It may be appropriate to abort a simulation if the value of the objective function (as accumulated to the current simulation time) exceeds the minimum objective function value, indicating that the run will not lead to a reduction in the objective function. (Note that aborting such simulations is not appropriate when evaluating the Jacobian matrix, or when using a global minimization algorithm.)
This option may reduce overall CPU time. It is the default option when evaluating trial steps in the Levenberg-Marquardt algorithm (the default can be overwritten by using command >>> NO ABORT).
It may be specifically effective if >>> GRID SEARCH is solely used to find the minimum (rather than mapping the entire response surface), or if >>> MONTE CARLO and >>> LATIN HYPERCUBE are used not for uncertainty propagation analysis, but as a random search algorithm.
>>> allow runs to ABORT as soon as OF>OFmin