Machine learning detection of dust impact signals observed by the Solar Orbiter

Abstract

Abstract. This article presents the results of automatic detection of dust impact signals observed by the Solar Orbiter – Radio and Plasma Waves instrument. A sharp and characteristic electric field signal is observed by the Radio and Plasma Waves instrument when a dust particle impacts the spacecraft at high velocity. In this way, ∼ 5–20 dust impacts are daily detected as the Solar Orbiter travels through the interplanetary medium. The dust distribution in the inner solar system is largely uncharted and statistical studies of the detected dust impacts will enhance our understanding of the role of dust in the solar system. It is however challenging to automatically detect and separate dust signals from the plural of other signal shapes for two main reasons. Firstly, since the spacecraft charging causes variable shapes of the impact signals, and secondly because electromagnetic waves (such as solitary waves) may induce resembling electric field signals. In this article, we propose a novel machine learning-based framework for detection of dust impacts. We consider two different supervised machine learning approaches: the support vector machine classifier and the convolutional neural network classifier. Furthermore, we compare the performance of the machine learning classifiers to the currently used on-board classification algorithm and analyze 2 years of Radio and Plasma Waves instrument data. Overall, we conclude that detection of dust impact signals is a suitable task for supervised machine learning techniques. The convolutional neural network achieves the highest performance with 96 % ± 1 % overall classification accuracy and 94 % ± 2 % dust detection precision, a significant improvement to the currently used on-board classifier with 85 % overall classification accuracy and 75 % dust detection precision. In addition, both the support vector machine and the convolutional neural network classifiers detect more dust particles (on average) than the on-board classification algorithm, with 16 % ± 1 % and 18 % ± 8 % detection enhancement, respectively. The proposed convolutional neural network classifier (or similar tools) should therefore be considered for post-processing of the electric field signals observed by the Solar Orbiter.

Modeling Solar Orbiter dust detection rates in the inner heliosphere as a Poisson process

“…In addition to L2 Solar Orbiter RPW data, this work makes use of the data product provided by Kvammen et al (2023), which is a result of a convolutional neural-network-classified time-domain-sampled data. It builds on a supervised classification algorithm trained using a randomly chosen subsample of manually labeled data.…”

“…This however does not spoil the assumption of Poisson distribution and it is accounted for in the analysis. Although no supervised classifier could get rid of human bias and error completely, these data provide the most reliable Solar Orbiter dust detection data available to date, as has been shown in Kvammen et al (2023). The data set consists of 4606 dust detections acquired over approximately 669 h within 457 days between 29 June 2020 and 16 December 2021.…”

“…We make use of the data that were publicly available on 01 September 2022. We refer to these data as TDS/TSWF-E/CNN and it is publicly accessible (see Kvammen et al 2023).…”

“…It is shown by Zaslavsky et al (2021) that this approach allows for an order of magnitude estimation of the radial component of dust velocity v dust;rad ≈ 50 km s −1 , which is in line with expectations. In the present work, we extend the approach to estimate continuous heliocentric-distance-dependent dust radial velocities, using the data product of Kvammen et al (2023) and taking into account more unknown variables that influence our estimates.…”

“…The procedure described in sec. 4 was also applied to the original TDS data, meaning impacts were identified onboard Solar Orbiter, described by Maksimovic et al (2020), which are different from TDS/TSWF-E/CNN data (Kvammen et al 2023) used otherwise in sec. 4.…”