Tianju Xue

In JAX-FEM, you may have variable parameters up to each quadrature. Is this function helpful for your case? https://github.com/tianjuxue/jax-am/blob/55115e49b4354ea220296f5f5d6f906854e39a7c/jax_am/fem/core.py#L712-L731

What benchmark FEM software did you use? I think this is highly likely due to difference choices of linear solver as well as associated tolerance setting. For example, on my...

For homogeneous property distribution your final linear system will be better conditioned, so the solving will be easier and it matches for 5 digits for different solvers. But for inhomogeneous...

Please allow us some time to implement this feature. @SNMS95 previously worked on this. There is an experimental version that you can play with https://github.com/tianjuxue/jax-am/blob/814384eb522b839d1537a168e2cc07b34553c012/jax_am/fem/autodiff_utils.py#L38 Should be able to provide...

JAX-FEM is well tested and compared with other software, e.g., FEniCS for stiffness matrix value, particularly for linear elastic problems. When compared with Abaqus, we indeed found a difference in...

Hi Jørgen, Thanks for your comments. In the paper, when we did the comparison, JAX-FEM used BICGSTAB without preconditioner to be relatively fair. We will mention the solver choice in...

Thanks for your interest. We purposely let users handle `atol` (rather than define it in a systematic way) since this is really problem dependent, and different users may have their...

I have just implemented a function to print out boundary info. Please use this to check if desired boundary conditions are imposed successfully. In you application code, `problem.print_BC_info()` will print...

I think this is a problem with PETSc not converging. If you look at the logging information, the solver did not converge at all. So gradient computation will definitely be...

JAX itself is not meant to be used on Windows. Same holds true for JAX-FEM. Please consider a linux system like Ubuntu or Mac OS will also be fine.