In the exploitation of ocean oil and gas, many offshore structures maybe damaged due to severe environment, an effective method for diagnosing structural damage is needed to locate the damage and evaluate its severity. And genetic algorithm has become one of the most important global optimization tools and is widely used in many fields in resent years because of its simple operation and strong robustness. Based on the natural frequencies and mode shapes of the structure, the damage diagnosis of jacket offshore platform is attributed to an optimization problem and studied using genetic algorithm. According to the principle that the structure stiffness of a certain direction can be affected only when the brace bar of corresponding direction is damaged, an improved objective function is proposed based on noise measurement and the characteristics of modal identification for offshore platform, it is used as fitness function of genetic algorithm. Both numerical simulation and physical model test results show that the improved method can locate the structural damage and evaluate its severity of jacket offshore platform, thus it can improve the robustness of evolutionary searching and the reliability of damage diagnosis.