Nonlinear constraint systems can be solved by combining consistency techniques and search. In this approach, the search space is reduced using local reasoning on constraints. However, local computations may lead to slow convergences. In order to handle this problem, we introduce a symbolic technique to combine nonlinear constraints. Such redundant constraints are further simplified according to the precision of interval computations. As a consequence, constraint reasoning becomes tighter and the solving process faster. The efficiency of this approach is shown using experimental results from a prototype.