space between Mars and Earth, is a tremendous
undertaking. Japanese attempts for ocean simulation resorted to a giant supercomputer that
occupied an area four times the size of a football field.
Parallel Radiation Analysis Enhancement
A promising key solution would be to restructure the computation to match the newly
developed resource of parallel, multithreading network cluster computers and
reconfigurable parallel FPGA platforms. Our preliminary investigation is to study the
current execution mode or platform of NASA space radiation code, to determine the
feasibility of finding a high-performance and parallel improvement that will bring us to
Mars sooner, at lower expense, and safely.
HZETRN Radiation Transport Analysis
Limited Speed and Accuracy--HZETRN is a Fortran program designed to model the
transport of high-Z and high-energy particles. NASA's MARIE (Martian Radiation Environment
Experiment) project uses HZETRN to predict the Mars radiation climate based on data
collected from the Mars Global Surveyor mission [MARIE].5 HZETRN is written in
Fortran77 for a VAX4000 machine, but is essentially machine and platform independent
[HZETRN, Appendix A]. This allows us to adapt HZETRN to parallel machine platform. The
current performance of the HZETRN program is apparently unsatisfactory. With a ratio of
about 1:3 CPU time needed for analysis vs. duration of data collection, analyzing 50 hours
of data requires 18 CPU hours on the current setup [TP-2000-210299].6 Data have
been collected from mid-March 2002 through October 2003. Assuming a similar level of
performance, as described in [TP-2000-210299], 4,212 CPU hours, or 175 CPU days, will be
needed to analyze about 11,700 hours (487.5 days) of the collected data. HZETRN also
underestimates radiation exposure by 15-30% [HZETRN, p. 32]. We need to investigate
hardware and software improvements to improve HZETRN's speed of execution, and thus allow
a more accurate radiation transport model that can accelerate the MARIE project and ensure
a higher level of confidence in the safety of the astronauts.7
FLUKA Monte Carlo Radiation Transport Simulation
FLUKA (FLUktuierende KAskede) Monte Carlo technique, developed since 1962, models space
radiation transport with much ease and accuracy with 3D capability and produces more
specific/detailed results.8 Similarly, a promising key solution would be to
restructure the computation to match the newly developed resource of (1) massively
parallel processors (MPP e.g., Cray T3E), (2) multithreading (e.g., Tera Multi-Threaded
Architecture MTA), (3) MPI, Open MP or PVM on Symmetric MultiProcessing (SMP) network
cluster (e.g., IBM SP-X or workstations), and (4) reconfigurable parallel FPGA computer
platforms (e.g., Hypercomputer). Computing with thousands of iterations but being
inherently parallel with minimal inter-thread communication in nature, Monte Carlo
research methodology is capable of producing near perfect parallel speedup and cost
efficiency shown in such similar applications as photon, charge, or magnetic transport.9
Radiation Impact--Dual-Use Energy
As earth's ozone depletion continues, space radiation study could lead to dual-use
countermeasures that will, in turn, protect human health from radiation and aging effects,
in general, such as treatment to slow cataract development. Other evolving critical
medical cures are being effected through proton cancer radiation treatment. With the help
of computer architects, scientists will be able to produce better radiation shielding to
make nuclear energy safer and more acceptable. This effort will, in turn, uncover the
unlimited resources of peaceful nuclear energy that can reduce world oil conflict and
global warming. With expertise in high-performance and parallel computing, university and
NASA-JSC researchers are devoted to protecting our astronauts and space gear to ensure the
success of our future space missions.
Approach--High-Performance/Parallel Space Radiation Transport Code
Preliminary Investigation for Improvement
Our study focuses on the radiation transport computation model describing the atomic and
nuclear reaction processes that alter GCR in their passage through Martian atmosphere,
materiel, and tissue. The current code solves the Botzmann equation using straight-ahead
and continuous slowing down approximation to calculate particle fluxes.4 A
preliminary study to restructure current space radiation HZTRN computation suggests the
following performance procedures:
To Provide the speed and accuracy needed for achieving NASA's goal in radiation study, reseachers need to complete the following tasks:
Preliminary Results--High-Performance HZETRN
High-Performance HZETRN--HZETRN models the path of space radiation particles by
solving the Boltzmann equation for the energy of a particle as it passes through various
mediums such as Mars' atmosphere or water shielding. Finding this solution requires
integration and interpolation. The current method of interpolation is Lagrange 3rd order.
We can pre-compute the intermediate steps of the Lagrange method and store them in a table
for reference rather than calculating them each time. Newton's divided difference method
yields good results when working with table data.
Depending on the application, other interpolation methods can also be evaluated for accuracy and speed. Integration is done by compound Simpson's with the midpoint interpolated again by Lagrange 3rd order [HZETRN, Appendix A]. This method of integration is a good tradeoff between speed and accuracy. Upon receipt of the HZETRN code, we will also analyze the code itself and will do a critical code execution analysis to find those sections of code that execute the most often (frequently) in order to find other numerical methods or functions that can be improved in accuracy and speed. Adapting HZETRN to parallel platform is also promising by distributing the workload by particles/types.
Language/Data Structure Options Explored--We could also upgrade the code to C++ or Fortran90. There is concern about using C++ in that "C math libraries are slower than the FORTRAN math libraries, except that a C++ wrapper is possibly acceptable by NASA, if it turns out to be warranted or needed," according to Dr. Singleterry of NASA Langley. Fortran90 is a version of Fortran that has been enhanced with emphasis on modernizing the language with features like dynamic memory allocation, more control structures, and operator overloading. Fortran90 is completely backwards compatible with Fortran77 for a smooth upgrade. We can improve the accuracy of HZETRN by updating the code with a simple improvement to increase the number of significant digits in two of the input data files. ATOMIC.DAT and NUCLEAR.DAT both have variables with only three significant digits in order to be compatible with generated data files [HZETRN, p.36].
Space Radiation Accuracy Saves Mission Cost--The high-charge-and-energy transport (HZETRN) model approximates the paths of high-Z and high-energy particles through space. It is used to model the effects and interaction of radiation with various shielding materials. These data are used to help determine the types of shielding necessary to protect astronauts and equipment during the missions to Mars. Any error in the method must be shown and compensated for in the physical shielding.
If the accuracy of the HZETRN code can be improved, overestimation of shielding can be reduced and the weight of the shielding can be reduced at a savings both in cost and in the initial construction since every pound counts on a space mission, especially a long-term mission to Mars. Improved radiation accuracy reduces excessive compensating weight of materiel, thus cutting mission costs.
Speedup Interpolation/Integration--Improved accuracy alone is not enough. The calculation must be done in a timely manner so that the results can actually be used. It does no good to improve accuracy if the results take too long to compute. As mentioned previously, the HZETRN project relies heavily on interpolation and integration. In fact, the integration routine incorporates interpolation. Therefore, if interpolation can be improved, both routines will execute faster and if both can be improved, the code as a whole will execute even faster. Because the HZETRN code uses 3rd order Lagrange interpolation and compound Simpson's method for integration, we are currently attempting to implement both in an FPGA format. Integration will be accomplished by taking one interval and splitting it into two regions, i.e., by interpolating two extra points in addition to the three points given. Parallel routines are also worth trying, e.g., Clarke method and Optimal interpolation for parallel interpolation, as well as Dynamic synchronous and Adaptive algorithm for parallel integration.
HZETRN Experience--Recently, we corresponded with Dr. Cucinotta who informed us that we could run the HZETRN code from the SIREST (<http://sirest.larc.nasa.gov>) system. We ran several test jobs and found that the TransportMARS application calculates the particle fluxes in a given environment on Mars. The application took 5 hours, 16 minutes, and 47 seconds to process the data from March 1, 1990 to March 7, 1990. (The process started at 14:46:30, Aug 23, 2004; ended 20:03:17, Aug 23, 2004). Without access to the computer itself, we cannot know how much CPU time this represents. The output is odd as well; it does not give a timeline and then the flux for each type of particle over that timeline; it gives a total particle flux over the whole time period. Because of NASA ‘s new security procedure, we are currently working with LaRC to obtain the actual code for analysis.
Conclusion--Safer/Cheaper/Sooner Mission
We performed a study on the current execution mode or platform for NASA space radiation
methods to verify the needs and to propose several options of high-performance and
parallel improvement that will bring us to Mars sooner, more economically, and more
safely. The study of the feasibility of applying parallel technique to enhance both HZETRN
and the FLUKA Monte Carlo radiation transport analysis/simulation code used by NASA
scientists will greatly enhance space radiation study, which can also prove vital in
studies of safe nuclear energy, anti-aging processes, and cancer treatment. This research
can bring about health and peace through new technology advancements.
Personnel
Professor Liwen Shih teaches High-Performance Computer Architecture, Machine Intelligence
and Parallel Processing. Dr. Shih and her students successfully improved the performance
of Scientific/Image modeling, including the speedup of Medical Image Texture Analysis from
8 days/image CPU time on a VAX Micro 3400 to 40 minutes/image on TMC CM5, and to within 15
seconds/image on a PC platform. They expect to apply their high-performance computing
expertise to projected space expeditions to the moon and Mars.
Acknowledgments
Dr. Francis A. Cucinotta of NASA-JSC informed us about feasibility of running the HZETRN
code from the SIREST (<http://sirest.larc.nasa.gov>)
system. Two researchers currently work with us and will continue to assist us in obtaining
radiation transport algorithms and documentation following the new NASA security
guidelines: Dr. Robert C. Singleterry Jr., NASA Aerospace Fellow, Research and Technology
Directorate, the Computational Structures & Materials Branch of the NASA Langley
Research Center, , and Dr. Premkumar Saganti, Associate Professor in Physics and Research
Professor of the NASA Center for Applied Radiation Research (CARR) at Prairie View A&M
University and Sr. Research Scientist of Space Radiation Health Project (SRHP) NASA-JSC,
on the NASA-JSC MARIE Team. The School of Science and Computer Engineering, UHCL, provides
us space and computer needs. Many national supercomputing centers, e.g., the Maui
High-Performance Computing Center (MHPCC), will provide us remote access accounts if we
need to use a large scaled computer system in future work.
References
1J. W. Wilson, F. F. Badavi, F. A. Cucinotta, J. L. Shinn, G. D. Badhwar, R.
Silberberg, C. H. Tsao, Lawrence W. Townsend, and R. K. Tripathi, "HZETRN:
Description of a Free-Space Ion and Nucleon Transport and Shielding Computer
Program," NASA Technical Paper 3495, May, 1995 <http://techreports.larc.nasa.gov/ltrs/PDF/NASA-95-tp3495.pdf>.
2F. A. Cucinotta, "Space Radiation Biology," NASA-M. D. Anderson
Cancer Center Mini-Retreat, Jan. 25, 2002 <http://advtech.jsc.nasa.gov/presentation_portal.shtm>.
3Space Radiation Health Project, May 3, 2005, NASA-JSC, March 7, 2005 <http://srhp.jsc.nasa.gov/>
4P. B. Saganti, F. A. Cucinotta, J. W. Wilson, L. C. Simonsen, and C. Zeitlin,
"Radition Climate Map for Analyzing Risks to Astronauts on the Mars Surface from
GCR," 2002 <http://marie.jsc.nasa.gov/Documents/Mars-Flux-Paper.pdf>.
5Martian Radiation Environment Experiment [MARIE], Oct. 27, 2003, NASA-JSC,
March 7, 2005 <http://marie.jsc.nasa.gov/>.
6M. S. Clowdsley, J. W. Wilson, J. H. Heinbockel, R. K. Tripathi, R. C.
Singleterry, Jr., and J. L. Shinn, "An Improved Elastic and Nonelastic Neutron
Transport Algorithm for Space Radiation," NASA/TP-2000-210299, July, 2000 <http://techreports.larc.nasa.gov/ltrs/PDF/2000/tp/NASA-2000-tp210299.pdf>.
7P. B. Saganti, F. A. Cucinotta, J. W. Wilson, and W. Schimmerling,
"Visualization of Particle Flux in the Human body on the Surface of Mars," March
31, 2003, <http://marie.jsc.nasa.gov/Documents/PS-Nara-Paper.pdf>.
8L. S. Pinsky, T. L. Wilson, A. Ferrari, P. Sala, F. Carminati, and R. Brun,
"Development of a Space Radiation Monte Carlo Computer Simulation Based on the Fluka
and Root Codes," First International Workshop on Space Radiation Research and 11th
Annual NASA Space Radiation Health Investigators' Workshop, Arona, Italy, May 27-31, 2000.
86-89.
9A. Majumdar, "Parallel Performance Study of Monte Carlo Photon Transport
Code on Shared-, Distributed-, and Distributed Shared-Memory Architectures," San
Diego Supercomputing Center, UC San Diego, Feb. 7, 2000 <http://www.sdsc.edu/~majumdar/ipdps00.pdf>.
Publications
Gilbert, T. and L. Shih. "High-Performance Martian Space Radiation Mapping,"
IEEE/ACM/UHCL Computer Application Conference, University of Houston-Clear Lake, Houston,
TX, April 29, 2005.
Kadari, A.. S. Kodali, T. Gilbert, and L. Shih. "Space Radiation Analysis with
FPGA," IEEE/ACM/UHCL Computer Application Conference, University of Houston-Clear
Lake, Houston, TX, April 29, 2005.
PDF (KB)
Table of Contents
Institute for Space Systems Operations - Y2004 Annual
Report
Copyright © 2005
|
