78 European Journal of Operational Research 40 (1989) 78-84
North-Holland
Theory and Methodology
Modeling learning effects
linear programming
via successive
Terry P. HARRISON
Management Science Department, The Pennsylvania State University, University Park, PA 16802, USA
J. Edward KETZ
Department of Accounting and MIS, The Pennsylvania State University, University Park, PA 16802, USA
Abstract: A learning effect occurs when the amount of labor required per unit of production decreases as
cumulative production increases. Learning effects occur in many situations, and this effect can be
especially significant in the startup of a new process. However, from a modeling viewpoint, the inclusion of
learning effects can result in a problem that is considerably more difficult to solve than when these effects
are ignored. A number of mathematical programming solution methods have been proposed for modeling
a learning effect. Unfortunately, these methods frequently require the implementation of sophisticated
algorithms. In this paper we develop an alternate solution strategy for modeling a learning effect based
upon the use of Successive Linear Programming (SLP). This approach is particularly attractive in that it
can be easily implemented, and only requires access to a linear programming package.
Keywords: Learning, mathematical programming, nonlinear programming
Introduction
Suppose that an enterprise can manufacture a number of products. One goal of management is to
maximize the contribution margin (revenues less variable costs), however the production process faces a
variety of resource constraints. When all of the underlying mathematical relationships are linear, then
linear programming is one of several techniques that may be used to address this optimization problem.
However, one of the assumptions of linear programming is that the coefficients describing the
technology of production are constant. On the other hand, if learning effects exist this assumption is
invalid. Learning effects imply that the technological coefficients are declining and that the contribution
margins per unit are increasing as cumulative output increases. Therefore, to obtain a better approxima-
tion to reality a mathematical programming model should incorporate these effects if learning is present.
The direct labour hours needed to produce a unit of output may decline as more units are produced,
which is precisely the learning effect. Learning processes are well-documented and occur with labor
efficiency, production technology, product redesign, product standardization, and scale effects. Two
comprehensive reviews are Nanda and Alder (1977) and Yelle (1979).
Received October 1987; revised April 1988
0377-2217/89/$3.50 © 1989, Elsevier Science Publishers B.V. (North-Holland)