A heuristic procedure for rounding posynomial geometric programming solutions to discrete values

While geometric programming is a useful computational approach for engineering design problems, it is often the case that some design variables specifying sizes must be rounded to readily available standard sizes, or variables specifying the number of component parts must be rounded to integers. To assist in deciding whether to round these variables up or down to the discrete admissible values, we present a technique to estimate the respective penalties which would be incurred.