The problem of constraints violation associated with numerical integra
tion of dynamical equations of motion subject to constraints is dealt
with. A new method for the suppression of constraints violation is pre
sented. The method involves the definition of old constraints and of n
ew constraints, the removal of the old constraints and the imposition
of the new constraints, and the evaluation, after each integration ste
p, of changes in the integration variables resulting from the removal-
imposition process. The method is shown to be superior to the classica
l Baumgarte's method for constraints violation suppression in three as
pects. First, the constraint violation measure number, a quantity defi
ned herein, is smaller. Second, the constraint violation measure numbe
r stabilizes after the first integration step and does not degrade wit
h time. Finally, the new method makes the process of trial and error,
required by Baumgarte's method for the determination of integration co
nstants, unnecessary. The method is discussed in connection with holon
omic and simple nonholonomic constraints.