Here we present a brief introduction to the methods used in MultiSimplex.
The Method Summary
Optimization of technical systems is the process of adjusting the control variables to find a level that achieved the best results (response). Usually a lot of conflicting responses to be optimized simultaneously. The lack of a systematic optimization approach is done by "trial and error" or by changing one control variable at a time while holding the rest constant. The method is not efficient in finding the right optimization.
In 1962 an efficient sequential optimization method called the basic simplex method presented by Spendley et al. This method will find a true optimization of the test response with less than non-systematic approaches or methods of one-variable-at-a-time. Simplex The method has been improved by active workers in the field, for what is called the modified simplex method (see eg, Nelder and Mead, 1965; Aberg and Gustavsson, 1982 and Betteridge et al, 1985). Two of the simplex method is the optimization algorithms used in software MultiSimplex.
The software also uses a modified matrix MultiSimplex first to start the optimization design. These D-optimal design has been shown to perform linear better than previous approaches in the normal experimental situation (Oberg, 1998).
Simplex algorithm can handle only one answer at a time, but usually there are many variables to optimize the response simultaneously. A method for forming a common response measure, the individual response variables, because it is needed.
Zadeh introduced the method as in 1965, with the concept of "fuzzy sets". Fuzzy set theory provides a flexible and efficient technique for handling the optimization criteria are different and contradictory (see Otto, 1988). Membership functions of fuzzy sets is the way to handle multiple responses in the software MultiSimplex.
The software MultiSimplex rests on a solid foundation combines two well-established and popular method. Together these two methods can simultaneously handle both control multiple variables and multiple response variables.
Please note:
Sequential simplex The method used in the software MultiSimplex should not be confused with the simplex method for linear programming (a method for solving linear programs with a progress from one extreme of the feasible polyhedron for a contiguous).
Philosophy Behind MultiSimplex
Reality is nonlinear and multivariate! MultiSimplex designed as a true means of multivariate nonlinear optimization. It seeks the optimal step by step, with a minimum of experiments.
The main principle behind MultiSimplex is to put you in charge of everything. MultiSimplex count from a purely mathematical considerations and do not have the intelligence itself. This is the experience and your skills as a professional job is important. In each step during the optimization, you can change both the optimization objectives and how the software operates. Optimization procedures that have been assigned will usually work fine, but there is always a reason to try a "flash of genius" (if there). In each step the software will also automatically check that you do not violate the basic principles for the algorithm, and warn you if you do it.
source
Categories: