Online Content

1. Hagenson,Norodd, "System Dynamics Combined with Monte Carlo Simulation"

Type:
Document
Date Created:
1990
Collection:
System Dynamic Society Records
Collecting Area:
University Archives
Collection ID:
ua435
Parent Record(s):
23d738ba88f8333bc39725f9cb5bd0b8, e3ebd4855fa658877b17f518e29a6ac9, and 386b7d5fcd17a644199db727913a7a7a
Thumbnail Path Ss:
/downloads/tt44q560j?file=thumbnail
System Modified Dtsi:
2020-01-02T18:44:30Z
Description:
The author suggests that a combination of System Dynamics (SD) thinking combined with Monte Carlo simulation models can yield new insight and be a useful tool. Systems with feedback loops often contain elements of uncertainty or randomness which can be modeled by Monte Carlo methods. On the other hand, feedback loop analysis could certainly benefit Mont Carlo simulation models. Studying single runs of SD models may yield considerable insight. But when a parameter is set to a constant or average value, variance is lost. Variance plays an important role in portraying any risk involved in a system.These points will be illustrated by an example from an analysis performed at NDRE where SD thinking applied to a Monte Carlo model was the key to solving an important question. The example concerns dimensioning Airfield Damage Repair (ADR) capacity on Norwegian airbases subject to hostile attacks. One key question was: How long time must the runway be open per day in order to obtain acceptable operating conditions for air defense fighter aircraft? Does there exist some minimum threshold?The main feedback loops concern damage on the runway and attribution between attacking aircraft, ground based air defense and defending air defense aircraft (depending on open runways). The elements of randomness concern the damage inflicted on the runway, and the repair time.It is shown that under certain conditions (too low repair capacity) there is a risk of defending aircraft either being pinned in or wining the battle. The feedback loop between defending aircraft and the runway state plays a key role along with the randomness in the early damage. The statistical distribution of the fraction of day open may over time develop into having one peak close to 0 (closed), one peak close to 1 (open), and little in between. The average value is merely a weighted average between two extremes.On the other hand, with sufficient repair capacity, the risk of being pinned in was eliminated. The effects were easily understood when thinking in terms of feedback loops, but the element of randomness was essential in order to recognize the threshold when the risk of being pinned in occurred.The author believes that a similar combination of techniques could benefit traditional SD models, too.