LARGE-SCALE MATHEMATICAL-PROGRAMMING SYSTEMS

This survey covers the state of the art of large scale mathematical programming systems (MPSs) for solving problems which can be modeled using linear programming or its extensions, such as mixed integer and some types of nonlinar programming. Such models occur frequently in chemical engineering and in other types of process, manufacturing and distribution applications. A typical MPS contains: (1) A Program control Language (PCL) which may be compiled and executed, or interpreted interactively. (2) Facilities for reading a model in standard external (MPS format) form, with several model options, such as alternative objective functions, right hand sides, etc. (3) The ability to select the appropriate model from among the above alternatives and 'SETUP' for the optimizer. (4) Provisions for model modification at several levels of detail (exernal, intermediate and internal). (5) A variety of LP algorithms and/or algorithmic options, which attempt to exploit the characteristics of particular models. (6) Extensive solution reporting capabilities (also in standard external form), with options to select from the flood of potential output. (7) External post-optimal analysis.