Definitions of some quantities inside newtonsolver

[zhangmuElias]

Definitions of some quantities inside newtonsolver

[keiyamamo]

[bc.apply(b, dvp_["n"].vector()) for bc in bcs]

This is to make sure that both b and dvp_["n"].vector() satisfies the boundary condition before we solve the linear system. After solving the linear system, we add the residual to the dvp_["n"]

dvp_["n"].vector().axpy(lmbda, dvp_res.vector())

However, the residual does not necessary satisfy the boundary condition and thus we need to make sure dvp_["n"] satisfies the boundary condition after that by calling

[bc.apply(dvp_["n"].vector()) for bc in bcs]

[zhangmuElias]

Ok! That's very clear, thank you.

[keiyamamo]

Hi @zhangmuElias

testfunction psi phi gamma mean virtual δ(Δd) δ(Δv) δ(Δp)

This part, I do not agree with you. They are test functions associated with the d,v,p itself, not the increments.

For each Newton iteration, we are solving

$$J_{F}(x_k)s = -F(x_{k})$$ $$x_{k+1} = x_{k} + s$$

where $J_{F}$ is the Jacobian (A in turtleFSI), $x_{k}$ is the solution we are looking for (dvp_["n"] in turtle), and s is a vector that represents each "Newton step" and the one we are actually solving for (dvp_res in turtle).

[zhangmuElias]

Dear @keiyamamo

Thank you! I also don't understand why in the newton solver, we need to apply so many times the boundary conditions.
They must have different kinds of functions?

Best regards
ZHANG