Needles are tools widely used in minimally invasive surgery. During such procedures the localization of the needle and its tip is a challenging situation because of the needle deformations due to its interactions with tissues. To tackle this problem, instrumented needles with sensors have been currently developed to allow needle reconstruction and tip localization. In conventional surgery this difficulty is overcome by medical imaging. The interest in using an instrumented needle resides in the possible dispense of medical imaging. This papers develops new methods to reconstruct needles in three dimensions and to find the locations of sensors which minimizes the error of reconstruction of the needle. A notable feature of our method is that input data are based on real needle data, that should assure a better representation of reality. Reconstructions simulated with 22 gauge 200 mm long needles show that the localization of the needle tip is more accurate by 18% to 52% with optimal sensors positions compared to equidistant sensors positions.