This paper proposes a failure path approach for time-variant system re
liability estimation of brittle structures. The significant failure se
quences are identified by the branch-and-bound method, and the damage
states in these sequences are grouped together into discrete damage le
vels ranging from intact structure to system collapse. The grouping is
based on the number of degrees of redundancy in the structure as it a
ccumulates different amounts of damage. The Markov chain assumption is
used to model the cumulative damage under time-varying loads. The exp
ressions for transition probabilities from one damage level to another
are derived. The effect of load overlapping under multiple load proce
sses is investigated. The method is applicable to large structures due
to the systematic failure sequence identification strategy employed.
The resultant probability transition matrix is found to be more compac
t and to represent the accumulation of damage better than existing met
hods. Two numerical examples are used to illustrate and examine the pr
oposed method.