Robust optimal sensor configuration using the value of information

Abstract

Sensing is the cornerstone of any functional structural health monitoring technology, with sensor number and placement being a key aspect for reliable monitoring. We introduce for the first time a robust methodology for optimal sensor configuration based on the value of information that accounts for (1) uncertainties from updatable and nonupdatable parameters, (2) variability of the objective function with respect to nonupdatable parameters, and (3) the spatial correlation between sensors. The optimal sensor configuration is obtained by maximizing the expected value of information, which leads to a cost-benefit analysis that entails model parameter uncertainties. The proposed methodology is demonstrated on an application of structural health monitoring in plate-like structures using ultrasonic guided waves. We show that accounting for uncertainties is critical for an accurate diagnosis of damage. Furthermore, we provide critical assessment of the role of both the effect of modeling and measurement uncertainties and the optimization algorithm on the resulting sensor placement. The results on the health monitoring of an aluminum plate indicate the effectiveness and efficiency of the proposed methodology in discovering optimal sensor configurations.

Bayesian optimal sensor placement for acoustic emission source localization with clusters of sensors in isotropic plates

“…Using utility theory [40], the optimal experimental design is accomplished by maximizing the expected information gain over all possible data generated from the prediction error model. It can be shown that asymptotically for large number of data and small prediction errors, the utility function that measures the expected information gain from the data can be simplified as follows [41,42,43,30]…”

“…Leyder et al [28] reported on optimal sensor configurations for the modal identification of a post-tensioned timber frame structure while keeping the number of necessary sensors to a minimum. More recently, a methodology for optimal sensor configuration based on the value of information was reported in [29,30], while Ercan et al [31,32] presented information entropy based optimal sensor placement strategies for virtual sensing. A thorough review of the different optimal sensor placement strategies for SHM applications can be found in [33,34].…”

Three-Way k-Means Model: Dynamic Optimal Sensor Placement for Efficient Environment Monitoring in Pig House

“…Utilizing the enhanced particle swarm optimization algorithm, the system iteratively seeks the optimal solution for the data set, which corresponds to the solution of the cluster center. Once the cluster center was determined, data points were assigned to the nearest cluster based on the nearest neighbor distance, calculated as shown in Formula (11). The resulting output was the data set's clustering set.…”

“…In step 5, calculate the distances from all samples within each particle to the cluster centers using Formula (11). Following the principle of proximity, assign each sample to the cluster whose center was closest, thereby effectively partitioning them into their respective clusters.…”

“…In tackling the challenge of the optimal sensor placement, researchers have harnessed a diverse array of algorithms and techniques for sensor configuration, encompassing genetic algorithms [7,8], heuristic algorithms [9], optimization algorithms [10], information theory [11,12], and machine learning [13]. This study is motivated by the objective of exploring the optimal positioning of temperature and humidity sensors within pig houses, with the overarching aim of orchestrating their layout in a manner that is both effective and cost-efficient for comprehensive indoor environment monitoring.…”