Accidental defects occur often during the construction of the structural sealant in a glass curtain wall structure. However, there are tens of thousands of glass panel units in a building, and the baseline models of different panel units are different from each other. In this paper, a two-stage construction defect identification method is proposed to identify a few units with structural sealant defect from a large number of units. In the first stage, the average driving-point accelerance (DA) was used as statistical baseline DA to calculate the accumulative difference of driving-point accelerance (ADDA). Then, the abnormal panel units are identified by the quantile estimation method. In the second stage, structural sealant construction defects are detected further by comparing the peak frequency difference between the statistical baseline DA and the abnormal DA. Several numerical examples and on-site experiments indicated that the probability density function of ADDA is agreement with the Weibull distribution, and the proposed method can effectively detect the structural sealant construction defects from many units without baseline model.
Read full abstract