Computationally efficient multi-type sensor placement for large-scale engineering structures

Andrzej Świercz

doi:10.1016/j.ymssp.2025.112615

Return

Computationally efficient multi-type sensor placement for large-scale engineering structures

2025-05-01

PRE

Andrzej Świercz *

Bartłomiej Błachowski

Piotr Olaszek

Jan Holnicki‐Szulc

Łukasz Jankowski

DOI：10.1016/j.ymssp.2025.112615

Original

Original request for help

Save

Abstract

En 中文

Performance of a structural health monitoring (SHM) system depends on the set of sensors distributed across the monitored structure. Optimal deployment of sensors on large-scale structures, such as tied-arch bridges, is a significant challenge. Condition assessment of a bridge is typically based on its displacement response under operational or diagnostic loads. However, direct displacement measurements require reference-based methods, which is problematic for bridges. Consequently, other sensor types that do not require reference points, such as accelerometers or inclinometers, are commonly used in practice. These sensors can indirectly provide displacement information but require sophisticated numerical integration and filtering techniques. Deploying a sensor network becomes even more challenging when it is heterogeneous and simultaneously utilizes sensors of various types. This paper proposes a sensor placement method for distributing such heterogeneous sensor networks. Two computationally efficient procedures are introduced, based on Kalman filtering and response estimation uncertainty. Their effectiveness is demonstrated using a realistic example of a tied-arch bridge located in Poland. One algorithm operates in a discrete greedy manner, while the other fuzzifies the sensor set to convert the originally discrete problem into a continuous one. Their numerical efficiency is related to the computationally inexpensive use of the cross-covariance matrix between the sensor responses and the target responses of interest. Compared to an existing multi-type sensor placement method, the proposed algorithms yield results of comparable quality with several times smaller computational cost.

Keywords:

Sensor networks

Optimal sensor placement

Kalman filter

Convex relaxation