Disruption prediction based on fusion feature extractor on J-TEXT

Predicting disruptions across different tokamaks is necessary for next generation device.Future large-scale tokamaks can hardly tolerate disruptions at high performance discharge,which makes it difficult for current data-driven methods to obtain an acceptable result.A machine learning method capable of transferring a disruption prediction model trained on one tokamak to another is required to solve the problem.The key is a feature extractor which is able to extract common disruption precursor traces in tokamak diagnostic data,and can be easily transferred to other tokamaks.Based on the concerns above,this paper presents a deep feature extractor,namely,the fusion feature extractor(FFE),which is designed specifically for extracting disruption precursor features from common diagnostics on tokamaks.Furthermore,an FFE-based disruption predictor on J-TEXT is demonstrated.The feature extractor is aimed to extracting disruption-related precursors and is designed according to the precursors of disruption and their representations in common tokamak diagnostics.Strong inductive bias on tokamak diagnostics data is introduced.The paper presents the evolution of the neural network feature extractor and its comparison against general deep neural networks,as well as a physics-based feature extraction with a traditional machine learning method.Results demonstrate that the FFE may reach a similar effect with physics-guided manual feature extraction,and obtain a better result compared with other deep learning methods.

An application of the shortest path algorithm for the identification of weak MHD mode

The identification of magnetohydrodynamic(MHD)modes is a crucial issue in the control of magnetically confined plasmas.This paper proposes a novel method for identifying the evolution of MHD modes from a signal with a low signal-to-noise ratio.The proposed method generates a weighted directed graph from the time-frequency spectrum and calculates the evolution of the mode frequency by solving the shortest path.This method addresses the limitations posed by the lack of data channels and the disturbance of noise in the estimation of mode frequency and yields much better results compared to traditional methods.It is demonstrated that,using this method,the evolution of an unlocked tearing mode was more accurately calculated on the J-TEXT tokamak.This method remains feasible even with a low signal-to-noise ratio of 0.5,as shown by its uncertainty.Furthermore,with appropriate parameters,this method can be applied to not only signals with MHD modes,but also to general signals with continuous modes.

Study of the spectrum effect on the threshold of resonant magnetic perturbation penetration on J-TEXT

The spectrum effect on the penetration of resonant magnetic perturbation(RMP) is studied with upgraded in-vessel RMP coils on J-TEXT.The poloidal spectrum of the RMP field,especially the amplitudes of 2/1 and 3/1 components,can be varied by the phase difference between the upper and lower coil rows,ΔΦ=Φ_(top)-Φ_(bottom),where Φ_(top)and Φ_(bottom)are the toroidal phases of the n=1 field of each coil row.The type of RMP penetration is found to be related to ΔΦ,including the RMP penetration of either 2/1 or 3/1 RMP and the successive penetrations of 3/1 RMP followed by the 2/1 RMP.For cases with penetration of only one RMP component,the penetration thresholds measured by the corresponding resonant component are close for variousΔΦ.However,the 2/1 RMP penetration threshold is significantly reduced if the 3/1 locked island is formed in advance.The changes in the rotation profile due to 3/1 locked island formation could partially contribute to the reduction of the 2/1 thresholds.

Overview of machine learning applications in fusion plasma experiments on J-TEXT tokamak

Machine learning research and applications in fusion plasma experiments are one of the main subjects on J-TEXT.Since 2013,various kinds of traditional machine learning,as well as deep learning methods have been applied to fusion plasma experiments.Further applications in the real-time experimental environment have proved the feasibility and effectiveness of the methods.For disruption prediction,we started by predicting disruptions of limited classes with a short warning time that could not meet the requirements of the mitigation system.After years of study,nowadays disruption prediction methods on J-TEXT are able to predict all kinds of disruptions with a high success rate and long enough warning time.Furthermore,cross-device disruption prediction methods have obtained promising results.Interpretable analysis of the models are studied.For diagnostics data processing,efforts have been made to reduce manual work in processing and to increase the robustness of the diagnostic system.Models based on both traditional machine learning and deep learning have been applied to real-time experimental environments.The models have been cooperating with the plasma control system and other systems,to make joint decisions to further support the experiments.

Machine learning application to predict the electron temperature on the J-TEXT tokamak

The reliability of diagnostic systems in tokamak plasma is of great significance for physics researches or fusion reactor.When some diagnostics fail to detect information about the plasma status,such as electron temperature,they can also be obtained by another method:fitted by other diagnostic signals through machine learning.The paper herein is based on a machine learning method to predict electron temperature,in case the diagnostic systems fail to detect plasma temperature.The fully-connected neural network,utilizing back propagation with two hidden layers,is utilized to estimate plasma electron temperature approximately on the J-TEXT.The input parameters consist of soft x-ray emission intensity,electron density,plasma current,loop voltage,and toroidal magnetic field,while the targets are signals of electron temperature from electron cyclotron emission and x-ray imaging crystal spectrometer.Therefore,the temperature profile is reconstructed by other diagnostic signals,and the average errors are within 5%.In addition,generalized regression neural network can also achieve this function to estimate the temperature profile with similar accuracy.Predicting electron temperature by neural network reveals that machine learning can be used as backup means for plasma information so as to enhance the reliability of diagnostics.

Magnetic diagnostics for magnetohydrodynamic instability research and the detection of locked modes in J-TEXT

Magnetohydrodynamic(MHD)instabilities are widely observed during tokamak plasma operation.Magnetic diagnostics provide important information which supports the understanding and control of MHD instabilities.This paper presents the current status of the magnetic diagnostics dedicated to measuring MHD instabilities at the J-TEXT tokamak;the diagnostics consist of five Mirnov probe arrays for measuring high-frequency magnetic perturbations and two saddle-loop arrays for low-frequency magnetic perturbations,such as the locked mode.In recent years,several changes have been made to these arrays.The structure of the probes in the poloidal Mirnov arrays has been optimized to improve their mechanical strength,and the number of in-vessel saddle loops has also been improved to support better spatial resolution.Due to the installation of high-field-side(HFS)divertor targets in early 2019,some of the probes were removed,but an HFS Mirnov array was designed and installed behind the targets.Owing to its excellent toroidal symmetry,the HFS Mirnov array has,for the first time at J-TEXT,provided valuable new information about the locked mode and the quasi-static mode(QSM)in the HFS.Besides,various groups of magnetic diagnostics at different poloidal locations have been systematically used to measure the QSM,which confirmed the poloidal mode number m and the helical structure of the QSM.By including the HFS information,the 2/1 resonant magnetic perturbation(RMP)-induced locked mode was measured to have a poloidal mode number m of~2.

Recent progress on the J-TEXT three-wave polarimeter-interferometer

The J-TEXT three-wave polarimeter-interferometer system(POLARIS),which measures timespace distribution of electron density and current density,has been optimized with both the optical system and the equilibrium reconstruction method.The phase resolution of a Faraday rotation angle has been improved from 0.1 to 0.06 degree in chords from–0.18 to 0.18 m(plasma minor radius),and the sawtooth oscillation behavior has been detected by Faraday rotation angle measurement.By combining the POLARIS measured data and the equilibrium and fitting code(EFIT),an upgraded equilibrium reconstruction method has been developed,which provides a more accurate temporal and spatial distribution of current density and electron density.By means of the optimized POLARIS and improved equilibrium reconstruction,variations of profiles with increasing density have been carried out,under both Ohmic and electron cyclotron resonance heating discharges.

The spectrum analysis of RMP coils with multiple connection modes and the design of high field side coils on J-TEXT

The phase difference Δξ between locked islands(2/1 and 3/1) has been found to influence the heat transport on the thermal quench during disruptions by numerical modeling [Hu Q et al 2019Nucl.Fusion 59,016005].To verify this experimentally,a set of resonant magnetic perturbation(RMP) coils is required to excite coupled magnetic islands with different Δξ.The spectrum analysis shows that the current RMP coils on J-TEXT can only produce sufficient 2/1 and 3/1RMP fields with a limited phase difference of Δξ∈[-75°,75°].In order to broaden the adjustable range of Δξ,a set of coils on the high field side(HFS) is proposed to generate 2/1 and 3/1 RMP fields with Δξ=180°.As a result,RMPs with adjustable Δξ∈[-180°,180°] and sufficient amplitudes could be achieved by applying the HFS coils and the low field side(LFS)coils.This work provides a feasible solution for flexible adjustment of the phase difference between m and m+1 RMP,which might facilitate the study of major disruptions and their control.

Tunable Multicolor Circularly Polarized Luminescence via Co-assembly of One Chiral Electron Acceptor with Various Donors

Chiroptical materials with multicolor and signinvertible circularly polarized luminescence(CPL)are important for advanced optical devices and display technologies.Here,a general strategy is developed to generate CPL with highly tunable emission bands and handedness through charge-transfer(CT)complexation and co-assembly of one chiral non-emissive tetranitrofluorene-based acceptor with various achiral purple to blue emissive donors.The resulting assemblies exhibit intense CPL with a rich array of colors(519-668 nm)and prominent dissymmetry factors(|glum|)in the range of 10^(-3)-10^(-2).Notably,the CPL sign can be readily inverted by slightly changing the molecular structure of achiral donors.Single-crystal analysis reveals that the donor and acceptor molecules are alternately and asymmetrically packed in a lamellar fashion through CT interactions,leading to efficient transfer of chirality.Furthermore,the refined packing is mediated by the intensity and manner of CT interactions,rendering an inversion of chirality.The chiral co-assembly not only occurs for planar achiral donor molecules,but is also accessible to nonplanar conjugated molecules such as[4]helicene derivatives.Thus,the CPL feature of the resulting products can be easily and broadly manipulated,aiming at advanced chiroptical systems.

Tunable construction of transition metal-coordinated helicene cages

We report a facile and tailored method to prepare globally twisted chiral molecular cages through tunable coordination of bis-bipyridine-terminated helicene ligands to a series of transition metals including Fe(Ⅱ), Co(Ⅱ), Ni(Ⅱ) and Zn(Ⅱ). This system shows an efficient remote transfer of stereogenecity from the helicene core to the bipyridine-metal coordination sites and subsequently the entire cages. While the Fe(Ⅱ), Co(Ⅱ) and Ni(Ⅱ)-derived M_(2)L_(3)(M for metal and L for ligand) cages exhibit quasi-reversible redox features, the Zn(Ⅱ) analogues reveal prominent yellow circularly polarized luminescence. Interestingly,with the addition of Na_(2)SO_(4), the Zn_(2)L_(3)cages reassemble into sextuple-stranded Zn_(6)L_(6)(SO_(4))_(4)cages in which three Zn_(2)L_(2) units are bound together by four sulfates and further coalesced by offset inter-ligandπ-π interactions.

Helical Conformation Tunability via Hydrogen Bonding in Supramolecular Frameworks

Hydrogen-bonded molecules and their dynamics are significantly important in chemistry and biology due to their widespread functionality.Besides their natural abundance and diversity,applications of molecules with dynamic conformations in artificial networks allow information storage and molecular motor design on the nanometer scale.Here,we report hydrogen-bonded molecular networks with tunable helical conformation on metal surfaces.The dynamics of helical conformation in two-dimensional hydrogenbond networks are triggered and resolved by scanning probe microscopy at the single-molecule level.In combinationwith theoretical calculations,the surfacespecific hydrogen bonds are identified as the origin of the dynamic helical conformation.Our results provide a distinctive access to molecular architecture with tunable helical conformation driven by hydrogenbond interaction on surfaces.