Statistical analysis of the maximum response of multi-degree linear system to a seismic excitation

The study shows that the seismic response of two separated pics system is badly estimated by Rosenblueth's formula. Errors may reach 20% in the worst cases (little damping, one of the oscillators in a transitory phase, level ratio of the oscillators is nearly equal to 1). The averaged maximum is under-estimated in this case. To take these effects into account, we purpose a modified formula using the H function depending on the different parameters of the problem. This function is tabulated by numerical simulations. This tabulation is nearly complete except the cases of different dampings and both oscillators in the transitory phase. The proposed formula applies well to multi-degree of freedom systems even when their frequencies are close. The maximum is lightly overestimated. (orig./HP)