8th September 2010 1:14
|
System Dynamics – Understanding and Improving Complex Systems
|
|
  |
|
Problem Statement
How to undestand and decribe complex systems. |
Improvement Approach
System dynamics modelling, a concept first introduced by MIT professor J. W. Forrester in the 1950s. There are several commercial tools with easy-to-use graphical user interfaces available in the field of system dynamics. |
Benefits
Various benefits |
|
|
|
In the early years, the lack of computing power limited many approaches but the recent advances in computers and simulation technology as well as in the large scale and nonlinear system theory has enabled the development of system dynamics theory and a large number of successful practical applications. The development of easy-to-use software packages with graphical user interfaces and with complex mathematics hidden to the background has further widened the system dynamics community. |
|
The models are used for system optimization, prediction and control. The optimization of different policies and structures of an organisation can be done when the system contains suboptimal elements, which can be changed at political will. Typical example is a minimisation of labour supply cycle oscillations (hiring-layoffs) by the introduction of a flexible workweek.
Prediction, on the other hand, is useful when one organisation’s actions cannot affect the outcome but the organisation can benefit from reacting quickly to changing circumstances (such as predicting the raw material market price and reacting to it). The pulp and paper prices contain three different oscillations: production, inventory and prices oscillate with 3 – 5 year cycle and capacity oscillates with a 12 – 20 year cycle. Overall economy oscillates with a 60 year cycle time (the long wave of economy). The model captures the reasons for oscillation, adjusts the historical data to the model and gives best predictions on, e.g., 1, 5 and 10 year scale. |
The model can, for instance, detect a turning point in pulp prices half a year ahead, and this prediction can be used for aiding decision making.
Model-based control design concentrates on finding the best strategy on managing the system under various conditions. Let us consider a case in which the industry demand of engineers on a certain field increases rapidly. However, it takes 4 – 6 years to graduate and a conventional strategy for increasing the number of graduates quickly causes significant oscillations in the system (not to mention lower graduate quality, motivational problems, etc.). A Smith predictor type of control design combined with feedforward compensators would perform much better under the same conditions. |
|
Typical results of this analysis might be something in line of “There is a danger of a death spiral if the subcontractor supplier chain is not secured” or “The detected oscillation in the stocks and orders is probably due to the time delay on the negative feedback loop of order processing”. |
|
|
