A signal processing method for structural healthmonitoring is applied to detect damage in theformer Imperial County Services (ICS) Building,caused by the 1979 Imperial Valley earthquake insouthern California. The building response wasrecorded by a 13-chanel array of accelerometers,and a description of the distribution of the damagethroughout the structure is available. The methodis based on detecting abrupt changes in the seismicvibration response by analysis of the finest detailcoefficients of a wavelet basis expansion of therecorded response. This method has beenpreviously proven to work for numericallysimulated response of simple models withpostulated damage, but not for real earthquakedata. The analysis in this paper critically examinesthe capabilities of this method to detect damage inreal data. The analysis shows that most of thedetected prominent abrupt changes are consistentwith the spatial distribution and severity of thereported damage. Other less prominent abruptchanges can be explained by high frequencyenergy pulses of the input motion that propagatedthrough the building. There are also fewprominent abrupt changes that remain unexplainedat this time. It is concluded that this method couldprovide useful information for structural healthmonitoring and for understanding the seismicresponse of structures and the occurrence ofdamage. Further investigations are needed of the
A signal processing method for structural healthmonitoring is applied to detect damage in theformer Imperial County Services (ICS) Building,caused by the 1979 Imperial Valley earthquake insouthern California. The building response wasrecorded by a 13-chanel array of accelerometers,and a description of the distribution of the damagethroughout the structure is available. The methodis based on detecting abrupt changes in the seismicvibration response by analysis of the finest detailcoefficients of a wavelet basis expansion of therecorded response. This method has beenpreviously proven to work for numericallysimulated response of simple models withpostulated damage, but not for real earthquakedata. The analysis in this paper critically examinesthe capabilities of this method to detect damage inreal data. The analysis shows that most of thedetected prominent abrupt changes are consistentwith the spatial distribution and severity of thereported damage. Other less prominent abruptchanges can be explained by high frequencyenergy pulses of the input motion that propagatedthrough the building. There are also fewprominent abrupt changes that remain unexplainedat this time. It is concluded that this method couldprovide useful information for structural healthmonitoring and for understanding the seismicresponse of structures and the occurrence ofdamage. Further investigations are needed of the