The factor levels for components in simplex designs are subject to the following constraints:
The proportions of all the components must sum to one.
The amounts of each component must be at least the specified lower bound and no greater than the specified total amount.
If you want to further constrain the factor levels by adding upper bounds to components, or by adding linear constraints (i.e., bounds on combinations of components), use an extreme vertex design.
There are three types of simplex designs:
Simplex Lattice: With this design, the blends in the experiment are determined by the specified number of levels of each component (i.e., the degree of design + 1). Since it includes all the reasonable combinations of components, this design is useful when the number of components is not large, and a higher polynomial equation is needed for optimization.
Simplex Centroid: By default, this design includes single-component blends (vertices) and all centroids up to the dimension g, where g is the number of components. Users can specify the degree of design (i.e., the dimension of centroids). Since it usually has fewer test runs than a simplex lattice design, this design is useful when the number of components is larger, but a lower polynomial equation will suffice for optimization.
Simplex Axial: This design includes vertices, blends with an absent component (edge points), center points and interior points between the center point and vertices (axial points). It is mainly used for screening components.