Engineering Optimization

32Identifying the Models, OF, DV, Convergence Criteria, and Constraints

32.1 Introduction

In routine applications, the legacy of solving the optimization and using the results will have given customers and implementers a good understanding of the context of the application and the appropriate choices for the models OF, DV, convergence criteria, thresholds, and constraints. However, for new applications, the implementer must design these procedures and must make the appropriate choices. The question is “How?” The answer is “trial and error.” Try something, and when you have the result, consider all aspects of the outcome. Then evaluate whether all aspects are satisfactory. If not satisfactory, improve the procedure.

That answer, the solution, and the method of getting optimization right is qualitative. Accordingly, this section has no mathematical relations. It is also fuzzy, providing a general direction, but not a specific rule to execute.

Although not rigorous, or sophisticated, or quantitative, the following material is important to application success. Perhaps this section is more important than being an expert in the optimization algorithm.

The user of a new application will want to know how to define the best way to create the model and to choose the DVs and the convergence criterion. I still revert to trial and error. It is a design problem. Sketch a possible definition of the application, get the solution, and explore it. Decide what are the desirables and undesirables ...