Feature-weighted Random Forest with Boruta for Fault Diagnosis of
Satellite Attitude Control Systems
Shaozhi Chen
1
, Xiaopeng Xi
2
, Maiying Zhong
3
, and Marcos E. Orchard
4
1,3
College of Electrical Engineering and Automation, Shandong University of Science and Technology,
Qingdao, Shandong, 266590, China
szchen@sdust.edu.cn
myzhong@sdust.edu.cn (corresponding author)
2
Advanced Center for Electrical and Electronic Engineering, Universidad T
´
ecnica Federico Santa Mar
´
ıa,
Valpara
´
ıso, Chile
xi.xiaopeng@usm.cl
4
Department of Electrical Engineering, Faculty of Mathematical and Physical Sciences, Universidad de Chile,
Santiago, 8370451, Chile
morchard@ing.uchile.cl
ABSTRACT
The performance of random forest (RF) based satellite atti-
tude control system (ACS) fault diagnosis methods is lim-
ited by uninformative features in high-dimensional data. To
solve this problem, we proposed a feature-weighted random
forest with Boruta (FWRFB) based fault diagnosis method
is proposed for fault diagnosis of ACSs. Firstly, a Boruta
feature selection algorithm is used to obtain a feature set
and determine significant feature weights. Subsequently, a
novel feature-weighted random forest (FWRF) algorithm is
designed, which utilizes feature-weighted random sampling
instead of simple random sampling to generate feature sub-
sets in the RF. The FWRFB effectively utilizes the feature
information while mitigating noise interference. Finally, a
FWRFB-based diagnostic module is developed for online
fault diagnosis of ACSs. The effectiveness of the proposed
method is verified by the ACS data from a semi-physical sim-
ulation platform.
1. INTRODUCTION
The satellite attitude control system (ACS) is crucial to guar-
antee the normal operation of onboard loads and even the in-
tegrity of the entire satellite (Yuan, Song, Pan, Song, & Ma,
2021). As the subsystem of satellites has a high fault rate,
sensor faults and actuator faults may occur in the ACS due
to the complex and changeable space environment. Faults
Shaozhi Chen et al. This is an open-access article distributed under the terms
of the Creative Commons Attribution 3.0 United States License, which per-
mits unrestricted use, distribution, and reproduction in any medium, provided
the original author and source are credited.
without immediate handling can lead to satellite performance
degradation or even cause on-orbit mission failure (Ji, Zhang,
& Liu, 2024). Therefore, the fault diagnosis of ACS is
crucial for improving the safety and stability of satellites
(Pourtakdoust, Mehrjardi, & Hajkarim, 2022).
A large amount of test data and telemetry data of ACSs
can be easily obtained. Thus, data-driven fault diagnosis
methods are more adequate and feasible to implement than
model-based approaches; particularly if we cannot rely on
prior knowledge or accurate model descriptions (Xiao & Yin,
2021). A variety of intelligent methods have been applied
to ACS fault diagnosis (Yang & Zhong, 2022). The random
forest (RF) based method is an ensemble method that com-
bines decision trees (DTs) to form a strong classifier, achiev-
ing high robustness and accuracy in handling large-scale data
(Wu, Chen, Qiu, & Zhou, 2024). It has been demonstrated
that the RF algorithm is a suitable method for the fault diag-
nosis of ACSs (Huang et al., 2021). The effectiveness of RF-
based methods usually depends on the classification power
of extracted features from original data (Papakonstantinou,
Daramouskas, Lappas, Moulianitis, & Kostopoulos, 2022).
However, the extracted features affect the fault diagnosis per-
formance differently, and some of them may encumber the
fault diagnosis performance improvement.
Feature subsets are constructed by the way of simple ran-
dom sampling in conventional RF algorithm (S. Chen, Yang,
Zhong, Xi, & Liu, 2023). Features enter the feature subset
of DT with same probability, which may restrain the function
of features with high feature importance (Sun et al., 2011).
1
