Abstract Control and performance optimization of modern manufacturing systems is characterized by an increasing degree of complexity that accompanies automation and flexibility. Because the functional relationship between typical performance measures and parameters of interest generally cannot be determined, analytic (queueing) models are only of limited help. Detailed analysis requires extensive simulation, an approach limited by its high, sometimes prohibitive, computational cost and infeasible when realtime decisions must be made. An alternative is offered by Perturbation Analvsis, a recently developed technique which efficiently estimates performance gradients with respect to parameters from a single interval of observing real on-line data. This technique can be used to exploit the data collection capabilities of computer integrated manufacturing systems and automatically provide valuable "sensitivity" information. It essentially creates a decision aid whose uniqueness lies in supplying on-line answers to "what if..." questions. An application of perturbation analysis is discussed, where emphasis is placed on optimizing control strategies for dynamic (i.e. state dependent) scheduling and routing of individual parts in a computer integrated manufacturing system.
[1]
J. Buzacott.
"Optimal" operating rules for automated manufacturing systems
,
1980
.
[2]
Xi-Ren Cao.
Convergence of parameter sensitivity estimates in a stochastic experiment
,
1984
.
[3]
J. A. Buzacott,et al.
Models for Understanding Flexible Manufacturing Systems
,
1980
.
[4]
Christos G. Cassandras,et al.
A new approach to the analysis of discrete event dynamic systems
,
1983,
Autom..
[5]
R. Suri.
Implementation of sensitivity calculations on a monte carlo experiment
,
1983
.
[6]
C. Cassandras,et al.
An event domain formalism for sample path perturbation analysis of discrete event dynamic systems
,
1985
.