In recent years, many studies have been done on how to learn from nature to make more efficient structures. Studies show that a clever arrangement of raw materials inside a composite could have a huge positive effect on mechanical properties. Researchers can use patterns inspired from nature to improve the mechanical response of composites made of synthetic materials. One of the unique sources for such natural inspiration is human cortical bone. This bone has a good response against all kinds of loads, especially torsional loads. Actually, studies prove that the geometric shape and position of soft material in the cortical bone which is called osteon have a beneficial effect on mechanical properties, including torsional properties. The aim of this project is to inspire the structure of human cortical bone to make torsional load-bearing composites. In the first stage, models are designed according to the bone structure and their shear modulus are calculated. Then, the structures are eva‎luated and the best ones are identified. All models consist of two hard and soft materials. The studied shapes of soft material at this stage are circular in different sizes, elliptical in concentric and radial orientations with different aspect ratios, helical and layered. In order to validate the obtained numerical results, two homogeneous soft and hard structures, as well as one concentric and one radial composites with the same aspect ratio, were considered and printed by a machine at the Technical University of Delft. Finally, the printed models are tested under torsional loads. According to the results, the error of the numerical calculations is estimated to be less than 6%, which is acceptable. The results verified that if the soft material is oriented along with the direction of torsion, e.g. layered or concentric elliptical structures, the shear modulus will increase dramatically. Furthermore, the analysis proves that as the number of layers increases and the soft material gets more homogeneously distributed, the shear modulus will increase as well. In order to search for more composites with highly good mechanical properties, some artificial intelligent methods are invoked and the final results are compared with results obtained from the first stage. In contrast to the first stage that the human cortical bone was inspired to design composites, here the algorithm was not restricted to no initial pattern and indeed started with a random distribution of materials. Genetic Algorithm Optimization was one of these artificial intelligent methods. The Python code of the Genetic Algorithm tries to generate and select the structures that have higher shear modulus. In the end, the most optimum structure has a shear modulus equal to 185.10 MPa. However, there might be some structures that are missed by the Genetic Algorithm. For this reason, a machine learning algorithm (Neural Network) was designed and deployed. For training the Neural Network algorithm, the data generated in each step of the Genetic Algorithm is recycled and fed into the Neural Network which includes nearly nine thousand different representations along with their shear modulus. The results show that the Neural Network has found composites with higher shear moduluses than the Genetic Algorithm. The results also confirm that the Genetic Algorithm and Neural Network are able to predict structures with nearly as good as the structures inspired by nature

احسان فروزمهر

محمد ميرزاعلي

مهدي كاروان، محمود فرزين