Vibration-based structural health monitoring under changing environmental conditions using Kalman filtering

Abstract A Kalman filtering based framework for structural damage assessment under changing environmental conditions is presented. The approach is based on the well-known property that the filtering residual is a realization of a white stochastic process when the filter is operating under optimal conditions. To decouple structural damage and environmental effects two additional properties of the filtering residual are employed: i) under global changes in the structure caused by environmental variations the residual remains a white process, and thus its spectral density is approximately constant; ii) local changes caused by structural damage induce peaks in the residual spectral density at the affected vibration frequencies, and thus the residual is a colored process. A Bayesian whiteness test is employed to discriminate between the two situations under finite length data conditions (damage detection), while a normalized damage measure based on the spectral moments of the residual spectral density is proposed as a quantitative damage-sensitive feature (damage quantification). The proposed approach is numerically verified in a continuous beam model of a bridge under different operating conditions, including a robustness assessment for non-uniform temperature fields. It is shown that the approach has the capability to decouple physical changes caused by structural damage and varying environmental conditions, providing robust damage measures for structural health monitoring applications.