This paper looks at a Multi-Period Renewal equipment problem (MPR). It is inspired by a specific real-life situation where a set of hardware items is to be managed and their replacement dates determined, given a budget over a time horizon comprising a set of periods. The particular characteristic of this problem is the possibility of carrying forward any unused budget from one period to the next, which corresponds to the multi-periodicity aspect in the model. We begin with the industrial context and deduce the corresponding knapsack model that is the subject of this paper. Links to certain variants of the knapsack problem are next examined. We provide a study of complexity of the problem, for some of its special cases, and for its continuous relaxation. In particular, it is established that its continuous relaxation and a special case can be solved in (strongly) polynomial time, that three other special cases can be solved in pseudo-polynomial time, while the problem itself is strongly NP-hard when the number of periods is unbounded. Next, two heuristics are proposed for solving the MPR problem. Experimental results and comparisons with the Martello&Toth and Dantzig heuristics, adapted to our problem, are provided.
