Equipment replacement decisions due to technological obsolescence under uncertainty.

摘要

We consider the problem of deciding whether to keep a piece of equipment or to replace it with a more advanced technology in an environment of technological obsolescence. We develop four related models of this problem under various conditions of technological change. The feature that distinguishes these models from previous equipment replacement models is that we allow the technological change to be stochastic and non-homogeneous in time.; In the first model we assume that revenue functions associated with technologies are known and constant with time, but appearance times of future technologies are uncertain. We consider the case where only one alternative technology is currently available on the market but others may appear in the future. The decision maker has to decide whether to keep his existing equipment or to replace it with the alternative technology. We develop a procedure for computing the optimal decision that makes minimal use of forecasted data. We also show that keeping the technology becomes more attractive as the likelihood of a breakthrough increases.; In the second model we give a procedure to find the optimal decision when two alternative technologies are already available. We show that "hedging" against future technological change by purchasing the cheaper alternative, becomes more attractive if an advanced technology is more likely to appear in the immediate future.; In the third model we generalize the first model to the case where revenue functions are not constant in time.; In the last model we address the case of equipment that undergoes Markovian deterioration. We show that a control limit structure exists for the optimal policy, and use this result to develop an efficient algorithm to compute the optimal decision. We also give conditions under which it is optimal to keep the current equipment at least as long as in the situation when the advanced technology were to never appear.; The solution techniques in all the models use the forecast horizon approach. This enables them to make the minimal use of forecasted data and incorporate this data iteratively.