Stellar operations12 September 2005
The National Compact Stellarator Experiment is a marriage of science, engineering and compute-optimised design in fusion research.
At the USA’s Princeton Plasma Physics Laboratory (PPPL), the National Compact Stellarator Experiment (NCSX) marks the beginning of a new era in the marriage of science, engineering and computer-optimsed design. “NCSX is really a test of our ability to design an optimum magnetic fusion configuration based on physics performance,” project director Hutch Neilson told NEI.
Magnetic fusion research has focused on finding the best shape for hot reacting plasmas and the magnetic fields that hold them in place. Dramatic advances in physics, engineering and computation over the last 20 years have yielded a promising new configuration – the compact stellarator, which combines features of the tokamak, the leading magnetic fusion configuration, and classic stellarators.
NCSX is the centrepiece of the US effort to optimise the physics and engineering and determine whether the compact stellarator offers the basis for an attractive future fusion power reactor. As a proof-of-principle experiment, it was designed to bring together in one device recent developments in both stellarators and tokamaks, said Michael Zarnstorff, head of the physics team and leader of the computer optimisation effort that was crucial to the design work.
Two Department of Energy (DoE) laboratories, PPPL in New Jersey and the Oak Ridge National Laboratory in Tennessee, led the design effort. Once NCSX moves into the experimental phase, an additional 20-30 partners, including laboratories and universities in the USA and abroad, are expected to participate in the 10-year research effort. First plasma is scheduled for May 2009.
Challenges Boost Potential
Although one of the strengths of the stellarator design (both classic and compact) is its inherent steady-state operation, NCSX will operate in one-to-two second pulses, rather than steady state, to minimise the expense of operations, Neilson said. However, steady state operation is not needed for the NCSX experimental research goals. Two much-larger classic stellarators will provide experience with near steady-state operation in stellarators: Japan’s Large Helical Device (LHD), which began operating in 1998, and Germany’s Wendelstein 7-X (W7X), which is under construction. The LHD recently announced experiments lasting 30 minutes.
If NCSX performs as expected, the next-step device will be a larger, more performance-oriented machine that will introduce the basic level of burning plasma science into the stellarator configuration. Understanding gained from the next-step tokamak, the International Thermonuclear Experimental Reactor (Iter), as well as from NCSX, LHD and W7X, will be incorporated into future stellarator designs.
The most challenging aspect of NCSX is its complex, three-dimensional geometry coupled with the need to maintain high dimensional accuracy in its magnets in order to achieve the magnetic field conditions required for good plasma confinement, said Philip Heitzenroeder, deputy project manager for engineering and technical representative for the modular coil winding forms. “State-of-the-art computer aided design (CAD) and computer aided manufacturing (CAM) were essential to handle the challenge presented by NCSX’s geometry.”
By necessity, the design team used a concurrent engineering approach in which the overall project needs – including physics, engineering and manufacturing – were developed simultaneously in a balanced approach. The approach “worked well and contributed significantly to the NCSX development effort,” said Heitzenroeder.
The $86.3 million NSCX will have a plasma major radius of 1.4m and a cross-section that varies periodically around the plasma three times. The nominal beta (ratio of plasma pressure to magnetic field) will average 4%, with an aspect ratio (ratio of the major radius to the average plasma radius) of 4.4. Magnetic field strength will range from 1.2 to 2.0 Tesla. Neutral beams and radiofrequency heating will ultimately supply up to 12MW of plasma heating.
NCSX’s twisting, helical magnetic field is dominantly produced by 18 modular coils in three coil shapes. In addition, the NCSX design includes 18 toroidal field coils and six pairs of poloidal field coils located symmetrically around the horizontal midplane to provide experimental flexibility and control the plasma location and shape. Trim coils will compensate assembly inaccuracies.
On 6 October 2004, PPPL awarded the first of two long-lead-time component contracts, worth a total of $12.5 million. The first award went to a team led by Energy Industries of Ohio for fabrication of the structural winding forms that will be used to mount the complex modular electromagnetic coils. The winding forms will be manufactured between October 2004 and September 2006. The second contract went to Major Tool and Machine of Indiana (which also is a member of the Energy Industries team) for fabrication of the vacuum vessel. Major Tool and Machine will deliver the vacuum vessel, which will be fabricated from commercially available Inconel 625, in segments between October and December 2005.
The modular coils and the vacuum vessel are the “most challenging and critical components of NCSX,” PPPL director Robert Goldston said at the time the awards were made.
Contracts are yet to be awarded for two other major components: the toroidal and poloidal coils. The PPPL-Oak Ridge team is finishing design of the toroidal coils and expects to issue a fabrication contract in April or May of 2005. This will be the only major contract (more than $1 million) awarded in 2005.
Fabrication of the poloidal coils will be the last major NCSX procurement, and may be issued either as one or several contracts. Awards are expected in late 2006 or early 2007.
The Promise of STELLARATOR Physics
The holy grail of fusion science is a magnetic configuration that is both inherently stable and in which the plasma temperature can be reliably sustained by power from the fusion reaction itself.
In recent years, the most widely studied fusion concept has been the tokamak, which bends the plasma into a toroidal (doughnut) shape for achieving reactor-level plasma parameters for a short time. The stellarator, invented in the 1950s by Princeton astrophysicist Lyman Spitzer, has offered a very fruitful line of complementary research.
In a tokamak, the cross section of the plasma remains the same around the torus, giving it a two-dimensional geometry. In stellarators, the cross section varies, depending on where the torus is ‘sliced,’ giving it a helical three-dimensional geometry. This third dimension allows physicists additional degrees of freedom that they can use to tailor the plasma shape to obtain desirable physical properties.
In the 1990s, researchers began to study new stellarator designs with much lower aspect ratios than classical stellarators. One Princeton fact sheet describes the plasmas produced by these new compact stellarators as looking more like truck tyres than the bicycle-tyre plasma shape produced by traditional stellarators.
Although compact stellarators have a three-dimensional geometry, their magnetic field structure is designed to be ‘quasi-symmetric,’ which makes them tokamak-like in their underlying physics and confinement performance characteristics. Researchers expect these new designs to have higher power density, and therefore more economical operation, than classical stellarators.
Allen Boozer, now at Columbia University, and German physicist Juergen Nuhrenberg developed the theory behind quasi-symmetry in the early 1980s, however its application to stellarators had to await the development of highly sophisticated computer codes and access to massively parallel computing capability.
Boozer’s work demonstrated that the magnetic configuration did not have to be symmetric visually, but rather that variations in magnetic field strength needed to be made symmetric in their flux coordinates. “If you could make the magnetic field strength quasi-symmetric, then you would have the same plasma confinement properties as actual symmetry,” Zarnstorff explained.
Physicists have developed three strategies for optimising stellarator performance: quasi-axial symmetry, which will be used by NCSX; quasi-helical symmetry, which is used on the Helically Symmetric Experiment at the University of Wisconsin; and quasi-poloidal symmetry which will be used on Wendelstein 7-X and on a proposed small stellarator at Oak Ridge National Laboratory. The related ‘sigma-optimised’ strategy is used on Japan’s JHD.
Each of these physics strategies has different implications for the rest of the machine’s physics properties, Zarnstorff said, pointing out that, worldwide, the stellarator community is conducting a very synergistic set of experiments. “We are exploring the physics consequences of making different choices; at this point, we are not sure which will be the best.
“What PPPL has done is design a device with the same transport properties as a tokamak, but with capabilities that cannot be achieved in a tokamak,” such as better plasma stability and less susceptibility to heat leakage. The compact stellarator, like other stellarators, is capable of creating self-sustaining plasmas, and has been designed to minimise ripple and plasma disruptions that plagued earlier designs. These advances only were possible because of huge recent developments in both theory and high-throughput computation.
“We are not done changing” either the theory or the computation ability, Zarnstorff stressed. “Ten years from now, we will know more things and be able to produce better, more comprehensive models.” These models will include the lessons learned from NCSX.
In every experiment, some things have worked and some things have not, but all results have led to new discoveries and revisions to the underlying theory, he explained. Some experiments have revealed entirely new phenomena, particularly in the area of plasma turbulence.
This new information will be incorporated into newer, more sophisticated computer models to allow design of future experiments, and ultimately, design of the next-step device.
Testing and confirmation of the physics theories will begin during the first year of operations.
Computer Support Comes Of Age
The use of massively parallel computers to simultaneously optimise three tremendously complex aspects of a fusion device – physics characteristics, machine design and manufacturability of components – is one of the major innovations of NCSX, said Neilson. “There is a lot of excitement about the power of advanced computation to advance fusion research.” NCSX will determine whether computer optimisation is in the league of importance with theory and experimentation. “We know it as a powerful tool, now the question is whether it is more than that,” whether it can significantly speed up progress in fusion energy. “Computation will not supplant experimentation, but it may allow you to take bigger steps and to more precisely target the results you want to achieve.”
NCSX could not have been designed and built without this unprecedented degree of computer optimisation, and simultaneous modeling of both the physics and machine engineering. It is a “much larger” step in stellarator development than anyone would have dared to take without the first-of-a-kind computer optimisation resources, Neilson explained.
The use of massively parallel computers and new computer codes allowed the design team to calculate the optimum shape of the coils and plasma, starting with a set of physics objectives. For example, researchers wanted a stable beta in the range of 4-4.5%. They also wanted to minimise ripple, a magnetic field imperfection that has plagued stellarators for years, and to minimise the aspect ratio to the extent possible with magnets that could actually be built.
Hundreds of thousands of plasma simulations were used to develop the best machine design. “This is the equivalent of doing hundreds of thousands of experiments over the course of a decade or more, only we were able to produce the simulations in about three years,” Neilson said. Physics optimisation, which began in 1998 and ran through 2002, was a very exciting period in NCSX development, he added.
Since NCSX requires a highly shaped vacuum vessel and modular coils, CAD and computer optimisation were “absolutely essential to not only develop these key components, but to be able to assure that the entire stellarator core, consisting of the vacuum vessel, modular coils, toroidal field coils, poloidal field coils, field correction coils and structure can be assembled” and operated properly, said Heitzenroeder.
He went on to explain that the only way to develop the extremely complex NCSX design was to use a “concurrent engineering” approach where the overall needs, including physics, engineering, and manufacturing, were developed simultaneously in a balanced way. “NCSX was driven to the concurrent engineering approach by necessity.” As examples, he said that it would not be possible to design the vacuum vessel without considering the tightly fitting modular coils and the other components and systems that must fit around it. In the same way, it would not be possible to design the winding forms without verifying that they could actually be fabricated at an acceptable cost and within a reasonable time period.
The design team called on computer resources not only from PPPL and Oak Ridge, but from other DoE and university laboratories, including Los Alamos National Laboratory, the National Energy Research Scientific Computing Center at the University of California-Berkeley, and computer centres at University of Wisconsin, and Columbia and Auburn universities. Every step along the way, they developed both the design and new computer codes to improve the design further. “We have been publishing papers every year… at the American Physical Society, at the biennial International Fusion Energy Conference, in a number of journals,” Neilson said.
Art + Physics = Magnet Design
Computer optimisation allowed the design team to push the state-of-the-art in magnet technology with the design and fabrication of the modular coils. The extremely complex geometry of these coils is characterised by very tight bends and very high accuracy requirements.
In order to design these coils in a way that could actually be built, as opposed to simply producing animated designs on a computer screen, the team had to develop a new level of sophistication in computer analysis and computer-aided design. The demanding coil requirements also required the fabrication team to develop new materials specifically for these magnets.
The energy industries group also had to devise a sophisticated casting process to make fabrication of the modular magnet forms possible. The process uses high-level computer flow solidification modeling of the casting to assure that the coils meet stringent requirements.
Heitzenroeder described the massive iterative team effort that went into the computer optimisation of the coil designs. The team needed “to develop designs that simultaneously would meet physics requirements such as plasma volume and shape; diagnostic access; magnetic field strength, geometry, uniformity and pulse characteristics; engineering constraints such as current density, stress, deflection, conductor bend radii and twist limitations; and manufacturing constraints such as shape, size, and tolerance limitations, and in the case of the winding forms, having adequate reach and access for the five-axis numerically controlled machining.”
All of this had to be achieved “along with being able to manufacture the parts on a reasonable schedule and at affordable costs,” Heitzenroeder added.
Developing the three-dimensional CAD and CAM programs required for the complex geometry of the modular coils proved to be very challenging and more time consuming than originally estimated, said Heitzenroeder.
Current filaments that satisfied physics requirements were used to develop the basic winding and structural support details. This required a number of iterations to resolve geometry problems such as coil overlaps, excessive twists, or excessively sharp bend radii.
Once coil geometry was established, surfaces were developed to provide a stiff structural ‘shell’ for the coil that would maintain deflections within limits. Ultimately, this shell was divided into 18 segments, one per modular coil, which permits the shell structure to also serve as a winding form for the coil.
With this design, the conductors are directly wound onto the winding form. The smaller shell segments also made it easier to develop a design with smooth surfaces, which aided the casting process for the modular coil winding forms.
Once the CAD models were developed with satisfactory surfaces, they were passed on to the industrial manufacturers for review. The manufacturers checked for adequate reach and access for machining tools and machining details. This information was then fed back into the design and used to refine the models further.
The refined models were used to develop computer models for the casting patterns. This required an overall adjustment of the dimensions to compensate for temperature, and ‘padding’ of cast surfaces to allow for machining cleanup.
The adjusted model was used to perform flow/solidification analyses to verify the feasibility of casting the winding forms and to determine casting risers and gating details.
“Because of the complexity, this was also an iterative process requiring multiple runs in which the risers and gating were modified until satisfactory casting pour parameters and uniform cool-down without shrinkage could be achieved,” Heitzenroeder said.
A strong magnet R&D programme, which included a series of increasingly complex winding studies and industrial fabrication of prototypes of a modular coil winding form, provided feedback and guidance throughout the development process. Heitzenroeder called this R&D “an extremely important part of the effort.”
The design team also had to develop a custom steel alloy for the modular coil winding forms, which could be useful in other fusion applications that require castings with low magnetic permeability. The alloy is a variant of CF8M stainless steel with a magnetic permeability of less than 1.01 and which develops the necessary mechanical properties without the need for water quenching.
DEMO Could Be a Stellarator-Like Device
In the future, fusion researchers will be able to design their experiments to an unprecedented degree on a computer-based ‘virtual NCSX’ before ever running tests on the real device, Zarnstorff said. These virtual experiments will be based on work done during the design optimisation, constantly enhanced by new information gained during NCSX operations. “Modelling beforehand is crucial to achieving the properties you really want to get,” and it will be much easier using the tools developed during the design phase.
NCSX, like many experimental fusion devices, will have an operating lifetime of about a decade. After that, “it makes more sense to build a new device [that incorporates the knowledge gained] than to struggle to modify the old machine,” Zarnstorff said.
In many ways, the international fusion community is thinking of NCSX, and stellarators in general, as a way of improving understanding of fusion physics, particularly of three-dimensional plasma physics, and attacking problems with an eye to the stage beyond Iter, the next-generation tokamak to be built by an international consortium. “Iter will explore burning plasma physics, but it is not viewed as a prototype reactor,” Zarnstorff explained.
Avoiding plasma disruptions and developing an easy and inexpensive way to sustain plasma by external coils goes beyond Iter to the next stage – a demonstration reactor, or Demo project. The understanding of fusion physics will be broadly applicable to both tokamaks and stellarators. “Depending on that understanding, Demo could be a stellarator-type facility.’
By Thecla Fabian