Although the number of meshes and their distribution are important parameters that affect both computation resources and the accuracy of the numerical results, meshing of most of all computational problems still depends highly on the users’ experiences. In this paper, a deterministic adjoint-based method of optimizing mesh distribution is proposed. The developed method is applied to a nuclear power plant safety analysis. The mesh optimization was performed with 1D steady state cylindrical nuclear fuel problem first. Radial and axial mesh distributions are optimized respectively. With no surprise, the optimized mesh system performs superior than the same number of uniformly meshed system. However, it was unexpected that the optimized mesh retains generality and therefore, the optimized mesh system can be still the best mesh system for given number of meshes under different condition or even during transient analysis. The authors applied the optimized mesh distribution to nuclear system safety analysis. A large pressurized water reactor cold leg guillotine break (LBLOCA) scenario was analyzed and the consequence of different mesh systems is investigated and discussed. From this preliminary study the usefulness and implication of the adjoint based mesh optimization method for the nuclear safety analysis is uncovered.