Abstract
Errors caused by numerical time integration procedures in the earthquake response of a ground are quantitatively discussed. Newmark's β method (β=1/4) and Wilson's θ method (θ=1.37) are investigated with several treatments of unbalanced forces. The unbalanced forces are carried into next time integral, neglected, or made small by iteration before carrying next time integral. Accuracy of the analyses is discussed by means of amplification and response spectrum at the ground surface, which are important indices in evaluating the ground response in the engineering practice. As the results, we strongly suggest the importance to evaluate errors included in the numerical integration procedure in practice, which can be done by comparing the solution with that under small time increment analysis.
