Whole-system Simulation of Solid RocketsIs Goal of ASCI Center at IllinoisMay 14, 1998
The types of models that will be used initially by CSAR researchers to simulate the individual components of a solid-propellant rocket, in itself a challenging problem.
In addition to its traditional role in basic research, computational simulation is playing an increasingly important role in the design and evaluation of a wide variety of manufactured products. Designing "virtual" products on computers has a number of advantages:
It is often faster and less expensive than building mock-ups or prototype versions of real products.
- It reduces the need to deal with potentially hazardous materials or processes.
- It permits a much more thorough exploration of the design space, as design parameters can be varied at will.
This is not to say, however, that effective computational simulation is easier than more conventional design methods. If anything, an even wider array of capabilities is required:
Accurate mathematical models, usually expressed in the form of equations, must be developed to describe the components and processes involved.
- Accurate, efficient algorithms must be developed for solving these equations numerically.
- Software must be developed to implement these algorithms on computers.
- Very high-speed, high-capacity computers are usually required to run the resulting simulation scenarios and produce numerical results.
- Computed results must be displayed in some comprehensible form, often through graphical visualization.
- Computed results must be validated by comparison with theory or experiment in known situations to ensure that the predictions can be trusted in new, previously untested situations.
Thus, computational simulation is an inherently multidisciplinary activity that requires considerable expertise in the relevant science and engineering, as well as in applied mathematics, numerical analysis, and computer science. Moreover, in the design of a complex, multicomponent system, such as an automobile or an airplane, many individual scientific and engineering disciplines come into play, such as fluid mechanics, solid mechanics, chemistry, and electronics. Simulation of such multicomponent devices requires a faithful capturing not only of the behavior of each individual component, but also of the interactions among components. To accomplish such a simulation, a sizable team of people with diverse technical expertise must work together in concert. The organizational challenges in managing a large-scale simulation effort often rival the scientific and technological challenges.
In Place of Conventional Testing
The U.S. Department of Energy established the Accelerated Strategic Computing Initiative (ASCI) to advance the state of the art in computational simulation of complex, multicomponent systems and to provide the computational hardware and infrastructure necessary to carry out very large-scale simulations. From DOE's perspective, the specific motivation is to ensure the safety and reliability of the U.S. nuclear weapons stockpile in an era when empirical testing of such weapons has been banned by international treaties. Replacing conventional testing by computational simulation, however, requires a giant leap in both simulation methodology and computational capacity. In addition to large research efforts at the DOE national laboratories, the ASCI program has also funded new centers of excellence in computational simulation at five U.S. universities. Each of the centers is focused on simulation in the context of a different physical problem, but they all share the theme of integrated, multidisciplinary research.
The Center for Simulation of Advanced Rockets (CSAR) at the University of Illinois at Urbana-Champaign (UIUC) is focused on detailed, integrated, whole-system simulation of solid-propellant rockets under both normal and abnormal operating conditions. A multidisciplinary team of engineers, physical scientists, and computer scientists is developing and implementing the necessary mathematical models, algorithms, and software to build a simulated "virtual rocket" that will make it possible to explore a variety of issues in rocket design, including analysis of various potential failure modes.
Comprehensive simulation provides a way to investigate technical issues in rocket design that is safer and less expensive than experimental trial and error. Solid rocket manufacturers have long recognized the current and potential benefits of computer simulation in rocket design. Recent breakthroughs in computational capability--most notably the teraflops--plus ASCI computing platforms, along with improved algorithms and software technology--have set the stage for a quantum advance in the state of the art in solid rocket performance simulation. Such a simulation capability will also have direct benefits for closely related technologies, such as gas generators used for fire suppression systems and automobile air bags, as well as many other technological design problems that involve complex components and require similar levels of system integration. Moreover, the computational capabilities developed to support this effort are applicable to a wide variety of important problems in computational science and engineering, such as fluid dynamics, combustion, and the failure of materials.
From Crude 1-D Modeling to Detailed 3-D Simulation
The basic idea behind a solid-propellant rocket motor is simple: Thrust arises from pressurization of a vented chamber by mass injection due to burning of the propellant. The detailed behavior is quite complicated, however, as the combustion rate depends on the chamber pressure, as well as on the surface area and storage temperature of the propellant. The particular shape of the solid propellant, called the propellant grain, determines the burning surface area, which in turn affects how the thrust varies over time: Progressive, regressive, and neutral profiles are possible. The propellant grain is usually not just a simple cylinder; slots and fins have often been created in the interior cavity to increase the surface area. The propellant surface regresses as propellant is consumed, however, so that the shape and area of the burning surface change dynamically with time.
The coupling and feedback between these variables can lead to instabilities. For example, the burning rate increases with the chamber pressure, and the chamber pressure increases with the burning rate. Relatively small defects can thus lead to catastrophic failure. A crack in the propellant, for example, causes an abrupt change in the surface area, and hence in the burning rate, which in turn causes an abrupt change in the pressure. Pressurization of the crack causes it to grow rapidly, possibly leading to premature burn-through to the rocket casing and catastrophic failure. Yet another potential failure mode is a possible transition from deflagration (normal surface burning) to detonation (abnormal for a solid rocket propellant), in which, perhaps due to pre-existing damage or compaction of the propellant, energy is released throughout a volume of the propellant, with fatal consequences.
Rocket design is further complicated by manufacturing and transportation constraints. Large boosters are manufactured in segments that are then assembled at the launch site, and the joints between segments are a potential source of failure. The technical disciplines involved in simulating a solid-propellant rocket include:
Ignition and combustion of composite energetic materials
- Solid mechanics of the propellant, case, insulation, and nozzle
- Fluid dynamics of the interior flow and exhaust plume
- Quantum chemistry and shock physics of energetic materials
- Damage and aging of components
- Analysis of potential failure modes
These problems are characterized by very high energy densities, extremely diverse length and time scales, dynamically changing geometries, complex interfaces, and reactive, turbulent, multi-phase flows. Because of the complexity of the components and their interactions, conventional rocket design techniques have largely been limited to relatively crude, quasi-one-dimensional modeling based on gross thermomechanical properties, and have relied heavily on experimental trial and error. We hope to remedy this situation by developing fully three-dimensional, high-resolution simulations of the major rocket components and their interactions, with sufficient fidelity to examine potential instabilities and failure modes in detail.
Simulation of each component is a challenging problem in its own right, both in terms of developing realistic models and in the computational capacity required to solve them accurately. Understanding the overall behavior of the entire rocket requires an integrated, whole-system simulation of the complex coupling and interactions among the various components. Interactions among component models must be supported at multiple levels: physical, mathematical, numerical, and software.
Algorithms and software must be developed to provide the following capabilities:
Geometric interface between different components and meshes
- Numerical interpolation between discretizations that differ in methodology and resolution
- Software interface between different data structures and code modules
- Data sharing among component codes
- Orchestration (e.g., resource allocation and management, load balancing)
Emphasis from the Outset on System Integration
To investigate the many difficult interface details and evaluate strategies for handling them, we are developing a prototype system simulation code: the GEN1 (generation 1) code, which allows simple component models to be integrated into a full-rocket simulation and supports simple component interactions. It also allows the component models to be refined to the extent possible given the level of interaction supported. Thus, the GEN1 code will evolve into a reasonably credible rocket simulation, but with limited capabilities. The GEN1 integration exercise will inform the development of the programming environments and orchestration language, and provide guidance for the design of a fully functional interface code. When implemented, this new interface will form the core of the GEN2 code, which will allow for more general modeling and the subscale models required for many failure scenarios. It will thus be a fully capable rocket simulation tool.
High-resolution, three-dimensional simulation of an entire solid rocket motor easily requires terascale computational capacity, which can be achieved only through the exploitation of parallel computers with several hundred to several thousand processors. Implementing a complex engineering code that scales efficiently to such levels of parallelism is a daunting task for any major component of a rocket, and is far more challenging for an integrated, whole-system simulation. Development of an integrated system simulation code of this scale requires solutions to a broad spectrum of problems in computer science, from algorithmic issues in numerical computation and large-scale parallelization to software engineering issues in large-scale code and data management and component integration. Research in computer science within CSAR includes the following topics:
Programming environments: an integrated environment to support multilingual, multiparadigm parallel programming
- Compilers: automatic parallelization and code optimization for shared-memory nodes within a distributed-memory architecture
- Performance analysis: symbolic performance scalability predictions and real-time adaptive resource control
- Parallel input/output: instrumentation and automated analysis of I/O patterns and development of adaptive I/O policies
- Computational mathematics: solvers and preconditioners for very large systems of equations
- Computational geometry: mesh generation for complex three-dimensional shapes, with good quality and adaptivity
The central goal of CSAR is to develop an integrated rocket simulation tool capable of detailed, whole-system simulation of solid-propellant rockets under both normal and abnormal operating conditions. These simulations must be sufficiently accurate and reliable that they can be used to design new solid-propellant rockets in untested situations, to investigate the causes of accidents and remedies for avoiding them, and to replace at least some costly experimental testing of rockets. Accomplishing this central goal requires advances across an array of technical issues in the individual disciplines involved, including reliable predictions in turbulent flow, combustion, detonation, material fracture, and other potential failure modes. It also requires significant improvements in system integration to deal with the complex interactions among components, especially in a highly parallel computing environment. Finally, substantial advances are required in the computational infrastructure for supporting complex, large-scale simulations on highly parallel computer architectures.
These ambitious goals cannot be achieved overnight. Full simulations of such complexity require a sequence of incremental developments over an extended period of time. From the outset, however, our emphasis is on system integration rather than separate threads of development that might eventually come together at some point in the future. Rapid exploration of critical system integration issues entails the use of simplified-but fully integrated-models and interfaces initially, to be followed by successively refined models and interfaces as experience is gained.
We have established a staged approach featuring two major generations of code development, GEN1 and GEN2. GEN1 is a prototype whole-system simulation code employing relatively simple component models and interactions; its main purpose is to provide feedback for subsequent redesign. GEN2 is intended to be a fully capable rocket simulation tool featuring detailed component models and complex interactions, and supporting sub-scale simulations. The design, implementation, and validation of GEN1 are expected to span years one through three; similarly, GEN2 will span years three through five of the five-year program initially funded. Additional features beyond our initial scope that could be incorporated later include the exterior aerodynamics of the rocket, chemical interaction of the exhaust with the atmosphere, and electronic instrumentation and controls.
Michael Heath is director of the Center for Simulation of Advanced Rockets at the University of Illinois at Urbana-Champaign.