Neural Network Based Identification of Energy Conversion Regions and Bursty Bulk Flows in Cluster Data

Vlad Constantinescu,Octav Marghitu

doi:10.3389/fspas.2020.00051

Vlad Constantinescu, Octav Marghitu

Open Access

https://doi.org/10.3389/fspas.2020.00051

Copy DOI

Abstract

Neural networks (NN) provide a powerful pattern recognition tool, that can be used to search large amounts of data for certain types of 'events'. Our specific goal is to make use of NN in order to identify events in time series, in particular energy conversion regions (ECRs) and bursty bulk flows (BBFs) observed by the Cluster spacecraft in the magnetospheric tail. ECRs are regions where E·J≠0 is rather well defined and observed on time scales from a few minutes to a few tens of minutes (E is the electric field and J the current density). BBFs are high speed plasma jets, known to make a significant contribution to magnetospheric dynamics. Not surprisingly, ECRs are often associated with BBFs. The manual examination of the Cluster plasma sheet data from the summer of 2001 provided start-up sets of several ECRs and, respectively, BBFs, used to train feed-forward back-propagation NNs. Subsequently, larger volumes of Cluster data were searched for ECRs and BBFs by the trained NNs. We present the results obtained and discuss the impact of the signal-to-noise ratio on these results.

Highlights

As sensor resolutions and sampling frequencies increase, data available from space missions is steadily increasing
In sections Neural Network Identification of ECRs: Comparison With an Algorithmic Approach and Neural Network Identification of BBFs, we present the application of this algorithm on two types of time series data provided by instruments onboard Cluster spacecrafts, namely energy conversion regions (ECRs) and bursty bulk flows (BBFs)
Using the software developed to search for ECRs, we considered testing the functionality of the selection tool provided by the neural network implementation on other time series

Summary

INTRODUCTION

As sensor resolutions and sampling frequencies increase, data available from space missions is steadily increasing. In sections Neural Network Identification of ECRs: Comparison With an Algorithmic Approach and Neural Network Identification of BBFs, we present the application of this algorithm on two types of time series data provided by instruments onboard Cluster spacecrafts, namely energy conversion regions (ECRs) and bursty bulk flows (BBFs). These two examples illustrate the cases of low and high signal to noise ratio, for ECRs and BBFs, respectively, and provide qualitative information on the impact of this parameter, as well as a test bed for an upcoming quantitative assessment. Data are multiplied with the corresponding weights and the input of each neuron is computed as the sum of the weighted input values: NEURAL NETWORKS

Introduction

CONCLUSIONS

Findings

DATA AVAILABILITY STATEMENT