Sensor Placement Optimization for Numerical Model Reduction using Genetic Algorithm

Resumo

Numerical reduction techniques are crucial for numerical and experimental model compatibility.
Besides reducing computational costs, finite element (FE) models generally have significant number of degrees of
freedom (DOFs) when compared to experimental models. Therefore, sensor placement techniques based on
effective independence (EI), condition number of the modal matrix (CN) and the sum of the off-diagonal terms of
the modal assurance criteria (off-MAC) matrix aim to provide optimal sensors position setup for future accurate
experimental tests. Thus, this work presents a comparative study among the mentioned sensor placement
techniques for selecting the candidate numerical DOFs to reduce the free-free beam FE model through the Guyan
technique. In addition, it is applied genetic optimization algorithm (GA) under CN and MAC techniques to reach

optimal solutions. A sensitivity analysis of the optimal responses from CN and MAC was held, along with an F-
test, which ranked the relevant DOFs for sensor placement. The results showed that root-mean-square-error

(RMSE) between the reduced FE and full FE models was less than 5%. MAC values were above 0.86. Finally, it
was identified that the three methods need a DOF’s selection of spatial constraints to circumvent possible problems
of the modes poor spatial resolution of the reduced FE model.