Topology optimization of structure seeks to achieve the best material distribution in the Pre-determined design domain. In this paper, the effect of design parameters contains length scale parameter and evolutionary volume ratio in improved bi-directional evolutionary structural optimization method with soft kill approach is discussed. The main aim of this method is searching for the stiffest structure with a given volume of material using finite element method. At each iteration of finite element analysis, sensitivity number is calculated for each individual element in design domain and then converted to the nodal sensitivity number. With Filter Scheme and using length scale, an improved sensitivity numbers is defined. This number is used as a criterion for rating each element in design domain and determining the addition and elimination (remove) of elements. To increase the convergence of the optimization process, the accuracy of the new elemental sensitivity numbers is improved by considering the sensitivity history. This method is convergent and mesh-independent and there are no checkerboard pattern and local solutions in optimal topologies. Using three design samples, a cantilever and classical beam and Michell type structure, affecting factors will be discussed on the final design of the structure. Change of length scale parameter makes various schemes in final structures in which with increasing this parameter, more iteration is needed for convergent solution. Reducing evolutionary volume ratio forms different and even asymmetric topologies. Better final topologies are obtained with higher evolutionary volume ratios.