1
Frequency domain tensor-based 1D-convolutional neural network and
multilinear principal component analysis for machinery fault detection
Ayantha Senanayaka
1
, Qing Liu
2
, Nayeon Lee
3
, Sungkwang Mun
4
, Amin Amirlatifi
5
, Joe Jabour
6
, Thomas Arnold
7
, Maria
Seale
8
1,2,3,4
Center for Advanced Vehicular Systems, Mississippi State University, Starkville, MS, 39762, US
aus20@cavs.msstate.edu
ql90@cavs.msstate.edu
nayeon@cavs.msstate.edu
sungkwang@cavs.msstate.edu
5
Swalm School of chemical Engineering, Mississippi State University, Starkville, MS, 39762, US
amin@che.msstate.edu
6,7,8
U.S. Army Engineer Research and Development Center, Vicksburg, MS, 39180, US
joseph.e.jabour@erdc.dren.mil
maria.a.seale@erdc.dren.mil
thomas.l.arnold@erdc.dren.mil
ABSTRACT
Challenges in detecting machinery faults, particularly in
multivariate sensor environments, necessitate advanced
feature extraction and classification techniques. This study
introduces a novel approach that combines Multilinear
Principal Component Analysis (MPCA) with a 1D-
Convolutional Neural Network (1D-CNN) for efficient fault
detection. By constructing Frequency Domain (FD) tensors
from multivariate sensor data and applying MPCA for
dimensionality reduction, our methodology enhances the
capabilities of a 1D-CNN in feature learning and fault
classification. The efficacy of this approach is validated
through experiments on a Machinery Fault Simulator (MFS)
with acoustic and vibration sensors, demonstrating notable
improvements in fault detection accuracy compared to
benchmark methods. The study results demonstrate that the
proposed approach exhibits high accuracy in identifying
machine fault conditions and outperforms the benchmark
methods. The findings of this study have significant
inferences for machine fault detection and fill the gap of more
effective and reliable techniques in this domain.
Keywords: Predictive maintenance, Prognostic health
monitoring, Real-time fault diagnosis, Condition monitoring,
Rotating machinery faults, Multilinear principal component
analysis, 1D-convolutional neural network
1. INTRODUCTION
The process of identification of malfunctions or faults in
machinery systems is critical for maintaining proper
equipment health. The primary objective of this process is to
minimize downtime, lower maintenance costs, and ensure
safe and efficient machinery operations (Nallusamy &
Majumdar, 2017). Industrial practitioners utilize various
methods and tools for monitoring and diagnosing machinery
faults. Among these practices, data-driven techniques have
proven to be the most efficient and effective compared to
visual inspections or regular tests (Gonzalez-Jimenez et al.,
2021). In the last two decades, advancements in sensing
devices have led to a revolution in their capacity for sensing
and computation efficiency, enabling real-time monitoring
and diagnosis to improve the system's health and ensure
productivity (Javaid et al., 2021; Kalsoom et al., 2020).
Researchers have also found that multi-sensor information
can achieve more accurate fault diagnosis, providing
comprehensive information on the machinery system's
operation compared to using single sensor information. This
is known as multi-sensor fusion (Liton Hossain et al., 2018).
Acoustic, vibration, pressure, temperature, and current trends
are frequently used signals for multi-sensor fusion (Mallegni
et al., 2022). After collecting data from multi-sensors,
extracting the essential features from the data is the next vital
step. This process helps ML algorithms to identify
