In this paper, a new stochastic programming approach is presented to address chemical process optimization problems under uncertainty. The novel algorithm, named as delayed sampling approach, solves an equivalent deterministic optimization model transformed from the stochastic optimization problem between two stochastic simulations. The sampling numbers are reduced considerably and the computational burden is then alleviated remarkably. A complex crude distillation unit is modeled and optimized using the new stochastic approach. Savings of up to 80% in CPU time has been achieved without significant loss of solution precision compared to the conventional stochastic optimization method.