Source author record

the DIII-D Team

the DIII-D Team appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

physics.plasm-ph Machine Learning physics.app-ph

Catalog footprint

What is connected

3works

3topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2022arXiv

Mitigation of plasma-wall interactions with low-Z powders in DIII-D high confinement plasmas

Experiments with low-Z powder injection in DIII-D high confinement discharges demonstrated increased divertor dissipation and detachment while maintaining good core energy confinement. Lithium (Li), boron (B), and boron nitride (BN) powders were injected in high-confinement mode plasmas ($I_p=$1 MA, $B_t=$2 T, $P_{NB}=$6 MW, $\langle n_e\rangle=3.6-5.0\cdot10^{19}$ m$^{-3}$) into the upper small-angle slot (SAS) divertor for 2-s intervals at constant rates of 3-204 mg/s. The multi-species BN powders at a rate of 54 mg/s showed the most substantial increase in divertor neutral compression by more than an order of magnitude and lasting detachment with minor degradation of the stored magnetic energy $W_{mhd}$ by 5%. Rates of 204 mg/s of boron nitride powder further reduce ELM-fluxes on the divertor but also cause a drop in confinement performance by 24% due to the onset of an $n=2$ tearing mode. The application of powders also showed a substantial improvement of wall conditions manifesting in reduced wall fueling source and intrinsic carbon and oxygen content in response to the cumulative injection of non-recycling materials. The results suggest that low-Z powder injection, including mixed element compounds, is a promising new core-edge compatible technique that simultaneously enables divertor detachment and improves wall conditions during high confinement operation.

preprint2019arXiv

Deep convolutional neural networks for multi-scale time-series classification and application to disruption prediction in fusion devices

The multi-scale, mutli-physics nature of fusion plasmas makes predicting plasma events challenging. Recent advances in deep convolutional neural network architectures (CNN) utilizing dilated convolutions enable accurate predictions on sequences which have long-range, multi-scale characteristics, such as the time-series generated by diagnostic instruments observing fusion plasmas. Here we apply this neural network architecture to the popular problem of disruption prediction in fusion tokamaks, utilizing raw data from a single diagnostic, the Electron Cyclotron Emission imaging (ECEi) diagnostic from the DIII-D tokamak. ECEi measures a fundamental plasma quantity (electron temperature) with high temporal resolution over the entire plasma discharge, making it sensitive to a number of potential pre-disruptions markers with different temporal and spatial scales. Promising, initial disruption prediction results are obtained training a deep CNN with large receptive field (~30k), achieving an $F_1$-score of ~91% on individual time-slices using only the ECEi data.

preprint2016arXiv

Statistical analysis of $m/n$ = 2/1 locked and quasi-stationary modes with rotating precursors at DIII-D

A database has been developed to study the evolution, the nonlinear effects on equilibria, and the disruptivity of locked and quasi-stationary modes with poloidal and toroidal mode numbers $m=2$ and $n=1$ at DIII-D. The analysis of 22,500 discharges shows that more than 18% of disruptions are due to locked or quasi-stationary modes with rotating precursors (not including born locked modes). A parameter formulated by the plasma internal inductance $l_i$ divided by the safety factor at 95% of the poloidal flux, $q_{95}$, is found to exhibit predictive capability over whether a locked mode will cause a disruption or not, and does so up to hundreds of milliseconds before the disruption. Within 20 ms of the disruption, the shortest distance between the island separatrix and the unperturbed last closed flux surface, referred to as $d_{edge}$, performs comparably to $l_i/q_{95}$ in its ability to discriminate disruptive locked modes. Out of all parameters considered, $d_{edge}$ also correlates best with the duration of the locked mode. Within 50 ms of a locked mode disruption, average behavior includes exponential growth of the $n=1$ perturbed field, which might be due to the 2/1 locked mode. Even assuming the aforementioned 2/1 growth, disruptivity following a locked mode shows little dependence on island width up to 20 ms before the disruption. Separately, greater deceleration of the rotating precursor is observed when the wall torque is large. At locking, modes are often observed to align with a residual error field. Timescales associated with the mode evolution are also studied and dictate the response times necessary for disruption avoidance and mitigation.