Article ID: | iaor1998616 |
Country: | Netherlands |
Volume: | 27 |
Issue: | 3 |
Start Page Number: | 265 |
End Page Number: | 278 |
Publication Date: | Dec 1996 |
Journal: | Engineering Optimization |
Authors: | Mehrez Abraham, Goldstein Zvi |
Keywords: | manufacturing industries, programming: markov decision, innovation, maintenance, repair & replacement, programming: dynamic |
A replacement problem is presented in which two technologies are involved. One technology is immediately available whereas the second, considered a technological breakthrough, is expected at some uncertain time in the future. The technological environment is dynamic. Therefore, special attention is given to the availability and reliability of data. Decision horizon stopping rules are suggested when long time horizons are considered. The algorithm is different from currently known standard procedures. It does not rely on the monotonicity properties of the optimal solution that occur when the horizon is extended, nor on action elimination considerations based on non-homogeneous Markov decision analysis. Rather, a special procedure is suggested in which an auxiliary problem binds the solution of the original problem. Interrelationships between the two problems lead to the desired horizon results. The procedure detects much shorter forecast horizons than the Bess and Sethi (1988) procedure without increasing the computational effort needed.