This paper derives a computationally efficient algorithm to determine optimal sequential sensor placement for state estimation in linear structural systems subjected to unmeasured excitations and noise contaminated measurements. The proposed algorithm is developed within the context of the Kalman filter and it minimizes the variance of the state estimate among all possible sequential sensor locations. The paper investigates the effects of measurement type, covariance matrix partition selection, spatial correlation of excitation and model selection on optimal sensor placement. The paper shows that the sequential approach reaches the optimal sensor placement as the number of sensor increases.