A bio-inspired structural health monitoring system based on ambient vibration

Abstract

A structural health monitoring (SHM) system based on naïve Bayesian (NB) damageclassification and DNA-like expression data was developed in this research. Adapted fromthe deoxyribonucleic acid (DNA) array concept in molecular biology, the proposedstructural health monitoring system is constructed utilizing a double-tier regressionprocess to extract the expression array from the structural time history recordedduring external excitations. The extracted array is symbolized as the variousgenes of the structure from the viewpoint of molecular biology and reflects thepossible damage conditions prevalent in the structure. A scaled down, six-story steelbuilding mounted on the shaking table of the National Center for Research onEarthquake Engineering (NCREE) was used as the benchmark. The structuralresponse at different damage levels and locations under ambient vibration wascollected to support the database for the proposed SHM system. To improve theprecision of detection in practical applications, the system was enhanced by anoptimization process using the likelihood selection method. The obtained arrayrepresenting the DNA array of the health condition of the structure was first evaluatedand ranked. A total of 12 groups of expression arrays were regenerated from acombination of four damage conditions. To keep the length of the array unchanged,the best 16 coefficients from every expression array were selected to form theoptimized SHM system. Test results from the ambient vibrations showed thatthe detection accuracy of the structural damage could be greatly enhanced bythe optimized expression array, when compared to the original system. Practicalverification also demonstrated that a rapid and reliable result could be givenby the final system within 1 min. The proposed system implements the idea oftransplanting the DNA array concept from molecular biology into the field of SHM.