160
6 Experimental Validation of the Methodology
four weeks, the energy consumption behavior of the ten machines was documented. Thus, three different resolutions of the energy consumption data are now
available for the simulation and optimization experiments:
• 15-minute resolution
The 15-minute resolution data is exported from the energy data portal. The
data in the energy portal is based on the machine controller data.
• 1-minute resolution
The 1-minute resolution is exported from the original machine controller.
• 1-second resolution
The 1-second resolution is exported from the additionally installed controller.
Since the energy data collection is done in a separate system that has no interface
to the ERP or MES systems, the combination of energy and production data proved to be difficult. While the energy consumption data is collected automatically,
the documentation of production quantities, production rejects, interruptions, and
maintenance times is a purely manual process, usually done at the end of a shift.
Thus, these data do not have a usable timestamp, which, in turn, makes automatic assignments to the data from the energy measurements impossible. To solve
this data issue, all machine states in manufacturing were manually triggered and
documented with exact timestamps and associated production quantities.
All machines went through a controlled state simulation 5 to be able to precisely
allocate the energy consumptions to the respective machine states. This method
is very time consuming, as it causes the shutdown of a production line for the
duration of the tests of all states. Each of the ten machines was operated in each
machine state for a total of 30 minutes in order to generate the energy database
required to set up the simulation. In this way, a clean, error-free database was
generated that can be used for the optimization of the total energy consumption
and occurring power peak loads.
Besides the simulation of machine states using the real production machines,
the load profile clustering (LPC) algorithm described in section 3.2.3 has been
tested, using only electrical load profiles and processing times to identify different load level. The automated extraction of state-based machine information from
available load profile data in combination with a manual processing cluster assignment saves time compared to the manual machine state simulation and leads
5 The controlled state simulation has been carried out by a Bosch employee for both production
lines. Every machine was manually put into every possible machine state and has been run
for about 30 minutes in every state. The energy load profiles generated in this time can thus
be assigned exactly to the individual states.
6 Experimental Validation of the Methodology
four weeks, the energy consumption behavior of the ten machines was documented. Thus, three different resolutions of the energy consumption data are now
available for the simulation and optimization experiments:
• 15-minute resolution
The 15-minute resolution data is exported from the energy data portal. The
data in the energy portal is based on the machine controller data.
• 1-minute resolution
The 1-minute resolution is exported from the original machine controller.
• 1-second resolution
The 1-second resolution is exported from the additionally installed controller.
Since the energy data collection is done in a separate system that has no interface
to the ERP or MES systems, the combination of energy and production data proved to be difficult. While the energy consumption data is collected automatically,
the documentation of production quantities, production rejects, interruptions, and
maintenance times is a purely manual process, usually done at the end of a shift.
Thus, these data do not have a usable timestamp, which, in turn, makes automatic assignments to the data from the energy measurements impossible. To solve
this data issue, all machine states in manufacturing were manually triggered and
documented with exact timestamps and associated production quantities.
All machines went through a controlled state simulation 5 to be able to precisely
allocate the energy consumptions to the respective machine states. This method
is very time consuming, as it causes the shutdown of a production line for the
duration of the tests of all states. Each of the ten machines was operated in each
machine state for a total of 30 minutes in order to generate the energy database
required to set up the simulation. In this way, a clean, error-free database was
generated that can be used for the optimization of the total energy consumption
and occurring power peak loads.
Besides the simulation of machine states using the real production machines,
the load profile clustering (LPC) algorithm described in section 3.2.3 has been
tested, using only electrical load profiles and processing times to identify different load level. The automated extraction of state-based machine information from
available load profile data in combination with a manual processing cluster assignment saves time compared to the manual machine state simulation and leads
5 The controlled state simulation has been carried out by a Bosch employee for both production
lines. Every machine was manually put into every possible machine state and has been run
for about 30 minutes in every state. The energy load profiles generated in this time can thus
be assigned exactly to the individual states.