Optimal switchover times between two activities utilizing the same resource (Q4533452)

The paper formulates and analyzes extensions of a problem of Bellman in nonlinear programming. The problem involves determining the pattern of usage of one machine in two production activities, specifically, extracting an ore from two mines. Bellman's formulation of the problem assumed that fractional parts of the machine could be used in the production activities and the fraction could be changed instantaneously. Moreover, the operation of the machine was assumed to be costless. The authors consider this problem with the restriction that the equipment may be switched from one activity to the other at most once during a finite horizon. Denoting by \(u_i(t)\) the fraction of time spent on the \(i\)th activity, where \(i=1,2,x_i(t)\) the amount of ore remaining in the \(i\)th mine, \(x^0_i\) the initial stock of ore in the \(i\)th mine, \(p(t)\) the probability that the machine survives until time \(t\), \(q_i(t)\) the constant failure rate of the machine if used in the \(i\)th mine and \(r_i\) the constant production rate if the machine is used in the \(i\)th mine, the problem may be stated as the following nonlinear optimization problem: \[ \text{Max}\,f_T\equiv \int^T_0 p(s) \left[ \sum_ir_iu_i(s) x_i(s)\right] ds \] subject to \[ \begin{aligned} u_1(t) & =1\text{ for }0\leq t\leq t_0\text{ and }= 0\text{ for }t_0<t\leq T;\;u_2(t)=1-u_1(t)\\ \dot x_i(t) & =-r_iu_i(t)x_i(t) \text{ and }x_i(0)=x^0_i,\text{ for each }i=1,2\text{ and all }t\text{ in }[0,T]\\ \dot p(t) & =-p(t)\left [\sum_iq_i u_i(t)\right] \text{ for all }t \in[0,T] \text{ and }p(0)=1. \end{aligned} \] The authors show that there is an unique global maximum which can be found analytically under some specific parameter values. They go on to analyze an extension of the problem where a cost of operation is incorporated and the planning horizon is a decision variable.