30
2 Simulation-Based Optimization
simulation-based optimization 16 , as well as challenges occurring when combining
both techniques. The section closes with a summary on optimization packages
interfaced with simulation software.
2.3.1 Coupling and Interaction Modes of Simulation
and Optimization
As described in section 2.1, the process of incrementally and interactivly varying
model parameters and comparing the system’s behavior against a starting situation
is referred to as simulation. Simulation supports the analysis of complex system
relationships, illustrates the effects of different environments on a system, enables the testing of options in a model before process changes are brought into
the real system, and allows the analysis of dynamic processes in case of time
adjustments [Bi+2004, p. 14]. The complexity of the optimization problems often
inhibits a solution by means of exact procedures. Therefore, heuristic optimization
methods are frequently applied, based on the combination of metaheuristics with
simulation models [VS2010, p. 93]. There is no common proceeding in industry for modeling, simulation and optimization coupling so far [VD2016a, p. 4].
Therefore, the coupling of simulation and optimization is not limited to one predominant method. It can be classified according to different criteria. Decisive
for the type of coupling are the interdependence of the simulation and optimization procedures used, the relation of subordinate or superordinate order as well
as the temporal sequence of the calculations [MK2011, p. 42; VD2016a, p. 3].
Generally, two types of hierarchical and two types of sequential coupling can be
distinguished. Respectively, this results in four coupling options for simulation
and optimization, which are shown in Figure 2.9. The sequential coupling requires the complete execution of each method. The results of the execution phase
of the first method form the input variables for the execution of the following
method. Besides the definition of the input variables for the second method, the
execution of phases is working independently [VD2016a, p. 3].
If the simulation results are used as starting values for the optimization, there
is usually a fully formulated optimization problem existing besides the simulation
model which includes unavailable parameters. These parameters are determined
during the simulation run and are made available for the optimization (Figure 2.9,
16 The terms “simulation-based optimization”, “simulation optimization”, and “optimization
via simulation” are not clearly differentiated in literature. This book consistently uses the term
“simulation-based optimization”, except for direct citations of reference literature.
2 Simulation-Based Optimization
simulation-based optimization 16 , as well as challenges occurring when combining
both techniques. The section closes with a summary on optimization packages
interfaced with simulation software.
2.3.1 Coupling and Interaction Modes of Simulation
and Optimization
As described in section 2.1, the process of incrementally and interactivly varying
model parameters and comparing the system’s behavior against a starting situation
is referred to as simulation. Simulation supports the analysis of complex system
relationships, illustrates the effects of different environments on a system, enables the testing of options in a model before process changes are brought into
the real system, and allows the analysis of dynamic processes in case of time
adjustments [Bi+2004, p. 14]. The complexity of the optimization problems often
inhibits a solution by means of exact procedures. Therefore, heuristic optimization
methods are frequently applied, based on the combination of metaheuristics with
simulation models [VS2010, p. 93]. There is no common proceeding in industry for modeling, simulation and optimization coupling so far [VD2016a, p. 4].
Therefore, the coupling of simulation and optimization is not limited to one predominant method. It can be classified according to different criteria. Decisive
for the type of coupling are the interdependence of the simulation and optimization procedures used, the relation of subordinate or superordinate order as well
as the temporal sequence of the calculations [MK2011, p. 42; VD2016a, p. 3].
Generally, two types of hierarchical and two types of sequential coupling can be
distinguished. Respectively, this results in four coupling options for simulation
and optimization, which are shown in Figure 2.9. The sequential coupling requires the complete execution of each method. The results of the execution phase
of the first method form the input variables for the execution of the following
method. Besides the definition of the input variables for the second method, the
execution of phases is working independently [VD2016a, p. 3].
If the simulation results are used as starting values for the optimization, there
is usually a fully formulated optimization problem existing besides the simulation
model which includes unavailable parameters. These parameters are determined
during the simulation run and are made available for the optimization (Figure 2.9,
16 The terms “simulation-based optimization”, “simulation optimization”, and “optimization
via simulation” are not clearly differentiated in literature. This book consistently uses the term
“simulation-based optimization”, except for direct citations of reference literature.