This paper aims at contributing to the sim2real challenge in soft robotics. We present a method for automatic finite element model calibration based on real data using quadratic programming optimisation. The method is generic and evaluated in this study to fit mechanical parameters from anisotropic materials. We show that we are able to optimise the mechanical properties of a given structure along its shape to achieve a given configuration goal. We show the methods interest for calibration by taking reference points from a real world robot and use them in our optimisation process as goals the simulation has to match. Our process will minimise the errors introduced by manufacturing, imperfect models or even mechanical fatigue/plasticity.