Sie sind auf Seite 1von 2


The optimization problem in which the performance measure is a function of one variable is the most elementary type of optimization problem. Yet, it isof central importance to optimization theory and practice, not only because it is a type of problem that the engineer commonly encounters in practice, but also because single-variable optimization often arises as a sub problem within the iterative procedures for solving multivariable optimization problems .Because of the central importance of the single-variable problem, it is not surprising that a large number of algorithms have been devised for its solution. Basically, these methods can be categorized according to the nature of the function (x) and the variable x involved as well as the type of information that is available about (x). 2.1 PROPERTIES OF SINGLE-VARIABLE FUNCTIONS A function (x), in the most elementary sense, is a rule that assigns to every choice of x a unique value y _ (x). In this case, x is called the independent variable and y is called the dependent variable. Mathematically, consider a set S _ R, where R is the set of all real numbers. We can define a correspondence or a transformation that assigns a single numerical value to every point x _ S. Such a correspondence is called a scalar function defined on the set S. When the set S _ R, we have an unconstrained function of one variable. When S is a proper subset of R, we have a function defined on a constrained region. For example,

is an unconstrained function, while

is a constrained function. In engineering optimization, we call the objective function and S the feasible region, the constraint set, or the domain of interest of x. Most physical processes can be modeled or described by continuous functions, that is, functions that are continuous at every point x in their domain. However, it is not at all unusual in engineering applications to encounter discontinuous functions. For instance, if we construct a function that measures the cost of heating per Btu at different delivery temperatures, we quite naturally arrive at the broken curve shown in Figure

2.1. The cost is a discontinuous function of the delivery temperature; however, the temperature itself can assume all values over the range of, say, 2003000_F. Of course, it is not necessary that the domain of the independent variable x assume all real values over the range in question. It is quite possible that the variable could assume only discrete values. For instance, if we construct a function that gives the cost of commercial pipe per foot of length at different pipe diameters, we quite naturally arrive at the sequence of discrete points shown in Figure 2.2. Pipe is made only in a finite number of sizes. Note: It is important to remember the following properties of continuous functions: 1. A sum or product of a continuous function is continuous. 2. The ratio of two continuous functions is continuous at all points where the denominator does not vanish. Clearly, depending upon whether the function to be optimized is continuous or discontinuous or depending upon the nature of its domain, different methods will have to be employed to conduct the search for the optimum. A method that is effective for continuous functions will probably not be effective for discontinuous functions, although the converse may be possible. In addition to these properties, functions can also be classified according to their shape or topology within the range of interest.