The unconfined random field may lead to higher variability especially when the variability of dissolution foundation is high

and may lose the geological information and practice experience. So the Bayes constraint random field model is put forward. The constraint random field is built by introducing Bayes formula to modify the distribution characteristics of parameters of dissolution

to reflect the new geological and samples’ location information. Then the effect of the dissolution system in the dam foundation is studied through stochastic finite element method (SFEM). The results show that the space variability of dissolution can be considered more objectively by Bayes constraint random field model rather than the unconfined one; also

the variance of the structural responses can be reduced effectively. In the probability analysis process

the Monte Carlo response surface coupled method (MC-RSM) is recommended for the random simulation of complex engineering

which can reduce the simulation time relative to the direct Monte Carlo method which can only repeat the process of the same mechanical mechanism.