this is for holding javascript data
Richard Otis edited Numerical_Approach.md
over 8 years ago
Commit id: e7d308b22f951ef5f3b3731e72be971ef0e7b052
deletions | additions
diff --git a/Numerical_Approach.md b/Numerical_Approach.md
index a3b36ca..53bb0a6 100644
--- a/Numerical_Approach.md
+++ b/Numerical_Approach.md
...
\[ F(T, P, f^{i}, y^i_{s, j}) = \sum_i f^{i}G^{i}_{m}(T, P, y^i_{s, j})\]
\[ L(T, P, f^{i}, y^i_{s, j}, \lambda_k) = F(T, P, f^{i}, y^i_{s, j}) - \sum_k \lambda_k c_k(T, P, f^{i}, y^i_{s, j})\]
\[ \left[ \begin{array}{ccc}
W_k & -A^T_k
\\\ \\
A_k & 0 \end{array} \right]\left[\begin{array}{ccc}
p_k
\\\ \\
p_\lambda \end{array} \right] =
\left[\begin{array}{ccc}
-\nabla F_k + A^T_k \lambda_k
\\\ \\
-c_k \end{array} \right]\]
\(p\) is the Newton step direction for the next iteration. \(W_k\) is the Hessian matrix of \(L\) with respect to all degrees of freedom except \(\lambda_k\). \(A_k\) is the Jacobian matrix of the constraints, \([\nabla c_1, \nabla c_2, ..., \nabla c_m]^T\).
The key advantage of this approach is you can optimize the degrees of freedom of all phases, including phase fractions and chemical potentials, simultaneously.