Damage detection using autocovariance functions

• ACFs are proven to be valid features for damage detection. • ACFs are particularly advantageous in wireless sensor networks. • A method is proposed to identify which changes are detectable. • Variations in excitation statistics do not affect detectability. • A diagram is presented to assist in selecting key parameters of the SHM system. Using wireless sensor networks in structural health monitoring systems is cost-efficient, because cables are not needed. On the other hand, issues occur with time synchronization between sensor nodes and transmission of long data records. To address these issues, autocovariance functions (ACFs) are proposed as damage-sensitive features, because they can be estimated independently in each sensor node and time-synchronization is not necessary. Data transmission to the central data repository is also more efficient, since ACFs are much shorter than the measurement period. The application of ACFs for damage detection lacks both theoretical justification and practical implementation guidelines. The objective of this paper is to prove the validity of ACFs for damage detection and provide practical instructions on choosing values for two design variables: the number of sensors and the number of time shifts given the number of active modes. Their values determine which changes are detectable. With appropriate design variables, data redundancy can be achieved, and damage can be detected using subspace methods. A diagram is provided to assist in selecting these two parameters. It is also proved that with a proper choice of design variables, the variability of the excitation statistics has no effect on redundancy. The use of ACFs for structural change detection was experimentally validated with a laboratory structure and the KW51 Bridge. A subspace-guided damage detection procedure was applied, in which it is not necessary to determine the subspace dimensions.