Abstract
The modal parameters extracted from a structure by accelerometers can be used for damage assessment as well as model updating. To extract modal parameters from a structure, it is important to place accelerometers at locations with high modal displacements. Sensor placement can be restricted by practical considerations, and installation might be conducted more based on engineering judgement rather than analysis. This leads to the question of how important the optimal sensor placement is, and if fewer sensors suffice to extract the modal parameters.
In this work, an offshore wind substation (OSS) from the Wikinger offshore wind farm (owned by Iberdrola) is instrumented with 12, 3-axis accelerometers. This sensor setup consists of 6 sensors in a permanent campaign where sensors were placed based purely on engineering judgement, as well as 6 sensors in a temporary campaign, placed based on a placement analysis. An optimal sensor placement study was conducted using a finite element model of the structure in the software package FEMtools, resulting in optimal layouts. The temporary campaign sensors were placed such that they, in combination with the permanent campaign, can be used to complete the proposed layouts.
Samples for each setup are processed using the software ARTeMIS modal to extract the mode shapes and natural frequencies through the Stochastic Subspace Identification (SSI) technique. The frequencies found by this approach are then clustered together using a k-means algorithm for a comparison within clusters.
The modal assurance criterion (MAC) values are calculated for each result and compared to the finite element model from which the optimal sensor placement study was conducted. This is to match mode shapes between the two and thus determine the importance of off diagonal MAC elements in the sensor optimization process. MAC values are also calculated relative to a cluster-averaged set of eigenvectors to determine how they vary over the 1.5 months.
The results show that for all sensor layouts, the three lower frequency modes are consistently identified. The most optimized sensor layout, consisting of only 3 sensors, was able to distinguish an additional, higher frequency mode which was never identified in the 6-sensor permanent layout. However, the reduced sensor layout shows slightly more scatter in the results than the 6-sensor layout. There is a higher signal to noise ratio in the temporary campaign which results in scatter. We conclude that with an optimized placement of accelerometers, more modes can be identified and distinguished. However, off diagonal elements in the original MAC matrix, as well as loss of sensor degrees of freedom, can result in additional scatter in the measurements. Some of these findings can be extended to other offshore jacket structures, such as those of wind turbines, in that it gives a better understanding of the consequence of an optimal sensor placement study.
A Multiobjective Perspective to Optimal Sensor Placement by Using a Decomposition-Based Evolutionary Algorithm in Structural Health Monitoring