This paper presents a new numerical approach using a Genetic algorithm (GA) to search for the singularity-free cylindrical space of a 4-RRR planar redundant parallel manipulator and investigates the effects of the joint position (namely the length ratios of two links) of each leg on the singularity-free cylindrical space. A previous method investigated the maximal singularity-free zone in a 3-dimensional (3-D) space within a given workspace. The method in this paper is improved by optimizing the maximal singularity-free zone in a 2-dimensional (2-D) plane while considering the whole workspace. This improvement can be helpful for reducing the searching time and for finding a larger singularity-free zone. Furthermore, the effect of the joint position of each leg on the maximal singularity-free zone is studied in this paper, which reveals a larger singularity-free zone than before. This result shows that changing the joint positions of one or two legs may be more practical than changing the joint positions of more legs. The approach in this paper can be used to analyze the maximal singularity-free zone of any other three-degree-of-freedom (3-DOF) planar parallel mechanisms and will be useful for the optimal design of redundant parallel manipulators.