Exploiting Jacobian Sparsity in a Large-scale Distillation Column.
Keywords
Abstract
The evaluation of a sparse Jacobian matrix arising from an industrial distillation column is described using automatic differentiation. In contrast to divided differencing, the derivatives are computed without truncation error. Rather than generating the full sparse Jacobian matrix, all its nonzero entries are computed in a compressed fashion, leading to a significant reduction in the computation time.