An MCDM framework for dynamic systems

A multi-criteria formulation for continuous and discrete-time dynamic systems is presented. The fundamental problem in the design of dynamic systems is the trade-off between response speed (e.g., the time to reach final value and the raise time) and response smoothness (e.g., the overshoot, the undershoot, and the transient dampening). Separate optimisation of criteria is impossible, thus the problem is inherently multiobjective. In a general dynamic system, this is accomplished by adjusting a number of technical parameters in accordance with some ad hoc practice. Previous multi-criteria approaches have been modelled as weighted sums of criteria, with shortcomings in terms of sensitivity analysis and preference articulation. The proposed framework enables the decision maker to design a most preferred system, with full knowledge of local trade-off ratios in terms of chosen criteria. Combining analytical techniques with simulation, the formulation makes the optimisation process transparent to the decision maker, working entirely in decision space. The framework is demonstrated on a dynamic production-inventory model.