A control model for a three-level man-machine production system under rando
m disturbances is considered. The system comprises the factory level, sever
al sections and multiple production units. The factory is faced with manufa
cturing several different products with planned target amounts at the given
due date. Each unit can manufacture all kinds of products by utilizing dif
ferent types of renewable resources. In order to introduce control actions
all production units have to be inspected at control points to observe the
output of the products. The problem is to develop control models at each pr
oduction level, in order to minimize the factory's average total expenses u
nder a chance constraint. The optimal variables are the starting time of th
e manufacturing process and the capacities of all types of resources to be
hired throughout the planning horizon. Two models are developed at the fact
ory level: a search coordinate descent model for optimized variables and th
e reallocation model to redistribute both the resources and the planned tar
get among the sections. At the section level a hierarchical "tree" of optim
al reallocation models to redistribute both the resources and the target am
ounts among the subordinated elements at the adjacent lower level will be c
onsidered. At the unit level control actions boil down to determining contr
ol points and production speeds. All the developed optimal control models a
re imbedded in a three-level simulation model which provides representative
statistics in order to check the fitness of the developed factory model. A
numerical example based on simulation is presented. Practical applications
on a real man-machine plant are outlined. (C) 2001 IMACS. Published by Els
evier Science B.V. All rights reserved.