We　consider　a　box-constrained　continuous　global　minimization　problem.　A　new　definition　of　filled　function,　namely　that　of　globally　concavized　filled　function,　is　proposed.　A　new　two-parameter　class　of　globally　concavized　filled　functions　A(x,k,p)　is　constructed,　which　has　the　same　global　minimizers　as　the　problem　on　the　solution　domain　if　p　is　large　enough.　The　minimization　of　A(x,k,p)　can　escape　successfully　from　a　previously　converged　local　minimizer　by　taking　increasing　values　of　k.　A　dynamic　globally　concavized　filled　function　method　is　designed　based　on　these　functions　and　the　convergence　property　is　proved.　Numerical　experiments　on　a　set　of　standard　testing　functions　show　that　the　resulting　method　is　competitive　with　some　well-known　global　minimization　methods.　©　2008　Springer　Science+Business　Media,　LLC.