An on-line control model for a two-level man-machine production system is c
onsidered. The system comprises a section with several production units and
manufactures different products with planned target amounts by utilizing s
everal types of non-consumable resources. Each unit may operate at several
possible speeds which are subject to random disturbances. The unit's output
can be measured only at pre-given control points. At the unit level, contr
ol actions result in determining control points and production speeds. At t
he section level, production control centres on reallocating resources and
target amounts among the units. The performance of the two-level control al
gorithm depends mainly on solving the reallocation problem, as the latter t
akes more computational time than other control actions. Thus, speeding up
the reallocation algorithm results in enhancing the control model's perform
ance. We suggest a newly developed reallocation model, together with a corr
esponding reallocation algorithm. Two objectives are embedded in the model:
ii! to maximize the delivery performance of the slowest unit so that it ca
n meet its target on time, on condition that only maximal speeds are used;
and (ii) to maximize the delivery performance of the latest product so that
it is manufactured on time under the same conditions.
A subsidiary iterative switching algorithm is developed by following both o
bjectives. A generalized reallocation algorithm, which is a combination of
the iterative algorithm and the coordinate descent method, is suggested. Su
ch a reallocation algorithm significantly enhances the control model's perf
ormance and enables large-size man-machine production systems to be control
led under random disturbances. Extensive experimentation has been carried o
ut to verify the efficiency of the reallocation algorithm. The algorithm ha
s been used successfully in a real industrial plant.