This paper presents the link mass optimization of a serial robot manipulator based on minimum joint torque requirements that is primary concern in the industrial robot applications. The optimization of link mass globally minimizes joint torques weighted by inverse of inertia matrix. Genetic algorithm (GA) was used to optimize energy produced by robot manipulator. The influences of GA parameters (population sizes and mutation rates) on the solution of this problem were examined by varying these parameters. The rigid body dynamics of a cylindrical three-link serial robot manipulator is used as an optimization model. A fifth order polynomial used for actuating the joints from initial position to the goal position in a smooth manner. The link masses are used as design variables limited to upper, and lower bounds.