I'm working on a project involving FEM modeling of fractured femurs to assess the performance of fracture fixation devices. Bone is a highly complex material, with bulk elastic properties that vary depending primarily on the local density of the bone. The particular problem in my research is the development of a patient specific model by converting data from CT scans to a map of the material properties in the bone. This is a common procedure, but with most software, it renders it impossible to alter the actual geometry of the bone, which I will need to do in order to study the fracture behaviour.
Generating the geometry of the bone from CT scan is quite simple. So is pre-processing the scan images in slices and using voxel intensities to determine bone density and thus elastic behaviour, stored as a pointcloud (i.e. x-y-z spatial data paired with material property, defined in a grid/array according to scan slices). The problem that I am having is mapping these properties onto the geometry/mesh in a way that will allow me to edit the geometry of the bone (i.e. introduce a fracture, screws, and locking plate). Is there a way of defining the variation in the material properties in a way that will also allow me to edit the bone geometry?
I have tried to define the material properties as "temperature dependant" and then use temperature as a surrogate field by which to apply the point-cloud map of properties. This approach has apparently been used successfully before in research, but details of the implementation weren't given. In small scale tests this is successful, but it does not work on the scale at which I need. ABAQUS seems to give up after the number of points in the point-cloud exceeds 10,000. Using 4mm voxel size, this requires at least 20,000 points defined across the volume of bone.
I have also considered using the UTEMP subroutine, but don't know the Fortran coding to implement this. If this would be a workable solution to my problem, I will happily learn and use this approach.
This answer will not fix your problem, but will hopefully point you in the right direction.
Generally, there are two methods for creating an FEA model for a bone:
The first method entails using a CT image (bigger pixels) and then calculating material properties (i.e. Young's modulus) from the pixel intensity. We can then create an element for each pixel and the elements will have different materials. This can be done using BoneMat or its FOSS alternative py_bonemat_abaqus. It allows for fewer elements but the model will be homogeneous.
The second method requires using a micro-CT image. At that scale, all of the details will be captured and you will have a homogeneous model with all of the intricacies of the cortical and cancellous bone reflected in the geometry. The good news is that all elements will have the same properties as the cortical bone, which is available in the literature. This method usually uses a voxel-based mesh, but You may be able to use software such as mimics. I have developed a free and open-source software named VCAMS as part of my PhD thesis, but it will not be publicly available for a few weeks. You can send me an email from my website for more information.
The main problem for you will be incorporating screws and fixtures in the model. A voxel mesh may not be compatible with that due to imperfect contact, but you won't know unless you test it. Furthermore, these methods create an orphan mesh, which cannot be edited, only deleted via boolean operations.