Unsupervised Fault Detection in a Controlled Conical Tank
Joaqu
´
ın Ortega
1
, Camilo Ram
´
ırez
2
, Tom
´
as Rojas
3
, Ferhat Tamssaouet
4
, Marcos Orchard
5
, Jorge Silva
6
1,2,3,5,6
Information and Decision Systems Group, University of Chile, Santiago, 8370451, Chile
joaquin.ortega@ug.uchile.cl
camilo.ramirez@ug.uchile.cl
tomas.rojas.c@ug.uchile.cl
morchard@ing.uchile.cl
josilva@ing.uchile.cl
4
PROMES-CNRS, UPVD, Rambla de la Thermodynamique, Tecnosud, Perpignan, 66100, France
ferhat.tamssaouet@univ-perp.fr
ABSTRACT
Current trends in the Industrial Internet of Things (IIoT) have
increased the sensorization of systems, thus increasing data
availability to apply data-driven fault detection and diagnosis
techniques to monitor these systems. In this work, we show
the capabilities of an information-driven method for detecting
and quantifying faults in a subsystem common among a broad
range of industries: the conical tank. Our main experiment
consists of using a simple black-box model (multi-layer per-
ceptron – MLP) to capture the dynamics of a PID-controlled
conical tank built in Simulink and then induce pump failures
of different severities; the derived data-driven indicators that
we developed increase with the severity of the fault validating
its usefulness in this controlled setting. A complementary ex-
periment is carried out to enrich our analysis; this consists of
simulating an open-loop discrete-time version of the conical
tank to explore a range of fault severity and analyze the distri-
bution of the indicators across this range. All our results show
the applicability of the data-driven fault monitoring method in
conical tanks subjected to either open- or closed-loop opera-
tion.
1. INTRODUCTION
Fault detection and identification (FDI) and fault diagnosis
are essential elements in ensuring the reliability and safety
of systems, including those used in industrial processes such
as conical tanks. FDI involves recognizing the presence of
faults in a system, whereas fault diagnosis goes a step fur-
ther by determining their location and nature (Abid, Khan, &
Iqbal, 2021). These tasks are essential, as undetected faults
Joaqu
´
ın Ortega et al. This is an open-access article distributed under the
terms of the Creative Commons Attribution 3.0 United States License, which
permits unrestricted use, distribution, and reproduction in any medium, pro-
vided the original author and source are credited.
can lead to system failure, reduced efficiency, safety risks,
and increased operating costs. Early detection and accurate
diagnosis of faults can prevent these problems and ensure
continuous and safe operation. Common methods for FDI
and fault diagnosis include model-based approaches, signal-
processing techniques, and data-driven methods. Model-
based approaches use mathematical models to detect devia-
tions, signal processing analyzes output signals for anoma-
lies, and data-driven methods leverage machine learning and
historical data to identify patterns and correlations (Abid et
al., 2021).
The majority of the works in the literature assume open-loop
systems when identifying faults. Indeed, closed-loop control
can degrade FDI and fault diagnosis performance. This is be-
cause the system’s robustness can mask early or minor faults,
lowering detection rates. Additionally, the feedback mech-
anism can cause faults to propagate and couple within the
system, making fault identification more challenging (Sun,
Wang, He, Zhou, & Gu, 2019; Talebi & Khorasani, 2012;
Costa, Angelov, & Guedes, 2015). This raises a caveat, as, in
most industrial and real-world settings, systems are subjected
to a control loop.
Several strategies were developed to deal with system degra-
dation. These strategies can be related to fault mitigation or
failure prevention. In fault mitigation, the failure is taken into
account in the design stage, which tends to increase the fault
resilience of systems. Hence, domains such as system re-
configuration, fault tolerance (Amin & Hasan, 2019), self-
repairing systems (Yang & Kwak, 2022), and self-healing
(Ghosh, Sharman, Rao, & Upadhyaya, 2007) can be gath-
ered under the name of fault mitigation. Despite the out-
standing achievements of fault mitigation, failures cannot be
eliminated; therefore, it is necessary to consider them as un-
avoidable events that have to be prevented. In practice, fail-
1
