A multi-objective optimization method for system reliability allocation was proposed. Firstly, a constrained multi-objective optimization model for system reliability allocation was established with the objective to maximize the system reliability, minimize the cost, volume and mass. Secondly, aiming at the characteristic of the model, a non dominated sorting genetic algorithm with elitist strategy (NSGA II) was presented and designed. In the algorithm, an object-oriented technique was introduced to map each individual to each corresponding object and a population to an array of objects, each individual was encoded by each corresponding vector made up of the selection indices of every function unit, the tournament selection strategy was used to implement the selection operation based on the ranks and congestion degrees of the individuals, a swap way of two-point gene segment was used to implement the crossover operation, a one point mutation way was used to implement the mutation operation and the Pareto operation was implemented based on the ranks and congestion degrees of the individuals. Finally, the effectiveness of the proposed research was validated by case study.