Physics-assisted Latent Space Dynamics Learning for Stiff Collisional-radiative Models

High-dimensional time series are commonly encountered in modern control systems, especially in autonomous systems. In this work, a novel parsimonious latent state space (LaSS) model is proposed to characterize the latent dynamics, achieving general latent dynamic modeling with dimension reduction. The LaSS model is optimized by alternating estimations of the dimension reduction projection and the latent state space model. Precisely, the latent state dynamics are estimated by stochastic subspace identification methods. Furthermore, the canonical correlation analysis (CCA) objective is employed to acquire the optimal predictability for the extracted latent variables. The proposed LaSS-CCA algorithm is tested on a real industrial case for its effectiveness.

To control the H(0) atom production profile in the H(-) ion sources is one of the important issues for the efficient and uniform surface H(-) production. The purpose of this study is to construct a collisional radiative (CR) model to calculate the effective production rate of H(0) atoms from H2 molecules in the model geometry of the radio-frequency (RF) H(-) ion source for Linac4 accelerator. In order to validate the CR model by comparison with the experimental results from the optical emission spectroscopy, it is also necessary for the model to calculate Balmer photon emission rate in the source. As a basic test of the model, the time evolutions of H(0) production and the Balmer Hα photon emission rate are calculated for given electron energy distribution functions in the Linac4 RF H(-) ion source. Reasonable test results are obtained and basis for the detailed comparisons with experimental results have been established.

To address the significant temporal characteristics and the scarcity of labelled data in complex industrial processes, specifically targeting the latent temporal dynamics that are difficult to observe directly during process evolution, this paper proposes a latent temporal feature dual‐head coordination network (LTFDHCN). First, a posterior generation (PG) module is constructed, utilizing a bi‐directional long short‐term memory (Bi‐LSTM) network to extract bidirectional temporal context and infer the posterior distribution in the latent space. Second, a prior prediction (PP) module is designed to establish a dynamic temporal model based on the latent variable of the previous time step. This mechanism predicts the prior distribution for the current moment, breaking the traditional assumption in variational autoencoders (VAE) that latent variables follow a static prior, thereby capturing deeper latent temporal dynamics. The Kullback–Leibler (KL) divergence is employed to minimize the discrepancy between the posterior and dynamic prior distributions, enforcing the latent space to encode temporal dependencies. Furthermore, a dual‐head coordination (DHC) strategy is proposed. Leveraging a shared latent space and decoder backbone, the model simultaneously executes unsupervised data reconstruction and supervised quality prediction. This strategy effectively overcomes the label scarcity issue by explicitly mining the latent information embedded in unlabelled data. The model is optimized via a weighted objective function that integrates KL divergence, reconstruction and prediction loss. Simulation experiments on a debutanizer column and a thermal power generation process demonstrate that the proposed method outperforms baseline methods, including VAE, VAE‐LSTM, VRNN, and temporal convolutional networks (TCN)‐VAE, verifying its effectiveness and superiority in handling complex industrial data.

In plasmas the population densities are affected by electron collisions and by photon emission and absorption. In low temperature plasmas such as divertor or process plasmas, emissions from impurities are important for plasma diagnostics and plasma modeling. We study the density effect on population densities, effective ionization and recombination rate coefficients and emission lines from carbon ions. Especially we studied recombination processes for L-shell carbon ions because atomic data for dielectronic recombination for L-shell ions have recently become available. We made collisional radiative (CR) models for carbon L-shell ions including both ionization/excitation and recombination processes. Our CR model can calculate the line emission associated with recombination as well as that associated with excitation. In this work we apply these models to measured spectra from LHD (Large Helical Device) in the National Institute for Fusion Science. The spectra in recombining phase are measured in LHD and they are compared with our CR model.

Comparison of charge state distribution in commercially available sulfated cyclodextrins used as chiral resolving agents in capillary electrophoresis

An algorithm for diagnosing the electron density and temperature of helium plasma at atmospheric pressure has been developed based on a revised helium collisional-radiative (CR) model. Atomic collision processes are included, and part of the atomic data of electron collision processes in the conventional CR model has been updated to expand its valid pressure. The algorithm uses eight emission lines in the visible-wavelength range as inputs to determine the electron density, electron temperature, and number density of the two metastable states by fitting the number density of the states corresponding to the emission lines. The algorithm has a considerably small theoretical error. In the microwave-discharged low-pressure helium plasma experiment, the results obtained with the algorithm agreed well with the results obtained with the probe method. The electron density and temperature of the atmospheric-pressure helium plasma obtained with the algorithm agreed well with the results of the continuum spectrum analysis.

To examine the influence of radiative transitions on the performance of nonequilibrium seed-free argon plasma magnetohydrodynamic (MHD) generators, we developed MHD numerical simulation technique with a collisional-radiative (C-R) model, where neutral argon atoms with 45 discrete effective electronic excitation levels, singly ionized argon ions, and free electrons were considered under a two-temperature, low-magnetic Reynolds number MHD approximation. Furthermore, a so-called escape factor was implemented in the C-R model to consider the effects of radiation trapping by ground-state neutral atoms. Using the developed technique, we computed the performance characteristics of a nonequilibrium seed-free argon plasma MHD generator. The numerical results showed that the generator performance gets higher as the escape factor decreases, i.e., the optical thickness for radiative transitions to the ground state of neutral atoms increases. The numerical results also suggested that when the escape factor is <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1.0\times 10^{-3}$ </tex-math></inline-formula> or less, the generator performance is almost the same as that for the escape factor of 0 corresponding to completely optical-thick plasma case for the radiative transitions to the ground state of neutral atoms. Furthermore, realistic escape factor values for resonance lines in the power generation channel were evaluated using a well-known estimation model. Consequently, the estimated values had the order of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$10^{-4}$ </tex-math></inline-formula> . We also showed that under such relatively optically thick plasma cases, a collisional (C) model and a widely used global ionization-recombination rate model with no explicit consideration of electronically excited-state neutral atoms can reproduce the generator performance predicted by the C-R model with relatively good accuracy.

Collisional-radiative modeling with multitemperature thermodynamic models

A deep Koopman operator‐based modelling approach for long‐term prediction of dynamics with pixel‐level measurements

Summary form only given. X-pinch experiments using 188 alloy and combined Mo/W wires were implemented at the Cornell 450 kA XP facility. The spatially resolved, time-gated and integrated X-ray line spectra have been recorded using different crystal spectrometers (time-gated ADP and time-integrated mica crystal spectrometers). In particular, K-shell and L-shell X-ray line spectra from Cr, Co, and Ni ions as well as M-shell spectra from W ions have been recorded through different filters in the same shot from the 188 wire alloy X-pinch. In addition, L-shell spectra of Mo ions and M-shell spectra of W ions have been studied produced from X-pinches composed from two different wires of Mo and W. Non-LTE collisional-radiative (CR) atomic kinetic models of Ni and Co have been developed earlier to identify the useful K- and L-shell diagnostic features in X-pinch plasmas from nichrome and conichrome alloys produced at the same Cornell XP device. These models along with the Mo model were used in the present work to interpret X-ray spectra from 188 alloys and to isolate M-shell spectra of W ions. The newly developed non-LTE CR model of W has been applied to identify and diagnose the spectral features of W ions in both types of X-pinches. The distinct features, similarities, and differences of these M-shell spectra of W ions have been analyzed and compared for both types of X-pinches. The future development of this work is discussed.

Physical systems governed by advection-dominated partial differential equations (PDEs) are found in applications ranging from engineering design to weather forecasting. They are known to pose severe challenges to both projection-based and non-intrusive reduced order modeling, especially when linear subspace approximations are used. In this work, we develop an advection-aware (AA) autoencoder network that can address some of these limitations by learning efficient, physics-informed, nonlinear embeddings of the high-fidelity system snapshots. A fully non-intrusive reduced order model is developed by mapping the high-fidelity snapshots to a latent space defined by an AA autoencoder, followed by learning the latent space dynamics using a long-short-term memory (LSTM) network. This framework is also extended to parametric problems by explicitly incorporating parameter information into both the high-fidelity snapshots and the encoded latent space. Numerical results obtained with parametric linear and nonlinear advection problems indicate that the proposed framework can reproduce the dominant flow features even for unseen parameter values.

We develop a collisional radiative (CR) model of high-Z multiple charge ions based on computational atomic data. We use an algorithm to determine a set of atomic energy levels in the model with significant population in the plasmas and contribute to the dielectronic recombination (DR) processes. The model is validated through participation to the non-LTE kinetics workshops. At the nLTE workshop, an alternative method for the analysis of the spectrum from plasmas based on genetic algorithm (GA) is also discussed. An application of the present model to calculate the ion abundance and radiative power loss of Gd and Nd plasmas is also shown.

SPIDER is the full scale prototype for the ITER Heating Neutral Beam (HNB) source. The production of negative ions in the source mostly relies on the conversion of H/D atoms and positive ions on a 1.6 m <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times0.8$ </tex-math></inline-formula> m converter surface, which is covered with cesium in order to reduce its work function and maximize the conversion efficiency. In SPIDER, optical emission spectroscopy (OES) is used to study the light emitted by the plasma. In particular, the intensities of the cesium emission lines are tracked and used to reconstruct the distribution of its excited states. By coupling these measurements with a collisional radiative (CR) model, estimates of the cesium ground-state density and plasma parameters can be obtained in various parts of the source which are unreachable by other diagnostics. This work describes such a model and its use in conjunction with OES during the first operation with cesium in SPIDER. The electron density at the bottom of the source was found to be much larger than the rest, as a consequence of vertical drifts caused by the horizontal filter field, which would likely indicate a lower negative-ion availability for the beam. The effect of increasing the cesium evaporation rate on the negative-ion and electron densities was also investigated.

A collisional radiative (CR) model has been developed for the diagnostics of Ne–Ar mixture plasma. The 40 fine‐structure resolved energy levels of neon corresponding 2p53s, 2p53p, 2p53d, 2p54s, and 2p54p along with ground state (2p6) and ion state (2p5) have been considered. All the essential population and de‐population channels viz. electron impact excitation‐de‐excitation, ionization, two/three‐body recombination, radiative decay, self‐absorption correction, diffusion, etc., have been considered in the model. The quenching of relatively higher lifetime states 2p53s due to interaction with Ar atoms through associative and Penning ionization is also included in the model. The applicability of the CR model is demonstrated by performing the optical diagnostic of rf induced inductively coupled Ne–Ar mixture plasma. The plasma parameters viz. electron temperature, electron density, and population distribution of levels at various mixture conditions are extracted by coupling the present CR model predictions with emission and absorption measurements by Boffard et al. [J. Phys. D: Appl. Phys., 45, 382001 (2012) and private communication]. The results are also compared with earlier optical models and Langmuir probe measurements. The self‐absorption correction (escape) factors and electron impact excitation rates at various Ne–Ar mixture concentrations are also presented.

Charge state distribution of 16O from the 4He(12C,16O)γ reaction of astrophysical interest studied both experimentally and theoretically