A Bayesian Bi-objective Approach for Sensor Allocation in Complex Systems

Abstract

This article addresses the challenge of sensor allocation in complex engineering systems, particularly focusing on the implications of Industry 4.0. It explores the intricacies of these systems, defined as interwoven networks of events, interactions, and uncertainties, and emphasizes the pivotal role of monitoring a system to ensure their safe and efficient operation. Drawing from the growing availability of data and advancements in machine learning, the paper delves into the critical task of sensor placement to gather comprehensive information about the system’s state. The proposed methodology models a bi-objective sensor allocation problem, considering both the information gained through sensor allocation and the associated installation costs, aiming to optimize system health monitoring while considering various sensor trade-offs. This work presents a framework for addressing this challenge, leveraging known probabilities of subsystem failures, and employing an evolutionary optimization algorithm, NSGA-II, to identify Pareto-optimal solutions for effective sensor deployment in complex systems.