Evaluation Metrics for Collaborative Fault Detection and Diagnosis in Cyber-Physical Systems
Artigo de Conferência
Visão geral
Pesquisas
Identidade
Informação adicional documento
Ver Todos
Visão geral
resumo
Cyber-physical systems (CPS) rapidly expand within
industrial contexts in a new era of digitalization, processing
power, and inter-device communication capabilities. These advancements
integrate technologies such as the Internet of Things
(IoT), artificial intelligence (AI), and cloud and edge computing,
granting processes and operations a high degree of autonomy.
In addition, these interconnections foster collective intelligence
arising from information exchange and collaboration between
components, often outperforming individual capabilities. This
collective intelligence manifests in fault detection and diagnosis
(FDD) tasks within CPS, as it significantly improves the flexibility,
performance, and scalability. However, the inherent complexity
of CPS poses challenges in determining the best configuration
of the collaboration parameters, such as when and how to
collaborate, wherein incorrect adjustments may lead to decision
errors and compromise the system’s performance. With this in
mind, this paper proposes seven metrics to evaluate collaboration
performance for fault detection and diagnosis in multi-agent
systems (MAS)-based CPS, evaluating when the collaboration
is beneficial or when the collaboration parameters need to be
adjusted. The experiments focus on collaborative fault detection
in temperature and humidity sensors within warehouse racks,
where the proposed evaluation metrics point out the impact of
collaboration on the detection task, as well as possible actions to
be adopted to improve the agent’s performance.
