 |
Stellarator-Tools
|
|
Stellarator-Tools
|
Overview of the architecture of V3FIT.
Dedesign goals of V3FIT are to provide an easily extensible code for the purposes of fully three dimensional equilibirum reconstruction. To achieve this, V3FIT uses object oriented programming techniquies to create abstract away the implementation details. This abstraction allows extension of code without altering large parts of the code base.
This page provided an overview of the V3FIT architecture and describes how the various subsystems work together in a broad general sense. For specifics, consult the application programming interface (API) documentation.
V3FIT is written to be object oriented with five major classes. Two of the class, the equlibrium and signal classes, are designed to be subclassed and provide an abstract interface. While the model, parameter and reconstruction classes are designed as self contained objects to encapsilate the V3FIT functionality. An overview of the V3FIT object oriented programming pattern is documented in the Class Anatomy page.
The equilibrium class provides an abstract interface to abstract away the details of the equilibrium from the rest of V3FIT. All equilibrium solvers are implemented as a subclass of this class. The methods of the equilibrium class should be generally applicible to all equilibrium solvers and all access to the equilibrium should be performed through the equilibrium class. Currently, there are two subclasses of the equilibrium class implemented. The equilibrium coding is contained in the equilibrium module.
The VMEC equilibrium is an fully three dimensional equilibrium solver that assumed closed nested flux surfaces. The VMEC equilibrium is implemented in the vmec_equilibrium module.
The VACUUM equilibrium is a simple model that assumes there is no plasma. As a result, the VACUUM equilibrium is only useful for reconstructing based on magnetic signals. The vacuum equilibirum computes the magnetic fields generated external coils. This simplicity extends into the equilibrium coding as well. Since the vacuum doesn't contain the concept of temperature, density, pressure and other quanitied that are only relievent in the presence of a plasma, many methods maybe inherited from the equilibium class. The VACUUM equilibrium is implemented in the vacuum_equilibrium module.
The model class provides the primary interface signals to access the equilibrium model. The model contains everything that is not provided directly from the equilibrium but generic every equilibrium. The model class stores an instance of the equilibrium. The model is implemented in the model module.
The parameter class encapliates all the information and procdures to get, set, increment and contraint a reconstruction parameter. The parameter class conatines an ID obtained from the model that represents a reconstructable quantitiy. This quanitily may be subjected to a set of constraints. Access to the underlying quanitiy is handled through the model object's setter and getter routines. The parameter class is implemented in the v3fit_params module.
The signal class provides an abstract interface to abstract away the details of the signals from the rest of V3FIT. Signals are further subclassed into four categories. Diagnostics refer to experimentally measured quanities. Geometric signals refer to the shapes and positions of equilibrium. Priors refer a priori knowlegde about the system. Combination allow combinig any privious signal in to composite signal. The signal class is implemented in the signal module.
The diagnostic class is a generic super class to experimentally measured quantities. This class is subclassed by magnetic, sxrem, intpol, thomson and extcurz diagnostics. The diagnostic class is implemented in the diagnostic module.
The geometric class is a generic super class to geometric quantities. This class is subclassed by the limiter class. The geometric class is implemented in the geometric module.
The prior class is a generic super class to prior knowlegde quantities. This class is subclassed by the prior_gaussian class. Prior maybe defined on any quanitiy that me be simply looked up. If additional computation is required the signal should be added as a diagnostic or geometric depending on context. The prior class is implemented in the prior module.
Combination classes allow signals to be combined in multiple ways. The combination signal contains an array of signal pointers. Signals are combined by calling the get signal method on each pointed. The prior class is implemented in the signal module. Note, since this signal need to use the signal interface, the rule of one class per module needs to be broken in order to avoid a circular module dependency.
The reconstruction class encapsilated the reconstruction algrothium. The reconstruction class is implemented in the reconstruction module.
Other sections of the code are also implented as objects as well. The most important of these auxiliary objects is the context class. This class is a single object that contains all the memory and objects used by V3FIT. This context can be used to store, copy and reset the state of V3FIT. The context class is implemented in the v3fit_context module.
1.9.1