In view of the drawbacks that it is prone to the phenomenon of premature convergence andits local search ability is poor, the genetic algorithm was improved by 2 operations of the fish algo-rithm:the rear end behavior which can speed up the optimal and the cluster behavior which has anexcellent ability to jump out of the local extremum. Taking the gate-shaped frame for an example，the improved genetic algorithm was applied to the reliability analysis of the frame structures. The opti-mization model was established with the unit section area and the external load as the design variablesand the reliability indices as the objective function. The results of the optimization model calculatedby the JC method and the improved genetic algorithm were similar at the same checking point underdifferent loadings，but the improved genetic algorithm performed efficiently with less constraints.