[en] This paper describes a program for developing analytical techniques to evaluate the response of nuclear material shipping containers to severe accidents. Both lumped-mass and finite element techniques are employed to predict shipping container and shipping container-carrier response to impact. The general impact problem is computationally expensive because of its nonlinear, three-dimensional nature. This expense is minimized by using approximate models to parametrically identify critical cases before more exact analyses are performed. The computer codes developed for solving the problem are being experimentally substantiated with test data from full-scale and scale-model container drop tests. 6 figures, 1 table