Systems and methods for nuclear fusion reactor design

Abstract

Disclosed herein are methods, systems, media, and techniques for training machine learning models to generate nuclear fusion reactor configurations. Further disclosed herein are systems, methods, computer-readable media, and techniques for configuring a stellarator, including: (a) obtaining one or more parameters of a target stellarator; (b) generating a plurality of target stellarator approximations based at least in part on the one or more parameters of the target stellarator, wherein at least a subset of the plurality target stellarator approximations: (i) have a toroidal profile, and (ii) comprise a plurality of concentric toroids; and (c) presenting at a graphical user interface a representation of the plurality of target stellarator approximations.

Description

SYSTEMS AND METHODS FOR NUCLEAR FUSION REACTOR DESIGNCROSS-REFERENCE

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 563,825 filed March 11, 2024, U.S. Provisional Application No. 63 / 673,635, filed July 19, 2024, U.S.Provisional Application No. 63 / 675,633, filed July 25, 2024, and U.S. Provisional Application No. 63 / 727,583, filed December 3, 2024.BACKGROUND

[0002] A stellarator is a plasma device that relies primarily on external magnets to confine a plasma. Scientists researching magnetic confinement fusion aim to use stellarator devices as a vessel for nuclear fusion reactions. The selection of a configuration for a nuclear fusion reactor may incorporate physics and engineering considerations.SUMMARY

[0003] Disclosed herein are methods, systems, media, and techniques for the training and inference of machine learning models for the design of nuclear fusion reactors. Current approaches to designing nuclear fusion reactors are computationally expensive yet lacking in robustness and do not provide for inverse design. Further, these approaches are not amenable to machine learning due to the challenge of using or representing computational or mathematical models in a manner suitable for input into a machine learning model. As such, there is a long felt need in the field of nuclear fusion for machine learning strategies to facilitate reactor design.

[0004] The selection of a configuration for a nuclear fusion reactor is a non-trivial problem. Physics and engineering considerations must be balanced to ensure both adequate conditions for fusion to occur while factoring in the boundaries imposed by real-world engineering. To determine a configuration worthy of reducing to a functioning device, mathematical optimization of nuclear fusion reactors is typically considered of paramount importance.However, the optimization of nuclear fusion reactors is a complex mathematical problem that requires exploration of a large param eter / soluti on space and powerful computers often running for hours or days to complete one calculation. Given the high dimensionality of the optimization space, existing techniques follow a trial -and-error approach, wherein costly optimization algorithms based on first-principles physics are launched from a plurality of distinct starting points. These starting points are then subject to mathematical optimization techniques that are neither guaranteed to find an optima nor are guaranteed to provide a useful result. This approachis both time consuming and computationally demanding while providing no clear route to a reasonable reactor design.

[0005] The instant techniques forego reliance on trial-and-error optimization of mathematical models and instead provide techniques for the generation of nuclear fusion reactor configurations from desired physics properties. For example, a user may provide shape metrics, magnetohydrodynamic (MHD) stability metrics, neoclassical transport metrics, energetic particle confinement metrics, turbulent transport metrics, coil design metrics, magnetic field metrics, or other desired properties of a nuclear fusion reactor to a machine learning model and receive a nuclear fusion reactor configuration satisfying the desired properties. In some cases, the configuration may comprise a 2D or 3D spatial representation of the nuclear fusion reactor. In some cases, the configuration may comprise a plurality of descriptors sufficient to model, fabricate components for, or assemble the nuclear fusion reactor. For example, the machine learning model may output a plurality of a nuclear fusion reactor cross sections, a pressure profile, a rotational transform profile, a current profile, a total magnetic flux, or any combination thereof such that a complete description of the nuclear fusion reactor configuration is attained. In some cases, the completeness of a configuration may be described by its applicability to MHD equilibrium solvers.

[0006] Accordingly, the methods, systems, media, and techniques disclosed herein provide teachings that leverage the power of machine learning to facilitate nuclear fusion reactor design. By training machine learning models to provide aspects of nuclear fusion reactor configurations from desired reactor properties, the present disclosure foregoes the limitations of existing techniques that introduce computational bottlenecks into the design of nuclear fusion reactors. Further, the techniques disclosed herein address existing problems encountered in machine learning aided design of nuclear fusion reactors. Conventional approaches aim to input into a machine learning model the governing physics of a mathematical model directly (e.g., physics informed neural networks) or indirectly via reduction of the governing physics to some other representation. However, this approach is fraught in the context of nuclear fusion reactor design as the problem is particularly complex and sensitive. The complexity consideration limits the ability to represent the problem in a machine learning amenable manner, while the sensitivity consideration limits the ability of the trained machine learning model to provide accurate predictions. For example, minor changes in the physical description of a nuclear fusion reactor can have drastic implications in the resultant reactor performance. Machine learning models struggle to capture such intricacies during the learning of a particular latent space, in which drastic differences in output for minor differences in input are typically penalized. As such, thepresent disclosure represents an advance in the process of nuclear fusion reactor design and allows for the computationally efficient inverse design of nuclear fusion reactors. Further, the rapid generation and evaluation of reactor designs using machine learning represents an advancement in the field of clean fusion energy as a whole.

[0007] In one aspect disclosed herein is a computer-implemented method for training a machine learning model to predict a configuration of a nuclear fusion reactor, comprising: (a) obtaining a dataset comprising a shape metric, a magnetohydrodynamic (MHD) stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric; and (b) using the dataset, training the machine learning model to predict the configuration of the nuclear fusion reactor.

[0008] In some embodiments, the dataset comprises a shape metric. In some embodiments, the dataset comprises an MHD stability metric. In some embodiments, the dataset comprises a neoclassical transport metric. In some embodiments, the dataset comprises an energetic particle confinement metric. In some embodiments, the dataset comprises a turbulent transport metric. In some embodiments, the dataset comprises a magnetic field metric. In some embodiments, the dataset comprises a coil design metric. In some embodiments, the method further comprises assembling the nuclear fusion reactor based on the configuration of the nuclear fusion reactor predicted in (b). In some embodiments, the method further comprises causing fabrication of a component of the nuclear fusion reactor based on the configuration of the nuclear fusion reactor predicted in (b). In some embodiments, the component comprises a coil, a blanket, a divertor, a first wall, a heat shield, an outer vessel, or a plasma vessel. In some embodiments, the method further comprises: (c) generating an updated configuration of the configuration of the nuclear fusion reactor using a fine-tuning module; (d) obtaining a second dataset comprising the updated configuration; and (e) using the second dataset, training the machine learning model to predict the updated configuration of the configuration of the nuclear fusion reactor. In some embodiments, the fine-tuning module comprises a physics solver or a second machine learning model. In some embodiments, the nuclear fusion reactor comprises a stellarator. In some embodiments, the nuclear fusion reactor comprises a tokamak. In some embodiments, the configuration of the nuclear fusion reactor comprises a magnetic confinement fusion (MCF) device. In some embodiments, the configuration of the nuclear fusion reactor comprises a pressure profile, a cross section of the nuclear fusion reactor, a total magnetic flux, a rotational transform profile, or a current profile. In some embodiments, a datum of the dataset comprises a feature portion comprising the shape metric, the MHD stability metric, the neoclassical transport metric, the energetic particle confinement metric, the turbulent transport metric, the magneticfi eld metric, or the coil design metric, and wherein the datum further comprises a label portion comprising the configuration of the nuclear fusion reactor. In some embodiments, the configuration of the nuclear fusion reactor comprises a two-dimensional or three-dimensional spatial representation of the nuclear fusion reactor. In some embodiments, the two-dimensional or three-dimensional spatial representation of the nuclear fusion reactor comprises a model of the nuclear fusion reactor, an image of the nuclear fusion reactor, a rendering of the nuclear fusion reactor, or a latent representation of the nuclear fusion reactor. In some embodiments, the dataset comprises a plurality of simulated cross sections of the nuclear fusion reactor. In some embodiments, a first cross section of the plurality of simulated cross sections comprises a first toroidal angle and a second surface of the plurality of simulated cross sections comprises a second toroidal angle, and wherein the first toroidal angle and the second toroidal angle differ by at least about 1 degree. In some embodiments, the method further comprises: (f) obtaining a parent dataset; (g) grouping the parent dataset into a plurality of groups based on a physics principle; and (h) clustering a group of the plurality of groups into a plurality of clusters using a clustering model, wherein a cluster of the plurality of clusters comprises the dataset. In some embodiments, the physics principle comprises a type of symmetry, a number of field periods, or an edge rotational transform.

[0009] In one aspect disclosed herein is a computer-implemented method for generating a configuration of a nuclear fusion reactor, comprising: (a) obtaining a dataset comprising a shape metric, a magnetohydrodynamic stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric; and (b) generating, using the dataset as input into a machine learning model, the configuration of the nuclear fusion reactor.

[0010] In some embodiments, the method further comprises, (c) performing operations (a) and (b) in an optimization framework. In some embodiments, the optimization framework comprises a loss function configured to optimize the configuration of the nuclear fusion reactor. In some embodiments, the optimization framework comprises reinforcement learning. In some embodiments, the optimization framework queries a computational model that outputs the configuration of the nuclear fusion reactor. In some embodiments, the dataset comprises a shape metric. In some embodiments, the dataset comprises an MHD stability metric. In some embodiments, the dataset comprises a neoclassical transport metric. In some embodiments, the dataset comprises an energetic particle confinement metric. In some embodiments, the dataset comprises a turbulent transport metric. In some embodiments, the dataset comprises a magnetic field metric. In some embodiments, the dataset comprises a coil design metric. In someembodiments, the method further comprises assembling the nuclear fusion reactor based on the configuration of the nuclear fusion reactor predicted in (b). In some embodiments, the method further comprises causing fabrication of a component of the nuclear fusion reactor based on the configuration of the nuclear fusion reactor predicted in (b). In some embodiments, the component comprises a coil, a blanket, a heat shield, an outer vessel, or a plasma vessel. In some embodiments, the configuration of the nuclear fusion reactor is used as input into a second machine learning model trained to fine-tune the configuration of the nuclear fusion reactor. In some embodiments, the machine learning model is selected by a model selection pipeline, wherein the model selection pipeline is configured to perform operations comprising: (i) classifying the dataset based on a physics principle, (ii) classifying the dataset based on a clustering model, and (iii) identifying the machine learning model based on (i) and (ii). In some embodiments, the physics principle comprises a type of symmetry, a number of field periods, or an edge rotational transform. In some embodiments, the nuclear fusion reactor comprises a stellarator. In some embodiments, the nuclear fusion reactor comprises a tokamak. In some embodiments, the configuration of the nuclear fusion reactor comprises a magnetic confinement fusion (MCF) device. In some embodiments, the configuration of the nuclear fusion reactor comprises one or more of a pressure profile, a total magnetic flux, a cross section of the nuclear fusion reactor, a rotational transform profile, or a current profile. In some embodiments, a datum of the dataset comprises a feature portion comprising the shape metric, the MHD stability metric, the neoclassical transport metric, the energetic particle confinement metric, the turbulent transport metric, the magnetic field metric, or the coil design metric, and wherein the datum further comprises a label portion comprising a nuclear fusion reactor cross section, a pressure profile, current profile, a rotational transform profile, or a total magnetic flux. In some embodiments, the configuration comprises a two-dimensional or three-dimensional spatial representation of the nuclear fusion reactor. In some embodiments, the two-dimensional or three-dimensional spatial representation of the nuclear fusion reactor comprises a model of the nuclear fusion reactor, an image of the nuclear fusion reactor, a rendering of the nuclear fusion reactor, or a latent representation of the nuclear fusion reactor. In some embodiments, the dataset comprises a plurality of simulated cross sections of the nuclear fusion reactor. In some embodiments, a first cross section of the plurality of simulated cross sections comprises a first toroidal angle and a second cross section of the plurality of simulated cross sections comprises a second toroidal angle, and wherein the first toroidal angle and the second toroidal angle differ by at least about 1 degree.

[0011] In one aspect disclosed herein is a computer-implemented method for generating a nuclear fusion reactor surrogate model, comprising: (a) extracting a plurality of input metrics from a nuclear fusion reactor dataset, wherein a metric of the plurality of metrics is extracted based on a correlation or anti-correlation with a desired property of the nuclear fusion reactor; and (b) generating the nuclear fusion reactor surrogate model by fitting a mathematical model or a machine learning model with the plurality of input metrics to output the desired property of the nuclear fusion reactor.

[0012] In some embodiments, the nuclear fusion reactor dataset comprises a simulation of a nuclear fusion reactor. In some embodiments, the method further comprises: (c) generating the simulation of the nuclear fusion reactor. In some embodiments, a datum of the nuclear fusion reactor dataset comprises the plurality of input metrics and the desired property of the nuclear fusion reactor. In some embodiments, the method further comprises: (d) selecting a subset of the nuclear fusion reactor dataset using the surrogate model to obtain an updated nuclear fusion reactor dataset. In some embodiments, (a)-(d) are repeated until a stopping condition is met. In some embodiments, the stopping condition is a number of iterations or an accuracy of the surrogate nuclear fusion reactor model. In some embodiments, the surrogate nuclear fusion reactor model forms a portion of a loss function used in training a machine learning model. In some embodiments, the surrogate nuclear fusion reactor model is used in an optimization framework. In some embodiments, a metric of the plurality of metrics is a shape metric, a magnetohydrodynamic (MHD) stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric. In some embodiments, the desired property of the nuclear fusion reactor is a shape metric, a magnetohydrodynamic (MHD) stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric. In some embodiments, the nuclear fusion reactor dataset describes a stellarator. In some embodiments, the nuclear fusion reactor dataset describes a tokamak.

[0013] In one aspect disclosed herein is a computer-implemented method, comprising: (a) providing a plurality of nuclear fusion reactor configurations, wherein a configuration of the plurality of nuclear fusion reactor configurations comprises a plurality of metrics; (b) generating a metric space based on the plurality of metrics as input into an optimization framework; and (c) determining an optimized nuclear fusion reactor configuration using the optimization framework.

[0014] In some embodiments, the plurality of nuclear fusion reactor configurations are stored in one or more database. In some embodiments, a nuclear fusion reactor configuration of theplurality of nuclear fusion reactor configurations comprises a shape metric, a magnetohydrodynamic (MHD) stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric. In some embodiments, a nuclear fusion reactor configuration of the plurality of nuclear fusion reactor configurations comprises a Fourier coefficient. In some embodiments, (b) comprises determining a range of values for each of the plurality of metrics. In some embodiments, the optimization framework is an active learning framework. In some embodiments, optimization framework uses Bayesian optimization. In some embodiments, an acquisition function of the Bayesian optimization comprises expected improvement, batch expected improvement, upper-confidence bound, probability of improvement, Thompson sampling, batch energy-entropy, expected hypervolume improvement, noisy expected hypervolume improvement, batch noisy expected hypervolume improvement, Pareto efficient global optimization (ParEGO), noisy Pareto efficient global optimization (NParEGO), batch noisy Pareto efficient global optimization (qNParEGO), or any combination thereof. In some embodiments, generating the optimized nuclear fusion reactor configuration comprises training a machine learning model using the optimization framework. In some embodiments, the method further comprises acquiring a nuclear fusion reactor configuration, wherein the acquiring the nuclear reactor configuration comprises simulating the nuclear fusion reactor configuration based on a sample of the metric space output by the optimization framework. In some embodiments, (c) comprises comparing a value of a desired physical property of the nuclear fusion reactor configuration with a provided metric of the nuclear fusion reactor configuration. In some embodiments, the desired physical property of the nuclear fusion reactor configuration is a shape metric, a magnetohydrodynamic (MHD) stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric In some embodiments, the optimization framework is configured to select a datum to acquire based at least in part on one or both of: (i) a likelihood to decrease uncertainty of the machine learning model; or (ii) a likelihood to generate the optimized nuclear fusion reactor configuration. In some embodiments, (c) comprises gathering data to train the machine learning model using the optimization framework until a stopping condition is reached. In some embodiments, the stopping condition is a number of iterations, an accuracy of the machine learning model, or achievement of the optimized nuclear fusion reactor configuration.

[0015] In one aspect disclosed herein is a method comprising using a trained machine learning model to generate a configuration of a nuclear fusion reactor.

[0016] Configurations of a fusion system may incorporate competing design limits. These design limits may make the parameter space for determining an effective (e.g., at least partially optimal) configuration quite wide. A fusion system may be a stellarator or a tokamak. The design space for a stellarator may be wider than that for a tokamak: A stellarator may vary in both toroidal and poloidal directions, and thus may not share the benefits afforded to tokamaks by symmetry. Fully resolving neutron transport (e.g., neutronics) responses may be computationally intensive. A workflow that reduces the parameter space prior to turning to higher fidelity models may facilitate rapid evaluation of concepts and fast design iteration.

[0017] Disclosed herein is a workflow where toroidal models of approximately similar (e.g., the same) scale as their stellarator counterpart may be evaluated using various radial builds to cover the design space of interest. This data may be then tabulated by total radial build length and neutron wall loading (NWL) so this data can be accessed while building a higher fidelity model. The NWL and available space for the radial build at different (e.g., phi, theta) locations may be evaluated in the stellarator geometry. These values may then be used to access the tabulated data, and may be passed to a parametric geometry tool for stellarators (e.g., ParaStell) to create a computer-aided design (CAD)-based neutronics model for use in Monte Carlo neutron and photon transport simulation code, such as OpenMC of the Direct Accelerated Geometry Monte Carlo (DAGMC) toolkit. This workflow may provide an improved (e.g., at least partially- optimized) 3D stellarator model based at least in part on plasma and magnet coil configurations.

[0018] In another aspect, disclosed herein is a non-transitory computer-readable media comprising machine-executable code comprising one or more instructions that, upon execution, implements any of the methods disclosed above or elsewhere herein on a computer, wherein said computer is configured to execute said one or more instructions.

[0019] Another aspect of the present disclosure provides a non-transitory computer readable medium comprising machine executable code that, upon execution by one or more computer processor, implements any of the methods above or elsewhere herein.

[0020] Another aspect of the present disclosure provides a system comprising one or more computer processor and computer memory coupled thereto. The computer memory comprises machine executable code that, upon execution by the one or more computer processor, implements any of the methods above or elsewhere herein.

[0021] Additional aspects and advantages of the present disclosure will become readily apparent to those skilled in this art from the following detailed description, wherein only illustrative embodiments of the present disclosure are shown and described. As will be realized, the presentdisclosure is capable of other and different embodiments, and its several details are capable of modifications in various obvious respects, all without departing from the disclosure. Accordingly, the drawings and description are to be regarded as illustrative in nature, and not as restrictive.INCORPORATION BY REFERENCE

[0022] All publications, patents, and patent applications mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent, or patent application was specifically and individually indicated to be incorporated by reference. To the extent publications and patents and patent applications incorporated by reference contradict the disclosure contained in the specification, the specification is intended to supersede or take precedence over any such contradictory material.BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The novel features of the inventive concepts are set forth with particularity in the appended claims. A better understanding of the features and advantages of the present inventive concepts will be obtained by reference to the following detailed description that sets forth illustrative embodiments, in which the principles of the inventive concepts are utilized, and the accompanying drawings (also “Figure” and “FIG.” herein), of which:

[0024] FIG. 1 shows an example of a computer-implemented method for training a machine learning model to output a nuclear fusion reactor configuration, per one or more embodiments herein;

[0025] FIG. 2 shows an example of a clustering technique as disclosed herein, per one or more embodiments herein;

[0026] FIG. 3 shows an example of a computer-implemented method for generating a nuclear fusion reactor configuration using a machine learning model, per one or more embodiments herein;

[0027] FIG. 4 shows an example of a computer-implemented method for generating a surrogate model for predicting a desired metric of a nuclear fusion reactor, per one or more embodiments herein;

[0028] FIG. 5 shows an example of a computer-implemented method for optimizing a configuration of a nuclear fusion reactor, per one or more embodiments herein;

[0029] FIG. 6 shows an example of a computing device with one or more processors, memory, storage, and a network interface, per one or more embodiments herein;

[0030] FIG. 7 shows an example of a web / mobile application provision system providing browser-based or native mobile user interfaces, per one or more embodiments herein;

[0031] FIG. 8 shows an example of a cloud-based web / mobile application provision system comprising an elastically load balanced, auto-scaling web server and application server resources as well synchronously replicated databases, per one or more embodiments herein;

[0032] FIG. 9 shows an example of a plurality of clusters of stellarator configurations, per one or more embodiments herein;

[0033] FIG. 10 shows an example of an illustrative use of a neural network, per one or more embodiments herein;

[0034] FIG. 11A shows an example of metrics along a flux-tube, per one or more embodiments herein;

[0035] FIG. 11B shows an example of a desired property along a flux-tube of a nuclear fusion reactor for surrogate and high-fidelity models, per one or more embodiments herein;

[0036] FIG. 12 shows an example of a surrogate model desired property output versus a desired property output by a high-fidelity computer modelling software, per one or more embodiments herein;

[0037] FIG. 13A illustrates an example of a section of a stellarator, per one or more embodiments herein;

[0038] FIG. 13B illustrates an example of complex geometry of stellarators, per one or more embodiments herein;

[0039] FIG. 14A illustrates an example solution to modeling the complex geometry of stellarators, per one or more embodiments herein;

[0040] FIG. 14B illustrates an example sample radial build modeling the thickness for each layer of a plurality of layers in a stellarator, per one or more embodiments herein;

[0041] FIGs. 15A, 15B illustrate example workflows for modeling the complex geometry of stellarators, per one or more embodiments herein;

[0042] FIG. 16 illustrates an example of a 2D representation in a toroidal or poloidal space of a minimum radial build with regions assigned to high-temperature shielding, per one or more embodiments herein;

[0043] FIG. 17A illustrates an example of predicted and calculated local responses for helium production in the vacuum vessel, per one or more embodiments herein; and

[0044] FIG. 17B illustrates an example of predicted and calculated local responses for fast fluence in the superconducting coils, per one or more embodiments herein.DETAILED DESCRIPTIONMACHINE LEARNING MODELS FOR NUCLEAR FUSION REACTOR DESIGN

[0045] Disclosed herein are methods, systems, media, and techniques for the training and inference of machine learning models for nuclear fusion reactor design. Nuclear fusion reactor design is a highly multi-dimensional, mathematically complex task when approached from a first-principles perspective. As such, the techniques disclosed herein alleviate bottlenecks in nuclear fusion reactor design and facilitate accelerated exploration of the nuclear fusion reactor design space. This design space, while describable via physics and math, is intractably complex when said physics and math are used directly as input into machine learning models. As such, the present disclosure provides alternative, efficient techniques for nuclear fusion reactor design and property prediction based on modem machine learning paradigms. Further disclosed herein are methods, systems, media, and techniques for inferencing machine learning models to generate nuclear fusion reactor configurations. Generally, training may comprise updating weights of a machine learning model to predict a configuration associated with one or more desired physical property of a nuclear fusion reactor and inferencing may comprise using the weights updated during training to generate configurations of nuclear fusion reactors. Further, the techniques disclosed herein may be agnostic to the type of nuclear fusion reactor. For example, the present disclosure may be used to generate configurations for magnetic confinement fusion devices including stellarators or tokamaks.Training

[0046] As shown in operations 100 of FIG. 1, training machine learning models for nuclear fusion reactor design may comprise the operation 110 of obtaining a dataset comprising a shape metric, a magnetohydrodynamic stability metric (MHD), a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric, and the operation 120 of training a machine learning model using the dataset to predict a configuration of a nuclear fusion reactor.Dataset

[0047] A dataset used as input into a machine learning model as disclosed herein may comprise a feature portion and optionally a label portion. In some cases, a fully or partially label-free dataset may be used in an unsupervised, weakly supervised, or semi -supervised training operation. For example, a clustering algorithm as disclosed herein may employ an unsupervised learning paradigm to generate a plurality of clusters from input data. In some cases, a fully labelled dataset may implement supervised learning to train a machine learning model to determine a specific (e.g., the label) output from an input. In some cases, weakly supervised or semi-supervised learning may be used when obtaining labels is expensive. In some cases, active learning may be used when obtaining labels is expensive. For example, the simulation of a nuclear fusion reactor using high-fidelity computer modelling software from which a dataset may be generated may be computationally expensive. As such, a dataset may comprise both labelled and label-free data. In some cases, data generated from machine learning models, computer modelling software, or both, may be stored in one or more database.

[0048] Herein, a dataset may comprise a shape metric, an MHD stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, a coil design metric, or any combination thereof. In some cases, a metric (e.g., shape metric, MHD stability metric, neoclassical transport metric, energetic particle confinement metric, turbulent transport metric, magnetic field metric, coil design metric) may comprise a physical property describing a nuclear fusion reactor. In some cases, the metric comprises a desired physical property. For example, a metric may comprise a threshold value for nuclear fusion reactor operation. Furthering the example, the metric may comprise a beta value or range of acceptable beta values describing an operation of the nuclear fusion reactor. In some cases, a beta value may comprise a ratio of plasma pressure to magnetic pressure.

[0049] The dataset may comprise a plurality of metrics describing desired physical properties of a nuclear fusion reactor. The desired physical properties may comprise considerations of physics or engineering. For example, a desired physical constraint for ideal plasma behavior may be slightly relaxed in favor of practical engineering considerations. The careful balance of the sensitive physics of fusion reactions and the practical limitations of engineering provides for a significant number of diverse combinations of features that may be combined to generate a training set for machine learning training. In some cases, the combination of one or more shape metric, one or more MHD stability metric, one or more neoclassical transport metric, one or more energetic particle confinement metric, one or more turbulent transport metric, one or more magnetic field metric, or one or more coil design metric may comprise the feature portion of onedatum of the dataset. In some cases, the feature portion of the datum comprises the desired physical properties of the nuclear fusion reactor. The datum of the dataset may be optionally associated with a label as described herein.

[0050] In some cases, a desired physical property may be context dependent. For example, a desired property of a plasma confinement device or nuclear fusion reactor may change if the reactor is to be used for neutron production instead of power generation. In the case of neutron production, a user may indicate a lower desire to maintain high efficiency, fusion power, or confinement quality (e.g., as relevant to turbulent transport, beta values). The desired physical property may be dependent on a type of reactor (e.g., tokamak versus stellarator). Generally, the desired physical properties may be a product of a combination of considerations comprising physics considerations, engineering considerations, goal (e.g., power or energy generation, power or energy efficiency, neutron production) considerations, or any combination thereof. For example, a user may desire a macroscopic performance metric such as a beta value. As such, the user may provide a beta value of at least about 0.01, 0.02, 0.03, 0.04, 0.05, 0.06, 0.07, 0.08, 0.09, 0.1, or more. Extending the example, the user may additionally provide descriptors of heat flux on a wall or divertor of the nuclear fusion reactor or other physical descriptors of the desired system. The machine learning models disclosed herein may be used to provide one or more configuration satisfying the desired physical properties provided by the user.

[0051] A dataset may comprise a shape metric. A shape may comprise general descriptors of a fusion device. For example, a shape metric may comprise a type of symmetry, number of field periods, a major radius, a minor radius, an aspect ratio, an angle, a radii, a thickness, a mass, or other aspect of a shape of a fusion device. For example, a user may prefer a 5 field period, quasi- helically symmetric stellarator configuration. In some cases, shape metric may comprise a quasi- isodynamic property, a quasi-axisymmetric property, a quasi-helically symmetric property. In some cases, a shape metric may comprise a layer description. For example, a nuclear fusion reactor may comprise a plurality of layers of materials with distinct purposes to manage, promote, control, or facilitate fusion in the plasma. In some cases, the shape metric may comprise a layer count, a layer thickness, a layer material, a layer order, a cross-sectional layer shape, or a layer type.

[0052] A dataset may comprise an MHD stability metric. Magnetohydrodynamics may generally be used to describe electrically conducting fluids and their behavior under the influence of magnetic fields. In some cases, a nuclear fusion reactor may be designed to maintain MHD equilibrium during operation of the nuclear fusion reactor. For example, a stellarator’s twisted torus shape may decrease aspects of MHD instability introduced by non-twisted configurations. Continuing the example, current driven instabilities present in tokamak devices may be of lower concern in stellarator devices. In some cases, an MHD stability metric may comprise a description of an ideal MHD behavior of a plasma in a nuclear fusion reactor. In some cases, an MHD stability metric may comprise a beta value. The beta value may comprise a ratio of plasma pressure to magnetic pressure. In some cases, an MHD stability metric may comprise a plasma pressure, a beta value, a magnetic pressure, a plasma current, a Mercier criterion, a ballooning stability threshold, an interchange stability threshold, a coefficient or exponent of a magnetohydrodynamic equation or equilibrium description, a kink stability threshold, or other descriptor of MHD stability or instability.

[0053] A dataset may comprise a neoclassical transport metric. In some cases, a nuclear fusion reactor may be designed to limit neoclassical transport losses. For example, neoclassical transport losses may comprise a substantial contributor to one or both of heat loss or particle loss in stellarators. In some cases, limiting neoclassical transport losses may increase fuel utilization efficiency. In some cases, neoclassical transport may comprise a form of heat loss induced by inter-particle collisions that expel heated particles from their orbits. In some cases, a neoclassical transport metric may comprise an effective ripple value, a collisionality, a particle diffusion, a coefficient or exponent of a neoclassical transport equation or description, or other descriptor of neoclassical transport.

[0054] A dataset may comprise an energetic particle confinement metric. In some cases, a nuclear fusion reactor may be designed to promote the retention of energetic particles in a plasma of a nuclear fusion reactor. For example, an energetic particle confinement metric may indicate a ratio or other statistic of energetic particles used for intended or favorable processes versus deleterious processes. To illustrate, a particle loss fraction may describe a ratio of energetic alpha particles generated by a fusion reactor that contribute to plasma maintenance - thereby promoting fuel efficiency - versus are lost to loss processes. In some cases, loss processes may include mechanisms that damage the nuclear fusion reactor, for example by colliding with and wearing on plasma facing nuclear fusion reactor components (e.g., a first wall). In some cases, an energetic particle confinement metric may comprise an energetic particle loss fraction. In some cases, the energetic particle may comprise an alpha particle. In some cases, an energetic particle loss fraction may comprise the fraction of alpha particles lost on an alpha particle thermalization time scale. In some cases, an energetic particle loss fraction may comprise the mean time before loss of an energetic particle. For example, the loss fraction may comprise the mean time before loss of an alpha particle. In some cases, an energetic particleconfinement metric may comprise a power load value. For example, an energetic particle confinement metric may comprise a power load due to alpha particle loss.

[0055] A dataset may comprise a turbulent transport metric. Turbulent transport may provide inefficiencies in plasma confinement and energy dissipation. In some cases, turbulent transport may be induced by deviations in an electric or magnetic field to which the plasma is exposed or provides. In some cases, turbulent transport may comprise small scale fluctuations in the density, temperature, or electromagnetic potentials of the plasma. In some cases, turbulent transport may comprise electrostatic turbulence or electromagnetic turbulence. In some cases, turbulent transport may comprise turbulent transport driven by modes for which electrons respond adiabatically. In some cases, turbulent transport may comprise turbulent transport driven by modes for which the electrons respond kinetically. In some cases, a turbulent transport metric may comprise a heat flux, a particle flux, a heat flux at a given density gradient, a heat flux at a given temperature gradient, a particle flux at a given density gradient, a particle flux at a given temperature gradient, a heat transport value, a coefficient or exponent of a turbulent transport equation or description, or other descriptor of turbulent transport.

[0056] A dataset may comprise a magnetic field metric. A magnetic field may be used to confine a plasma of a fusion fuel. In some cases, the fusion fuel may comprise a plasma of charged particles that may be manipulated or guided by the magnetic field. In some cases, the magnetic field may be used to increase an interaction among particles of the plasma to promote nuclear fusion. In some cases, a magnetic field metric may comprise an edge rotational transform, a rotational transform on axis, a maximum magnetic field on axis, a minimum magnetic field on axis, a volume averaged magnetic field, a magnetic mirror ratio, a magnetic field vector, a magnetic field strength, or a magnetic field gradient. In some cases, an edge rotational transform may comprise a value for the rotational transform of a nuclear fusion reactor at an edge of a plasma.

[0057] A dataset may comprise a coil design metric. A coil design metric may comprise a coil distance to plasma metric. In some cases, a coil design metric may comprise a maximum or minimum allowable distance between a plasma and a coil. In some cases, the coil design metric may comprise a coil geometry, a coil shape, a coil thickness, a coil count, or a coil separation. In some cases, a coil design metric may comprise an aggregate measure of the magnetic field conferred by the coils. For example, in some cases, a plurality of a coil configurations may be applied to a single nuclear fusion reactor configuration. As such, an aggregate descriptor of the coils may be used rather than a coil by coil description.

[0058] In some cases, a label of a dataset herein may comprise a configuration of a nuclear fusion reactor. A label may comprise an output of a machine learning model as disclosed herein. A configuration of a nuclear fusion reactor may comprise a two-dimensional (2D) or three- dimensional (3D) spatial representation of the nuclear fusion reactor. In some cases, the label of a dataset herein may comprise an image, a model, a rendering, a schematic, a graph, a plot, a latent representation, or other multi-dimensional representation of a shape, layout, organization, structure, architecture, outline, or spatial description of a nuclear fusion reactor. In some cases, a latent representation may comprise a transformation of an image, a model, a rendering, a schematic, a graph, a plot as a function of a passage of the data (e.g., image, model, etc.) through a neural network. In some cases, the label may comprise a visualization of the nuclear fusion reactor. In some cases, the visualization may comprise a plurality of viewpoints. For example, a label may comprise one or more rendering of the nuclear fusion reactor from different angles (e.g., top view, side view, cross-sectional view, etc.). In some cases, the label may comprise a parameterized version of the previously described labels. For example, a parameterized representation of a torus may comprise radii or angles describing the torus, rather than an image or other more visually explicit spatial description of the torus itself. In some cases, the label may comprise a spatial representation with an indication of scale. For example, a graph of a shape with axes describing dimensions of the shape may comprise a 2D representation of the shape. Generally, a label as described herein may comprise any set of values such that a spatial representation of the nuclear fusion reactor is apparent or optionally arrived at with trivial processing. For example, a complete set of descriptors for generating a 3D model of a torus may comprise radii or angles of the torus. As such, the 3D model is fully described by the complete set of descriptors and the 3D model and the complete set of descriptors are two representations of the same shape with the only distinction being arbitrary choice of representation.

[0059] In some cases, the label of the dataset may comprise an image of a nuclear fusion reactor. In some cases, the image may comprise a model, rendering, optical image, drawing, blueprint, sketch, or other spatial representation of the nuclear fusion reactor. Computationally, an image may be represented by a matrix of pixel values. In some cases, the matrix of pixel values may comprise one color dimension (e.g., greyscale) or a plurality of color dimensions (e.g., RGB, CMYK). In some cases, the image may comprise a color scale that correlates to a physics property of the nuclear fusion reactor. For example, a strength of magnetic field in an output of a simulation of a nuclear fusion reactor may comprise a color scale ranging from blue to red corresponding with low to high magnetic field flux values.

[0060] In some cases, the label of the dataset may comprise triangulation data. For example, a plurality of coordinate, magnetic field data pairs may comprise a description of a configuration of a nuclear fusion reactor as disclosed herein. In some cases, the triangulation data may be used to visualize the configuration of the nuclear fusion reactor in rendering or modelling software. For example, surface triangulation may be used to describe a surface in finite element method software.

[0061] Generally, a label, output of a machine learning model, or configuration of a nuclear fusion reactor herein may comprise any spatial description of the nuclear fusion reactor either explicitly (e.g., images, models, renderings, plots, etc.) or via a parameterized description of the spatial description (e.g., coordinates, widths, lengths, angles, etc.) such that the shape in one, two, or three dimensions is apparent.

[0062] In some cases, a label of a dataset or a configuration of a nuclear fusion reactor herein may comprise a cross section of a nuclear fusion reactor, a rotational transform profile, a current profile, a pressure profile, a total magnetic flux, or any combination thereof. For example, a stellarator configuration may be described by a plurality of cross sections of the stellarator, a pressure profile of the stellarator, and a total magnetic flux of the stellarator. In some cases, the total magnetic flux may comprise a scaling factor for the configuration of the nuclear fusion reactor. In some cases, the total magnetic flux may comprise a total toroidal magnetic flux. In some cases, the configuration of the nuclear fusion reactor may be associated with a plurality of metrics describing the plasma or operation of the fusion device. The plurality of metrics describing the plasma or operation of the fusion device may comprise the features or physics properties described herein.

[0063] In some cases, a current profile may comprise the bootstrap current of the plasma in a nuclear fusion reactor. In some cases, a rotational transform profile may indicate a degree to which magnetic field lines rotate about a magnetic axis. In some cases, the rotational transform profile may comprise an indication of the evolution of one or both of toroidal or poloidal angles. In some cases, a rotational transform profile may be referred to as a field line pitch or an iota profile. In some cases, a rotational transform profile may comprise a profile of the thread angle of magnetic field lines.

[0064] In some cases, a nuclear fusion reactor as disclosed herein may comprise a stellarator. In some cases, a nuclear fusion reactor as disclosed herein may comprise a tokamak. In some cases, a nuclear fusion reactor as disclosed herein may comprise an inertial confinement fusion (ICF) device. In some cases, a nuclear fusion reactor as disclosed herein may comprise a magneticconfinement fusion (MCF) device. Generally, a nuclear fusion reactor as disclosed herein may comprise any device configured to constrain a plasma or fusion fuel so as to generate fusion relevant conditions.

[0065] In some cases, the dataset as disclosed herein may comprise simulated nuclear fusion reactor designs. Generally, a simulated nuclear fusion reactor may comprise an output of a high- fidelity computer modelling software. In some cases, a simulation of a nuclear fusion reactor may comprise one or more mathematical function, boundary condition, or starting condition. In some cases, a simulation may implement finite element, finite difference, spectral, pseudo- spectral, or other solving technique. The simulated nuclear fusion reactor may be solved for in one, two, three, or more dimensions. For example, the simulation may comprise multidimensional, multi-scale, time-dependent, or any combination thereof of simulations of the nuclear fusion reactor. In some cases, the simulation may comprise a multi-physics computational model of the nuclear fusion reactor. In some cases, the dataset may comprise data output by an MHD equilibrium solver. In some cases, the dataset may comprise data based on a DESC, VMEC or SPEC output. In some cases, an equilibrium solver may comprise VMEC, DESC, HINT2, GVEC, SPEC or PIES. In some cases, an MHD stability solver may comprise TERPSICHORE or CAS3D. In some cases, a computer modelling software may be used to determine a neoclassical transport effect. In some cases, a computer modelling software for determining neoclassical transport effects may comprise SFINCS or NEO. In some cases, a computer modelling software may be used to determine an energetic particle confinement effect. In some cases, a computer modelling software for determining energetic particle confinement effects may comprise SIMPLE or ASCOT. In some cases, a turbulent transport solver may be used to determine a turbulent transport effect. In some cases, a turbulent transport solver may comprise GX or GENE.

[0066] In some cases, the dataset may comprise data from an operating nuclear fusion reactor. For example, physics properties associated with a functioning nuclear fusion reactor may be associated with descriptions or labels describing the nuclear fusion reactor as disclosed herein.Clustering

[0067] In some cases, operations 100 may optionally further comprise grouping a parent dataset into a plurality of groups based on a physics principle, clustering the groups of the plurality of groups into a plurality of clusters using a clustering model, and training a machine learning model for a cluster of the plurality of clusters. In other words, clustering operations using one or both of first-principles physics or machine learning may be implemented to provide a pluralityof clusters to train a plurality of machine learning models. Generally, a machine learning model of the plurality of machine learning models may be trained to provide a nuclear fusion reactor configuration for a subset of the possible configurations as implied by the desired physics or metrics. This may provide particular utility in model accuracy and robustness by distributing the general learning problem among a plurality of learners (the machine learning models) rather than training a global configuration predictor. In some cases, however, a single machine learning model may be used.

[0068] In some cases, global machine learning models may struggle to retain robustness or accuracy as the complexity or diversity of a training dataset increases. As such, a dataset may be distributed among a plurality of subsets for the development of a plurality of machine learning models. A model of the plurality of machine learning models may then be responsible for learning a subset of the parameter space for discriminative or generative tasks. This may provide particular utility in a method or system by foregoing limitations of machine learning models trained to generalize over a broad parameter space. For example, a machine learning model trained to predict an output for a sub-type of nuclear fusion reactor may capture local optima of the parameter space during training where a global nuclear fusion reactor configuration predictor may smooth over local optima in order to better approximate global optima or otherwise generalize to the entire parameter space.

[0069] Accordingly, the methods, systems, media, and techniques disclosed herein may implement a clustering operation to split a training dataset into a plurality of sub-sets. In some cases, there may be more than one clustering operation. For example, a physics rule-based clustering and a machine learning based clustering may be performed. In some cases, one or both of the physics rule-based or the machine learning based clustering may be performed. In some cases, a surrogate model as described herein may be used to facilitate clustering. For example, a surrogate model may be trained to output a metric describing a configuration which is subsequently used to cluster configurations.

[0070] A physics rule-based clustering may comprise a clustering ontology. In some cases, the clustering ontology may be implemented in computer readable media. The clustering ontology may be established based on a subject matter expert set of rules. For example, nuclear fusion reactors may be clustered based on a quasi-isodynamic property, a quasi -axisymmetric property, a quasi-helically symmetric property, a number of field periods, an edge rotational transform, or other shape or physics descriptor of a nuclear fusion reactor.

[0071] In some cases, a physics rule-based clustering may be configurable according to the desires of the user. As such, the rules described herein are for illustrative purposes as the enumeration of physical rules by which the nuclear fusion reactor configuration parameter space may be divided is vast. For example, in some cases a physics rule implemented in the clustering ontology may comprise a rule based on a shape metric, an MHD stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, a coil design metric, or any combination thereof. In some cases, a datum or sample of the dataset may be placed into a cluster described by the clustering ontology. The clustering ontology may be used in one of both of training or inference.

[0072] Additionally, or alternatively to the physics rule-based clustering ontology, a machine learning clustering may be implemented on a dataset as described herein. Machine learning clustering may implement an unsupervised learning paradigm. The machine learning clustering may generate a plurality of clusters based on similarities in the data to be clustered. Machine learning based clustering may provide particular utility by capturing order or latent organizational principles within a dataset that may not be readily apparent for or otherwise applicable to non-machine learning clustering techniques. As such, clustering of nuclear fusion reactor datasets may provide a plurality of sub-datasets. The sub-datasets may be used to train cluster-specific machine learning algorithms. In some cases, a clustering ontology may be used in addition to a machine learning clustering algorithm. In some cases, a clustering ontology may precede a machine learning clustering.

[0073] In some cases, grouping or clustering of nuclear fusion reactor configurations may be performed by clustering algorithms. For example, clustering algorithms may comprise K-means clustering, mean-shift clustering, affinity propagation clustering, hierarchical clustering, densitybased clustering (e.g., DBSCAN), BIRCH clustering, Gaussian mixture clustering, agglomerative clustering, or other clustering algorithm.

[0074] Clustering ontologies, clustering algorithms, or both, may be implemented to generate a plurality of clusters, groups, or sub-datasets. The clustering ontologies, clustering algorithms, or both may be leveraged in machine learning model training or machine learning model inference. For example, during training, each of the clusters may be used to train a machine learning model for nuclear fusion reactor configuration prediction. As such, during inference, features (e.g., physical properties of interest) may be exposed to the clustering used during training. This may be used to identify the appropriate neural network from which to receive a nuclear fusion reactor configuration. Generally, the clustering techniques may be used to establish a mapping between types of nuclear fusion reactor input and nuclear fusion reactor configuration prediction models.This may provide particular utility during training by sharing the learning task among a plurality of machine learning models while promoting individual model accuracy during inference.

[0075] In some cases, clustering may comprise operations 200 as shown in FIG. 2. In some cases, a clustering ontology (e.g., based on physics rules or principles), clustering machine learning model, or both, may be generated based on a database of nuclear fusion reactors 210. For example, the database of nuclear fusion reactors may comprise a plurality of stellarator configurations. Optionally, the database of nuclear fusion reactors may be exposed to physics based clustering 220. In some cases, physics based clustering 220 may comprise clustering a plurality of nuclear fusion reactor configurations based on one or more physics principle as disclosed herein. In addition, or alternative to the physics based clustering 220, machine learning based clustering 230 may be performed. For example, a plurality of clusters may be generated directly from the database of nuclear fusion reactors 210 or from the clusters generated by physics based clustering 220. In some cases, only physics based clustering 220 may be performed. Regardless of whether physics based clustering, machine learning based clustering, or both, is performed, the generated clusters may be used to train a plurality of configuration predicting models. In some cases, each cluster may be associated with a configuration predicting model. As such, operations 200 may comprise neural network training for each cluster 240. A cluster of the clusters may be used during operations 100 (e.g., operation 120) or 300 (e.g., operation 320) as disclosed herein to identify a machine learning model for use in predicting or generating a configuration of a nuclear fusion reactor.Secondary Training

[0076] In some cases, operations 100 may optionally further comprise a second training operation. The second training operation may comprise a reinforcement learning operation as described herein. For example, a deployed machine learning model may receive feedback from a simulation, user (reinforcement learning through human feedback), or a secondary machine learning model (reinforcement learning through artificial intelligence). The second training operation may be used to improve accuracy or robustness of the machine learning model. For example, outputs of the machine learning model (e.g., configurations of fusion devices) may be altered via MHD equilibrium code software to better align the configuration with a desired physical property. As such, the alteration may be stored and used for training the machine learning model in a second training operation. Such secondary training may provide particular utility by utilizing expensive calculations performed during the alterations as training data while maintaining the benefit of decreasing the overall reliance on expensive calculations for nuclear fusion reactor configuration generation. For example, the machine learning model may provide aconfiguration that is nearly optimal for a given desired physics input such that relatively minor modifications of the configuration may provide the optimal configuration. These minor alterations may be based on substantially fewer computations than a first-principles design of a nuclear fusion reactor configuration.

[0077] In some cases, the operations 100 may comprise a second training. The second training may comprise generating an updated configuration of the configuration of the nuclear fusion reactor. In some cases, the second training may comprise obtaining, forming, creating, or retrieving a second dataset comprising the updated configuration. In some cases, the second training may comprise using the second dataset to train a machine learning model to predict the updated configuration. In some cases, the second training may update a portion of the machine learning model parameters. In some cases, a portion of the machine learning model weights may be frozen during training and a second portion updated during training.Model Architecture

[0078] A machine learning model as disclosed herein may comprise a traditional (e.g., non- neural network based) or deep learning model. In some cases, a machine learning model may comprise a deterministic or stochastic machine learning model. In some cases, a machine learning model may comprise a classification, dimensionality reduction, clustering, regression, or generative machine learning model. In some cases, more than one machine learning model may be implemented on one or more local or remote computing device to perform reactor configuration prediction or generation as disclosed herein. In some cases, the machine learning model may comprise fully connected neural network layers, convolutional neural network layers, transformer layers, or residual layers.

[0079] In some cases, a single machine learning model may be used to predict the nuclear fusion reactor configuration or a portion of the nuclear fusion reactor configuration. For example, one or more machine learning model may be used to predict a nuclear fusion reactor configuration comprising a plurality of cross sections of the nuclear fusion reactor, a pressure profile of the nuclear fusion reactor, or a total magnetic flux. In some cases, additionally or alternatively, a nuclear fusion reactor configuration may comprise a current profile, a rotational transform profile, or any combination thereof. In some cases, a machine learning model may be trained for a particular task. For example, a machine learning model may be trained to predict a specific aspect of a nuclear fusion reactor configuration. In some cases, multi-task learning may be used to train a machine learning architecture to predict a plurality of the components of the nuclear fusion reactor configuration. For example, a shared plurality of weights of a machine learningmodel may be trained given the features describing a nuclear fusion reactor, and a portion of weights, or a head, may be trained to predict a component of the nuclear fusion reactor. For example, a multi-task fine-tuned model may comprise a portion of weights trained for a specific task. Continuing the example, the portion of weights trained for the specific task may be a regression head used to predict a total magnetic flux or another head appropriate for another task such as classification or generative tasks.

[0080] In some cases, a machine learning model as disclosed herein may comprise a transformer or large language model architecture. As such, a user may build a prompt describing the desired physical properties or features as described herein. For example, a multi-modal transformer model may take as input a plurality of desired physics properties (e.g., shape metric, MHD stability metric, neoclassical transport metric, energetic particle confinement metric, turbulent transport metric, magnetic field metric, coil design metric) and output a nuclear fusion reactor configuration. The nuclear fusion reactor configuration output in a multi-modal transformer model may comprise both image representations and written description of the image representations. For example, a plurality of nuclear fusion reactor cross sections may be output as images of nuclear fusion reactor cross sections while a natural language caption or table is provided to give numerical context for the cross sections (e.g., toroidal angles, widths, heights, etc.). In some cases, a generative machine model (e.g., transformer-based model) may directly output images, schematics, blueprints, renderings, models, or other 2D or 3D spatial representations of nuclear fusion reactor configurations.

[0081] In some cases, an output of a machine learning model may be one or more configuration of the nuclear fusion reactor. For example, a machine learning model may output multiple possible configurations from which a user may select. In some cases, the user may provide feedback to the machine learning model. For example, user selection, ranking, scoring, alteration of an output, or other feedback may be used to fine-tune the machine learning model (e.g., reinforcement learning). In some cases, the one or more configuration may be output by the same machine learning model.

[0082] In some cases, the features of the model may comprise an encoded representation of a nuclear fusion reactor configuration. As such, a decoder or decoder-like machine learning model may be trained to convert the input features into a nuclear fusion reactor configuration. For example, a decoder may be trained to output images, schematics, blueprints, renderings, models, or other 2D or 3D spatial representations of nuclear fusion reactor configurations. Generally, the machine learning architecture or learning paradigm chosen may be selected based on a desired type of output or modality of input.

[0083] Generally, a machine learning model herein may comprise k-nearest neighbors algorithm, learning vector quantization, linear regression, non-linear regression, least squares regression, partial least squares regression, logistic regression, stepwise regression, multivariate adaptive regression splines, ridge regression, principal component regression, least absolute shrinkage and selection operation (LASSO), least angle regression, canonical correlation analysis, factor analysis, independent component analysis, linear discriminant analysis, multidimensional scaling, non-negative matrix factorization, principal components analysis, principal coordinates analysis, projection pursuit, Sammon mapping, t-distributed stochastic neighbor embedding, AdaBoosting, boosting, gradient boosting, bootstrap aggregation, ensemble averaging, decision trees, conditional decision trees, boosted decision trees, gradient boosted decision trees, random forests, stacked generalization, Bayesian networks, Bayesian belief networks, naive Bayes, Gaussian naive Bayes, multinomial naive Bayes, hidden Markov models, hierarchical hidden Markov models, support vector machines, encoders, decoders, autoencoders, stacked auto-encoders, perceptrons, multi-layer perceptrons, artificial neural networks, feedforward neural networks, convolutional neural networks, recurrent neural networks, physics informed neural networks, residual neural networks, long short-term memory, deep belief networks, deep Boltzmann machines, deep convolutional neural networks, deep recurrent neural networks, large language models, transformer models, vision transformers, or generative adversarial networks. Further, a machine learning model as disclosed herein may be used in an active learning, reinforcement learning, or reward modelling framework.Inference

[0084] Disclosed herein are methods, systems, media, and techniques for generating a nuclear fusion reactor configuration using a machine learning model. In some cases, the machine learning model is trained according to the training operations disclosed elsewhere herein (e.g., operations 100). As shown in operations 300 of FIG. 3, the operations for generating a nuclear fusion reactor configuration may comprise the operation 310 of obtaining a dataset comprising a shape metric, a magnetohydrodynamic (MHD) stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric, and the operation 320 of generating, using the dataset as input into the machine learning model, the configuration of the nuclear fusion reactor.

[0085] In some cases, the dataset for generating a nuclear fusion reactor may comprise a shape metric, an MHD metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, a coil design metric, or any combination thereof. In some cases, the dataset for generating the nuclear fusion reactor may bethe same or similar to the dataset used for training. For example, a training dataset may comprise a plurality of samples describing a breadth of possible nuclear fusion reactor configurations. For training, the dataset may be agnostic to utility of the nuclear fusion reactor and may be constructed to provide a wide range of possible sample inputs and outputs to promote model generalization ability. In generating a configuration of a nuclear fusion reactor, the user may provide a plurality of desired metrics or desired physics properties with features similar to a datum of the training dataset. In some cases, a user may use the machine learning model to obtain a configuration of a nuclear fusion reactor according to desired physics properties of the nuclear fusion reactor.

[0086] In some cases, a configuration generated by a machine learning model may comprise a two-dimensional (2D) or three-dimensional (3D) spatial representation of the nuclear fusion reactor. In some cases, the configuration of the nuclear fusion reactor may comprise an image, a model, a rendering, a schematic, a graph, a plot, a latent representation, or other multidimensional representation of a shape, layout, organization, structure, architecture, outline, or spatial description of a nuclear fusion reactor. In some cases, a latent representation may comprise a transformation of an image, a model, a rendering, a schematic, a graph, a plot as a function of a passage of the data (e.g., image, model, etc.) through a neural network. In some cases, the configuration may comprise a visualization of the nuclear fusion reactor. In some cases, the visualization may comprise a plurality of viewpoints. For example, a configuration may comprise one or more rendering of the nuclear fusion reactor from different angles (e.g., top view, side view, cross-sectional view, etc.). In some cases, the configuration may comprise a parameterized version of the previously described labels. For example, a parameterized representation of a torus may comprise radii or angles describing the torus, rather than an image or other more visually explicit spatial description of the torus itself. In some cases, the configuration may comprise a spatial representation with an indication of scale. For example, a graph of a shape with axes describing dimensions of the shape may comprise a 2D representation of the shape. Generally, a configuration as described herein may comprise any set of values such that a spatial representation of the nuclear fusion reactor is apparent or optionally arrived at with trivial processing. For example, a complete set of descriptors for generating a 3D model of a torus may comprise radii or angles of the torus. As such, the 3D model is fully described by the complete set of descriptors and the 3D model and the complete set of descriptors are two representations of the same shape with the only distinction being choice of representation.

[0087] In some cases, the configuration of the nuclear fusion reactor may comprise an image of a nuclear fusion reactor. In some cases, the image may comprise a model, rendering, optical image, drawing, blueprint, sketch, or other spatial representation of the nuclear fusion reactor. Computationally, an image may be represented by a matrix of pixel values. In some cases, the matrix of pixel values may comprise one color dimension (e.g., greyscale) or a plurality of color dimensions (e.g., RGB, CMYK). In some cases, the image may comprise a color scale that correlates to a physics property of the nuclear fusion reactor. For example, a strength of magnetic field in an output of a simulation of a nuclear fusion reactor may comprise a color scale ranging from blue to red corresponding with low to high magnetic field flux values.

[0088] In some cases, the configuration of the nuclear fusion reactor may comprise triangulation data. For example, a plurality of coordinate, magnetic field data pairs may comprise a description of a configuration of a nuclear fusion reactor as disclosed herein. In some cases, the triangulation data may be used to visualize the configuration of the nuclear fusion reactor in rendering or modelling software. For example, surface triangulation may be used to describe a surface in finite element method software.

[0089] Generally, an output of a machine learning model, or configuration of a nuclear fusion reactor herein may comprise any spatial description of the nuclear fusion reactor either explicitly (e.g., images, models, renderings, plots, etc.) or via a parameterized description of the spatial description (e.g., coordinates, widths, lengths, angles, etc.) such that the shape in one, two, or three dimensions is apparent.

[0090] In some cases, a configuration of the nuclear fusion reactor may comprise a cross section of the nuclear fusion reactor, a pressure profile, a total magnetic flux, a current profile, a rotational transform profile, or any combination thereof. For example, a stellarator configuration may be described by a plurality of cross sections of the stellarator, a pressure profile of the stellarator, a current profile, and a total magnetic flux of the stellarator. In another example, a stellarator configuration may comprise a plurality of cross sections of the stellarator, a pressure profile of the stellarator, a total magnetic flux of the stellarator, and a rotational transform profile of the stellarator. In some cases, the plurality of cross sections may correspond to a plurality of toroidal angles. In some cases, a cross section may comprise a toroidal angle (e.g., represent a cross section at a toroidal angle of a torus). In some cases, a first cross section and a second cross section of the plurality of cross sections may differ by at least about 1 degree, 2 degrees, 3 degrees, 4 degrees, 5 degrees, 6 degrees, 7 degrees, 8 degrees, 9 degrees, 10 degrees or more.

[0091] In some cases, a nuclear fusion reactor as disclosed herein may comprise a stellarator. In some cases, a nuclear fusion reactor as disclosed herein may comprise a tokamak. In some cases, a nuclear fusion reactor as disclosed herein may comprise an inertial confinement fusion (ICF) device. In some cases, a nuclear fusion reactor as disclosed herein may comprise a magnetic confinement fusion (MCF) device. Generally, a nuclear fusion reactor as disclosed herein may comprise any device configured to constrain a plasma or fusion fuel so as to produce fusion relevant conditions.

[0092] In some cases, the machine learning model used for generating the configuration of the nuclear fusion reactor may be selected using the clustering techniques disclosed elsewhere herein. For example, a clustering ontology, a clustering machine learning model, or both may be used to classify a group of desired physical properties (e.g., an input during model inference) as being associated with a particular machine learning model. As such, the machine learning model selected for use with a given input may be selected based on a similarity between the input during inference and a portion of the dataset used to train the model.

[0093] In some cases, one or more configuration may be output by one or more machine learning model for one input. For example, a plurality of machine learning models may be trained as disclosed herein according to a clustering technique. As such, a user may receive one or more configuration for each of a portion of the plurality of machine learning models trained.Nuclear Fusion Reactor Database Interaction

[0094] In some cases, a database of nuclear fusion reactor configurations may be established. In some cases, the database may be established in part or entirely using an output of machine learning models as described herein. For example, a plurality of metrics may be used to determine a configuration such that the plurality of metrics and the configuration are stored in the database. In some cases, the database comprises a configuration for a stellarator. In some cases, the database may comprise metrics, configurations, or both used as input into or output from a computer modelling software. In some cases, a nuclear fusion reactor configuration of the database of nuclear fusion reactor configurations may be associated with a shape metric, an MHD stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, a coil design metric, or any combination thereof. In some cases, a nuclear fusion reactor configuration may comprise a portion of metrics as described but lack a particular desired physical property. In some cases, a desired physical property may be an MHD stability metric, a neoclassical transport metric, an energetic particle confinement metric, or a turbulent transport metric.

[0095] In some cases, a database of nuclear fusion reactor configurations may comprise at least about 1,000, 5,000, 10,000, 100,000, 500,000, 1,000,000, 10,000,000, 100,000,000, or more nuclear fusion reactor configurations. Accordingly, evaluating the database for a subset of nuclear fusion reactor configurations satisfying the desired physical property may be computationally expensive. This may be particularly true if computational modelling software is relied upon for the evaluation. For example, the use of GX software to determine a turbulent heat flux metric for each nuclear fusion reactor of the database to determine a subset of configurations satisfying a turbulent heat flux constraint may be computationally intractable. In another example, the use of ASCOT software to determine an alpha particle loss fraction or power loads due to alpha particle losses for each nuclear fusion reactor of the database to determine a subset of configurations satisfying constraints based on the loss fraction and / or power loads may be computationally intractable. Accordingly, the evaluation of an existing configuration for a metric not considered in determining the configuration may be of value when searching existing configurations for a desired physical property.

[0096] In some cases, a database of nuclear fusion reactor configurations may be evaluated using a surrogate model. In some cases, the surrogate model may be a machine learning model. In some cases, the surrogate model may be a mathematical model. In some cases, the mathematical function may be a power-law function. Generally, a surrogate model may comprise a plurality of metrics as described herein. For example, the plurality of metrics of the surrogate models may be operands of a function. To illustrate, the plurality of metrics may comprise variables of a power law-function to which coefficients or exponents are fit. In another example, the plurality of metrics of the surrogate model may be input features of a machine learning model. In some cases, a surrogate model may be fit according to plurality of outputs from a computer modelling software. For example, a subset of a nuclear fusion reactor configurations may be subject to computer modelling software to obtain high-fidelity outputs for a desired metric of interest. In some cases, these high-fidelity outputs may be used to fit the surrogate model. In some cases, the surrogate model may be used to evaluate other configurations in a database of nuclear fusion reactor configurations. In some cases, a training dataset may be established between a plurality of nuclear fusion reactor configurations and the outputs of a desired physical property from a computer modelling software for each of the plurality of nuclear fusion reactor configurations.

[0097] In some cases, a method for generating a surrogate model may comprise operations 400 as shown in FIG. 4. In some cases, operations 400 may comprise operation 410 of generating a plurality of desired physical properties based on a high fidelity simulation of a plurality ofconfigurations, operation 420 of extracting a plurality of input metrics of the plurality of configurations, and operation 430 of fitting a surrogate model comprising the plurality of input metrics using the plurality of values for the desired physical properties. In some cases, the method may optionally further comprise operation 440 of using the surrogate model to select a second plurality of configurations from the database of configurations. In some cases, a second surrogate model may be generated based on the second plurality of configurations from the database of configurations. In some cases, an iterative process of generating a surrogate model, selecting a subset of nuclear fusion reactor configurations from a database, generating a surrogate model from the subset, etc., may be established. In some cases, the iterative process may be used to obtain progressively more accurate surrogate models for a subset of nuclear reactor configurations. In some cases, after one or more iterations of surrogate model generation, the method may optionally comprise operation 450 of using the surrogate model to populate a database. In some cases, populating a database may comprise adding a new entry to the database. In some cases, populating a database may comprise updating existing entries of the database. For example, the surrogate model may be used to evaluate existing configurations for a physical property. As such, the database may be enriched by using the surrogate model to calculate the physical property for each member of the database.

[0098] In some cases, a surrogate model may be established for a subset of nuclear fusion reactor configurations. For example, a surrogate model may be established for quasi-isodynamic nuclear fusion reactors. In another example, a surrogate model may be established for quasi- helically symmetric nuclear fusion reactors. In another example, a surrogate model may be established for quasi-axisymmetric nuclear fusion reactors. In some cases, a surrogate model is established based in part on a number of field periods. In some cases, a surrogate model is established based on a type of MHD equilibrium. In some cases, a surrogate model may be established based on an output of a clustering as described herein. For example, each group of a plurality of groups established by a physics-rule based clustering may be associated with a surrogate model.

[0099] In some cases, extracting the plurality of input metrics may comprise a correlation between an input metric and a desired physical property. For example, a computer modelling software may take as input a large number (e.g., 10s, 100s, 1,000s) of metrics to solve for the value of a desired property. Generally, computer modelling software may be used to perform a simulation. Accordingly, the large number of metrics may be reduced to a subset of metrics based on a correlation with the desired property. In some cases, a property may be determined to have little or no impact on the desired property. Alternatively, a property may be determined tohave a large correlation or anti correlation with a desired property. In some cases, determination of the subset of metrics may be performed based on measuring an impact of a metric over a plurality of simulations. In some cases, determining the subset is performed by a regression analysis. In some cases, determining the subset is performed by a sensitivity analysis. In some cases, the subset is determined by design of experiment. In some cases, the subset is determined by a machine learning technique. In some cases, the subset is determined by a dimensionality reduction algorithm. In some cases, the subset may comprise 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or more metrics.

[0100] In some cases, a surrogate model may be used in an optimization framework. In some cases, the surrogate model may be used for reinforcement learning. In some cases, the surrogate model may be used in active learning. In some cases, the surrogate model may be used to establish a loss function for a mathematical optimization. In some cases, the surrogate model may be used to establish a loss function for a machine learning model.

[0101] In some cases, a surrogate model used in an optimization framework may communicate with a database of nuclear fusion reactor configurations. For example, a surrogate model may be used to select a subset of nuclear fusion reactor configurations from a database. In some cases, selecting the subset may comprise sorting or searching the database based on the proxy model. For example, a database comprising configurations of nuclear fusion reactors may be evaluated based on the surrogate model to determine a value of a desired property of interest. To illustrate, a turbulent heat flux metric may have a desired value for an operating nuclear fusion reactor; however, evaluating each configuration of a nuclear fusion reactor stored in a database for the turbulent heat flux metric may be computationally intractable. As such, a surrogate model may be fit or trained on a sample of nuclear fusion reactor configurations and subsequently used to predict the value of the turbulent heat flux metric for each nuclear fusion reactor in the database. The values predicted by the surrogate model may then be used to satisfy a threshold or constraint, for example, that a turbulent heat flux be below a value. Analogous scenarios can be contemplated for any metric described herein. For example, a surrogate model for energetic particle confinement metrics such as an energetic particle loss ratio may be used to search or sort a database of nuclear fusion reactor configurations. In some cases, upon selecting a subset of nuclear fusion reactor configurations, the surrogate model may be fit or trained on the subset of nuclear fusion reactor configurations. In this way, a surrogate model may be iteratively trained to determine a configuration with an optimal alignment with a desired metric. Such selection of configurations for training may provide particular utility by reducing the number of high fidelitysimulations of nuclear fusion reactors during selection of nuclear fusion reactor configurations for closer analysis.

[0102] In some cases, a stopping condition may be used when training a surrogate model. In some cases, the stopping condition may be a number of iterations or an accuracy of the machine learning model. For example, a surrogate model may be used to iteratively select subsets of a database of nuclear reactor configurations until the surrogate model achieves a desired accuracy.

[0103] In some cases, a surrogate model may be used to train machine learning models as described herein (e.g., as with operations 100). For example, a machine learning model may be used to output a nuclear fusion reactor configuration based on training on a dataset comprising known metrics (e.g., MHD equilibrium metrics, shape metrics, turbulent transport metrics, etc.) associated with a configuration. In some cases, one of the metrics used in training may be previously determined, readily determined (e.g., stored in a database), or user selected. For example, large databases of nuclear fusion reactors may exist that contain some metrics describing the reactors but may be missing others. Accordingly, a desired metric of the configuration may be unknown. As such, a nuclear reactor configuration as described herein may be evaluated via a surrogate model. In some cases, the surrogate model may provide a value for a desired physical property. In some cases, this may be used to guide training by providing a weighting to an output configuration. For example, a configuration output by a machine learning model may be predicted by the surrogate model to align poorly with a value for a desired physical property. In some cases, this alignment may be used in a loss function to aid in training the machine learning models to output configurations accounting for a particular metric. This may provide particular utility where a large database of other metrics describing configurations are known, but a desired, expensive to calculate value for a desired physical property is not.

[0104] In some cases, a surrogate model may be used to output the desired metric of interest at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, or more orders of magnitude faster than the computer modelling software. For example, a turbulent heat flux metric determined via computer modelling software may require up to several days of calculation for a powerful GPU. Alternatively, a surrogate model may determine the turbulent transport metric in a few seconds on a standard CPU. Analogous examples can be described for any metric described herein, for example surrogate models may be used to supplement or replace ASCOT software to determine an alpha particle loss fraction or power loads due to alpha particle losses.

[0105] In some cases, a surrogate model may evaluate at least about 1, 10, 100, 1,000, 10,000, 100,000, or more configurations per second, hour, or day. In some cases, the surrogate modelmay evaluate at least about 1 configuration per second using a computer comprising 8 GB of RAM.

[0106] In some cases, a surrogate model may comprise a mean squared error (MSE) of a percent difference between a value for a desired physical property output by a computer modelling software and the surrogate model of less than about 5%, 1%, 0.1%, 0.01%, 0.001%, or less.Optimization Framework

[0107] In some cases, a machine learning model for generating a nuclear fusion reactor configuration as disclosed herein may be implemented in a reinforcement learning, active learning, or other optimization framework. As such the machine learning model may be used to optimize nuclear fusion reactor configurations or the machine learning model itself over time.

[0108] In some cases, the optimization framework may be used to improve the accuracy or robustness of a nuclear fusion reactor configuration generating machine learning model. For example, a machine learning model may serve as a proxy model for more expensive calculations. As such, the machine learning model may provide an output that can be compared with computational or mathematical models of nuclear fusion reactors. In some cases, the output of the machine learning model may be periodically compared with a simulation or computational model of the nuclear fusion reactor. A loss function may be established comparing the output of the machine learning model with the simulation or computational model. The loss function may be used to generate a fine-tuning set that may be used to update the weights of the machine learning model based on performance of the machine learning model on post-training inference tasks as governed by the computational model or simulation. In some cases, a human may be used in addition or alternatively to the computational model or simulation in reinforcement learning.

[0109] In some cases, a configuration of a fusion device as disclosed herein may be subject to augmentation using other machine learning models or adjustments via computational modelling. For example, a configuration of a fusion device output by a machine learning model as disclosed herein may be altered for engineering, practicality, physics, or other considerations. Such configuration augmentation may be collected and used as training data for later machine learning model fine-tuning. This may provide particular utility by increasing the practicality, real-world applicability, or general accuracy of the configurations output by machine learning models disclosed herein.

[0110] In some cases, the augmentation of the configuration may be performed using computational modelling software. For example, VMEC, DESC, GVEC, HINT2, SPEC, PIES,TERPISCHORE, SFINCS, NEO, SIMPLE, ASCOT, GX, GENE, or other spectral, pseudo- spectral, finite element, finite difference, or other computation modelling software or paradigm may be used for modelling nuclear fusion reactors or nuclear fusion reactor (e.g., MHD) equilibrium, stability and transport conditions. Such augmentation from a configuration provided by a machine learning model disclosed herein may be performed with substantially lower computational or time overhead as compared to first principles-based optimization configuration determination.Configuration Active Learning[OHl] In some cases, a subset of highly similar configurations may comprise substantially different operating parameters. For example, a plurality of substantially different metrics used as input into a machine learning model herein may result in highly similar configurations. This may be due to the highly sensitive behavior of nuclear fusion reactors as a function of their configuration. In some cases, this phenomenon may also be due to the inadequacy of certain metric representations of nuclear fusion reactors, which may not effectively and robustly discriminate between configurations with substantially different operating parameters. This behavior may frustrate global solvers or optimizers aiming to predict a configuration for a desired property from a metric space that comprises all possible metric values for all possible metrics. Accordingly, herein is a method that may comprise active learning to obtain an optimized nuclear reactor configuration.

[0112] In some cases, a method for optimizing a nuclear fusion reactor configuration may be represented by the operations 500 shown in FIG. 5. In some cases, the operations 500 may comprise operation 510 of providing a plurality of nuclear fusion reactor configurations, operation 520 of generating a metric space based on the plurality of nuclear fusion reactors and using the metric space as input into an optimization framework, performing operation 530 of generating an optimized nuclear fusion reactor configuration using the optimization framework. In some cases, the nuclear fusion reactor configuration comprises a stellarator configuration.

[0113] In some cases, the plurality of nuclear fusion reactor configurations are selected based on a surrogate model as described herein. In some cases, the plurality of nuclear fusion reactor configurations are selected based on a similarity in the configurations of the plurality of nuclear fusion reactors. For example, a pressure profile, a cross section of the nuclear fusion reactor, a total magnetic flux, a rotational transform profile, a current profile, or a combination thereof may have similar values for a subset of a database of nuclear fusion reactor configurations. Insome cases, the plurality of nuclear fusion reactor configurations are selected from one or more databases.

[0114] In some cases, the active learning may comprise Bayesian optimization. In some cases, active learning may be used to train a machine learning, surrogate, or statistical model. In some cases, active learning may be used to train a Gaussian process model. In some cases, active learning may be used when data used to train a model or evaluate a nuclear fusion reactor configuration is expensive to obtain. For example, a computer modelling software as described herein may be powerful, but slow. Accordingly, active learning may be used to select a next simulation to perform with the computer modeling software. In some cases, the next simulation is selected based on an acquisition function. In some cases, an acquisition function comprises an expected improvement. In some cases, an acquisition function comprises a batch expected improvement. In some cases, an acquisition function comprises an upper confidence bound. In some cases, an acquisition function comprises a probability of improvement. In some cases, an acquisition function comprises Thompson sampling. In some cases, an acquisition function comprises batch energy-entropy. In some cases, an acquisition function comprises expected hypervolume improvement. In some cases, an acquisition function comprises noisy expected hypervolume improvement. In some cases, an acquisition function comprises batch noisy expected hypervolume improvement. In some cases, an acquisition function comprises Pareto efficient global optimization (ParEGO). In some cases, an acquisition function comprises noisy Pareto efficient global optimization (NParEGO). In some cases, an acquisition function comprises batch noisy Pareto efficient global optimization (qNParEGO). In some cases, the acquisition function may select a next simulation comprising a plurality of metrics used as input into the simulation. In some cases, the acquisition function may be used to sample the metric space. In some cases, the acquisition function may select metrics that are interpolated from metrics of a subset of configurations informing the active learning. For example, a convex hull of the Fourier representations of N distinct configurations may be established based on a subset of configurations of the database. Accordingly, the acquisition function may select (or sample) Fourier coefficients inside that convex hull, describing the boundary of a new configuration. N may be equal to 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30 or more. In some cases, the next simulation is selected based on a likelihood to decrease uncertainty of a model, a likelihood to provide an optimum, or a combination thereof. In some cases, an optimum is a minimum or maximum of a desired physical property. In some cases, active learning may provide particular utility by selecting a configuration to evaluate with a computer modelling software or surrogate model herein. Generally, active learning mayimplement an acquisition function that may balance exploration and exploitation to select the configuration to evaluate. Exploration may indicate the searching of the metric space that is not well defined by a model subject to the active learning. In some cases, exploitation may indicate the evaluation of a section of the metric space for an optimum. In some cases, after sampling the metric space to obtain a sample of the metric space, a high-fidelity computer modelling software or a surrogate model may take the sample of the metric space as input and output a value for a desired physical property. In some cases, the sample of the metric space and the value for the desired physical property may form a datum of a training dataset used to update a model subject to active learning.

[0115] In some cases, the active learning may be performed using Fourier coefficients describing a nuclear reactor configuration. In some cases, active learning may be performed based on a database of Fourier coefficients describing a plurality of nuclear fusion reactors. Accordingly, in some cases interpolation or extrapolation, e.g. via linear combinations, among Fourier coefficients describing a configuration may be used to determine an optimal configuration. In some cases, Fourier coefficients may be selected by an acquisition function of an active learning method. In some cases, the selected Fourier coefficients may be used as input into a computer modelling software. For example, Fourier coefficients may be used as input into VMEC computer modelling software. As such, the Fourier coefficients may be used to search a nuclear fusion reactor metric space to guide the selection and evaluation of particular configurations of the nuclear fusion reactor.In some cases, active learning may terminate based on meeting a stopping condition. In some cases, a stopping condition is a number of iterations, an accuracy of the machine learning model, or achievement of the optimized nuclear fusion reactor configuration. In some cases, achievement of the optimized nuclear fusion reactor configuration indicates a configuration satisfying a value for a desired property of interest has been found. In some cases, an accuracy of the machine learning model indicates that a model subject to active learning has a maximum degree of uncertainty. For example, the model may have different degrees of uncertainty for different sections of the metric space. As such, when a maximum of these uncertainties reaches a value, the active learning may be terminated as the model may be considered sufficiently trained. In some cases, an accuracy of the machine learning model may be a mean squared error (MSE) or other accuracy measure. In some cases, the model may be subsequently used for exploitation (configuration optimization) only. In some cases, however, an optimal configuration may be determined prior to the termination of model training. Generally, active learning mayprovide particular utility by facilitating the use of high-fidelity computer modelling software in configuration optimization.

[0116] In some cases, active learning may be used to populate a database. In the course of active learning, a plurality of high fidelity simulations may be performed. As such, the metrics and configurations explored may populate a database of nuclear fusion reactor configurations.Application of Configuration

[0117] In some cases, operations 100, operations 300, or operations 500 may optionally further comprise assembling the nuclear fusion reactor as predicted in operation 120, generated in operation 320, or optimized in operation 530. In some cases, the nuclear fusion reactor may be assembled based on an output of a surrogate model as generated in operations 400. For example, a machine learning model as disclosed herein may provide sufficient information, optionally with configuration fine-tuning, for the assembly, fabrication, or construction of a nuclear fusion reactor or nuclear fusion reactor component.

[0118] In some cases, assembly, fabrication, or construction may comprise 3D printing, lasercutting, milling, forging, casting, molding, or otherwise causing components of a fusion reactor to be physically produced. For example, an output of a machine learning model herein may be used as input into an automated or manually driven system for producing a component of a nuclear fusion reactor. The techniques disclosed herein may provide particular utility by decreasing the time necessary to design nuclear fusion reactors and their constituent components. As such, the outputs of machine learning models may be used as input (e.g., directly or after fine-tuning) into computer-aided design or other design software used to render and ultimately create physical nuclear fusion reactor components. In some cases, the output of the machine learning model may comprise a rendering or model directly used for producing a component herein. For example, a coil used in a stellarator or tokamak may be dependent on the configuration of the relevant device. As such, the configuration may be used to ultimately cause the fabrication of physical coils used to generate electro-magnetic fields in nuclear fusion reactors. As another example, the heat shield and plasma vessel of a stellarator are usually conformal to the plasma surface. As such, the spatial representation of the designed reactor may be used to determine the design of the heat shield or plasma vessel.

[0119] In some cases, the nuclear fusion reactor configuration output by a machine learning model herein may be used as input into a physics or computer-aided design software. For example, the configuration may be used as input into a plasma physics solver such as VMEC, GVEC, SPEC, HINT2, PIES, TERPISCHORE, SFINCS, NEO, SIMPLE, ASCOT, GX, GENE,or DESC or other modelling or simulation software. In another example, the configuration may be used as input into a computer-aided design software for the reduction of the configuration to physical components. For example, the configuration may be used to define, describe, render, model, or otherwise provide a spatial description for a nuclear fusion reactor or nuclear fusion reactor component such that the nuclear fusion reactor or nuclear fusion reactor component may be fabricated, constructed, assembled, printed, or otherwise formed from materials for implementation in a nuclear fusion reactor.

[0120] In some cases, the nuclear fusion reactor may comprise a stellarator, a tokamak, an inertial confinement device, or a magnetic confinement device. In some cases, a component of a nuclear fusion reactor may comprise an electrode, a coil, a planar coil, a non-planar coil, an inner vessel, an outer vessel, a plasma vessel, a vacuum vessel, a divertor, a blanket, a first wall, a shield, a heat shield, a support, a port, a solenoid, a poloidal coil, or a toroidal coil. In some cases, a component of a stellarator may comprise a divertor, a first wall, or a coil.Computer Systems

[0121] Referring to FIG. 6, a block diagram is shown depicting an exemplary machine that includes a computer system 600 (e.g., a processing or computing system) within which a set of instructions can execute for causing a device to perform or execute any aspect or methodology for static code scheduling of the present disclosure. The components in FIG. 6 are examples only and do not limit the scope of use or functionality of any hardware, software, embedded logic component, or a combination of two or more such components implementing particular embodiments. The computer system 600 may implement entirely or in part the training, clustering, generating, predicting, or other operations disclosed herein. For example, the computer system 600 may utilize the processor(s) 601 to implement model training or inference as described with operations 100 or operations 300. In another example, a user may interface with the computer system 600 via the one or more input device 633 or the network 630 to process, analyze, review, configure, run, or otherwise control or manage the computer- implemented methods for nuclear fusion reactor configuration generation herein.

[0122] Computer system 600 may include processor(s) 601, a memory 603, and a storage 608 that communicate with each other, and with other components, via a bus 640. The bus 640 may also link a display 632, one or more input device 633 (which may, for example, include a keypad, a keyboard, a mouse, a stylus, etc.), one or more output device 634, one or more storage device 635, and various tangible storage media 636. All of these elements may interface directly or via one or more interface or adaptor to the bus 640. For instance, the various tangible storagemedia 636 can interface with the bus 640 via storage medium interface 626. Computer system 600 may have any suitable physical form, including but not limited to one or more integrated circuit (ICs), printed circuit board (PCBs), mobile handheld device (such as mobile telephones or PDAs), laptop or notebook computer, distributed computer system, computing grid, or server.

[0123] Computer system 600 includes one or more processor 601 (e.g., central processing units (CPUs), general purpose graphics processing units (GPGPUs), or quantum processing units (QPUs)) that carry out functions. Processor(s) 601 optionally contains a cache memory unit 602 for temporary local storage of instructions, data, or computer addresses. Processor(s) 601 are configured to assist in execution of computer readable instructions. Computer system 600 may provide functionality for the components depicted in FIG. 6 as a result of the processor(s) 601 executing non-transitory, processor-executable instructions embodied in one or more tangible computer-readable storage medium, such as memory 603, storage 608, storage devices 635, or storage medium 636. The computer-readable media may store software that implements particular embodiments, and processor(s) 601 may execute the software. Memory 603 may read the software from one or more other computer-readable medium (such as mass storage device(s) 635, 636) or from one or more other source through a suitable interface, such as network interface 620. The software may cause processor(s) 601 to carry out one or more process or one or more step of one or more process described or illustrated herein. Carrying out such processes or steps may include defining data structures stored in memory 603 and modifying the data structures as directed by the software.

[0124] In some cases, the computer system 600 and the processor(s) 601 may be used to implement an active learning framework. In some cases, the processor(s) 601 may execute Bayesian optimization techniques. In some cases, Bayesian optimization may be implemented from a library such as BoTorch, Hyperopt, scikit-optimize, Ax-platform, or GPyOpt.

[0125] The memory 603 may include various components (e.g., machine readable media) including, but not limited to, a random access memory component (e.g., RAM 604) (e.g., static RAM (SRAM), dynamic RAM (DRAM), ferroelectric random access memory (FRAM), phasechange random access memory (PRAM), etc.), a read-only memory component (e.g., ROM 605), and any combinations thereof. ROM 605 may act to communicate data and instructions unidirectionally to processor(s) 601, and RAM 604 may act to communicate data and instructions bidirectionally with processor(s) 601. ROM 605 and RAM 604 may include any suitable tangible computer-readable media described below. In one example, a basic input / output system 606 (BIOS), including basic routines that help to transfer informationbetween elements within computer system 600, such as during start-up, may be stored in the memory 603.

[0126] Fixed storage 608 is connected bidirectionally to processor(s) 601, optionally through storage control unit 607. Fixed storage 608 provides additional data storage capacity and may also include any suitable tangible computer-readable media described herein. Storage 608 may be used to store operating system 609, executable(s) 610, data 611, applications 612 (application programs), and the like. Storage 608 can also include an optical disk drive, a solid-state memory device (e.g., flash-based systems), or a combination of any of the above. Information in storage 608 may, in appropriate cases, be incorporated as virtual memory in memory 603.

[0127] In one example, storage device(s) 635 may be removably interfaced with computer system 600 (e.g., via an external port connector (not shown)) via a storage device interface 625. Particularly, storage device(s) 635 and an associated machine-readable medium may provide non-volatile or volatile storage of machine-readable instructions, data structures, program modules, or other data for the computer system 600. In one example, software may reside, completely or partially, within a machine-readable medium on storage device(s) 635. In another example, software may reside, completely or partially, within processor(s) 601.

[0128] Bus 640 connects a wide variety of subsystems. Herein, reference to a bus may encompass one or more digital signal line serving a common function, where appropriate. Bus 640 may be any of several types of bus structures including, but not limited to, a memory bus, a memory controller, a peripheral bus, a local bus, and any combinations thereof, using any of a variety of bus architectures. As an example and not by way of limitation, such architectures include an Industry Standard Architecture (ISA) bus, an Enhanced ISA (EISA) bus, a Micro Channel Architecture (MCA) bus, a Video Electronics Standards Association local bus (VLB), a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCLX) bus, an Accelerated Graphics Port (AGP) bus, HyperTransport (HTX) bus, serial advanced technology attachment (SATA) bus, and any combinations thereof.

[0129] Computer system 600 may also include an input device 633. In one example, a user of computer system 600 may enter commands or other information into computer system 600 via input device(s) 633. Examples of an input device(s) 633 include, but are not limited to, an alphanumeric input device (e.g., a keyboard), a pointing device (e.g., a mouse or touchpad), a touchpad, a touch screen, a multi-touch screen, a joystick, a stylus, a gamepad, an audio input device (e.g., a microphone, a voice response system, etc.), an optical scanner, a video or still image capture device (e.g., a camera), and any combinations thereof. In some embodiments, theinput device is a Kinect, Leap Motion, or the like. Input device(s) 633 may be interfaced to bus 640 via any of a variety of input interfaces 623 (e.g., input interface 623) including, but not limited to, serial, parallel, game port, USB, FIREWIRE, THUNDERBOLT, or any combination of the above.

[0130] In particular embodiments, when computer system 600 is connected to network 630, computer system 600 may communicate with other devices, specifically mobile devices and enterprise systems, distributed computing systems, cloud storage systems, cloud computing systems, and the like, connected to network 630. Communications to and from computer system 600 may be sent through network interface 620. For example, network interface 620 may receive incoming communications (such as requests or responses from other devices) in the form of one or more packet (such as Internet Protocol (IP) packets) from network 630, and computer system 600 may store the incoming communications in memory 603 for processing. Computer system 600 may similarly store outgoing communications (such as requests or responses to other devices) in the form of one or more packet in memory 603 and communicated to network 630 from network interface 620. Processor(s) 601 may access these communication packets stored in memory 603 for processing.

[0131] Examples of the network interface 620 include, but are not limited to, a network interface card, a modem, and any combination thereof. Examples of a network 630 or network segment 630 include, but are not limited to, a distributed computing system, a cloud computing system, a wide area network (WAN) (e.g., the Internet, an enterprise network), a local area network (LAN) (e.g., a network associated with an office, a building, a campus or other relatively small geographic space), a telephone network, a direct connection between two computing devices, a peer-to-peer network, and any combinations thereof. A network, such as network 630, may employ a wired or a wireless mode of communication. In general, any network topology may be used.

[0132] Information and data can be displayed through a display 632. Examples of a display 632 include, but are not limited to, a cathode ray tube (CRT), a liquid crystal display (LCD), a thin film transistor liquid crystal display (TFT-LCD), an organic liquid crystal display (OLED) such as a passive-matrix OLED (PMOLED) or active-matrix OLED (AMOLED) display, a plasma display, and any combinations thereof. The display 632 can interface to the processor(s) 601, memory 603, and fixed storage 608, as well as other devices, such as input device(s) 633, via the bus 640. The display 632 is linked to the bus 640 via a video interface 622, and transport of data between the display 632 and the bus 640 can be controlled via the graphics control 621. In some embodiments, the display is a video projector. In some embodiments, the display is a head-mounted display (HMD) such as a VR headset. In further embodiments, suitable VR headsets include, by way of non-limiting examples, HTC Vive, Oculus Rift, Samsung Gear VR, Microsoft HoloLens, Razer OSVR, FOVE VR, Zeiss VR One, Avegant Glyph, Freefly VR headset, and the like. In still further embodiments, the display is a combination of devices such as those disclosed herein.

[0133] In addition to a display 632, computer system 600 may include one or more other peripheral output device 634 including, but not limited to, an audio speaker, a printer, a storage device, and any combinations thereof. Such peripheral output devices may be connected to the bus 640 via an output interface 624. Examples of an output interface 624 include, but are not limited to, a serial port, a parallel connection, a USB port, a FIREWIRE port, a THUNDERBOLT port, and any combinations thereof.

[0134] In addition, or as an alternative, computer system 600 may provide functionality as a result of logic hardwired or otherwise embodied in a circuit, which may operate in place of or together with software to execute one or more process or one or more step of one or more process described or illustrated herein. Reference to software in this disclosure may encompass logic, and reference to logic may encompass software. Moreover, reference to a computer- readable medium may encompass a circuit (such as an IC) storing software for execution, a circuit embodying logic for execution, or both, where appropriate. The present disclosure encompasses any suitable combination of hardware, software, or both.

[0135] Those of skill in the art will appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality.

[0136] The various illustrative logical blocks, modules, and circuits described in connection with the embodiments disclosed herein may be implemented or performed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination ofcomputing devices, e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessor in conjunction with a DSP core, or any other such configuration.

[0137] The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by one or more processor, or in a combination of the two. A software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium may be integral to the processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In the alternative, the processor and the storage medium may reside as discrete components in a user terminal.

[0138] In accordance with the description herein, suitable computing devices include, by way of non-limiting examples, server computers, desktop computers, laptop computers, notebook computers, sub-notebook computers, netbook computers, netpad computers, set-top computers, media streaming devices, handheld computers, Internet appliances, mobile smartphones, tablet computers, personal digital assistants, video game consoles, and vehicles. Computing devices may further comprise televisions, video players, and digital music players with optional computer network connectivity are suitable for use in the system described herein. Suitable tablet computers, in various embodiments, include those with booklet, slate, and convertible configurations, known to those of skill in the art.

[0139] In some embodiments, the computing device includes an operating system configured to perform executable instructions. The operating system is, for example, software, including programs and data, which manages the device’s hardware and provides services for execution of applications. A server operating system may include, by way of non-limiting examples, FreeBSD, OpenBSD, NetBSD®, Linux, Apple® Mac OS X Server®, Oracle® Solaris®, Windows Server®, and Novell® NetWare®. A suitable personal computer operating system may include, by way of non-limiting examples, Microsoft® Windows®, Apple® Mac OS X®, UNIX®, and UNIX-like operating systems such as GNU / Linux®. In some embodiments, the operating system is provided by cloud computing. A suitable mobile smartphone operating system include, by way of non-limiting examples, Nokia® Symbian® OS, Apple® iOS®, Research In Motion® BlackBerry OS®, Google® Android®, Microsoft® Windows Phone® OS, Microsoft® Windows Mobile® OS, Linux®, and Palm® WebOS®. Suitable media streaming device operating systemsmay include, by way of non-limiting examples, Apple TV®, Roku®, Boxee®, Google TV®, Google Chromecast®, Amazon Fire®, and Samsung® HomeSync®. Suitable video game console operating system may include, by way of non-limiting examples, Sony® PS3®, Sony® PS4®, Sony® PS5®, Microsoft® Xbox 660®, Microsoft® Xbox One, Microsoft® Xbox Series X, Microsoft® Xbox Series S, Nintendo® Wii®, Nintendo® Wii U®, Nintendo® Switch™, and Ouya®.

[0140] Another aspect of the disclosure herein describes a non-transitory, computer-readable medium comprising executable instructions, wherein when a processor, when executing the executable instructions, performs a method as described herein.

[0141] Methods and systems of the present disclosure can be implemented by way of one or more algorithms. An algorithm can be implemented by way of software upon execution by the central processing unit 605. The algorithm can, for example, implement methods for performing a non-classical computation described herein.Web Application

[0142] In some embodiments, a computer program includes a web application. In light of the disclosure provided herein, a web application, in various embodiments, utilizes one or more software framework and one or more database system. In some embodiments, a web application is created upon a software framework such as Microsoft® .NET or Ruby on Rails (RoR). In some embodiments, a web application utilizes one or more database system including, by way of non-limiting examples, relational, non-relational, object oriented, associative, XML, and document oriented database systems. In further embodiments, suitable relational database systems include, by way of non-limiting examples, Microsoft® SQL Server, mySQL™, and Oracle®. In some cases, a web application, in various embodiments, is written in one or more version of one or more language. A web application may be written in one or more markup language, presentation definition language, client-side scripting language, server-side coding language, database query language, or combinations thereof. In some embodiments, a web application is written to some extent in a markup language such as Hypertext Markup Language (HTML), Extensible Hypertext Markup Language (XHTML), or extensible Markup Language (XML). In some embodiments, a web application is written to some extent in a presentation definition language such as Cascading Style Sheets (CSS). In some embodiments, a web application is written to some extent in a client-side scripting language such as Asynchronous JavaScript and XML (AJAX), Flash® ActionScript, JavaScript, or Silverlight®. In some embodiments, a web application is written to some extent in a server-side coding language suchas Active Server Pages (ASP), ColdFusion®, Perl, Java™, JavaServer Pages (JSP), Hypertext Preprocessor (PHP), Python™, Ruby, Tel, Smalltalk, WebDNA®, or Groovy. In some embodiments, a web application is written to some extent in a database query language such as Structured Query Language (SQL). In some embodiments, a web application integrates enterprise server products such as IBM® Lotus Domino®. In some embodiments, a web application includes a media player element. In various further embodiments, a media player element utilizes one or more of many suitable multimedia technology including, by way of nonlimiting examples, Adobe® Flash®, HTML 5, Apple® QuickTime®, Microsoft® Silverlight®, Java™, and Unity®.

[0143] Referring to FIG. 7, in a particular embodiment, an application provision system comprises one or more database 700 accessed by a relational database management system (RDBMS) 710. Suitable RDBMSs include Firebird, MySQL, PostgreSQL, SQLite, Oracle Database, Microsoft SQL Server, IBM DB2, IBM Informix, SAP Sybase, Teradata, and the like. In this embodiment, the application provision system further comprises one or more application server 720 (such as Java servers, .NET servers, PHP servers, and the like) and one or more web server 730 (such as Apache, IIS, GWS and the like). The web server(s) optionally expose one or more web service via app application programming interfaces (APIs) 740. Via a network, such as the Internet, the system provides browser-based or mobile native user interfaces.

[0144] Referring to FIG. 8, in a particular embodiment, an application provision system alternatively has a distributed, cloud-based architecture 800 and comprises elastically load balanced, auto-scaling web server resources 810 and application server resources 820 as well synchronously replicated databases 830.Mobile Application

[0145] In some embodiments, a computer program includes a mobile application provided to a mobile computing device. In some embodiments, the mobile application is provided to a mobile computing device at the time it is manufactured. In other embodiments, the mobile application is provided to a mobile computing device via the computer network described herein.

[0146] In view of the disclosure provided herein, a mobile application may be created by using hardware, languages, and development environments. Additionally, mobile applications may be written in several languages. Suitable programming languages include, by way of non-limiting examples, C, C++, C#, Objective-C, Java™, JavaScript, Pascal, Object Pascal, Python™, Ruby, Rails, VB.NET, WML, and XHTML / HTML with or without CSS, or combinations thereof.

[0147] Suitable mobile application development environments are available from several sources. Commercially available development environments include, by way of non-limiting examples, AirplaySDK, alcheMo, Appcelerator®, Celsius, Bedrock, Flash Lite, .NET Compact Framework, Rhomobile, and WorkLight Mobile Platform. Other development environments are available without cost including, by way of non-limiting examples, Lazarus, MobiFlex, MoSync, and Phonegap. Also, mobile device manufacturers distribute software developer kits including, by way of non-limiting examples, iPhone and iPad (iOS) SDK, Android™ SDK, BlackBerry® SDK, BREW SDK, Palm® OS SDK, Symbian SDK, webOS SDK, and Windows® Mobile SDK.

[0148] Several commercial forums may be available for distribution of mobile applications including, by way of non-limiting examples, Apple® App Store, Google® Play, Chrome WebStore, BlackBerry® App World, App Store for Palm devices, App Catalog for webOS, Windows® Marketplace for Mobile, Ovi Store for Nokia® devices, Samsung® Apps, and Nintendo® DSi Shop.Standalone Application

[0149] In some embodiments, a computer program includes a standalone application, which is a program that is run as an independent computer process, not an add-on to an existing process, e.g., not a plug-in. A standalone applications may be compiled. A compiler is a computer program(s) that transforms source code written in a programming language into binary object code such as assembly language or machine code. Suitable compiled programming languages include, by way of non-limiting examples, C, C++, Objective-C, COBOL, Delphi, Eiffel, Java™, Lisp, Python™, Visual Basic, and VB .NET, or combinations thereof. Compilation is often performed, at least in part, to create an executable program. In some embodiments, a computer program includes one or more executable complied application.Web-Browser Plug-In

[0150] In some embodiments, the computer program includes a web browser plug-in (e.g., extension, etc.). In computing, a plug-in is one or more software component that add specific functionality to a larger software application. Makers of software applications support plug-ins to enable third-party developers to create abilities which extend an application, to support easily adding new features, and to reduce the size of an application. When supported, plug-ins enable customizing the functionality of a software application. For example, plug-ins are commonly used in web browsers to play video, generate interactivity, scan for viruses, and display particular file types. Several web browser plug-ins including, Adobe® Flash® Player, Microsoft®Silverlight®, and Apple® QuickTime®. In some embodiments, the toolbar comprises one or more web browser extension, add-in, or add-on. In some embodiments, the toolbar comprises one or more explorer bar, tool band, or desk band.

[0151] In view of the disclosure provided herein, several plug-in frameworks are available that enable development of plug-ins in various programming languages, including, by way of nonlimiting examples, C++, Delphi, Java™, PHP, Python™, and VB .NET, or combinations thereof.

[0152] Web browsers (also called Internet browsers) are software applications, designed for use with network-connected computing devices, for retrieving, presenting, and traversing information resources on the World Wide Web. Suitable web browsers include, by way of nonlimiting examples, Microsoft® Internet Explorer®, Mozilla® Firefox®, Google® Chrome, Apple® Safari®, Opera Software® Opera®, and KDE Konqueror. In some embodiments, the web browser is a mobile web browser. Mobile web browsers (also called microbrowsers, mini-browsers, and wireless browsers) are designed for use on mobile computing devices including, by way of nonlimiting examples, handheld computers, tablet computers, netbook computers, subnotebook computers, smartphones, music players, personal digital assistants (PDAs), and handheld video game systems. Suitable mobile web browsers include, by way of non-limiting examples, Google® Android® browser, RIM BlackBerry® Browser, Apple® Safari®, Palm® Blazer, Palm® WebOS® Browser, Mozilla® Firefox® for mobile, Microsoft® Internet Explorer® Mobile, Amazon® Kindle® Basic Web, Nokia® Browser, Opera Software® Opera® Mobile, and Sony® PSP™ browser.Software Modules

[0153] In some embodiments, the platforms, systems, media, and methods disclosed herein include software, server, or database modules, or use of the same. In view of the disclosure provided herein, software modules may be created using machines, software, and languages. The software modules disclosed herein are implemented in a multitude of ways. In various embodiments, a software module comprises a file, a section of code, a programming object, a programming structure, a distributed computing resource, a cloud computing resource, or combinations thereof. In further various embodiments, a software module comprises a plurality of files, a plurality of sections of code, a plurality of programming objects, a plurality of programming structures, a plurality of distributed computing resources, a plurality of cloud computing resources, or combinations thereof. In various embodiments, the one or more software modules comprise, by way of non-limiting examples, a web application, a mobileapplication, a standalone application, and a distributed or cloud computing application. In some embodiments, software modules are in one computer program or application. In other embodiments, software modules are in more than one computer program or application. In some embodiments, software modules are hosted on one machine. In other embodiments, software modules are hosted on more than one machine. In further embodiments, software modules are hosted on a distributed computing platform such as a cloud computing platform. In some embodiments, software modules are hosted on one or more machine in one location. In other embodiments, software modules are hosted on one or more machine in more than one location.Databases

[0154] In some embodiments, the platforms, systems, media, and methods disclosed herein include one or more database, or use of the same. In view of the disclosure provided herein, many databases are suitable for storage and retrieval of reactor device configurations or desired physical descriptors of such reactor device configurations, or any combination thereof. In various embodiments, suitable databases include, by way of non-limiting examples, relational databases, non-relational databases, object oriented databases, object databases, entityrelationship model databases, associative databases, XML databases, document oriented databases, and graph databases. Further non-limiting examples include SQL, PostgreSQL, MySQL, Oracle, DB2, Sybase, and MongoDB. In some embodiments, a database is Internetbased. In further embodiments, a database is web-based. In still further embodiments, a database is cloud computing-based. In a particular embodiment, a database is a distributed database. In other embodiments, a database is based on one or more local computer storage device.Data Transmissions

[0155] The subject matter described herein, including methods and systems as described herein and may be configured to be performed in one or more facility at one or more location. Facility locations are not limited by country and include any country or territory. In some instances, one or more step is performed in a different country than another step of the method. In some embodiments, one or more method step involving a computer system are performed in a different country than another step of the methods provided herein. In some embodiments, data processing and storage are performed in a different country or location than one or more step of the methods described herein. In some embodiments, one or more product or data are transferred from one or more of the facility to one or more different facility for analysis or further analysis. Data includes, but is not limited to, information regarding the stratification of a subject, and any data produced by the methods disclosed herein. In some embodiments of the methods andsystems described herein, the subject information is compiled, and a subsequent data transmission step will transmit or store the subject information.

[0156] In some embodiments, any step of any method described herein is performed by a software program or module on a computer. In additional or further embodiments, data from any step of any method described herein is transferred to and from facilities located within the same or different countries, including analysis performed in one facility in a particular location and the data shipped to another location or directly to an individual in the same or a different country. In additional or further embodiments, data from any step of any method described herein is transferred to or received from a facility located within the same or different countries, including analysis of a data input, such as queries, objects, properties, types, filters, tables, or any combination thereof, performed in one facility in a particular location and corresponding data transmitted to another location.COMPUTATIONAL REACTOR CONFIGURATION DESIGN

[0157] Provided herein are systems, methods, computer-readable media, and techniques for configuring a stellarator, including: (a) obtaining one or more parameters of a target stellarator; (b) generating a plurality of target stellarator approximations based at least in part on the one or more parameters of the target stellarator, wherein at least a subset of the plurality target stellarator approximations: (i) have a toroidal profile, and (ii) comprise a plurality of concentric toroids; and (c) presenting at a graphical user interface a representation of the plurality of target stellarator approximations.

[0158] In some cases, the complexity of fusion systems may present a challenge in determining the optimal configuration of a given conceptual design without searching a wide parameter space. This design space is even wider in a stellarator than in a tokamak as stellarators vary in both toroidal and poloidal directions, and therefore do not benefit from the symmetry of tokamaks. Therefore, a workflow may be beneficial that allows for the reduction of this parameter space much as possible prior to turning to higher fidelity models. This workflow may facilitate more rapid evaluation of concepts and therefore faster design iteration.

[0159] In some cases, workflows are disclosed herein for which simple toroidal models (which can be generated and evaluated quickly) are used to explore the parameter space, and these results are then used to inform a 3D model built using the parametric stellarator geometry tool (e.g., Parastell).

[0160] In one workflow, starting with plasma equilibrium data and the corresponding magnet filament data, the available radial build distance at an array of locations (e.g., theta, phi anglecombinations in flux coordinates) may be evaluated to set the limit of the radial build length at those locations. Next, in some cases, the neutron wall loading (NWL) may be evaluated using Monte Carlo particle simulation software (such as MCNP, OpenMC, or the like) and functionality in a parametric geometry tool for stellarators (e.g., ParaStell) to produce a map of NWL vs. phi, theta. These results (NWL and radial build length) may set the limits on the parameter space to be explored.

[0161] Next, in some cases, simple toroidal models may be generated using the native geometry available in OpenMC. These models may have similar major and minor radii to the stellarator configuration. In some cases, the radial build and NWL in these models may be varied parametrically to gather data on key neutronics responses across the parameter space. In some cases, this data may be tabulated by NWL and total radial build length so it can be accessed while building the stellarator model in a parametric stellarator geometry tool (e.g., Parastell).

[0162] In some cases, a parametric stellarator geometry tool (e.g., Parastell) provides for the generation of concentric layers based at least in part on the plasma equilibrium data. The thickness of each layer can be varied both toroidally and poloidally. Using the tabulated data, in some cases, a radial build that matches the available radial distance and local NWL may be chosen. This may result in a model which meets, or is close to meeting the same neutronics responses, which can then, in some cases, be iterated upon further.

[0163] In another workflow, a parametric model (e.g., toroidal model) of the selected blanket concept may be created (e.g., in OpenMC). A simple torus may be used for the toroidal model geometry. A dataset indexable by Neutron Wall Loading (NWL) and radial build length may then be created (e.g., an index). Maps of NWL and available radial build length versus toroidal position and poloidal position may then be produced via a parametric stellarator geometry tool (e.g, via Parastell, OpenMC, or Coreform Cubit). A parametric stellarator geometry tool (e.g., Parastell) may then be used to create a 3D stellarator model by selecting locally optimal radial builds from the parametric data and maps of NWL and radial build length. A prediction of the local responses in the 3D stellarator model may then be made using the locally optimized radial builds and the parametric data. OpenMC may then be used to transport neutrons in the 3D model for comparison against the predicted local responses.Computationally Configuring a Stellarator

[0164] In some cases, a stellarator structure may have a complex shape with twists, which may result in variation in both toroidal and poloidal directions. An example of a complex stellarator structure geometry is demonstrated in FIG. 13A. An example of a stellarator structure geometrymay be modeled by a parametric stellarator geometry tool (e.g., Parastell). The coils may be represented as a constant thickness layer (shown as the outermost layer).

[0165] The stellarator structure may have a plurality of layers, where each layer may be of a different thickness. In designing a stellarator, determining a thickness for each layer of the plurality of layers may improve performance of the stellarator. A stellarator configuration may have a different combination of thicknesses for each layer of the plurality of layers. Thus, performing many simulations of various stellarator configurations may be computationally expensive. The high computational expense in modeling a stellarator in 3D is illustrated and described further in FIG. 13B.

[0166] In some cases, the computational expense may be prohibitively high. In some cases, the computations may be intractable. To improve computability of the various stellarator configurations, a stellarator may be approximated as a toroid in a toroidal model.

[0167] Further, each layer of the stellarator may be approximated by concentric toroids atop a central toroid. For example, the central toroid may approximate the plasma within a stellarator. The concentric solid toroidal shell most closely around the central toroid may approximate the scrape-off standoff layer of the first wall of the stellarator. Then, the next outer toroidal shell may approximate a first wall structural material of the stellarator. Then, the next outer toroidal shell may approximate a breeder material as a component of the blanket. This may continue with each subsequent layer of the toroid approximating another layer of the stellarator blanket, thus generating a toroidal model. The layers of the stellarator blanket may comprise a first wall, a vacuum vessel, a high-temperature shield, a low-temperature shield, or magnets. The solution for automating a workflow that facilitates rapid, high-throughput, multi-parameter global optimization of a radial build may be illustrated and described further in FIG. 14A. Example sample radial builds modeling the thickness for each layer of the plurality of layers in a stellarator are demonstrated in FIGs. 14A, 14B. The radial build may be based on the FNSF DCLL concept. The thickness of the breeder, HTS, or LTS may vary. The solution may be implemented using software (e.g., Python code).

[0168] In some cases, the stellarator may be approximated by about 2 concentric toroids to about 100 concentric toroids. In some cases, the stellarator may be approximated by about 2 concentric toroids to about 5 concentric toroids, about 2 concentric toroids to about 10 concentric toroids, about 2 concentric toroids to about 25 concentric toroids, about 2 concentric toroids to about 50 concentric toroids, about 2 concentric toroids to about 100 concentric toroids, about 5 concentric toroids to about 10 concentric toroids, about 5 concentric toroids toabout 25 concentric toroids, about 5 concentric toroids to about 50 concentric toroids, about 5 concentric toroids to about 100 concentric toroids, about 10 concentric toroids to about 25 concentric toroids, about 10 concentric toroids to about 50 concentric toroids, about 10 concentric toroids to about 100 concentric toroids, about 25 concentric toroids to about 50 concentric toroids, about 25 concentric toroids to about 100 concentric toroids, or about 50 concentric toroids to about 100 concentric toroids. In some cases, the stellarator may be approximated by about 2 concentric toroids, about 5 concentric toroids, about 10 concentric toroids, about 25 concentric toroids, about 50 concentric toroids, or about 100 concentric toroids. In some cases, the stellarator may be approximated by at least about 2 concentric toroids, about 5 concentric toroids, about 10 concentric toroids, about 25 concentric toroids, or about 50 concentric toroids. In some cases, the stellarator may be approximated by at most about 5 concentric toroids, about 10 concentric toroids, about 25 concentric toroids, about 50 concentric toroids, or about 100 concentric toroids.

[0169] In some cases, to perform a computational simulation, a plurality of concentric toroids with different thicknesses may approximate a stellarator with a plurality of layers of different thicknesses. In some cases, a plurality of concentric toroids with different thicknesses may be indexed to provide a plurality of combinations of thicknesses to approximate a stellarator with a plurality of layers of different thicknesses. In some cases, at least one stellarator may be simulated in this manner. In some cases, a number of stellarators may be simulated in this manner. For example, as provided for at least in part by the concentric toroidal approximation, a large number of stellarators may be simulated. In some cases, the number of stellarators simulated may be about 2 stellarators to about 1,000 stellarators. In some cases, the number of stellarators simulated may be about 2 stellarators to about 5 stellarators, about 2 stellarators to about 10 stellarators, about 2 stellarators to about 25 stellarators, about 2 stellarators to about 50 stellarators, about 2 stellarators to about 100 stellarators, about 2 stellarators to about 500 stellarators, about 2 stellarators to about 1,000 stellarators, about 5 stellarators to about 10 stellarators, about 5 stellarators to about 25 stellarators, about 5 stellarators to about 50 stellarators, about 5 stellarators to about 100 stellarators, about 5 stellarators to about 500 stellarators, about 5 stellarators to about 1,000 stellarators, about 10 stellarators to about 25 stellarators, about 10 stellarators to about 50 stellarators, about 10 stellarators to about 100 stellarators, about 10 stellarators to about 500 stellarators, about 10 stellarators to about 1,000 stellarators, about 25 stellarators to about 50 stellarators, about 25 stellarators to about 100 stellarators, about 25 stellarators to about 500 stellarators, about 25 stellarators to about 1,000 stellarators, about 50 stellarators to about 100 stellarators, about 50 stellarators to about 500stellarators, about 50 stellarators to about 1,000 stellarators, about 100 stellarators to about 500 stellarators, about 100 stellarators to about 1,000 stellarators, or about 500 stellarators to about 1,000 stellarators. In some cases, the number of stellarators simulated may be about 2 stellarators, about 5 stellarators, about 10 stellarators, about 25 stellarators, about 50 stellarators, about 100 stellarators, about 500 stellarators, or about 1,000 stellarators. In some cases, the number of stellarators simulated may be at least about 2 stellarators, about 5 stellarators, about 10 stellarators, about 25 stellarators, about 50 stellarators, about 100 stellarators, or about 500 stellarators. In some cases, the number of stellarators simulated may be at most about 5 stellarators, about 10 stellarators, about 2 stellarators 5 stellarators, about 50 stellarators, about 100 stellarators, about 500 stellarators, or about 1,000 stellarators.

[0170] In some cases, once the number of stellarators are simulated (e.g., using concentric toroids as an approximation), one or more of the simulated stellarators may be analyzed. This analysis may evaluate the efficacy of the one or more simulated stellarators. An example workflow for determining an effective stellarator structure is illustrated in FIG. 15A.

[0171] In some cases, the analysis may aim to identify the smallest stellarator (e.g., smallest in width, smallest in height, smallest material consumption, etc.) that satisfies a threshold efficacy. In some cases, the threshold efficacy may be based at least in part on radiative gain patterns of the one or more simulated stellarators. For example, the threshold efficacy may include a minimum threshold of radiation on the outer edge of the one or more simulated stellarators. In another example, the threshold efficacy may include a standard for tritium breeding. In another example, the threshold efficacy may include a standard of reducing radiation damage within materials. In another example, the threshold efficacy may include a standard of radiation heating within materials. In another example, the threshold efficacy may include a standard for neutron fluence within materials. In another example, the threshold efficacy may include a standard for nuclear heating within materials. In another example, the threshold efficacy may include a standard for damage in units of displacements per atom within materials. In another example, the threshold efficacy may include a standard for helium production via transmutation within materials. The value of the systems, the methods, the computer-readable media, and the techniques disclosed herein to enabling quickly narrowing a parameter search field to find effective (e.g., optimal values) may be illustrated and described further in FIG. 15B.

[0172] In some cases, the analysis may involve searching the index for at least one combination of layer thicknesses (e.g., radial build) that, at each toroidal-poloidal coordinate, may meet at least one design parameter or threshold efficacy. The design parameter or threshold efficacy may be minimizing helium production in the vacuum vessel, minimizing predicted fast fluencein the superconducting magnet coils, or fitting within the radial space availability of the stellarator configuration geometry. The analysis may generate at least one 2D representation in a toroidal or poloidal space of spatial 3D neutronics distributions (e.g., predicted NWL values) and at least one 2D representation in a toroidal or poloidal space of available radial distance values (FIG. 15B, panel (c)). Determining the available radial distance values may involve indexing radial distance availability in both toroidal and poloidal angles (e.g., theta and phi). Space constraints, imposed from the outer limit by magnet geometry and from the inner limit by plasma geometry, may cause radial space availability for building blanket structures between these limits to vary in toroidal and poloidal angles across the stellarator configuration geometry.

[0173] The analysis may overlay these 2D representations. An overlay utilizing predicted NWL values may indicate a radial build that results in the coldest corresponding neutronics distributions.

[0174] In some cases, if an acceptable radial build is found, a minimum radial build may be selected (FIG. 15A). In some cases, if multiple acceptable radial builds are found within the index, the build with the thickest breeder material layer may be selected. If the build with the thickest breeder material layer is selected, any extra radial space may also assigned to the breeder material layer.

[0175] In some cases, if an acceptable build is not found (FIG. 15A), a minimum radial build may be selected and any extra radial distance may assigned to high temperature shielding. For example, regions where helium production cannot be adequately minimized may be designated to comprise high temperature shielding (FIG. 16). Regions where helium production may not be adequately minimized may not have sufficient available radial space to build mitigating shielding structures.

[0176] In some cases, once a minimum radial build is selected, the minimum radial build may be modeled in a higher fidelity (e.g., 3D stellarator) model. The predicted local responses for various neutronics responses determined by a toroidal model may be compared to the calculated local responses determined by the higher fidelity model to determine the accuracy of the initial analysis. In one example, predicted and calculated local responses for helium production in the vacuum vessel may be compared (FIG. 17A). In another example, predicted and calculated local responses for fast fluence in the superconducting coils may be compared (FIG. 17B). Contour lines may be used to mark a maximum desired value for the respective local response. In FIG.17 A, the maximum desired value is noted as 0.2 APPM / FPY; in FIG. 17B, 10Al 8 n / cm2perFPY.

[0177] A toroidal model may underpredict the calculated responses determined by the 3D stellarator model. A toroidal model may predict the calculated responses determined by the 3D stellarator model. A toroidal model may overpredict the calculated responses determined by the 3D stellarator model. Both the toroidal model and the 3D stellarator model may indicate that a stellarator configuration combined with a combination of thicknesses for each layer of the plurality of layers will exceed response limits.

[0178] A toroidal model may predict neutronics responses with parametric data on the order of minutes or hours. A 3D stellarator model may predict neutronics responses with parametric data on the order of days. In some cases, the present invention may increase optimization workflow by decreasing a time for modeling one or more neutronics responses. For example, a toroidal model described herein may predict a neutronics response faster than a full stellarator model. Thus, the workflow described herein may automates the search for feasible radial builds based on neutronic performance. The workflow described herein may expedite graduation to higher fidelity neutronics models. The workflow described herein may quickly identify promising configurations. The workflow described herein may also eliminate configurations that exceed neutronics limits.

[0179] In some cases, the method may be applied to multiple configurations and blanket concepts. In some cases, the method may be applied to a detailed representation of the coils. In some cases, the method may select an “optimal” radial build. In some cases, the method may determine the available radial build distance.DEFINITIONS AND ADDITIONAL CONSIDERATIONS

[0180] Unless defined otherwise, all terms of art, notations and other technical and scientific terms or terminology used herein are intended to have the same meaning as is commonly understood by one of ordinary skill in the art to which the claimed subject matter pertains. Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. In some cases, terms with commonly understood meanings are defined herein for clarity or for ready reference, and the inclusion of such definitions herein should not necessarily be construed to represent a substantial difference over what is generally understood in the art.

[0181] As used in this specification and the appended claims, the terms “machine learning,” “machine learning techniques,” “machine learning algorithm,” “machine learning operation,” and “machine learning model” generally refer to any system or analytical or statistical procedure that may progressively improve computer performance of a task.

[0182] As used herein, the singular forms “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise. The use of the alternative (e.g., “or”) should be understood to mean either one, both, or any combination thereof of the alternatives.

[0183] Whenever the term “at least,” “greater than,” or “greater than or equal to” precedes the first numerical value in a series of two or more numerical values, the term “at least,” “greater than” or “greater than or equal to” applies to each of the numerical values in that series of numerical values. For example, greater than or equal to 1, 2, or 3 is equivalent to greater than or equal to 1, greater than or equal to 2, or greater than or equal to 3.

[0184] Whenever the term “no more than,” “less than,” or “less than or equal to” precedes the first numerical value in a series of two or more numerical values, the term “no more than,” “less than,” or “less than or equal to” applies to each of the numerical values in that series of numerical values. For example, less than or equal to 3, 2, or 1 is equivalent to less than or equal to 3, less than or equal to 2, or less than or equal to 1.

[0185] Certain inventive embodiments herein contemplate numerical ranges. When ranges are present, the ranges include the range endpoints. Additionally, every sub range and value within the range is present as if explicitly written out.

[0186] The term “about” or “approximately” may mean within an acceptable error range for the particular value, which will depend in part on how the value is measured or determined, e.g., the limitations of the measurement system. For example, “about” may mean within 1 or more than 1 standard deviation, per the practice in the art. Alternatively, “about” may mean a range of up to 20%, up to 10%, up to 5%, or up to 1% of a given value. Where particular values are described in the application and claims, unless otherwise stated the term “about” meaning within an acceptable error range for the particular value may be assumed.

[0187] As used herein, “or” is intended to mean an “inclusive or” or what is also known as a “logical OR,” wherein when used as a logic statement, the expression “A or B” is true if either A or B is true, or if both A and B are true, and when used as a list of elements, the expression “A, B or C” is intended to include all combinations of the elements recited in the expression, for example, any of the elements selected from the group consisting of A, B, C, (A, B), (A, C), (B, C), and (A, B, C); and so on if additional elements are listed. As such, any reference to “or” herein is intended to encompass “and / or” unless otherwise stated.

[0188] It will be understood that when an element such as a layer, region, or substrate is referred to as being “on” or extending “onto” another element, it may be directly on or extend directly onto the other element or intervening elements may also be present. In contrast, when anelement is referred to as being “directly on” or extending “directly onto” another element, there are no intervening elements present. Likewise, it will be understood that when an element such as a layer, region, or substrate is referred to as being “over” or extending “over” another element, it may be directly over or extend directly over the other element or intervening elements may also be present. In contrast, when an element is referred to as being “directly over” or extending “directly over” another element, there are no intervening elements present. It will also be understood that when an element is referred to as being “connected” or “coupled” to another element, it may be directly connected or coupled to the other element or intervening elements may be present. In contrast, when an element is referred to as being “directly connected” or “directly coupled” to another element, there are no intervening elements present.

[0189] Relative terms such as “below” or “above” or “upper” or “lower” or “horizontal” or “vertical” may be used herein to describe a relationship of one element, layer, or region to another element, layer, or region as illustrated in the Figures. It will be understood that these terms and those discussed above are intended to encompass different orientations of the device in addition to the orientation depicted in the Figures.

[0190] It will be understood that, although the terms “first,” “second,” “third,” etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element may be termed a second element, and, similarly, a second element may be termed a first element, without departing from the scope of the present disclosure.

[0191] While preferred embodiments of the present invention have been shown and disclosed herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. It is not intended that the invention be limited by the specific examples provided within the specification. While the invention has been described with reference to the aforementioned specification, the descriptions and illustrations of the embodiments herein are not meant to be construed in a limiting sense. Numerous variations, changes, and substitutions will now occur to those skilled in the art without departing from the invention. Furthermore, it shall be understood that all aspects of the invention are not limited to the specific depictions, configurations or relative proportions set forth herein which depend upon a variety of conditions and variables. It should be understood that various alternatives to the embodiments of the invention disclosed herein may be employed in practicing the invention. It is therefore contemplated that the invention shall also cover any such alternatives, modifications, variations, or equivalents. It is intended that the following claims define the scope of the invention and thatmethods and structures within the scope of these claims and their equivalents be covered thereby.

[0192] It should be noted that various illustrative or suggested ranges set forth herein are specific to their example embodiments and are not intended to limit the scope or range of disclosed technologies, but, again, merely provide example ranges for frequency, amplitudes, etc. associated with their respective embodiments or use cases. Where values are described as ranges, it will be understood that such disclosure includes the disclosure of all possible subranges within such ranges, as well as specific numerical values that fall within such ranges irrespective of whether a specific numerical value or specific sub-range is expressly stated.

[0193] It should be understood that, unless a term is expressly defined in this patent, there is no intent to limit the meaning of that term, either expressly or by implication, beyond its plain or ordinary meaning, and such term should not be interpreted to be limited in scope based at least in part on any statement made in any section of this patent (other than the language of the claims). To the extent that any term recited in the claims at the end of this patent is referred to in this patent in a manner consistent with a single meaning, that is done for sake of clarity only so as to not confuse the reader, and it is not intended that such claim term be limited, by implication or otherwise, to that single meaning.

[0194] Throughout this specification, plural instances may implement components, operations, or structures described as a single instance. Although individual operations of one or more methods are illustrated and described as separate operations, one or more of the individual operations may be performed concurrently, and nothing requires that the operations be performed in the order illustrated. Structures and functionality presented as separate components in example configurations may be implemented as a combined structure or component.Similarly, structures and functionality presented as a single component may be implemented as separate components. These and other variations, modifications, additions, and improvements fall within the scope of the subject matter herein.

[0195] Additionally, certain embodiments are disclosed herein as including logic or a number of routines, subroutines, applications, or instructions. These may constitute either software (e.g., code embodied on a machine-readable medium) or hardware. In hardware, the routines, etc., are tangible units capable of performing certain operations and may be configured or arranged in a certain manner. In example embodiments, one or more computer systems (e.g., a standalone, client or server computer system) or one or more hardware modules of a computer system (e.g., a processor or a group of processors) may be configured by software (e.g., an application orapplication portion) as a hardware module that operates to perform certain operations as disclosed herein.

[0196] In various embodiments, a hardware module may be implemented mechanically or electronically. For example, a hardware module may comprise dedicated circuitry or logic that is permanently configured (e.g., as a special-purpose processor, such as a field programmable gate array (FPGA) or an application-specific integrated circuit (ASIC) to perform certain operations. A hardware module may also comprise programmable logic or circuitry (e.g., as encompassed within a general-purpose processor or other programmable processor) that is temporarily configured by software to perform certain operations. It will be appreciated that the decision to implement a hardware module mechanically, in dedicated and permanently configured circuitry, or in temporarily configured circuitry (e.g., configured by software) may be driven by cost and time considerations.

[0197] Accordingly, hardware modules may encompass a tangible entity, be that an entity that is physically constructed, permanently configured (e.g., hardwired), or temporarily configured (e.g., programmed) to operate in a certain manner or to perform certain operations disclosed herein. Considering embodiments in which hardware modules are temporarily configured (e.g., programmed), each of the hardware modules need not be configured or instantiated at any one instance in time. For example, where the hardware modules comprise a general-purpose processor configured using software, the general-purpose processor may be configured as respective different hardware modules at different times. Software may accordingly configure processor, for example, to constitute a particular hardware module at one instance of time and to constitute a different hardware module at a different instance of time.

[0198] Hardware modules may provide information to, and receive information from, other hardware modules. Accordingly, the described hardware modules may be regarded as being communicatively coupled. Where multiple of such hardware modules exist contemporaneously, communications may be achieved through signal transmission (e.g., over appropriate circuits and buses) that connect the hardware modules. In embodiments in which multiple hardware modules are configured or instantiated at different times, communications between such hardware modules may be achieved, for example, through the storage and retrieval of information in memory structures to which the multiple hardware modules have access. For example, one hardware module may perform an operation and store the output of that operation in a memory device to which it is communicatively coupled. A further hardware module may then, at a later time, access the memory device to retrieve and process the stored output. Hardware modules may also initiate communications with input or output devices, and mayoperate on a resource (e.g., a collection of information). Elements that are described as being coupled and or connected may refer to two or more elements that may be (e.g., direct physical contact) or may not be (e.g., electrically connected, communicatively coupled, etc.) in direct contact with each other, but yet still cooperate or interact with each other.

[0199] The various operations of example methods disclosed herein may be performed, at least partially, by one or more processors that are temporarily configured (e.g., by software) or permanently configured to perform the relevant operations. Whether temporarily or permanently configured, such processors may constitute processor-implemented modules that operate to perform one or more operations or functions. The modules referred to herein may, in some example embodiments, comprise processor-implemented modules.

[0200] Similarly, the methods or routines disclosed herein may be at least partially processor- implemented. For example, at least some of the operations of a method may be performed by one or more processors or processor-implemented hardware modules. The performance of certain operations may be distributed among the one or more processors, not only residing within a single machine, but deployed across a number of machines. In some example embodiments, the processor or processors may be located in a single location (e.g., within a home environment, an office environment or as a server farm), while in other embodiments the processors may be distributed across a number of locations.

[0201] The performance of certain operations may be distributed among the one or more processors, not only residing within a single machine, but deployed across a number of machines. In some example embodiments, the one or more processors or processor-implemented modules may be located in a single geographic location (e.g., within a home environment, an office environment, or a server farm). In other example embodiments, the one or more processors or processor-implemented modules may be distributed across a number of geographic locations.EXAMPLES

[0202] The following examples are included for illustrative purposes only and are not intended to limit the scope of the inventive concepts.Example 1: Stellarator Configuration Prediction

[0203] A user seeking to develop a stellarator devises a set of descriptors that balance engineering and physics considerations for a practical stellarator. The set of descriptors comprises a set of shape metrics, a set of MHD stability metrics, a set of neoclassical transport metrics, a set of energetic particle confinement metrics, a set of turbulent transport metrics, a setof magnetic field metrics, and a set of coil design metrics. The set of descriptors are fed into a user interface of a computer system implementing the computer-implemented methods disclosed herein. The set of descriptors are first classified according to a plurality of physics-based rules as illustrated in FIG. 7. Following the classification based on physics-based rules, the set of descriptors is further classified based on a machine learning clustering algorithm. The machinelearning clustering algorithm provides a classification for the set of descriptors that is mapped to a machine learning model suitable for the prediction of a stellarator configuration for the set of descriptors provided by the user.

[0204] Upon retrieval of the appropriate machine learning model, the set of descriptors are used as an input into the machine learning model as shown in FIG. 8. The user applies the descriptors, or metrics, as input into the machine learning model, in this case a neural network, which is trained to output a pressure profile and a plurality of stellarator cross sections at a plurality of toroidal angles. Additionally, the neural network outputs a total magnetic flux, a rotational transform profile, and a current profile of the stellarator. The plurality of stellarator cross sections, the pressure profile, the rotational transform profile, the current profile, and the total magnetic flux serve as a full description of the configuration of the stellarator. The user applies the output of the neural network as input into a software for performing calculations on the stellarator configuration or for modelling the stellarator configuration in three dimensions.Example 2: Training a Surrogate Model

[0205] A surrogate model is established for a group of quasi-isodynamic (QI) stellarators. The surrogate model is trained to output a desired physical property (K) along a direction z for use in selecting a subset of nuclear fusion reactor configurations from a database. High-fidelity simulations are performed using computer modelling software for a plurality of configurations of the database. Four metrics (r, s, x, and ) used as input into computer modelling software are retained for their high correlation or anti-correlation with the desired physical property. An illustrative variation of these metrics along a magnetic field line is shown in FIG. 11 A. The metrics are used in a power-law expression, as shown below:

[0206] The power-law expression is fit (trained) using a curve fitting algorithm to determine the coefficients (a, b, c, d, and e). The fit is performed by fitting the local desired property value along the flux-tube. For the relevant QI configurations, the obtained coefficients are independent of the field period and lead to the following equation:

[0207] Equation 2 is used to define the surrogate model for reactor optimization, where the surrogate model for reactor optimization is described by the following equation:Equation 3. Y[ metric max(< Yt(z') >z) a

[0208] In the surrogate model for reactor optimization (Equation 3), <>zdenotes the average along the flux-tube and max(< V,(z) >z) denotes the maximum value among all the flux tubes. aThe surrogate model is then evaluated for all of the configurations in the database. An output of the surrogate model given in Equation 2 of one configuration along an electromagnetic field line is shown in FIG. 11B. A general comparison between the output of the surrogate model for reactor optimization (Equation 3) and the high-fidelity simulations is shown in FIG. 12.

Claims

CLAIMSWhat is claimed is:

1. A computer-implemented method for training a machine learning model to predict a configuration of a nuclear fusion reactor, comprising: a. obtaining a dataset comprising a shape metric, a magnetohydrodynamic (MHD) stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric; and b. using the dataset, training the machine learning model to predict the configuration of the nuclear fusion reactor.

2. The computer-implemented method of claim 1, wherein the dataset comprises a shape metric.

3. The computer-implemented method of claim 1 or 2, wherein the dataset comprises an MHD stability metric.

4. The computer-implemented method of any one of claims 1-3, wherein the dataset comprises a neoclassical transport metric.

5. The computer-implemented method of any one of claims 1-4, wherein the dataset comprises an energetic particle confinement metric.

6. The computer-implemented method of any one of claims 1-5, wherein the dataset comprises a turbulent transport metric.

7. The computer-implemented method of any one of claims 1-6, wherein the dataset comprises a magnetic field metric.

8. The computer-implemented method of any one of claims 1-7, wherein the dataset comprises a coil design metric.

9. The computer-implemented method of any one of claims 1-8, further comprising assembling the nuclear fusion reactor based on the configuration of the nuclear fusion reactor predicted in (b).

10. The computer-implemented method of any one of claims 1-9, further comprising causing fabrication of a component of the nuclear fusion reactor based on the configuration of the nuclear fusion reactor predicted in (b).

11. The computer-implemented method of any one of claims 1-10, wherein the component comprises a coil, a blanket, a divertor, a first wall, a heat shield, an outer vessel, or a plasma vessel.

12. The computer-implemented method of any one of claims 1-11, further comprising:c. generating an updated configuration of the configuration of the nuclear fusion reactor using a fine-tuning module; d. obtaining a second dataset comprising the updated configuration; and e. using the second dataset, training the machine learning model to predict the updated configuration of the configuration of the nuclear fusion reactor.

13. The computer-implemented method of claim 12, wherein the fine-tuning module comprises a physics solver or a second machine learning model.

14. The computer-implemented method of any one of claims 1-13, wherein the nuclear fusion reactor comprises a stellarator.

15. The computer-implemented method of any one of claims 1-13, wherein the nuclear fusion reactor comprises a tokamak.

16. The computer-implemented method of any one of claims 1-15, wherein the configuration of the nuclear fusion reactor comprises a magnetic confinement fusion (MCF) device.

17. The computer-implemented method of any one of claims 1-16, wherein the configuration of the nuclear fusion reactor comprises a pressure profile, a cross section of the nuclear fusion reactor, a total magnetic flux, a rotational transform profile, or a current profile.

18. The computer-implemented method of any one of claims 1-17, wherein a datum of the dataset comprises a feature portion comprising the shape metric, the MHD stability metric, the neoclassical transport metric, the energetic particle confinement metric, the turbulent transport metric, the magnetic field metric, or the coil design metric, and wherein the datum further comprises a label portion comprising the configuration of the nuclear fusion reactor.

19. The computer-implemented method of any one of claims 1-18, wherein the configuration of the nuclear fusion reactor comprises a two-dimensional or three-dimensional spatial representation of the nuclear fusion reactor.

20. The computer-implemented method of claim 19, wherein the two-dimensional or three- dimensional spatial representation of the nuclear fusion reactor comprises a model of the nuclear fusion reactor, an image of the nuclear fusion reactor, a rendering of the nuclear fusion reactor, or a latent representation of the nuclear fusion reactor.

21. The computer-implemented method of any one of claims 1-20, wherein the dataset comprises a plurality of simulated cross sections of the nuclear fusion reactor.

22. The computer-implemented method of any one of claims 1-21, wherein a first cross section of the plurality of simulated cross sections comprises a first toroidal angle and a second surface of the plurality of simulated cross sections comprises a second toroidalangle, and wherein the first toroidal angle and the second toroidal angle differ by at least about 1 degree.

23. The computer-implemented method of any one of claims 1-22, further comprising: f. obtaining a parent dataset; g. grouping the parent dataset into a plurality of groups based on a physics principle; and h. clustering a group of the plurality of groups into a plurality of clusters using a clustering model, wherein a cluster of the plurality of clusters comprises the dataset.

24. The computer-implemented method of claim 23, wherein the physics principle comprises a type of symmetry, a number of field periods, or an edge rotational transform.

25. A computer-implemented method for generating a configuration of a nuclear fusion reactor, comprising: a. obtaining a dataset comprising a shape metric, a magnetohydrodynamic stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric; and b. generating, using the dataset as input into a machine learning model, the configuration of the nuclear fusion reactor.

26. The computer-implemented method of claim 25, further comprising, c. performing operations (a) and (b) in an optimization framework.

27. The computer-implemented method of claim 26, wherein the optimization framework comprises a loss function configured to optimize the configuration of the nuclear fusion reactor.

28. The computer-implemented method of claim 26 or 27, wherein the optimization framework comprises reinforcement learning.

29. The computer-implemented method of any one of claims 26-28, wherein the optimization framework queries a computational model that outputs the configuration of the nuclear fusion reactor.

30. The computer-implemented method of any one of claims 25-29, wherein the dataset comprises a shape metric.

31. The computer-implemented method of any one of claims 25-30, wherein the dataset comprises an MHD stability metric.

32. The computer-implemented method of any one of claims 25-31, wherein the dataset comprises a neoclassical transport metric.

33. The computer-implemented method of any one of claims 25-32, wherein the dataset comprises an energetic particle confinement metric.

34. The computer-implemented method of any one of claims 25-33, wherein the dataset comprises a turbulent transport metric.

35. The computer-implemented method of any one of claims 25-34, wherein the dataset comprises a magnetic field metric.

36. The computer-implemented method of any one of claims 25-35, wherein the dataset comprises a coil design metric.

37. The computer-implemented method of any one of claims 25-36, further comprising assembling the nuclear fusion reactor based on the configuration of the nuclear fusion reactor predicted in (b).

38. The computer-implemented method of any one of claims 25-37, further comprising causing fabrication of a component of the nuclear fusion reactor based on the configuration of the nuclear fusion reactor predicted in (b).

39. The computer-implemented method of claim 38, wherein the component comprises a coil, a blanket, a heat shield, an outer vessel, or a plasma vessel.

40. The computer-implemented method of any one of claims 25-39, wherein the configuration of the nuclear fusion reactor is used as input into a second machine learning model trained to fine-tune the configuration of the nuclear fusion reactor.

41. The computer-implemented method of any one of claims 25-40, wherein the machine learning model is selected by a model selection pipeline, wherein the model selection pipeline is configured to perform operations comprising: i. classifying the dataset based on a physics principle, ii. classifying the dataset based on a clustering model, and iii. identifying the machine learning model based on (i) and (ii).

42. The computer-implemented method of claim 41, wherein the physics principle comprises a type of symmetry, a number of field periods, or an edge rotational transform.

43. The computer-implemented method of any one of claims 25-42, wherein the nuclear fusion reactor comprises a stellarator.

44. The computer-implemented method of any one of claims 25-42, wherein the nuclear fusion reactor comprises a tokamak.

45. The computer-implemented method of any one of claims 25-44, wherein the configuration of the nuclear fusion reactor comprises a magnetic confinement fusion (MCF) device.

46. The computer-implemented method of any one of claims 25-45, wherein the configuration of the nuclear fusion reactor comprises one or more of a pressure profile, a total magnetic flux, a cross section of the nuclear fusion reactor, a rotational transform profile, or a current profile.

47. The computer-implemented method of any one of claims 25-46, wherein a datum of the dataset comprises a feature portion comprising the shape metric, the MHD stability metric, the neoclassical transport metric, the energetic particle confinement metric, the turbulent transport metric, the magnetic field metric, or the coil design metric, and wherein the datum further comprises a label portion comprising a nuclear fusion reactor cross section, a pressure profile, current profile, a rotational transform profile, or a total magnetic flux.

48. The computer-implemented method of any one of claims 25-47, wherein the configuration comprises a two-dimensional or three-dimensional spatial representation of the nuclear fusion reactor.

49. The computer-implemented method of claim 48, wherein the two-dimensional or three- dimensional spatial representation of the nuclear fusion reactor comprises a model of the nuclear fusion reactor, an image of the nuclear fusion reactor, a rendering of the nuclear fusion reactor, or a latent representation of the nuclear fusion reactor.

50. The computer-implemented method of any one of claims 25-49, wherein the dataset comprises a plurality of simulated cross sections of the nuclear fusion reactor.

51. The computer-implemented method of claim 50, wherein a first cross section of the plurality of simulated cross sections comprises a first toroidal angle and a second cross section of the plurality of simulated cross sections comprises a second toroidal angle, and wherein the first toroidal angle and the second toroidal angle differ by at least about 1 degree.

52. A method comprising using a trained machine learning model to generate a configuration of a nuclear fusion reactor.

53. A computer-implemented method for generating a nuclear fusion reactor surrogate model, comprising: a. extracting a plurality of input metrics from a nuclear fusion reactor dataset, wherein a metric of the plurality of metrics is extracted based on a correlation or anti-correlation with a desired property of the nuclear fusion reactor; b. generating the nuclear fusion reactor surrogate model by fitting a mathematical model or a machine learning model with the plurality of input metrics to output the desired property of the nuclear fusion reactor.

54. The computer-implemented method of claim 53, wherein the nuclear fusion reactor dataset comprises a simulation of a nuclear fusion reactor.

55. The computer-implemented method of claim 53 or 54, further comprising: c. generating the simulation of the nuclear fusion reactor.

56. The computer-implemented method of any one of claims 53-55, wherein a datum of the nuclear fusion reactor dataset comprises the plurality of input metrics and the desired property of the nuclear fusion reactor.

57. The computer-implemented method of any one of claims 53-56, further comprising: d. selecting a subset of the nuclear fusion reactor dataset using the surrogate model to obtain an updated nuclear fusion reactor dataset.

58. The computer-implemented method of any one of claims 53-57, wherein (a)-(d) are repeated until a stopping condition is met.

59. The computer-implemented method of claim 58, wherein the stopping condition is a number of iterations or an accuracy of the surrogate nuclear fusion reactor model.

60. The computer-implemented method of any one of claims 53-59, wherein the surrogate nuclear fusion reactor model forms a portion of a loss function used in training a machine learning model.

61. The computer-implemented method of any one of claims 53-60, wherein the surrogate nuclear fusion reactor model is used in an optimization framework.

62. The computer-implemented method of any one of claims 53-61, wherein a metric of the plurality of metrics is a shape metric, a magnetohydrodynamic (MHD) stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric.

63. The computer-implemented method of any one of claims 53-62, wherein the desired property of the nuclear fusion reactor is a shape metric, a magnetohydrodynamic (MHD) stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric.

64. The computer-implemented method of any one of claims 53-63, wherein the nuclear fusion reactor dataset describes a stellarator.

65. The computer-implemented method of any one of claims 53-64, wherein the nuclear fusion reactor dataset describes a tokamak.

66. A computer-implemented method, comprising: a. providing a plurality of nuclear fusion reactor configurations, wherein a configuration of the plurality of nuclear fusion reactor configurations comprises a plurality of metrics;b. generating a metric space based on the plurality of metrics as input into an optimization framework; and c. determining an optimized nuclear fusion reactor configuration using the optimization framework.

67. The computer-implemented method of claim 66, wherein the plurality of nuclear fusion reactor configurations are stored in one or more database.

68. The computer-implemented method of claim 66 or 67, wherein a nuclear fusion reactor configuration of the plurality of nuclear fusion reactor configurations comprises a shape metric, a magnetohydrodynamic (MHD) stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric.

69. The computer-implemented method of any one of claims 66-68, wherein a nuclear fusion reactor configuration of the plurality of nuclear fusion reactor configurations comprises a Fourier coefficient.

70. The computer-implemented method of any one of claims 66-69, wherein (b) comprises determining a range of values for each of the plurality of metrics.

71. The computer-implemented method of any one of claims 66-70, wherein the optimization framework is an active learning framework.

72. The computer-implemented method of claim 71, wherein optimization framework uses Bayesian optimization.

73. The computer-implemented method of claim 72, wherein an acquisition function of the Bayesian optimization comprises expected improvement, batch expected improvement, upper-confidence bound, probability of improvement, Thompson sampling, batch energy-entropy, expected hypervolume improvement, noisy expected hypervolume improvement, batch noisy expected hypervolume improvement, Pareto efficient global optimization (ParEGO), noisy Pareto efficient global optimization (NParEGO), batch noisy Pareto efficient global optimization (qNParEGO), or any combination thereof.

74. The computer-implemented method of any one of claims 66-73, wherein generating the optimized nuclear fusion reactor configuration comprises training a machine learning model using the optimization framework.

75. The computer-implemented method of any one of claims 66-74, further comprising acquiring a nuclear fusion reactor configuration, wherein acquiring the nuclear reactor configuration comprises simulating the nuclear fusion reactor configuration based on a sample of the metric space output by the optimization framework.

76. The computer-implemented method of any one of claims 66-75, wherein (c) comprises comparing a value of a desired physical property of the nuclear fusion reactor configuration with a provided metric of the nuclear fusion reactor configuration.

77. The computer-implemented method of any one of claims 66-76, wherein the desired physical property of the nuclear fusion reactor configuration is a shape metric, a magnetohydrodynamic (MHD) stability metric, a neoclassical transport metric, an energetic particle confinement metric, a turbulent transport metric, a magnetic field metric, or a coil design metric.

78. The computer-implemented method of any one of claims 66-77, wherein the optimization framework is configured to select a datum to acquire based at least in part on one or both of: i. a likelihood to decrease uncertainty of the machine learning model; or ii. a likelihood to generate the optimized nuclear fusion reactor configuration.

79. The computer-implemented method of any one of claims 66-78, wherein (c) comprises gathering data to train the machine learning model using the optimization framework until a stopping condition is reached.

80. The computer-implemented method of any one of claims 66-79, wherein the stopping condition is a number of iterations, an accuracy of the machine learning model, or achievement of the optimized nuclear fusion reactor configuration.

81. A computer-implemented method for configuring a stellarator, comprising: a. obtaining one or more parameters of a target stellarator; b. generating a plurality of target stellarator approximations based at least in part on said one or more parameters of said target stellarator, wherein at least a subset of said plurality target stellarator approximations: (i) have a toroidal profile, and (ii) comprise a plurality of concentric toroids; and c. presenting at a graphical user interface a representation of said plurality of target stellarator approximations.

82. The computer-implemented method of claim 81, wherein said one or more parameters of said target stellarator comprise one or more of: height, width, or length.

83. The computer-implemented method of any one of claims 81 and 82, wherein said one or more parameters of said target stellarator comprise layer information.

84. The computer-implemented method of claim 83, wherein said layer information comprises a number of layers or one or more types of layers.

85. The computer-implemented method of any one of claims 81-84, wherein said toroidal profile is a torus profile and said plurality of concentric toroids are a plurality of concentric torii.

86. The computer-implemented method of any one of claims 81-85, wherein said representation of said plurality of target stellarator approximations comprises a simulation of said plurality of target stellarator approximations under fusion conditions.

87. The computer-implemented method of claim 86, wherein said representation of said plurality of target stellarator approximations comprises performance data of said simulation of said plurality of target stellarator approximations under said fusion conditions.

88. The computer-implemented method of any one of the preceding claims, further comprising: d. obtaining at least one selected target stellarator approximations of said target stellarator approximations.

89. The computer-implemented method of claim 88, further comprising: e. performing one or more simulations of said at least one selected target stellarator approximations.

90. The computer-implemented method of claim 89, wherein said one or more simulations of said at least one selected target stellarator approximations do not approximate said at least one selected target stellarator approximations as a toroid.

91. A computer-implemented system comprising a computing device comprising at least one processor, a memory, and a computer program including instructions executable by said computing device to create an application configured to perform any one of the methods of claims 1-90.

92. A non-transitory computer-readable storage media encoded with a computer program including instructions executable by one or more processors to perform operations comprising a method of any one of claims 1-90.

Images