168
6 Experimental Validation of the Methodology
6.3
Optimization Experiments
To increase the energy efficiency in the Bosch production scenario, the previously defined optimization experiments were used. As in the fictional case study,
the reduction of the total energy consumption of the production lines has the
highest priority. For this reason, two separate optimization experiments were performed. Firstly, the optimizer to minimize the total energy consumption of the
two production lines was run, followed by the consumption peak optimization
experiment.
6.3.1 Optimization Parameter Adaptations
As already briefly indicated, there is a very strong dependence of the machines among one another in the production lines considered. No safety stocks are
provided between the machines to decouple the processing steps, so failures of
individual machines in the line cannot be intercepted by material buffers but lead
to a shutdown of the entire line with a delay of only one to two machine cycles.
Due to the strong coupling of the machines, it does not make sense to optimize
and evaluate the idle and standby optimizer parameters per machine. For a holistic
optimization approach, the standby and idle optimizer parameters are introduced
per line, with one exception: for a number of reasons, the washing machines
are technically unable to change machine states as quickly as the remaining four
machines of the respective line.
In addition, the parts are buffered in front of the washing machines until the
required lot size is reached and the cleaning process can be started. For this reason, L1_wash and L2_wash have their own idle optimizer parameters to take into
account the technical requirements of the real washing machines (Figure 6.8).
In total, the number of optimization parameters in the optimization experiment is reduced from 30 to eight, which has a positive effect on the experiment
runtime but does not alleviate the quality of results. Six of the eight optimization parameters were used for the total consumption optimization and the optimal
configuration of the other two parameters is determined in the peak optimization
runs.
6 Experimental Validation of the Methodology
6.3
Optimization Experiments
To increase the energy efficiency in the Bosch production scenario, the previously defined optimization experiments were used. As in the fictional case study,
the reduction of the total energy consumption of the production lines has the
highest priority. For this reason, two separate optimization experiments were performed. Firstly, the optimizer to minimize the total energy consumption of the
two production lines was run, followed by the consumption peak optimization
experiment.
6.3.1 Optimization Parameter Adaptations
As already briefly indicated, there is a very strong dependence of the machines among one another in the production lines considered. No safety stocks are
provided between the machines to decouple the processing steps, so failures of
individual machines in the line cannot be intercepted by material buffers but lead
to a shutdown of the entire line with a delay of only one to two machine cycles.
Due to the strong coupling of the machines, it does not make sense to optimize
and evaluate the idle and standby optimizer parameters per machine. For a holistic
optimization approach, the standby and idle optimizer parameters are introduced
per line, with one exception: for a number of reasons, the washing machines
are technically unable to change machine states as quickly as the remaining four
machines of the respective line.
In addition, the parts are buffered in front of the washing machines until the
required lot size is reached and the cleaning process can be started. For this reason, L1_wash and L2_wash have their own idle optimizer parameters to take into
account the technical requirements of the real washing machines (Figure 6.8).
In total, the number of optimization parameters in the optimization experiment is reduced from 30 to eight, which has a positive effect on the experiment
runtime but does not alleviate the quality of results. Six of the eight optimization parameters were used for the total consumption optimization and the optimal
configuration of the other two parameters is determined in the peak optimization
runs.