Elsevier

Fusion Engineering and Design

Volume 138, January 2019, Pages 254-266
Fusion Engineering and Design

A locked mode indicator for disruption prediction on JET and ASDEX upgrade

https://doi.org/10.1016/j.fusengdes.2018.11.021Get rights and content

Abstract

The aim of this paper is to present a signal processing algorithm that, applied to the raw Locked Mode signal, allows us to obtain a disruption indicator in principle exploitable on different tokamaks. A common definition of such an indicator for different machines would facilitate the development of portable systems for disruption prediction, which is becoming of increasingly importance for the next tokamak generations. Moreover, the indicator allows us to overcome some intrinsic problems in the diagnostic system such as drift and offset. The behavior of the proposed indicator as disruption predictor, based on crossing optimized thresholds of the signal amplitude, has been analyzed using data of both JET and ASDEX Upgrade experiments. A thorough analysis of the disruption prediction performance shows how the indicator is able to recover some missed and tardy detections of the raw signal. Moreover, it intervenes and corrects premature or even wrong alarms due to, e.g., drifts and/or offsets.

Introduction

Rotating magneto-hydro-dynamic (MHD) modes in tokamaks are often observed to slow down as their amplitude grows up. When a critical threshold is reached, they stop rotating, or, as it is commonly said, they lock in a certain toroidal and poloidal position leading very often to disruption or, in any case, to a degradation of confinement. The slowing down of the mode propagation velocity can be detected directly measuring the voltage induced by the oscillating field perturbation in tangential field pick up coils or Mirnov coils. Before the mode locking, the measured signal exhibits a growing amplitude accompanied by a reduction in frequency until the oscillation disappears, that is when the mode locks. When the mode is locking the amplitude of its radial component can be evaluated through the measure of the voltage induced in the saddle coils. Such amplitude underlies the analysis developed in this work. In a tokamak device, the mode locking is the most frequent precursor of disruptions, even if, in JET-ILW campaigns, an increased rate of disruptions driven by core radiation peaking or impurities has been observed [1], [2]. Usually, mode locking appears at a later stage in the chain of events characterizing the disruptive process. Its process can be caused by intrinsic error fields or by the deceleration of rotating precursor modes [1]. The analysis of these instabilities can be exploited to develop disruption predictors provided that they have an amplitude large enough to be detected before the thermal quench.

As reported in [1,3,4], most disruptions on JET present precursor locked modes that can be due to either error fields or initially rotating modes. However, for a non-negligible number of them, they manifest too late to intervene. On ASDEX Upgrade (AUG), it is more common to see modes that lock prior to the thermal quench, but also modes that are still rotating at the time of the thermal quench. Such difference can be justified by the lower intrinsic error fields in ASDEX Upgrade [5] compared to JET. Also on AUG, in several percent of cases, the instability manifests itself too close to the disruption time to allow any intervention. An analysis of the locking of the mode before the disruption in the databases of JET and AUG is reported in Section 4.

On JET and AUG, protections against locked mode disruptions already exist, which use the amplitude of the locked mode signal as a threshold. Such protection systems are machine dependent: at AUG the raw signal, available in real time for disruption mitigation purposes, provides a measurement in Volt, whereas at JET such signal is in Tesla.

However, the development of portable systems for disruption prediction is becoming of increasingly importance for the next tokamak generations. Referring to JET and AUG, in this paper, a common definition of a locked mode indicator is proposed and the viability to identify a common threshold value is investigated. To this purpose, referring to the JET and AUG experiments, a multi-machine analysis has been performed. In particular, two databases have been built containing about six hundreds of disrupted and non-disrupted discharges in JET with ITER-Like Wall (ILW) from 2012 to 2014 and roughly the same number of shots at AUG from 2009 to 2015. The AUG raw signal has been calibrated in Tesla to be compared with that of JET. Moreover, in order to handle dimensionless quantities both the signals at JET and AUG have been normalized as suggested in [3].

Furthermore, in both machines, the raw signals sometimes show offsets and/or drifts, due to no compensation of the poloidal fluxes generated by the plasma current and the current driving the OH coils. Moreover, active coils for error field correction could introduce off-set that have to be carefully handled. Hence, to avoid the effect of spurious trips in this signal, the trigger level used to request the pulse termination is set to a rather high value. This procedure has the clear drawback of delaying the request to safely shut down the plasma even when, by monitoring the signal itself, the presence of a locked mode could have been detected earlier.

Usually, the roots of drift and off-set are well known, and the correction could be trivial. However, the analysis of their causes could be time consuming and its results do not enhance the disruption prediction capability of the available signal. Hence, removing the drift and the off-set from the point of view of the real time data processing, without getting into the diagnostic limits and machine settings, is sufficient. In this paper, an algorithm for drift and off-set removal has been implemented and applied to the normalized locked mode signal. The algorithm does not require any preliminary analysis, and it can be applied also to those shots not affected by drift and off-set preserving the signal information one is interested in.

Finally, as proposed in the literature [6,7], the resulting locked mode signal has been processed to exploit the content information of both the time and frequency domains resulting in an indicator fitting both devices.

In fact, some algorithms, mostly based on the frequency content of the raw locked mode signal, have been tested at JET for minimizing the effect of these issues so as to provide an earlier alarm than the current locked mode protection system [6]. These algorithms demonstrated to perform better than those that trigger an alarm only when the locked mode amplitude exceeds a prefixed threshold. Also in [7] time-frequency analysis was used to detect anomalies in the locked mode signal that triggers the disruption alarms on JET. Such predictor required a limited number of past experiments to set the alarm threshold.

In the present paper, the validity of the proposed indicator has been assessed by using it as disruption predictor.

Finally, a deep analysis of the prediction errors has been performed to understand the intrinsic limits of the use of only the locked mode signal as a disruption predictor.

The remainder of this paper is organized as follows. Section 2 reports details of the locked mode diagnostics at JET and AUG. The rationale of the locked mode indicator is presented in Section 3. The statistical analysis of the adopted databases is presented in Section 4. Results of the locked mode indicator as disruption predictor, and a reasoned analysis of the obtained results are discussed in Section 5. In Section 6 conclusions are drawn.

Section snippets

Locked mode diagnostics

On JET, the locked mode amplitude is measured by a set of 2 × 4 saddle flux loops, located at radial positions, above and below the middle plane, and mounted on the outside of the vacuum vessel at the low-field side (LFS) of the plasma. They are positioned at a 90° angle to each other. The locked mode amplitude, measure of the predominant odd n = 1 mode, comes from several elaborations of these flux loops signals, such as integration and the appropriate compensation of the poloidal fluxes

Locked mode indicator

Generally speaking, an indicator can synthesize more signals in order to describe a complex phenomenology, or can be based on a single signal when the signal itself is intrinsically representative of a disruptive behaviour as, for example, in the case of the locked mode. In most circumstances, in order to maximize the information content in the signal, it can be required to remove noise or unwanted spikes, or simply to extract the trend of the signal filtering transient phenomena. In this

Databases

In order to statistically assess the effectiveness and robustness of the proposed indicator, it has been tested as a disruption predictor with both JET and AUG data. For the sake of comparison, the performance of the prediction systems, obtained by optimizing a threshold on the actual locked mode trigger, i.e., the raw signal normalized with respect to the plasma current Ipla on JET (BLM/Ipla) in [T/MA], and the raw signal (LM-raw) in [V] at AUG, have been computed. Moreover, also the

Locked mode signal for disruption prediction

Presently, on JET and AUG, the protection against disruptions uses the amplitude of the locked mode signal as a trigger. In particular, the alarm threshold is set on the BML/Ipla signal in [mT/MA] at JET and on the LM-raw signal in [V] at AUG. Hence, there is no correlation between the two trigger values. In a view of a machine independent disruption prediction approach, dissimilarities between the two locked mode triggers can be overcome taking into account, for both machines, the raw locked

Analysis of prediction errors

This section reports the results of the analysis performed to identify the causes of FAs, MAs, and TDs for both the normalized raw signal and the LM indicator. The following analysis shows that some intrinsic problems in the raw signal such as drifts, offsets, and error field pick up coils (EFCC), can cause false alarms and incongruous alarms, which are not related to physical reasons.

The analysis has been made making reference to the different predictions performed by the two considered

Conclusions

The main aim of the paper was to evaluate the advantages and the limits of the proposed locked mode indicator if used in a disruption predictor instead of the raw signal, highlighting the possibility of improving the mode locking detection through a suitable processing algorithm exploiting both information in time and frequency domain. The paper, not only presents the performance of a locked mode based disruption predictor but also analyses the wrong prediction causes and investigates the

Acknowledgements

This work has been carried out within the framework of the EUROfusion Consortium and received funding from the EURATOM research and training programme 2014–2018 under grant agreement No 633053. The views and opinions expressed herein do not necessarily reflect those of the European Commission.

References (12)

There are more references available in the full text version of this article.

Cited by (13)

View all citing articles on Scopus
1

See the author list “A. Kallenbach et al., Nucl. Fusion 57 102015”

2

See the author list “H. Meyer et al 2017 Nucl. Fusion 57 102014”

3

See the author list of “X. Litaudon et al 2017 Nucl. Fusion 57 102001”

View full text