Optimal Sensor Configuration for Fatigue Life Prediction in Structural Applications

Abstract

Structural health monitoring is spreading widely across engineering domains. Its added value is not restricted to observing structural behavior, but crosses over to enabling the assessment of structural integrity under varying operating conditions. Damage prognosis is one vital demand from structural health monitoring solutions. Many methods have been developed to update damage predictions based on sensor data, nonetheless the selection and positioning of sensors to alleviate the prediction errors remains a question under investigation. In this work, an optimal sensor placement method is proposed for fatigue damage prediction in structures. An optimization problem is formulated to minimize the a-posteriori damage estimation error based on a Kalman filter. The derivation of the objective function is presented, along with a discussion of algorithm-related issues. Finally, the mentioned damage prediction approach is applied to two structures to verify the adequacy of the sensor configurations proposed by the method.

Information theoretic-based optimal sensor placement for virtual sensing using augmented Kalman filtering

“…In most of the approaches, the scalar measure is chosen to be the trace of the reconstruction error covariance matrix. The use of a scalar measure (usually the trace) of the steady-state covariance of the reconstruction error is borrowed from optimal sensor placement strategies proposed in structural dynamics applications for response reconstruction based on Kalman-type filter methods subjected to known loads [44][45][46]. The size of the reduction of the measure as a function of the number of sensors is used to select a reasonable number of sensors.…”

Advanced optimal sensor placement for Kalman-based multiple-input estimation

“…The selection of the optimal type and location of sensors is a common issue in experimental design and inverse identification problems. Recently, Mohamed et al [26] proposed an optimal sensor placement method for damage identification. The problem is formulated to minimize the a posteriori KF estimation error in the discrete-time formulation and it selects the optimal sensors using a recursive gradient-based screening starting from the full model.…”