IMAS. Integrated modeling and analysis system for the solution of optimal control problems

An Integrated Modeling and Analysis System (IMAS) for the solution of control problems is presented. IMAS provides algorithms for inverse modeling (data assimilation, parameter estimation, optimization), for sensitivity Studies, and the estimation of error covariances. IMAS is written in Fortran 90 and makes use of the new features of the language (abstract data types, modules). It provides the tangent-linear and adjoint model to a given Fortran program semi-automatically by an operator overloading technique. A limited-memory quasi-Newton algorithm for solving bound constrained optimization problems is included. Means to explore the (approximated) Hessian or inverse Hessian (to retrieve the diagonal or a column in case of large matrices) are provided. Thus confidence limits, variances, and covariances of the retrieved parameters can be obtained. For easy applicability an interactive user interface integrates the nonlinear, tangent-linear, and adjoint model, as well as the optimization algorithm and test facilities into one system. The theoretical and technical foundations of IMAS are outlined. The advantages of the system are demonstrated by an example program. (C) 1997 Elsevier Science B.V.