Deconvolution is a time domain technique to compute the impulse response function of dynamic systems directly from input-output time series. Since no FFT and IFFT is performed, data are leakage free also when they're not periodic and no window function is applied. The classical deconvolution expression (i.e. an integral equation) reveals two main limits: a) the numerical computation of the impulse response passes through the solution of a severely ill conditioned problem; b) the system initial conditions (displacement and velocity) have to be null. In this paper we propose an alternative formulation which should by-pass both these limits. Numerical results are reported to support the theoretical hypothesis and highlight the peculiarities of the method.
展开▼
