Developing an expert system to diagnose computer program outputs

Reported by the author is an expert system which enhances the capability of a nonlinear optimization program. Included are discussions on the knowledge acquisition aspect and the advantageous use of neural networks to accomplish knowledge acquisition. The expert system diagnoses the output of the NPSOL optimization package and seeks to find errors in the user-supplied inputs. These errors result from poor modeling formulation of the problem and poor parameter settings when running the program. A prototype system is installed on a Unix machine