Abstract Two impo rtan t aspect s in disp lacem en t back analysis, calcu lat ion eff iciency and reliab ility, are
discu ssed. F irst ly the erro r back2p ropagat ion neu ral netwo rk (BP netwo rk) is u sed to sub st itu te the t im e2
con sum ing f in ite elem en t analysis, and secondly genet ic algo rithm is u sed fo r the op t im izat ion of ob ject ive
funct ion, m ak ing back2analyzed so lu t ion s irrelevan t to in it ial values. The feasib ility is verif ied by back
analysis of a th ree2m edia slope.
|
|
Received: 22 December 1999
|
|
|
|
