A special purpose program for generalized geometric programming (GP) has been developed, with comparative studies of various existing methods. It is based on a modified cutting plane algorithm with double condensation and linearization. The convergence to a local solution is assured if an initial feasible solution is available to initialize the algorithm. A routine finding the initial feasible point is also included.
Several test problems are taken to check the comparative performance of the method with two general purpose optimization programs, IDESIGN and LINRM. The first of these is based on a gradient projection type method while the latter on a sequential quadratic programming approach by Pshenichny. In general the developed program has shown better performance than the other, which is expected because it is specialized only for GP problems. The efficiency of the 3 programs compared, however, is shown highly dependent on the characteristics of the given problems, such as the number of constraints and the number of monomial terms. Therefore, the program remained to be checked against other GP solvers in the literature.
