Search

Bayesian optimization is a powerful optimization tool for problems where native first-order derivatives are unavailable. Recently, constrained Bayesian optimization (CBO) has been applied to many engineering applications where constraints are essential. However, several obstacles remain with current CBO algorithms that could prevent a wider adoption. We propose CBO algorithms using merit functions, such as the penalty merit function, in acquisition functions, inspired by nonlinear optimization methods, e.g., sequential quadratic programming. Merit functions measure the potential progress of both the objective and constraint functions, thus increasing algorithmic efficiency and allowing infeasible initial samples. The acquisition functions with merit functions are relaxed to have closed forms, making its implementation readily available wherever Bayesian optimization is. We further propose a unified CBO algorithm that can be seen as extension to the popular expected constrained improvement (ECI) approach. We demonstrate the effectiveness and efficiency of the proposed algorithms through numerical experiments on synthetic problems and a practical data-driven engineering design problem in the field of plasma physics.

Due to their cost, experiments for inertial confinement fusion (ICF) heavily rely on numerical simulations to guide design. As simulation technology progresses, so too can the fidelity of models used to plan for new experiments. However, these high-fidelity models are by themselves insufficient for optimal experimental design, because their computational cost remains too high to efficiently and effectively explore the numerous parameters required to describe a typical experiment. Traditionally, ICF design has relied on low-fidelity modeling to initially identify potentially interesting design regions, which are then subsequently explored via selected high-fidelity modeling. In this paper, we demonstrate that this two-step approach can be insufficient: even for simple design problems, a two-step optimization strategy can lead high-fidelity searching towards incorrect regions and consequently waste computational resources on parameter regimes far away from the true optimal solution. We reveal that a primary cause of this behavior in ICF design problems is the presence of low-fidelity optima in distinct regions of the parameter space from high-fidelity optima. To address this issue, we propose an iterative multifidelity Bayesian optimization method based on Gaussian Process Regression that leverages both low- and high-fidelity modelings. We demonstrate, using both two- and eight-dimensional ICF test problems, that our algorithm can effectively utilize low-fidelity modeling for exploration, while automatically refining promising designs with high-fidelity models. This approach proves to be more efficient than relying solely on high-fidelity modeling for optimization.

Transfer learning is a promising approach to creating predictive models that incorporate simulation and experimental data into a common framework. In this technique, a neural network is first trained on a large database of simulations, then partially retrained on sparse sets of experimental data to adjust predictions to be more consistent with reality. Previously, this technique has been used to create predictive models of Omega and NIF inertial confinement fusion (ICF) experiments that are more accurate than simulations alone. In this work, we conduct a transfer learning driven hypothetical ICF campaign in which the goal is to maximize experimental neutron yield via Bayesian optimization. The transfer learning model achieves yields within 5% of the maximum achievable yield in a modest-sized design space in fewer than 20 experiments. Furthermore, we demonstrate that this method is more efficient at optimizing designs than traditional model calibration techniques commonly employed in ICF design. Such an approach to ICF design could enable robust optimization of experimental performance under uncertainty.

An experimental program is currently underway at the National Ignition Facility (NIF) to compress deuterium and tritium (DT) fuel to densities and temperatures sufficient to achieve fusion and energy gain. The primary approach being investigated is indirect drive inertial confinement fusion (ICF), where a high-Z radiation cavity (a hohlraum) is heated by lasers, converting the incident energy into x-ray radiation which in turn drives the DT fuel filled capsule causing it to implode. Previous experiments reported DT fuel gain exceeding unity [O.A. Hurricane et al., Nature 506, 343 (2014)] and then exceeding the kinetic energy of the imploding fuel [S. Le Pape et al., Phys. Rev. Lett. 120, 245003 (2018)]. We report on recent experiments that have achieved record fusion neutron yields on NIF, greater than 100 kJ with momentary fusion powers exceeding 1PW, and have for the first time entered the burning plasma regime where fusion alpha-heating of the fuel exceeds the energy delivered to the fuel via compression. This was accomplished by increasing the size of the high-density carbon (HDC) capsule, increasing energy coupling, while controlling symmetry and implosion design parameters. Two tactics were successful in controlling the radiation flux symmetry and therefore the implosion symmetry: transferring energy between laser cones via plasma waves, and changing the shape of the hohlraum. In conducting these experiments, we controlled for known sources of degradation. Herein we show how these experiments were performed to produce record performance, and demonstrate the data fidelity leading us to conclude that these shots have entered the burning plasma regime.

The design space for inertial confinement fusion (ICF) experiments is vast and experiments are extremely expensive. Researchers rely heavily on computer simulations to explore the design space in search of high-performing implosions. However, ICF multiphysics codes must make simplifying assumptions, and thus deviate from experimental measurements for complex implosions. For more effective design and investigation, simulations require input from past experimental data to better predict future performance. In this work, we describe a cognitive simulation method for combining simulation and experimental data into a common, predictive model. This method leverages a machine learning technique called transfer learning, the process of taking a model trained to solve one task, and partially retraining it on a sparse dataset to solve a different, but related task. In the context of ICF design, neural network models trained on large simulation databases and partially retrained on experimental data, producing models that are far more accurate than simulations alone. We demonstrate improved model performance for a range of ICF experiments at the National Ignition Facility, and predict the outcome of recent experiments with less than ten percent error for several key observables. We discuss how the methods might be used to carry out a data-driven experimental campaign to optimize performance, illustrating the key product -- models that become increasingly accurate as data is acquired.

Inertial confinement fusion (ICF) experiments are designed using computer simulations that are approximations of reality, and therefore must be calibrated to accurately predict experimental observations. In this work, we propose a novel nonlinear technique for calibrating from simulations to experiments, or from low fidelity simulations to high fidelity simulations, via "transfer learning". Transfer learning is a commonly used technique in the machine learning community, in which models trained on one task are partially retrained to solve a separate, but related task, for which there is a limited quantity of data. We introduce the idea of hierarchical transfer learning, in which neural networks trained on low fidelity models are calibrated to high fidelity models, then to experimental data. This technique essentially bootstraps the calibration process, enabling the creation of models which predict high fidelity simulations or experiments with minimal computational cost. We apply this technique to a database of ICF simulations and experiments carried out at the Omega laser facility. Transfer learning with deep neural networks enables the creation of models that are more predictive of Omega experiments than simulations alone. The calibrated models accurately predict future Omega experiments, and are used to search for new, optimal implosion designs.

In this work, sequence-to-sequence (seq2seq) models, originally developed for language translation, are used to predict the temporal evolution of complex, multi-physics computer simulations. The predictive performance of seq2seq models is compared to state transition models for datasets generated with multi-physics codes with varying levels of complexity - from simple 1D diffusion calculations to simulations of inertial confinement fusion implosions. Seq2seq models demonstrate the ability to accurately emulate complex systems, enabling the rapid estimation of the evolution of quantities of interest in computationally expensive simulations.

In this work a novel, automated process for constructing and initializing deep feed-forward neural networks based on decision trees is presented. The proposed algorithm maps a collection of decision trees trained on the data into a collection of initialized neural networks, with the structures of the networks determined by the structures of the trees. The tree-informed initialization acts as a warm-start to the neural network training process, resulting in efficiently trained, accurate networks. These models, referred to as "deep jointly-informed neural networks" (DJINN), demonstrate high predictive performance for a variety of regression and classification datasets, and display comparable performance to Bayesian hyper-parameter optimization at a lower computational cost. By combining the user-friendly features of decision tree models with the flexibility and scalability of deep neural networks, DJINN is an attractive algorithm for training predictive models on a wide range of complex datasets.

The first nonlinear gyrokinetic simulations of electron internal transport barriers (e-ITBs) in the National Spherical Torus Experiment show that reversed magnetic shear can suppress thermal transport by increasing the nonlinear critical gradient for electron-temperature-gradient-driven turbulence to three times its linear critical value. An interesting feature of this turbulence is nonlinearly driven off-midplane radial streamers. This work reinforces the experimental observation that magnetic shear is likely an effective way of triggering and sustaining e-ITBs in magnetic fusion devices., Comment: 4 pages, 5 figures