Search

In FTU, the density profile in the presence of large MARFE becomes more and more peaked with increasing density, forming strong density gradients close to the radial region affected by the MARFE. The temperature at the edge drops to few eV, driving a drop of the whole profile. The estimated particle source cannot justify the change of the density gradient, which instead is well-explained by a change of the pinch. A thermo-diffusion term well-describes the pinch evolution and the experimental behavior of the density at those radii where temperature measurements are reliable.

Identification of the mechanisms for the H-mode density limit in machines with fully metallic walls, and their scaling to future devices is essential to find for these machines the optimal operational boundaries with the highest attainable density and confinement. Systematic investigations of H-mode density limit plasmas in experiments with deuterium external gas fuelling have been performed on machines with fully metallic walls, JET and AUG and results have been compared with one another. Basically, the operation phases are identical for both tokamaks: the stable H-mode phase, degrading H-mode phase, breakdown of the H-mode with energy confinement deterioration usually accompanied by a dithering cycling phase, followed by the l -mode phase. The observed H-mode density limit on both machines is found close to the Greenwald limit (n/n GW = 0.8-1.1 in the observed magnetic configurations). The similar behavior of the radiation on both tokamaks demonstrates that the density limit (DL) is neither related to additional energy losses from the confined region by radiation, nor to an inward collapse of the hot discharge core induced by overcooling of the plasma periphery by radiation. It was observed on both machines that detachment, as well as the X-point MARFE itself, does not trigger a transition in the confinement regime and thus does not present a limit on the plasma density. It is the plasma confinement, most likely determined by edge parameters, which is ultimately responsible for the transition from H-to L-mode. The measured Greenwald fractions are found to be consistent with the predictions from different theoretical models [16,30] based on MHD instability theory in the near-SOL. For complete list of authors see http://dx.doi.org/10.1016/j.nme.2017.01.005

The neohypoplastic model [12] eliminates several shortcomings of the hypoplastic vW-HP model [17]. The most important improvements are presented here. The general form of the constitutive equation \(\dot{\sigma } = {\bar{\mathsf {E}}}:(\dot{\varepsilon } + \mathbf{m}Y \Vert \dot{\varepsilon }\Vert )\) is slightly modified by two additional terms in brackets. The functions \(\mathsf{E}(\sigma ,e), \mathbf{m}(\sigma ,e)\) and \(Y(\sigma ,e)\) are completely reformulated in order to deal with numerous problems of the old model: perpetuum mobile of the second kind, too small dilatancy during triaxial tension paths, poor predictions of peak stress and inconsistent behaviour of the model at the upper density limit. A new state variable \(\mathbf{z}\) similar to the one from the Sanisand model [3] is introduced. The hitherto used state variables: the stress and the void ratio are preserved, of course. Moreover, a new kind of nonlinearity is proposed: the rotation of the deviatoric stress response. The problems connected to the extension intergranular-strain extension are not discussed here.

The boundary region of the RFX-mod reversedfield pinch experiment is strongly influenced bythe value of the electron density. With increasing density the magnetic topology in the plasmacore and in the plasma edge changes, together with the statistical properties of the edge turbulentfluctuations. Using the density normalized to the Greenwald density(n/nG), two density regimesare identified: the low density regime at n/nG0.4-0.5, without a clear threshold between them. At low density the plasma isdominated by the(m, n)=(1,-7)magnetic mode, which gives the same shape to the magneticboundary; at higher density a strong(0, 1)mode rises. When increasingn/nG, edgefluctuationsdevelop strong non-Gaussian tails, turbulent structures become larger and their radial velocitybecome measurable with the gas puff imaging diagnostic whenn/nG>0.3-0.4. Thisphenomenology is described in detail for the two density regimes of RFX-mod.

A radiative collapse predictor has been developed using a machine-learning model based on high-density plasma experiments in the Large Helical Device (LHD). Concurrently, the physical background of radiative collapse was discussed based on the distinct features extracted by a sparse modeling, which is one of the frameworks of data-driven science. Electron density, CIV and OV line emissions, and electron temperature at the plasma edge have been extracted as the key parameters of radiative collapse. Those parameters are relevant to the physical knowledge that the major cause of radiative collapse is the enhancement of radiative loss by light impurities in the plasma-edge region. Using these four parameters, the likelihood of occurrence of radiative collapse has been estimated. The behavior of plasma at the edge—in particular, the carbon impurities outside the last closed flux surface—has been evaluated using EMC3-EIRENE code for the phase with increasing likelihood, that is, the plasma is getting close to the collapse. It is shown that the radiation caused by the C3+ ion, which corresponds to the CIV emission, is enhanced in the region where electron temperature is around 10 eV.

Disruption prediction using a long short-term memory (LSTM) algorithm has been developed on EAST, due to its inherent advantages in time series data processing. In the present work, LSTM is used as the model and the AUC (area under receiver operation characteristic curve) is used as the evaluation index. When the model is trained on data from the plasma current flattop phase and tested on data from the same period multiple times, the highest AUC is 0.8646 and the training time is about 6900 s per epoch. For comparison, the last 1000 ms of the flattop phases are intercepted as short time sequences. When the model is trained on data from short time sequences and tested on data from the same period, the highest AUC is increased to 0.9379 and the training time is restricted to 36 s per epoch. When the best model trained on the short time sequences is applied to the flattop phase for testing, the AUC is up to 0.9189. The experiment results show that it is possible for LSTM to train the model on data from short time sequences and migrate the model to the entire flattop phase, with a shorter training time and higher AUC value.

Efficient lower hybrid current drive (LHCD) is demonstrated at densities up to n[over ¯]_{e}≈1.5×10^{20} m^{-3} in diverted plasmas on the Alcator C-Mod tokamak by operating at increased plasma current-and therefore reduced Greenwald density fraction. This density exceeds the nominal "LH density limit" at n[over ¯]_{e}≈1.0×10^{20} m^{-3} reported previously, above which an anomalous loss of current drive efficiency was observed. The recovery of current drive efficiency to a level consistent with engineering scalings is correlated with a reduction in density shoulders and turbulence levels in the far scrape-off layer. Concurrently, rf wave interaction with the edge and/or scrape-off-layer plasma is reduced, as indicated by a minimal broadening of the wave frequency spectrum measured at the plasma edge. These results have important implications for sustaining steady-state tokamak operation and indicate a pathway forward for implementing efficient LHCD in a reactor.

Macroscopic two-dimensional (2D) images of flowing weakly magnetized lithium (Li+) ion beams are systematically presented. When the beam density approaches the Brillouin density limit, the 2D shape of the flowing Li+ ion beam deforms gradually. Also, the images clearly show not only the m = 1 shift mode but also higher modes of the flowing Li+ ion beams. These observed modes are analyzed by using a Fourier transform method.

The problem of the thermal equilibrium and density limit in tokamak-reactor are analyzed. The empirical Greenwald criterion determining the critical plasma density may be exceeded in tokamaks with auxiliary heating. One may expect that the fusion power may increase the critical plasma density. The thermal balance in tokamak-reactors is discussed. The critical density is defined by the equality of heating (auxiliary plus fusion) power and radiation losses at the plasma edge. The influence of fusion power input as well as auxiliary heating on the critical density is studied. The analytic model is based on some simplifying assumptions. The auxiliary heating as well as the thermonuclear one is assumed to be localized at the center of the plasma column. The slab geometry is assumed for simplicity. It is shown that the critical density rises up increasing the fusion output drastically using the feedback control.

High-density discharges on JET with ITER-like Wall (ILW) have been analysed with the aim of establishing a mechanism for the H-mode density limit (DL) and compared with experiments in the JET carbon material configuration. The density limit is up to 20% higher in the JET-ILW than in the JET-C machine. The observed H-mode density limit is found close to the Greenwald limit. It is sensitive to the main plasma shape and is almost independent of the heating power. It has been observed that the transition from H-mode to L-mode is not always an abrupt event but may exhibit a series of H–L–H transitions, the so-called “dithering H-mode”. It was observed that detachment, as well as the X-point MARFE itself, does not trigger the H–L transition and thus does not present a limit on the plasma density and that it is the plasma confinement which is ultimately responsible for the H-mode DL.

We analyzed broad-band X-ray and radio data of the magnetar SGR J1935+2154 taken in the aftermath of its 2014, 2015, and 2016 outbursts. The source soft X-ray spectrum, Comment: 17 pages, 7 figures, 5 tables, submitted to ApJ, comments welcome

Edge shear flow and its effect on regulating turbulent transport have long been suspected to play an important role in plasmas operating near the Greenwald density limit $ n_G $. In this study, equilibrium profiles as well as the turbulent particle flux and Reynolds stress across the separatrix in the HL-2A tokamak are examined as $ n_G $ is approached in ohmic L-mode discharges. As the normalized line-averaged density $ \bar{n}_e/n_G $ is raised, the shearing rate of the mean poloidal flow $ \omega_{\rm sh} $ drops, and the turbulent drive for the low-frequency zonal flow (the Reynolds power $ \mathcal{P}_{Re} $) collapses. Correspondingly, the turbulent particle transport increases drastically with increasing collision rates. The geodesic acoustic modes (GAMs) gain more energy from the ambient turbulence at higher densities, but have smaller shearing rate than low-frequency zonal flows. The increased density also introduces decreased adiabaticity which not only enhances the particle transport but is also related to a reduction in the eddy-tilting and the Reynolds power. Both effects may lead to the cooling of edge plasmas and therefore the onset of MHD instabilities that limit the plasma density.

The commonly adopted scaling for the maximum achievable plasma density in tokamak fusion devices, the so-called 'Greenwald limit', refers to the line-averaged density along a central chord and depends only on the average plasma current density. However, the Greenwald limit has been exceeded in tokamak experiments in the case of peaked density profiles, indicating that the edge density is the real parameter responsible for the density limit. Furthermore, the Greenwald limit has been obtained for fixed density profiles, so that the scaling can be very different when introducing density profile dependencies on plasma parameters. Dedicated density limit experiments were performed in recent years on the Frascati Tokamak Upgrade, exploring the high density domain in a wide range of values of plasma current, toroidal magnetic field and edge safety factor. New data were collected in the latest experimental campaigns, extending the study of the density limit towards lower values of toroidal magnetic field and plasma current. These experiments confirmed the edge nature of the density limit, as a Greenwald-like scaling was obtained for the maximum achievable line-averaged density along a peripheral chord, while a clear scaling of the maximum achievable line-averaged density along a central chord with the toroidal magnetic field only was found and successfully interpreted as due to interplay between the peripheral Greenwald limit and the specific density profile behavior when approaching the density limit. In particular, an analytical relation between the peripheral and the central density limit was derived for the first time, with the introduction of a generalized parabolic density profile with the peaking factor dependent on the plasma parameters. © 2017 IOP Publishing Ltd.

An analytic method to assign optimal coarse-grained charges based on electrostatic potential matching is presented. This solution is the infinite size and density limit of grid-integration charge-fitting and is computationally more efficient by several orders of magnitude. The solution is also minimized with respect to coarse-grained positions which proves to be an extremely important step in reproducing the all-atom electrostatic potential. The joint optimal-charge optimal-position coarse-graining procedure is applied to a number of aggregating proteins using single-site per amino acid resolution. These models provide a good estimate of both the vacuum and Debye–Huckel screened all-atom electrostatic potentials in the vicinity and in the far-field of the protein. Additionally, these coarse-grained models are shown to approximate the all-atom dimerization electrostatic potential energy of 10 aggregating proteins with good accuracy.

In this chapter, we establish weak convergence of the densities for the perturbed Navier–Stokes system endowed with the artificial pressure ( 11.2). Following the arguments used in Sect. 6.3 we need the renormalized formulation of the continuity equation ( 5.4). Unfortunately, the regularization technique of DiPerna and Lions [26] is no longer applicable as the limit density ϱ is not (known to be) square integrable. Instead, we exploit another piece of information hidden in the effective viscous flux identity ( 6.1).

The thermal pressure coefficients in liquid n-Pentadecane (C15), n-Heptadecane (C17), n-octadecane (C18) and n-nonadecane (C19) was measured using pVT data. The measurements were carried out at pressures up to 150 MPa in the temperature range from 293 to 383 K. The experimental results have been used to evaluate various thermophysical properties such as thermal pressure coefficients up to 150 MPa with the use of density and temperature data at various pressures. New parameters of the linear isotherm regularity, the so-called LIR equation of state, are used to calculate of thermal pressure coefficients of n-Pentadecane (C15), n-Heptadecane (C17), n-octadecane (C18) and n-nonadecane (C19) dense fluids. In this paper, temperature dependency of linear isotherm regularity parameters in the form of a first order has been developed to second and third order and their temperature derivatives of new parameters are used to calculate thermal pressure coefficients. The resulting model predicts accurately thermal pressure coefficients from the lower density limit at the Boyle density at the from triple temperature up to about double the Boyle temperature. The upper density limit appears to be reached at 1.4 times the Boyle density. These problems have led us to try to establish a function for the accurate calculation of the thermal pressure coefficients based on the linear isotherm regularity theory for different fluids.

A revised formulation of the perpendicular diffusive transport model in 2D multi-fluid edge codes is proposed. Based on theoretical predictions and experimental observations a dependence on collisionality is introduced into the transport model of EDGE2D–EIRENE. The impact on time-dependent JET gas fuelled ramp-up scenario modelling of the full transient from attached divertor into the high-recycling regime, following a target flux roll over into divertor detachment, ultimately ending in a density limit is presented. A strong dependence on divertor geometry is observed which can mask features of the new transport model: a smoothly decaying target recycling flux roll over, an asymmetric drop of temperature and pressure along the field lines as well as macroscopic power dependent plasma oscillations near the density limit which had been previously observed also experimentally. The latter effect is strongest for scenarios with strike points on vertical targets and vanishes in case of asymmetric divertor configurations.

Discrete element method (DEM) was used to study the effect of mill speed and lifter face angle on the characteristics of charge group. Based on the bin algorithm, a position density limit is proposed to study the motion of the charge. The result shows that the angle of charge group is mainly affected by mill speed and the measured result proves that the predicted equation is reasonable.

This study demonstrates that any temperature dependency introduced in the term describing the inter-molecular repulsive forces of analytical EOS models leads to prediction of the infinite value for the isochoric heat capacity at the density limit conditions. The temperature-dependent hard sphere diameters and co-volumes necessarily result in the non-physical negative infinite values of this property at the infinite pressure, which in turn leads to the non-physical predictions of other properties, such as the sound velocities. In addition, this study demonstrates that the inconsistency existing at the imaginary state of infinite pressure may have a very strong negative impact on predicting the data of real fluids at the experimentally available pressures. Therefore making the repulsion compressibility factor temperature-dependent seems highly non-recommended.

Routine reaction to approaching disruptions in tokamaks is currently largely limited to machine protection by mitigating an ongoing disruption, which remains a basic requirement for ITER and DEMO [1]. Nevertheless, a mitigated disruption still generates stress to the device. Additionally, in future fusion devices, high-performance discharge time itself will be very valuable. Instead of reacting only on generic features, occurring shortly before the disruption, the ultimate goal is to actively avoid approaching disruptions at an early stage, sustain the discharges whenever possible and restrict mitigated disruptions to major failures. Knowledge of the most relevant root causes and the corresponding chain of events leading to disruption, the disruption path, is a prerequisite. For each disruption path, physics-based sensors and adequate actuators must be defined and their limitations considered. Early reaction facilitates the efficiency of the actuators and enhances the probability of a full recovery. Thus, sensors that detect potential disruptions in time are to be identified. Once the entrance into a disruption path is detected, we propose a hierarchy of actions consisting of (I) recovery of the discharge to full performance or at least continuation with a less disruption-prone backup scenario, (II) complete avoidance of disruption to sustain the discharge or at least delay it for a controlled termination and, (III), only as last resort, a disruption mitigation. Based on the understanding of disruption paths, a hierarchical and path-specific handling strategy must be developed. Such schemes, testable in present devices, could serve as guidelines for ITER and DEMO operation. For some disruption paths, experiments have been performed at ASDEX Upgrade and TCV. Disruptions were provoked in TCV by impurity injection into ELMy H-mode discharges and in ASDEX Upgrade by forcing a density limit in H-mode discharges. The new approach proposed in this paper is discussed for these cases. For the H-mode density limit sensors used so far react too late. Thus a plasma-state boundary is proposed, that can serve as an adequate early sensor for avoiding density limit disruptions in H-modes and for recovery to full performance. © 2017 Max-Planck-Institut for Plasmaphysics.

A simple functional form for a general equation of state based on an effective near-neighbor pair interaction of an extended Lennard-Jones (12,6,3) type is given and tested against experimental data for a wide variety of fluids and solids. Computer simulation results for ionic liquids are used for further evaluation. For fluids, there appears to be no upper density limitation on the equation of state. The lower density limit for isotherms near the critical temperature is the critical density. The equation of state gives a good description of all types of fluids, nonpolar (including long-chain hydrocarbons), polar, hydrogen-bonded, and metallic, at temperatures ranging from the triple point to the highest temperature for which there is experimental data. For solids, the equation of state is very accurate for all types considered, including covalent, molecular, metallic, and ionic systems. The experimental pvT data available for solids does not reveal any pressure or temperature limitations. An analysis of the importance and possible underlying physical significance of the terms in the equation of state is given.

Density limits with different fuelling methods have been compared in HL-2A, i.e. direct gas puffing and supersonic molecular beam injection (SMBI) from outer midplane, and divertor gas fuelling. The maximum densities for low current discharges are 3.4×1019 m−3, 4.3×1019 m−3 and 4.7×1019 m−3 for the 3 kinds of fuelling methods. The corresponding density ratios to Greenwald density limit are 0.9, 1.1, 1.2, respectively. The behavior of density limit disruption is analyzed as well.

The scanning method with GM survey meter is useful for detection of surface radiological density. The detection ability is varied with several parameters; those are the distance between the sample and the window of GM tube, the scanning speed and the background dose rates. We made an experimental system. In that system, a radiation source can be moved at any constant velocity with a linear scanning device. The obtained data shows that the change of count rate has some patterns due to each scanning velocity. The count rate increased with the source coming closer, and decreased with the source moving away. The maximum count rate under the scanning condition, those are 5 mm distance and 50 mm/s scanning velocity, point source on a stainless steel plate was 23% of the count rate on the fixed source. The detection ability in various scanning conditions was evaluated from the measured data. As a result, it was confirmed to be able surely to detect the lower level of the surface radiological density limit by means of the scanning condition applied to the routine work, 5 mm distance and 50 mm/s scanning velocity.

Cold bubble (CB) structures were observed in START density limit studies for the first time in a low aspect ratio tokamak. They seem related to minor and major disruption processes, clearly identified here as a trigger to those events. Enormous discrepancies on the CB velocities in several devices are reported. This shows that the physical mechanisms related to the time scales for its propagation should be revised. Several models related to CB formation and its role in the disruptive process or just in a plasma with the presence of sawteeth qualitatively predict a great part of the observations.

We have searched for HI 21cm absorption toward the two brightest radio AGN at high redshift, J0924--2201 at $z = 5.20$, and J0913+5919 at $z=5.11$, using the Giant Meter Wave Radio Telescope (GMRT). These data set a 3$\sigma$ upper limit to absorption of $< 30%$ at 40 km s$^{-1}$ resolution for the 30 mJy source J0913+5919, and $< 3%$ for the 0.55 Jy source J0924--2201 at 20 km s$^{-1}$ resolution. For J0924--2201, limits to broader lines at the few percent level are set by residual spectral baseline structure. For J0924--2201 the column density limit per 20 km s $^{-1}$ channel is: N(HI) $< 2.2\times 10^{18} \rm T_s$ cm$^{-2}$ over a velocity range of -700 km s$^{-1}$ to $+1180$ km s$^{-1}$ centered on the galaxy redshift determined through CO emission, assuming a covering factor of one. For J0913+5919 the column density limit per 40 km s$^{-1}$ channel is: N(HI) $< 2.2\times 10^{19} \rm T_s$ cm$^{-2}$ within $\pm 2400$ km s$^{-1}$ of the optical redshift. These data rule out any cool, high column density HI clouds within roughly $\pm 1000$ km s$^{-1}$ of the galaxies, as are often seen in Compact Steep Spectrum radio AGN, or clouds that might correspond to residual gas left over from cosmic reionization.

We present an overview of the HALOGAS (Hydrogen Accretion in LOcal GAlaxieS) Survey, which is the deepest systematic investigation of cold gas accretion in nearby spiral galaxies to date. Using the deep HI data that form the core of the survey, we are able to detect neutral hydrogen down to a typical column density limit of about 10$^{19}$ cm$^{-2}$ and thereby characterize the low surface brightness extra-planar and anomalous-velocity neutral gas in nearby galaxies with excellent spatial and velocity resolution. Through comparison with sophisticated kinematic modeling, our 3D HALOGAS data also allow us to investigate the disk structure and dynamics in unprecedented detail for a sample of this size. Key scientific results from HALOGAS include new insight into the connection between the star formation properties of galaxies and their extended gaseous media, while the developing HALOGAS catalogue of cold gas clouds and streams provides important insight into the accretion history of nearby spirals. We conclude by motivating some of the unresolved questions to be addressed using forthcoming 3D surveys with the modern generation of radio telescopes., 4 pages, 1 figure. To be published in the Proceedings of IAU Symposium 309

With a consideration of the local lateral freedom of molecules, a general equation of state for columnar liquid crystals is derived, which is in agreement with the one obtained by Wensink [1] in the density limit. It is also shown that an additional negative pressure, which is not negligible, tends to be generated for thick disc-like liquid-crystalline molecules even in a high-density regime because of the cavities they create as a result of the allowed lateral motion.

The results of experiments to attain high stored energy and density in the Compact Helical System (CHS) are reported. The experiments have been carried out under the maximum neutral beam heating power and highest magnetic field strength of the CHS. With the help of the reheat mode, we have so far reached a stored energy of 9.4 kJ and a density limit expressed as n c = 0.65·(P abs ·B t /V p ) 0.5 for the CHS. In the high-density regime, the confinement of CHS plasma is limited by radiation collapse. A multichannel H α light detector system shows an asymmetric feature in the poloidal cross section and indicates that confinement degradation in the high-density regime begins at the inboard side where the CHS plasma is close to the vacuum vessel wall.

In this work, we discuss the preliminary analysis of some disruptive plasma discharges with a moveable graphite limiter on the HT-7 tokamak, which has been developed in order to investigate the influence of the horizontal plasma position on the density limit. What we have observed is that the m=2, MHD component dominates during almost the whole discharge duration, when the density limit is approached at values of the edge safety factor 3 < q(a) < 3.5. For the poloidal-magnetic-field fluctuations observed at the Mirnov coils, Mirnov-coil analysis shows that the m = 2 mode is unstable, which finally causes disruption.

A moveable graphite limiter has been developed in order to investigate the influence of the horizontal plasma position on the density limit in the HT-7 tokamak. The local recycling and density limit behaviour has been studied in this Letter, which shows that if plasma is shifted outwards the recycling is low, for inwards shift is high. For these set of ohmic discharges a maximum Greenwald number, N GW > 1.2 , was achieved, when the plasma column was repositioned more than Δ / / = + 2.25 cm outwards. It is important that lower local recycling discharge leads lower radiation power and higher density.

We review the high parton density limit of QCD and show how the Color Glass Condensate arises at small xbj. We discuss the applications of the Color Glass Condensate formalism to proton (deuteron) nucleus collisions at high energy colliders, such as RHIC and LHC.

Integrated core–pedestal–SOL modelling is performed for ITER using the MMM95 model for energy transport, modified with additional magnetic shear stabilization to produce a pedestal. The additional stabilization term is calibrated against experiment. The operating space of ITER predicted by the model with neoclassical accumulation of intrinsic carbon impurity is determined. The relationship to an SOL-based density limit is investigated. A reasonable window for operation, with Q larger than 10 is found.

The prediction of major disruptions caused by the density limit, the plasma current ramp-down with high internal inductance li, the low density locked mode and the β-limit has been investigated in JT-60U. The concept of 'stability level', newly proposed in this paper to predict the occurrence of a major disruption, is calculated from nine input parameters every 2 ms by the neural network and the start of a major disruption is predicted when the stability level decreases to a certain level, the 'alarm level'. The neural network is trained in two steps. It is first trained with 12 disruptive and six non-disruptive shots (total of 8011 data points). Second, the target output data for 12 disruptive shots are modified and the network is trained again with additional data points generated by the operator. The 'neural-net disruption predictor' obtained has been tested for 300 disruptive shots (128 945 data points) and 1008 non-disruptive shots (982 800 data points) selected from nine years of operation (1991–1999) of JT-60U. Major disruptions except for those caused by the β-limit have been predicted with a prediction success rate of 97–98% at 10 ms prior to the disruption and higher than 90% at 30 ms prior to the disruption while the false alarm rate is 2.1% for non-disruptive shots. This prediction performance has been confirmed for 120 disruptive shots (56 163 data points), caused by the density limit, as well as 1032 non-disruptive shots (1004 611 data points) in the last four years of operation (1999–2002) of JT-60U. A careful selection of the input parameters supplied to the network and the newly developed two-step training of the network have reduced the false alarm rate resulting in a considerable improvement of the prediction success rate.