Aplicação de métodos estatísticos multivariados na avaliação de vibrações sísmicas

The present paper analyses multivariate statistical methods applied to evaluate blasting seismic vibration in quarries. Principal components analysis and multigroups discriminate analysis were used. Blast pattern and properties of rock body were the variables used in the present study. Data consider the peak particle velocity, in mm/s and frequency, in Hertz, collected by the seismograph monitoring in an urban quarry at Campinas municipality, São Paulo, Brazil. In view of the great number of variables, it was impossible to correlate them without the multivariate analyses. The results obtained with principal component analysis shower distance as the most important factor in the attenuation of ground vibration. Discriminate analysis shower the charge weight per delay as the most important variable to separate the four groups of data, followed by thickness of soil overburden and distance. Properties of the rock body (orientation, frequency and angle of geological discontinuities and soil overburden) have influence in the ground vibration, increasing or reducing its level, depending on the direction of wave propagation. The analyses gave a better comprehension of the ground vibration process correlating physics data with measured parameters.