Anisotropy Studies, Neural Network-based Materials

As the biomaterial to replace and renovate the hard tissue of human beings, the bone renovation materials are widely used. When human body suffer from suddenness, the material character of skeleton and the mechanical properties of bone tissue must be understood if the appropriate renovation materials could be found. This thesis makes a new method to investigate and develop the mechanical properties from biomechanical opinion. The impress modulus which can be acquire from the analysis of nanoidentation load-displacement data could be input to the neural networks and the nine material elastic parameters can be output.The mechanical properties of bone have been extensively studied at the macrostructural and microstructural level. Conventional mechanical testing has a lower limit of a few hundreds of microns at best. Microidentation measures a dimension of a few tenths to hundreds of microns, while nanoidentation can study on a 1um or smaller length scale. Since many important microstructural components of bone have dimensions of only a few microns or less, the nanoidentation can be used to investigate the mechanical properties of ostional, interstitial, and trabecular lamellar bone at the largely unexplored micro and submicron level. As an example, the nanoidentation method has been use to examine variations in the individual lamellar properties within osteons.Nanoidentation, developed over the last 10 years, is now used widely in the materials science community for probing the mechanical properties of thin films, small volumes, and small microstructural features. From analysis of nanoidentation load-displacement data, it is possible to derive values of the elastic modulus, hardness, and properties associated within time-dependent deformation such as viscoelasticity . At present ,microstructurally based theories of bone behavior are difficult to test due to a lack of microscopic mechanical property data. But nanoidentation techniques can be used to acquire parameters on microstructral level, such result would fill a gap in the present knowledge.So far, the material parameters got from the nanoidentation load-displacement data, are obtained by Oliver-Pharr. But the precondition of this method is that the material must be the elastic isotropy. For the elastic anisotropy, because of the difference in the each direction of mechanical properties, actually ,the result of the nanoidentation load-displacement data is the synthesis of the mechanical properties in every direction. How to analyze this kind of synthesis or obtain the mechanical properties in each direction from the synthesis is not uncertain in theory so far. We will use the superority of the neural networks to get the properties of the elastic anisotropy from nanoidentation load-displacement data. Whether in shape or in mechanical character, bone is elastic anisotropy.In order to acquire the training sample, 12 impress modulus must be induced from 9 actual elastic parameters. With the help of the University of Memphis Biomedical Engineering skeleton Tissue Lab, a set of elastic parameters are provided. 12 impress modulus are induced from a Fortran program, thereby we can get the training samples to train the neural networks. Once the neural networks are accomplished, the impress modulus calculated from nanoidentation load-displacement data by Oliver-Pharr method can be input to the neural networks to get the 9 elastic parameters.