Abstract for:Rethinking of Continuous Models with Approximate Discrete Calculations in Simulation Software

Most system dynamics software implement a common calculation process based on the Euler method

of numerical integration. This calculation is a continuous models’ discrete approximation and has an

assumption that the definition of derivatives has no change during one time-unit. When a modeler

sets a time-unit too long which reflects plural decisions for a flow variable during one time-unit, its

simulation result differs with the analytical result of identical numerical models. The discretization

of the calculation process brings about this difference. In addition, discrete approximation does not

mean that simulation software can more “easily” deal with discrete calculation. A typical issue

appears when a modeler uses impulse functions whose implementations are different among system

dynamics software. The same meaning definitions can cause different results. This paper shows that

system dynamics modelers should pay significant attention to discretization and for discrete

function’s implementation in continuous system dynamics models and simulation software.