sustainability

Article
Energy Consumption of Beverage-Bottling Machines
Isabel Anna Osterroth 1, * and Tobias Voigt 2

1 TUM School of Life Sciences, Technical University of Munich, Alte Akademie 8, 85354 Freising, Germany
2 Brewing and Beverage Technology, Technical University of Munich, Weihenstephaner Steig 22,
85354 Freising, Germany; [email protected]
* Correspondence: [email protected]

Abstract: Sustainability is a megatrend influencing the beverage industry. Knowledge of the con-
sumption behavior and suitable metrics are required for energy optimization strategies. Machine
efficiency and energy consumption are intermixed in common parameters, e.g., customary specifi-
cations refer to the energy consumption for a specific number of products (e.g., kWh/1000 fillings).
This does not reflect the influence that inevitable breakdown times have on the energy consumption
(e.g., malfunction, lack, and tailback situations within the material flow). While specific energy
performance indicators are useful as a benchmark, it does not provide reliable information to verify
plant specifications, or to have a source-related cost allocation as a basis for a weak point analysis. In
this work, energy and operational data were analyzed, in order to find a generic description of the
operational-state related consumption behavior. Therefore, empirical data on the effective electrical
energy and operational state data were collected on machine level of two representative bottling
plants and for additional single machines. In the frequency distributions of the discrete values of the
measured electrical energy data, three main peaks were found. These can be correlated to operational
states such as state-related energy demand level. The change from one demand level to another was
found to be reproducible.






Keywords: energy consumption; bottling; energy performance indicators
Citation: Osterroth, I.A.; Voigt, T.
Energy Consumption of
Beverage-Bottling Machines.
Sustainability 2021, 13, 9880. 1. Introduction
https://doi.org/10.3390/su13179880
Packaging is the last step in food and beverage production and has an important
influence on the product quality, shelf life, and marketing aspects. The packaging of
Academic Editor: Lin Li
beverages and liquid food is performed by a complex network of machines, which are
interlinked with buffering transport units for primary or secondary packages (e.g., bottles
Received: 18 July 2021
Accepted: 25 August 2021
and pallets) [1]. The packaging process comprises unpacking and cleaning of the packages,
Published: 2 September 2021
filling them with the product and sealing of the package, examination of the packed product,
and packaging in a secondary package or transport package. The legal developments,
Publisher’s Note: MDPI stays neutral
economic, social, and technical trends, and rising energy costs have shifted the focus
with regard to jurisdictional claims in
of industry to energy. In particular, companies working in the field of the cost-driven
published maps and institutional affil- beverage and food industry increasingly analyze the energy and media consumption and
iations. are looking for optimization strategies. Olajire found, that despite significant technological
improvements over the last years, energy consumption, water consumption, and water
usage and waste remain major environmental challenges in the brewing industry [2]. The
cost of energy is already an important factor in a company’s cost structure in the food
Copyright: © 2021 by the authors.
and beverage industry, which is why monetary considerations have an influence on this
Licensee MDPI, Basel, Switzerland.
development. Recent social and political drivers (e.g., the UN Sustainable Development
This article is an open access article
Goals [3]) accelerate the demand for a more energy efficient production process. The
distributed under the terms and leading worldwide breweries and beverage companies have recently published ambitious
conditions of the Creative Commons targets to become carbon neutral in the nearer future (see e.g., sustainability goals of
Attribution (CC BY) license (https:// Anheuser-Busch InBev with 25% reduction in CO2 emissions across the value chain until
creativecommons.org/licenses/by/ 2025 [4], Heineken with 0% CO2 emission until 2030 [5], and Coca Cola with net-zero
4.0/). carbon emissions by 2050 [6]). Suitable tools and metrics are required to fulfill these targets.

Sustainability 2021, 13, 9880. https://doi.org/10.3390/su13179880 https://www.mdpi.com/journal/sustainability

Sustainability 2021, 13, 9880 2 of 19

1.1. Beverage-Packaging Technology: Energy Consumption of Bottling Machines

Beverages and liquid foods are packaged in industrial bottling plants consisting of
an arbitrary number of cooperating machines and aggregates (manufacturing stations),
which depend on the product and packaging type. These are all interlinked with buffering
transport elements (mainly conveyors). Various containers, such as bottles (disposable
and returnable bottles, PET bottles, and glass bottles), cans, barrels, or kegs, are used for
packaging and for transporting different beverages (beer, wine, alcohol-free beverages,
etc.). Owing to the modular design of the systems, the different requirements of beverage
filling can be met by a variety of machines. According to Manger [7], the components of a
beverage filling line are essentially the same:
• Machines unloading and loading pallets, as well as unpacking the reusable contain-
ers made of glass and plastic from the transport boxes (e.g., unloading machines,
unpacking machines, and new glass slippers);
• Machines ensuring the cleaning and control of the outer packaging (e.g., crate washing
machines);
• Machines ensuring the cleaning and inspection of the packaging material (returnable
bottles) (e.g., bottle and empty bottle inspectors, bottle-cleaning machines);
• Machines filling and sealing the packaging (e.g., bottle filling machines with a com-
bined capper);
• Machines labeling and checking the packing materials (e.g., labeling machines);
• Packing machines (e.g., shrink packers, palletizers);
• Conveyors for containers, packs, and pallets (e.g., chain conveyors with a drive engine).
• In addition, the following are used when needed:
• Machines for bottle manufacturing (e.g., stretch-blow molders for PET bottles);
• Machines improving the biological shelf life (e.g., pasteurizers);
• Machines for cleaning and disinfection (e.g., CIP systems);
• Sorting systems of returnable containers (e.g., selective unpacking machines);
• Machines for producing soft drinks (e.g., beverage mixing and carbonizing plants).
The requirements for individual plant components are related not only to the container
type and size and the packaging material (glass, PET, aluminum, or steel), but also to the
beverage properties (presence of CO2 and viscosity). During the filling and packaging
of the beverages and liquid foods, different media are consumed by different processes.
The typical consumption of beverage-bottling, which has a significant commercial effect,
includes the following:
• Electrical power and compressed air;
• Thermal power;
• Media such as cold water, lye, additives, and lubricants.
In order to operate a bottling plant, all bottling and packaging machines require
electrical power, in addition to cleaning machines (bottle-cleaning machines and crate
washers), and pasteurizers require thermal energy (see Table 1). In addition, some machines
use compressed air, usually created by electrically driven compressors (mainly packers and
palletizers). Water is mainly necessary for the cleaning machines of primary and secondary
packaging materials (bottles, crates) and during the pasteurization process.
Electrical power is usually used as the main drive for machines moving containers,
drive pumps, to perform their basic mechanical functions as well as for the production
of compressed air. The consumption of electrical energy is usually linked directly to the
function performed. Any interruption of the function or reduction of the output power also
leads to reduced consumption in electrically driven motors and pumps. Thermal energy is
usually used to heat various media (e.g., heating of products, heating of the lye baths in
bottle-cleaning machines, and supply of hot water). The consumption of thermal energy by
a machine can be decoupled from its operational state since the media might be heated even
in the case of a malfunction in the machine. Compressed air is used during beverage filling
either as an energy carrier (stretch-blow molding of PET bottles, transport of empty PET
Sustainability 2021, 13, 9880 3 of 19

bottles), for signal transmission (pusher), or for cleaning (blow-off, drying). Compressed
air is a comparatively energy-intensive energy carrier and is produced using electrical
power. Freshwater is usually used to clean machines and is also used in bottle-cleaning
machines (cleaning of returnable containers). Other media, such as lye, belt lubricants,
and additives, are used mainly in bottle-cleaning machines and during the transportation
of containers.
Table 1. Energy and media consumption of bottling machines according to the minimum and order
related specifications of beverage filling lines, own table according to [8].

Electrical Power Compressed Air Heat Water

Stretch-blow molder × ×
Depalletizer × ×
Unpacker × ×
Crate washer × × ×
Bottle-cleaning machine × × ×
Labeling machine × ×
Pasteurizer × × ×
Packing machine/tray packer × ×
Palletizer × ×

Over the last years, mainly industry associations and consulting companies have
published consumption values and benchmarks. The data are predominantly related to
the total production processes or single-process units (e.g., packaging). Most of the data
are published as total consumption for the production or differentiated into thermal and
electrical energy, which is mainly related to the product volume (e.g., 1 hl of sales beer).
The cost of supplying energy to a brewery differs worldwide from 3% to 10% of the total
budget [9–11]. Over the last 10 years, the required volume of water to produce 1 hl of beer
decreased from 5.0–5.2 hl to 4.2–4.3 hl [9,12–14]. Moreover, 16.5–30% of the total heat and
12–35% of the total electrical power are needed for the bottling process [15–17]. Hauser
and Shellhammer analyzed the sustainability challenges in beer production and found
that packaging has a large share on the environmental impact of beer production [18]. In
contrast to breweries, only a limited amount of consumption data has been published
for manufacturers of nonalcoholic beverages. No detailed machine-based consumption
data and analyses of beverage-packaging machines have been published in the scientific
literature, with the exception of one report [19] in the early 1980s. No energy data correlated
to the operational state of single machines have been published either.
For in-depth analyses and optimization measures, a detailed consideration of the
energy consumption on the machine level is lacking. Measurements on bottling and
packaging machines in industrial applications related to this research indicated that there is
a significant difference between the installed load and the average measured consumption,
which indicates a saving potential. The influence of the machine efficiency on the energy
consumption is not considered in commonly used key performance indicators. This does
not consider, e.g., the influence of breakdown times, which, because of equipment failure,
lack, and tailback situations, inevitably occur during the production. While considering
the energy consumption of a defined amount of product is useful for the comparison of
the performance of two plants and management decisions, it does not provide reliable
information to verify specified machine-based consumption data or for a source-related
cost allocation as a basis for a weak point analysis and machine, process, and automation
optimization. Some manufacturers have started to specify a consumption level related
to the nominal speed. However, consumption values for planned or unplanned stops
of machines were missing in the past. With VDMA 8751:2019-03 Packaging machinery
(incl. filling machinery)—Specification and measurement of energy and utility consumption [20],
recently, a normative directive was published to close this gap.
Osterroth et al. have published a summary of bottling related energy KPIs and a
survey for the German beverage bottling industry. It was found in this publication that
Sustainability 2021, 13, 9880 4 of 19

the available and published data are not yet detailed enough for a modeling approach [21]
that could be used as a tool for energy optimization of bottling plants. The available data
are not suitable for any comparison, as the survey approaches and system boundaries
vary greatly. No correlation between the current machine or process state and the energy
consumption has been investigated, and considerations to reduce the energy and media
consumption have been limited locally to individual system components or to a high-level
view so far. With detailed knowledge regarding the consumption behavior and with a
generic model mapping this behavior, the main consumers can be detected, identifying
optimization potential in the production process, the machines, and the plant automation.

1.2. Models Describing the Energy Consumption in Industrial Manufacturing Lines

Some models that describe the energy consumption of machines in manufacturing
lines can be found in the scientific literature: Dietmair and Verl published a generic
energy consumption model for decision making and energy efficiency optimization in
manufacturing for plants, machines, and components based on a statistical discrete event
formulation [22]. Lees et al. described a utility consumption model for real-time load
identification in a brewery [23]. Several fuzzy and neurofuzzy models have been developed
for electrical load forecasting (e.g., [24,25]). One example of the use of mathematic modeling
is the energy optimization of refrigeration systems in breweries. Xu et al. developed a case-
based reasoning (CBR) energy consumption model for cutting periods in CNC lathes [26].
Cataldo et al. worked on the modeling, simulation, and evaluation of energy consumption
for a manufacturing production line [27]. For tooling machines, a calculation method
was developed by Kuhrke using “energy blocks”, which are consumption patterns based
on functional changes in the machine operations, comparable to operational states. For
those machine tools, for mass production, a state-based model was defined, describing
the detailed single phases of the production process, based on energy blocks [28]. This
model was published in VDMA 34179 as a measurement instruction determining the
energy and resource demand of machine tools for mass production [29] and was updated
in 2019 [30]. Braun et al. discussed a state-based energy consumption modeling approach
for the example of wireless sensors and wireless local area networks, considering the
consumption patterns for single states and state transitions [31].
Owing to their high interlinking level and high output (up to 120,000 packages per
hour), packaging and bottling machines differ from other manufacturing plants. It was
found that state models from other industries cannot be transferred to bottling plants
directly. Customary specifications refer to the energy consumption for a specific number
of products (e.g., kWh/1000 fillings). Osterroth et al. published a state related simulation
model for bottling plants to close this scientific gap [32], which is completed by the here-
presented data analysis. As part of this work, the energy behavior of food and beverage
packaging machines were analyzed systematically, considering the recent operational state
of the machines for the first time.

1.3. Purpose
The data presented in this paper were unpublished as of now, extend the existing non-
sufficient database, complete the normative direction VDMA 8751:2019 and the published
simulation approach, and give additional insights in the state-related energy consumption
behavior for a larger number of considered machines. It is assumed that a detailed generic
model can be used as the basis for the in-depth analysis of the consumption behavior
for identifying inefficient production times and for conducting further research on the
modeling and simulation of the energy consumption of food and packaging plants. The
purpose of this study is to gain fundamental knowledge regarding the operational state
related consumption behavior of packaging and bottling machines based on detailed
empirical data from industrial plants. Detailed empirical electrical energy data on machine
level shall be analyzed to identify main consumer and to provide generic statements for
future modeling and forecasting approaches and optimization measures.
Sustainability 2021, 13, 9880 5 of 19

2. Data Acquisition
Owing to the lack of reliable data, electrical energy consumption data and operational
state data of 20 machines in industrial applications were considered to characterize the con-
sumption behavior of bottling and packaging machines in the food and beverage industry.

2.1. Acquisition of Energy Data

The electrical power consumption was recorded with active power meters of the
highest available sample rate (1 or 2 s) as effective electrical power (kW). Depending on the
technical possibilities, integrated (Janitza UMG 96RM-PN, measurement accuracy: ±0.2%)
or mobile measuring instruments (Fluke 435 II, measurement accuracy: ±1%) were used.
All active power meters were state of the art and calibrated. The measured discrete values
were recorded in a MS SQL database with a timestamp.

2.2. Acquisition of Operational State Data

Physical machine state changes of production machines and lines can be described by
established state models mainly used for machine control and automatization. Beverage
bottling plants and packaging machines are generally described by some state models in
scientific research and industrial application, as discussed below.
The American National Standards Institute (ANSI) and the Instrumentation, Systems
and Automation Society (ISA) published the international standard ANSI/ISA-S 88 [33],
addressing batch process control in which the models, terminology modes, and states
of single process units or machines are defined. These definitions are not part of the
standard but are generally defined. The model describes operating modes (automatic,
semi-automatic, and manual) and operating conditions (idle, running, complete, paused,
holding/held, restating, stopping/stopped, aborting/aborted). In the operating state
model, a focus is placed on the transitions between the final states. For the application to
packaging machines, a technical report is published [34].
The “Weihenstephan Standard” defines a communication protocol, data tags to acquire
relevant production data and procedures for efficiency analysis (e.g., OEE indicators,
cost transparency), tracking and tracing of batches, transport units and orders, and the
monitoring of production processes and the product quality. It includes a state model,
describing the physical machine states with its operating mode (off, automatic, semi-
automatic, manual), its program (e.g., production, start up, clean), and its operational
state [35]. The “Weihenstephan Standard” state definitions are harmonized with the
definitions of the OMAC Packaging Machine Language Working Group, which has also
published a uniform state model to describe the different operation states and user actions
with packaging machines [36].
For the analysis of the production and performance of packaging and beverage bot-
tling lines a time model is established in accordance with DIN 8743 [37] The machine
working time tm is defined as the theoretically available time (24 h/7 days a week). De-
ducting the idle time (e.g., weekend, holidays), the machine work time is defined. The
operating time is defined as the machine working time reduced by scheduled downtimes
(e.g., cleaning/maintenance). During the machine working time breakdowns will occur
due to internal and external failures. The machine running time is the working time re-
duced by the breakdown time. The time model is correlated with the physical machine
(see Figure 1).
The state data were collected time discrete every second or every two seconds (as
available) by the manufacturing execution systems (MES) in accordance with the industrial
“Weihenstephan Standard” based on the time model according to DIN 8743 [37]. If no MES
was available, the data were collected by the Weihenstephan Test Tool, collecting the state
information directly on the PLC of the machine or manually (handwritten notes). The
state data were recorded event discrete with a start time and an end time of the state. The
state information is based on a state model and indicates whether a machine is producing
(operating state “operating”) or waiting to produce in a suspended state due to an internal
Sustainability 2021,13,
Sustainability2021, 13,9880
x FOR PEER REVIEW 6 6ofof1921

causeFor the analysis

(“failure”, of the
“held”, production
“emergency and performance
stop”) of packaging
or external cause and beverage
(“lack”, “tailback”, bot-
“idle”,
tling lines a There
“prepared”). time model
was nois data
established
availablein for
accordance
the machinewithmode
DIN (on/off)
8743 [37]andTheprogram
machine
working
(e.g., time tm is“Maintenance”,
“Production”, defined as the theoretically
“Cleaning”). available time (24 h/7 can
The state information days
beaassigned
week). De-
to
ducting
time the idle
intervals time (e.g., weekend,
in accordance holidays),
with the time model the machine
of DIN workBefore
8743 [37]. time is defined. the
analyzing The
operating
state time
data, the is defined
correct as the machine
acquisition working
of the machine timewas
states reduced byatscheduled
verified least threedowntimes
times for
each
(e.g.,state by checking the stateDuring
cleaning/maintenance). information in the recording
the machine working database in parallelwill
time breakdowns withoccur
the
actual
due tomachine
internalbehavior. The failures.
and external data acquisition system
The machine does not
running record
time is theinformation
working timeabout
re-
the states
duced byofthe
thebreakdown
bottle, cratetime.
and pallet conveying
The time model systems, and the
is correlated inspection
with machines.
the physical machineIt
is(see
for Figure
this reason
1). that they were not considered in this work.

Figure1.1.Time
Figure Timemodel
modelaccording
accordingtotoDIN
DIN8743
8743[37]
[37]and
andexample
examplestates
states(grey).
(grey).

2.3. Considered
The stateBottling
data werePlants and Data
collected time discrete every second or every two seconds (as
available) by the manufacturing execution
To obtain representative data, five bottling systems (MES)
plants usingintypical
accordance with the
packaging indus-
types (re-
turnable glass and PET bottles) and products (water, soft drinks, and beer) were consideredIf
trial “Weihenstephan Standard” based on the time model according to DIN 8743 [37].
nothe
for MES wasanalysis.
data available,
Allthe data were
machines collected
of two by plants
bottling the Weihenstephan
were analyzed, Test
and Tool, collecting
for the other
the state
ones, onlyinformation directly on(bottle
the main consumers the PLC of the machine
cleaning machineor ormanually (handwritten
stretch-blow moulder)notes).
were
The state
taken into data wereAdditionally,
account. recorded event discrete with
a packaging a start
machine fortime
tray and
packsanwas
endanalyzed
time of the state.
having
aThe statemachine
control information
fromisa based on packaging
different a state model and indicates
application. whether aTable
The following machine is pro-
2 shows a
ducing (operating state “operating”)
summary of the considered machines. or waiting to produce in a suspended state due to an
internal cause (“failure”, “held”, “emergency stop”) or external cause (“lack”, “tailback”,
Table 2. Summary
“idle”, of data
“prepared”). acquisition
There was noondata
the considered
available formachines.
the machine mode (on/off) and pro-
gram (e.g., “Production”, “Maintenance”,
Electrical
“Cleaning”).
Sample Number of
The state information can be as-
Packaging
signed to time intervals in accordance with the time model of DIN Productana-
8743 [37]. Before
Consumption Data Rate Machines Type
lyzing the state data, the correct acquisition of the machine states was verified at least
active power meter, returnable water, soft
Bottling
three plantfor
times 1 each state by checking the 2 sstate information
10 inglass
the recording database
integrated bottle drinks in
parallel with the active
actualpower
machine behavior. The data acquisition
meter, PETsystem
bottle doeswater,
not record
soft
Bottling plant 2 2s 9
information about theintegratedstates of the bottle, crate and pallet conveying (single use)systems, and the
drinks
inspection active
machines. It power meter,
is for this reason that returnable
Bottling plant 3 1 s they were 1 not considered in this work. beer
mobile glass bottle
2.3. Considered active power meter, returnable water, soft
Bottling plant 4 Bottling Plants and Data 1s 1
mobile glass bottle drinks
To obtain representative data, five bottling plants using typical
active power meter, packaging
PET bottle types
water, soft(re-
Bottling plant 5 10 s 1
turnable glass and PETmobile
bottles) and products (water, soft drinks, anduse)
(single beer) were consid-
drinks
Packaging
ered active power
for the data analysis. All meter,
machines of two bottling plants were analyzed, and for the
1s 1 tray packs various
machine
other ones,Aonly the mainmobile
consumers (bottle cleaning machine or stretch-blow moulder)
were taken into account. Additionally, a packaging machine for tray packs was analyzed
3.having
Results
a control machine from a different packaging application. The following Table 2
3.1. Development
shows a summary of Analysis Methods machines.
of the considered
For the in-depth analysis of the measured energy consumption data of packaging and
bottling machines, the frequency distribution of the measured discrete effective electrical
power values was plotted as a histogram for every single machine (see Figure 2). The class
width of the histogram was defined depending on the number of the single values and the
distribution. It was assumed that peaks represent a so-called energetic demand level. As
Sustainability 2021, 13, 9880 7 of 19

Sustainability 2021, 13, x FOR PEER REVIEW 8 of 22

the height of the peaks is assumed to be the result of the machine use (occurring operation
states) during the measurements, the peaks were qualitatively evaluated.

Figure 2.
Figure Example of
2. Example of frequency
frequency distribution
distribution of
of the measured effective power values discrete.

The first
The first analysis
analysis results
results plotting
plotting the
the electrical
electrical energy
energy and
and the
the occurring
occurring operational
operational
states indicated
states indicated aa correlation
correlation between
between the
the consumption
consumption and
and the
the operational states (see
operational states (see
Figure 3).
Figure 3).
For the purpose of illustration and analyzing the correlation between the energy
consumption and operational states, a diagram was developed showing the measured
effective electrical power plotted as a time-discrete 2D line plot with colored event-discrete
intervals (areas) in a background layer representing the occurring operational states. The
boundaries of the intervals are described by the start time and end time of the operational
state ([start_timestate n; end_timestate n], see Figure 4). Therefore, every defined opera-
Sustainability 2021, 13, x FOR PEER REVIEW 8 of 21
tional state was assigned a color (see Table 3). Similar operational states were summarized
to one color, for example, lack and lack in branch line.

Figure 3. Example of the measured effective electrical energy raw data and empirical operational
state raw data plotted over time for different machines (Machine ID: 1001, 3001, etc.).

For the purpose of illustration and analyzing the correlation between the energy con-
sumption and operational states, a diagram was developed showing the measured effec-
tive electrical power plotted as a time-discrete 2D line plot with colored event-discrete
intervals (areas) in a background layer representing the occurring operational states. The
Figure 3. Example of the measured effective electrical energy raw data and empirical operational
boundaries
Figure of theofintervals
3. Example the are described electrical
measured by the start timerawanddata
endand
time of the operational
state raw data plotted over time for effective
different machines energy
(Machine ID: 1001, empirical
3001, etc.). operational
state ([start_timestate n; end_timestate n], see Figure 4). Therefore, every
state raw data plotted over time for different machines (Machine ID: 1001, 3001, etc.). defined opera-
tional state was assigned a color (see Table 3). Similar operational states were summarized
to one
Forcolor, for example,
the purpose lack and and
of illustration lackanalyzing
in branch the
line.correlation between the energy con-
sumption and operational states, a diagram was developed showing the measured effec-
tive electrical power plotted as a time-discrete 2D line plot with colored event-discrete
intervals (areas) in a background layer representing the occurring operational states. The
boundaries of the intervals are described by the start time and end time of the operational
state ([start_timestate n; end_timestate n], see Figure 4). Therefore, every defined opera-
tional state was assigned a color (see Table 3). Similar operational states were summarized
to one color, for example, lack and lack in branch line.
 sumption and operational states, a diagram was developed showing the measured effec-
tive electrical power plotted as a time-discrete 2D line plot with colored event-discrete
intervals (areas) in a background layer representing the occurring operational states. The
boundaries of the intervals are described by the start time and end time of the operational
Sustainability 2021, 13, 9880 state ([start_timestate n; end_timestate n], see Figure 4). Therefore, every defined opera-
8 of 19
tional state was assigned a color (see Table 3). Similar operational states were summarized
to one color, for example, lack and lack in branch line.

Figure 4. Example
Figure plotplot
4. Example forfor
evaluating
evaluatingthe
thecorrelation betweenthe
correlation between the operational
operational state
state and and energy
energy con-
consumption.
sumption.
Table 3. Operational states and related background colors in the plot.
Table 3. Operational states and related background colors in the plot.
Operational State Color Coding
Operational State Color Coding
Prepared
Prepared Orange
Orange
Lack Blue
Lack
Tailback Blue
Yellow
Tailback
Lacking branch line Yellow
Blue
Tailback
Lacking branch
branch line
line Yellow
Blue
Operating
Tailback branch line Green
Yellow
Equipment failure Red
External failure Dark orange
Held Grey
Idle Grey

3.2. Electrical Energy Consumption of Packaging and Bottling Machines

For the considered bottling plants, the average consumption for single machines was
calculated. Owing to the limitations in the available measurement instruments, no data
on the conveyor system of the returnable glass bottling plant were available. For the
PET bottling plant, the total value of the bottle conveyor and the total value of the pallet
transport were available.

3.2.1. Selected Absolute Consumption Values on the Example of a PET Bottling Plant
(Plant 2)
Table 4 shows the selected measured electrical energy consumption for all considered
machines on the example of the PET bottling plant (Plant 2) for three time periods of
24 h, each during a production phase. The stretch-blow molder was identified as the
main consumer with an average consumption of 157 kW, which is more than 50% of the
total consumption of the line (245 kW on average). The conveyor systems had an average
consumption of 6% of the total consumption (4% for the bottle conveyor and 2.2% for the
pallet conveyor).
Sustainability 2021, 13, 9880 9 of 19

Table 4. Measured electrical energy consumption for the example of a PET bottling plant for three time periods (ti ) of 24 h
each (sorted by main consumers), average consumption, standard deviation, and percentual share on total consumption.

Avg. Standard
∑t1 [kWh] ∑t1 [kWh] ∑t1 [kWh] Consumpt. Deviation Share [%]
[kW] [kW]
Stretch-blow molder A 3019.0 4281.0 3998.6 156.9 22.5 64
Labeler A 343.6 449.7 513.2 18.1 2.9 7.4
Shrink packer A 326.1 448.5 446.6 17.0 2.4 6.9
Cooler A
267.6 385.9 373.6 14.3 2.2 5.8
(stretch-blow molder)
Palletizer A 259.2 365.8 356.3 13.6 2.0 5.6
Conveyor (bottles) A 189.2 258.8 254.4 9.8 1.3 4.0
Filler A 157.8 203.7 200.2 7.8 0.9 3.2
Conveyor (pallets) A 107.0 141.9 135.5 5.3 0.6 2.2
Handle application A 46.9 62.2 61.2 2.4 0.3 1.0
Conveyor
1.2 1.3 1.3 0.1 <1 <1
belt lubrication A
Preform feed A
1.2 1.2 1.2 0.1 <1 <1
(stretch-blow molder)
Total consumption,
4718.8 6600.0 6342.0 245.3
ti (24 h each)

Figure 5 shows a comparison of the installed load and the specified consumption
for nominal speed as well as the measured average consumption and the consumption
during production times for the example of the selected machines of bottling plant 2. The
figure shows a wide deviation between the specified values and the measured values. For
Sustainability 2021,the
13, xstretch-blow molder and filler block, including the cooler and preform feed, less than
FOR PEER REVIEW 10 o
50% of the installed load was measured. Peak loads were not found with this interval of
measurement (one value per second).

500

400

300
kW

200

100

0
strech blow moulder conveyor (total) labeler pelletiser
and filler block

installed load specified consumption at nominal speed

average measured consumption measured consumption production

Figure 5.
Figure 5. Installed Installed
load, load, consumption
consumption nominal
nominal speed, speed,
measured measured
average average consumption,
consumption, and measuredand measured during
consumption
production.consumption during production.

Additionally, the energy consumption

Additionally, during longer
the energy consumption unproductive
during times (e.g., week-
longer unproductive times (e.g., wee
ends) was analyzed.
ends) was analyzed. Table 5 shows a summary of the averageconsumption
Table 5 shows a summary of the average measured measured consumpti
during a time period
during of >24 period
a time h, as well as the
of >24 percentage
h, as well as theofpercentage
consumption compared to
of consumption the
compared to t
average value.average
While the main
value. consumer,
While the main the stretch-blow
consumer, molder, has molder,
the stretch-blow a low percentage
has a low percenta
of consumptionof during these times,
consumption duringthe labeler
these times,and
thethe fillerand
labeler havethea filler
higher percentage
have of
a higher percentage
consumption during these times.
consumption duringFor thetimes.
these stretch-blow molder, the filler’s
For the stretch-blow molder, and
thethe labeler’s
filler’s and the labele
detailed operational state
detailed data were
operational available.
state data were available.

Table 5. Measured average power consumption during an unproductive time period of >24 h.

Average
Standard [%] of Avera
Consumption
Deviation [W] Consumptio
[W]
Labeler 4112 2341 33
Sustainability 2021, 13, 9880 10 of 19

Table 5. Measured average power consumption during an unproductive time period of >24 h.

Average Standard [%] of Average

Consumption [W] Deviation [W] Consumption
Labeler 4112 2341 33
Filler 2020 385.1 16
Stretch-blow molder 1715 62.4 14
Conveyor (pallets) 1082 25.4 9
Conveyor (bottles) 1058 465.7 9
Handle application 1024 836.4 8
Palletizer 581 2.7 5
Handle application 407 3.4 3
Cooler (stretch-blow molder) 231 4.8 2
Preform feed (stretch-blow molder) 49 1 <1
Conveyor belt lubrication 48 1.6 <1
Total consumption 12,327

Table 6 summarizes the specific energy consumption for those machines for a time
period of one production week and specifies the occurring operational states as well as the
consumption during these times.

Table 6. Specific energy consumption of the main consumer for a time period of one week and an
analysis of the consumption related to the operational behavior of the machines. No state data are
available for the shrink packer and palletizer.

Stretch-Blow Molder Filler Labeler

[Wh/1000 [Wh/1000 [Wh/1000
[h] [h] [h]
Bottles] Bottles] Bottles]
Production 119 4290.2 117 212.8 121 581.6
Prepared 4.6 12.4 5.2 8.1 3.5 3.3
Lack 0.1 1.3 0.0 0.0 4.6 4.4
Tailback 6.8 145.8 7.8 9.9 4.8 4.6
Equipment failure 8.1 195.6 4.9 6.5 2.5 2.2
Held 29.2 48.7 33 37.4 31.7 25.2
Total 4694.0 274.8 621.2
Availability of the
71.0% 69.8% 72.0%
machine

Figure 6 shows the typical consumption structures for a PET bottling plant (A) for
single-use bottles and a glass bottling plant (B) for returnable glass bottles. The main
consumers of these types of bottling plants are the stretch-blow molders (PET) and the
bottle-cleaning machines (glass).

3.2.2. Discussion of the Absolute Consumption Values

Both considered lines have a main consumer, which has a significant share on the
total consumption of the total bottling plant (stretch-blow molder: >60% for PET bottles;
bottle-cleaning machine: >45%), which should be the focus of optimization considerations.
The installed load and the specified consumption differ from the measured consump-
tion during production times by up to 60%. The common considerations using these values
for an estimation of the energy consumption will have a limited accuracy. For optimization
purposes and business decision criteria, these values should be verified by measurements
of the actual consumption. For modeling and forecasting of the energy consumption,
the implemented parameter should be verified at least by single measurements on the
considered plant.
 71.0% 69.8% 72.0%
chine

Figure 6 shows the typical consumption structures for a PET bottling plant (A) for
single-use bottles and a glass bottling plant (B) for returnable glass bottles. The main con-
Sustainability 2021, 13, 9880 11 of 19
sumers of these types of bottling plants are the stretch-blow molders (PET) and the bottle-
cleaning machines (glass).

Figure 6. Consumption
Figure 6. Consumption structure of aa PETPET bottling
bottling plant
plant (A)
(A) and
and aa glass
glass bottling
bottling plant
plant (B).
(B). Own
Own measurements for nn == 1
measurements for
production
productionline
lineofofeach
eachtype,
type,showing thethe
showing main consumers:
main consumers:stretch-blow molders
stretch-blow andand
molders bottle-cleaning machines.
bottle-cleaning The average
machines. The av-
erage consumption
consumption of the stretch-blow
of the stretch-blow moldermolder is 157
is 157 kW, kW, whereas
whereas that of that of the bottle-cleaning
the bottle-cleaning machine machine is 49 kW.
is 49 kW.

3.2.2.ItDiscussion
was foundofthat
thethere is a correlation
Absolute Consumptionbetween
Valuesthe consumption behavior (see, e.g.,
TableBoth
7) and the occurring operational states. A simple
considered lines have a main consumer, which consideration of the specified
has a significant share oncon-
the
sumption
total consumption of the total bottling plant (stretch-blow molder: >60% for PETa bottles;
values will not result in an accurate forecast, as the measurements have proven
deviation from the actual consumption value. The availability of the machine, defined as
bottle-cleaning machine: >45%), which should be the focus of optimization considerations.
the share of production time to the total time, and the resulting differing consumption
The installed load and the specified consumption differ from the measured consump-
during the occurring operational states will influence the specific energy consumption.
tion during production times by up to 60%. The common considerations using these val-
Therefore, in the following chapter, the operational-state-related energy consumption
ues for an estimation of the energy consumption will have a limited accuracy. For optimi-
behavior will be analyzed in order to provide a reliable analysis of the influence of the
zation purposes and business decision criteria, these values should be verified by meas-
operational state and the detailed characterization of the consumption behavior.
urements of the actual consumption. For modeling and forecasting of the energy con-
sumption, the implemented parameter should be verified at least by single measurements
Table 7. Analysis of the consumption peaks of the considered machines.
on the considered plant.
Number of
Plant Machine Correlation to the Operational State
Main Peaks
(1) Related to inactive times
Depalletizer 3 (2) Related to downtime, multimodal peak
(3) Extended peak, related to running time
(1) Related to inactive times
(2) Related to downtime, multimodal peak
Unpacker 3
(3) Extended multimodal peak, related to running time,
not clearly defined (see the example in Figures 9 and 10)
(1) Related to inactive times
(2) Multimodal peak, related to downtime (highest peak
Selective depacker 3
due to extended downtimes)
(3) Related to running time, extended peak
Bottling plant 1
(1) Related to inactive times
Bottle-cleaning (2) Minor peak related to heating-up processes, main
3
machine peak related to running time
(3) Related to operation
(1) Multimodal peak, related to inactive times
(2) Related to downtime, multimodal peak
(3) Related to running time
Note. The speed of the machine was reduced, resulting
Filler 4
in a reduced energy consumption for >1 min after the
beginning of a downtime due to technological reasons
(emptying of the machine). This results in a multimodal
peak (2)
Sustainability 2021, 13, 9880 12 of 19

Table 7. Cont.

Number of
Plant Machine Correlation to the Operational State
Main Peaks
(1) Multimodal peak, related to inactive times
Labeler 3 (2) Multimodal peak, related to downtime
(3) Related to running time, extended peak
(1) Related to inactive times
Packer 3 (2) Related to downtime
(3) Related to running time, extended peak
Mixer 3 No direct correlation found
(1) Related to inactive times
Palletizer 3 (2) Related to downtime
(3) Related to running time, extended peak
(1) Related to inactive times/downtime
Crate washer 2
(2) Related to running time
(1) Related to inactive times
Stretch-blow molder 3 (2) Related to downtime
(3) Related to running time
(1) Related to inactive times
Labeler 3 (2) Related to downtime
(3) Related to running time
(1) Related to inactive times
(2) Related to downtime
Shrink packer 4
(3) Related to downtime
(4) Related to running time
(1) Related to inactive times
Bottling plant 2
Packer 3 (2) Related to downtimes
(3) Related to running time
(1) Related to inactive times, multimodal peak
(2) Related to downtime, multimodal, not clearly defined
Filler 4 (3) (a, b, c) Three main peaks all related to running time,
different products with different machine speeds
produced on the machine during the measurement time.
(1) Related to inactive times
Handle application 3 (2) Related to downtime, multimodal peak
(3) Related to running time
(1) Related to inactive times
Bottle-cleaning (2) Minor peak related to heating-up processes, major
Bottling plant 3 3
machine peak related to downtime
(3) Related to running time
(1) Related to inactive times
Bottle-cleaning (2) Minor peak related to heating-up processes, major
Bottling plant 4 3
machine peak related to downtime
(3) Related to running time
(1) Related to downtime
(2) Related to running time
Bottling plant 5 Stretch-blow molder 2
Note. The measurement time was only 8 h; no inactive
times were considered.
(1) Related to inactive times, related to downtime
(2) Related to running time
Note. The measurement time was only 8 h; no inactive
Packaging machine A 2
times were considered. There was no significant change
in consumption owing to the variation of
materials/production parameters.
 It was found that there is a correlation between the consumption behavior (see, e.g.,
Table 7) and the occurring operational states. A simple consideration of the specified con-
sumption values will not result in an accurate forecast, as the measurements have a
proven deviation from the actual consumption value. The availability of the machine, de-
fined as the share of production time to the total time, and the resulting differing con-
Sustainability 2021, 13, 9880 sumption during the occurring operational states will influence the specific 13 of 19
energy con-
sumption. Therefore, in the following chapter, the operational-state-related energy con-
sumption behavior will be analyzed in order to provide a reliable analysis of the influence
of the operational state and the detailed characterization of the consumption behavior.
The energy consumption of the conveyor
The energy consumption of theelements
conveyorwas measured
elements to be 6.2%
was measured to of
be the
6.2% of the
total consumption
total for the PET bottling
consumption for the PETplant and should
bottling ideally
plant and be ideally
should considered in futurein future
be considered
measurements.measurements.
Owing to theOwinglarge to
number
the largeof single
numberdrives, thedrives,
of single measurement requiresrequires
the measurement
some effort. If possible, theIfmeasurement
some effort. can be summarized
possible, the measurement to one or two
can be summarized areas.
to one Because
or two areas. Because
of the limited
of the limited number number
of devices of devices for measurement,
for measurement, no values areno values are
available foravailable for the return-
the returnable
able glass bottling
glass bottling plant in this work. plant in this work.

3.3. Operational-State-Related
3.3. Operational-State-Related Energy Consumption
Energy Consumption Behavior Behavior
3.3.1. Analysis3.3.1.
of Empirical
Analysis Energy Consumption
of Empirical Data in Correlation
Energy Consumption to Operational
Data in Correlation to Operational
State Data State Data
For all machines,
For the frequency
all machines, thedistributions of the measured
frequency distributions discrete effective
of the measured elec- elec-
discrete effective
trical power values (kW), values
trical power plotted(kW),
as histograms, were analyzed.
plotted as histograms, The following
were analyzed. figures figures
The following
show an exampleshowofana example of a bottle-cleaning
bottle-cleaning machine,
machine, which is which
one ofisthe
onemain
of theconsumers
main consumers
in in the
bottling
the bottling process forprocess for returnable
returnable glass according
glass bottles bottles according
to the to the measurement
measurement shownshown
in in Fig-
Figure 6. ure 6.
Figure 7 shows Figure 7 shows the
the measured measured
discrete discrete
effective effectivepower
electrical electrical powerplotted
values, values,asplotted
a as a
histogram, for 42 days of measurement. Three main peaks were
histogram, for 42 days of measurement. Three main peaks were identified. Peak 1 indicatesidentified. Peak 1 indi-
cates low consumption, which is significantly larger than Peaks 2 and 3. This is the result
low consumption, which is significantly larger than Peaks 2 and 3. This is the result of
of the longer idle times (weekends, nights) during the measurement. Peak 2 shows re-
the longer idle times (weekends, nights) during the measurement. Peak 2 shows reduced
duced consumption and Peak 3 shows the highest consumption in kilowatt. The peaks are
consumption and Peak 3 shows the highest consumption in kilowatt. The peaks are not all
not all clearly delineated, but they are merged into each other (Peaks 2 and 3).
clearly delineated, but they are merged into each other (Peaks 2 and 3).

(1)
(2) (3)

Figure 7. Histogram of the measured

Figure 7. Histogram effective
of the electrical
measured powerelectrical
effective data of bottle-cleaning
power data of machine A over a period
bottle-cleaning of 42 A
machine days:
three main peaks (Peaks 1, 2, and 3).
over a period of 42 days: three main peaks (Peaks 1, 2, and 3).
While comparing the consumption patterns of the machines shown in the 2D line
While comparing the consumption patterns of the machines shown in the 2D line
plots of the effective power data with colored state information in the background and a
plots of the effective power data with colored state information in the background and a
detailed frequency distribution histogram (right part of Figure 8), a correlation between
the identified peaks and the operational states can be proven. Figure 8 shows the details
of the 2D line plot mapping the electrical energy consumption pattern of production
time interrupted by a tailback, lack, or equipment failure situation for a bottle-cleaning
machine (Machine A). The effective electrical power value changes with the operational
state changes of the machine.
The two peaks in the histogram (Figures 7 and 8, right) fit the value of the almost
constant performance level that the machine reaches after a state change. Different types
of peaks related to the running times (see the time model in Figure 1) were found for the
machines. The peaks of the machines with a steady operational behavior (bottle filling
machines, bottle-cleaning machines, thermoforming machines, etc.) were high and narrow,
 interrupted by a tailback, lack, or equipment failure situation for a bottle-cleaning ma-
chine (Machine A). The effective electrical power value changes with the operational state
changes of the machine.

Sustainability 2021, 13, x FOR PEER REVIEW 13 of 21

Sustainability 2021, 13, 9880 14 of 19

detailed frequency distribution histogram (right part of Figure 8), a correlation between
whereas theypeaks
the identified were and
flat the
andoperational
extended states
for cycling
can beand intermittently
proven. operating
Figure 8 shows machines
the details
(packer,
of the 2Dpalletizer). Figurethe
line plot mapping 9 electrical
shows this on the
energy example of
consumption a frequency
pattern distribution
of production time his-
togram for an
interrupted byunpacking machine.
a tailback, lack, Figure 10failure
or equipment showssituation
the correlation between the histograms
for a bottle-cleaning ma-
chine2D
and (Machine A).for
line plots Thethe
effective electrical
unpacker. The power valuevalues
alternating changesfor
with
thethe operational
effective state power
electrical
changesthe
during of the machine. state resulted in a flat and extended peak.
operational

Figure 8. Bottle-cleaning machine A. Details of a typical consumption pattern and related opera-
tional states (left) and a histogram of the measured values for the operating time (here: a time period
of 14 days, right in the figure), adapted from [32], Isabel Osterroth, 2017.

The two peaks in the histogram (Figures 7 and 8, right) fit the value of the almost
constant performance level that the machine reaches after a state change. Different types
of peaks related to the running times (see the time model in Figure 1) were found for the
machines. The peaks of the machines with a steady operational behavior (bottle filling
machines, bottle-cleaning machines, thermoforming machines, etc.) were high and nar-
row, whereas they were flat and extended for cycling and intermittently operating ma-
chines (packer, palletizer). Figure 9 shows this on the example of a frequency distribution
Figure
histogram
Figure for
8. Bottle-cleaning
8. Bottle-cleaning
an unpackingmachinemachine
machine. A.A.Figure
Detailsof
Details 10ofashows
atypical
typical
theconsumption pattern
correlationpattern
consumption between and
andthe related
histo-
related operational
opera-
states
grams and(left)
tional states andplots
(left)
2D line a histogram
and a histogramof the measured
of
for the unpacker. the measured values for
values
The alternating for the operating
the
valuesoperating time (here:
time (here:
for the effective a a time period of
time
electricalperiod
of 14
power
14 days,
during
days, right
the in
right inthe
thefigure),
figure),
operational adapted
state from
resulted
adapted fromin[32], Isabel
a flat
[32], andOsterroth,
Isabel extended
Osterroth,2017.
peak.
2017.

The two peaks in the histogram (Figures 7 and 8, right) fit the value of the almost
constant performance level that the machine reaches after a state change. Different types
(1)running times (see the time model in Figure 1) were found for the
of peaks related to the
machines. The peaks of the machines with a steady operational behavior (bottle filling
machines, bottle-cleaning machines, thermoforming machines, etc.) were high and nar-
row, whereas they were flat and extended for cycling and intermittently operating ma-
chines (packer, palletizer). Figure 9 shows this on the example of a frequency distribution
histogram for an unpacking (3)shows the correlation between the histo-
machine. Figure 10
(2)
grams and 2D line plots for the unpacker. The alternating values for the effective electrical
power during the operational state resulted in a flat and extended peak.

(1)

Figure 9. Histogram of the effective electrical power data of the intermittently operating machine
Figure 9. Histogram of the effective electrical power data of the intermittently operating machine
(unpacker) over a period of 14 days: three main peaks with one flat and extended peak (Peak 3)
(unpacker) over a period of 14 days: three main peaks with one flat and extended peak (Peak 3)
between 1 and 4 kW.
between 1 and 4 kW. (3)
(2)

Figure 9. Histogram of the effective electrical power data of the intermittently operating machine
(unpacker) over a period of 14 days: three main peaks with one flat and extended peak (Peak 3)
between 1 and 4 kW.

Figure 10. Unpacker A. Details of a typical consumption pattern and related operational states (left),
and a histogram of the measured values for the operating time (right).

“Lack”, “tailback”, “equipment failure”, and “held” show similar values in the con-
sumption pattern. This could be verified in a more detailed analysis of the consumption
 Figure 10. Unpacker A. Details of a typical consumption pattern and related operational states (left),
and a histogram of the measured values for the operating time (right).
Sustainability 2021, 13, 9880 15 of 19
“Lack”, “tailback”, “equipment failure”, and “held” show similar values in the con-
sumption pattern. This could be verified in a more detailed analysis of the consumption
behavior during the single states. Figure 11 shows the average values of the effective
behavior
power during
of single the single
states for anstates.
example Figure
time11 showsofthe
period 4 h.average values ofdifference
No significant the effective power
between
of single states for an example time period of 4 h. No significant
the total consumption values of “lack” and “tailback” was found. The occurring equip-difference between the
total consumption values of “lack” and “tailback” was found. The
ment failures were shorter than any lack and tailback situation in this time period. Asoccurring equipment
failures were
illustrated shorter8,than
in Figure the any
valuelack
of and tailback
effective situation
power in thisdepending
decreases time period. onAs
theillustrated
time of
in Figure 8, the value of effective power decreases depending on
the chance until a constant value is reached. Owing to the short duration of the statethe time of theequip-
chance
untilfailure,
ment a constant valueconstant
the lower is reached.
valueOwing to the power
of effective short duration of the state
was not reached. Thisequipment
resulted
failure, the lower constant value of effective power was not reached.
in a higher average effective power value. Any differences were due to the fact that equip-This resulted in a
higher average effective power value. Any
ment failure is shorter than lack/tailback events. differences were due to the fact that equipment
failure is shorter than lack/tailback events.

Figure11.
Figure Averageeffective
11.Average effectivepower
power(kW)
(kW)andandstandard
standarddeviation
deviation(kW)
(kW)for foroccurring
occurringoperational
operational
statesfor
states forbottle-cleaning
bottle-cleaningmachine
machineAAover
overananexample
exampletime
timeperiod
periodofoft =t =4 4h.h.

Forallallother
For otherinvestigated
investigatedmachines,
machines,except
exceptfor
forthe
themixer,
mixer,clearly
clearlydefinable
definablepeaks
peakswere
were
found. The number of peaks identified and the description of the peaks for all machines
found. The number of peaks identified and the description of the peaks for all machines
are summarized in Table 7. A minimum of two peaks were found for all machines. For
are summarized in Table 7. A minimum of two peaks were found for all machines. For
most machines, three main peaks were found.
most machines, three main peaks were found.
3.3.2. Consumption Behavior during State Transitions
Evaluating the frequency distributions of the packaging and bottling machines, it
should be noted that the observed peaks are not clearly distinguished, but rather they
merge into each other. The measured data shows some machine state changes taking
place immediately (e.g., packer). On the other hand, in fillers or bottle-cleaning machines,
the states change stepwise by state transitions according to the machine’s function (e.g.,
bringing the product out of the machine, stepwise switching of the pumps for technological
reasons; see Figure 12). State transitions are reproducible, as shown in the example in
Figure 12 for two machines and 20 state transitions (change from operating to equipment
failure). No significant correlation was found between the duration of a downtime (0–180 s)
and the energy consumption within the first 60 s of the following running time (see
Figure 13). The measurement methods have been shown to be unsuitable for short-term
peak loads resulting from the powering up.
Sustainability 2021, 13, 9880 16 of 19
Sustainability 2021, 13, x FOR PEER
Sustainability 2021, REVIEW
13, x FOR PEER REVIEW 17 of 21 17

Figure
Figure 12.
12. Reproducible
Figure 12. state
Reproducible transition
Reproducible
state from
transitionstate operating
fromtransition to
to equipment
operatingfrom operating failure
equipment to for
for nn == failure
equipment
failure 20
20 chances:
for n = 20 cha
chances:
average
average of the effective electrical energy, standard deviation, and maxima and minima (gray) of the
of the effective
average electrical
of the energy,
effective standard
electrical deviation,
energy, and
standard maxima
deviation,and minima
and maxima (gray)
and of
minima
the (gray)
measured values for a
measured bottle filling
values for machine
a bottle (right)
filling and a
machine bottle-cleaning
(right) and a machine (left).
bottle-cleaning
measured values for a bottle filling machine (right) and a bottle-cleaning machine (left). machine (left).

Figure 13. Average effective

Figure Figure 13. Average
consumption
13. Average (kWh)
effective effective consumption
over the(kWh)
consumption first 60over
s of the(kWh)
first 60over
production, of the
s as wellfirst 60 standard
as the
production, sas
ofwell
production,
deviation
as as(kWh)
welldeviation
the standard as the (kWh)
sorted by the duration
sorted byofthestandard
theduration
downtime deviation
of before (kWh) sorted
the measurement
the downtime by
before theof the duration of
a labeling machine.
measurement the downtime
No significant
of a labeling before the measurement
machine. correlation between
No significant of between
correlation
the duration ofthe
theduration a
stop andof labeling
the energy
the machine. No
consumption
stop and significant
the energywithin correlation
the firstwithin
consumption between
60 s of the
production the
first 60 swasduration
found.
of productionof the stop and
was found. the energy
consumption within the first 60 s of production was found.
Figure 14 shows Figurethe14influence
shows the of the machine
influence speed
of the on thespeed
machine energyonconsumption.
the energy consump
WhileFigure 14While
the machineshows themachine
speed
the influence of thewas
was reduced
speed bymachine
more speed
than
reduced on
60%
by morethe energy
in the
thanfirst
60%consumption.
step
in of
thestate While
firsttransi-
step of state tr
the machine
tion, the energyspeed was reduced
tion,consumption by more than
only reducedonly
the energy consumption 60%
by 17.5%.in the
reduced first
In by step
the17.5%. of state
secondInstep, transition,
the machine
the second the the mac
step,
energywas
speed consumption
further
speedreduced
wasonly reduced
to zero,
further by 17.5%.
but
reduced still In the
more
to zero, than
but second
60%
still morestep,
of the the
than machine
initial
60% of thespeed
energy was
was con-
initial energy was
furtherby
sumed reduced
thesumed to zero,
machine. butmachine.
The
by the still more
influence ofthan
The the 60% ofof
machine
influence thetheinitial
speed onenergy
machine the speedwasonconsumed
electrical energy by energy
con-
the electrical
the machine. The influence of the machine speed on the electrical energy consumption is
sumption is machine-specific, as it depends as
sumption is machine-specific, onitthe type ofon
depends consumer
the type within the machine
of consumer within the mac
machine-specific, as it depends on the type of consumer within the machine (e.g., electrical
(e.g., electrical(e.g.,
drives, pumps,
electrical or compressed
drives, pumps, orair), which is air),
compressed not necessarily
which is notcorrelated to correlate
necessarily
drives, pumps, or compressed air), which is not necessarily correlated to the machine speed.
the machine speed.
the machine speed.
 Sustainability2021,
Sustainability 13,x 9880
2021,13, FOR PEER REVIEW 1817ofof2119

Figure 14. Bottle filling machine 1. Reproducible machine speed and influence on the consumption
Figure 14. Bottle filling machine 1. Reproducible machine speed and influence on the consumption
behavior after state transitions from production to equipment failure (n = 20 state transitions). A
behavior after state transitions from production to equipment failure (n = 20 state transitions). A
machinespeed
machine speedreduction −64%
reductionofof−64% results
results in in a reduction
a reduction of of effective
effective electrical
electrical power
power ofof −17.5%.
−17.5%.
3.3.3. Discussion of the Operational-State-Related Energy Consumption Behavior
3.3.3. Discussion of the Operational-State-Related Energy Consumption Behavior
The measured effective electrical energy data proves a correlation between the con-
The
sumption measured
behavior effective
and theelectrical
machine’s energy data proves
operating state. An a correlation between the con-
operational-state-related ap-
sumption behavior and the machine’s operating state. An operational-state-related
proach to modeling suits all machines well, except for mixers. A mixer, showing defined ap-
proach
peaks to modeling
in the frequency suitsdistribution
all machines notwell, except
directly for mixers.
matching A mixer, showing
the operational states, is defined
a process
peaks in the frequency distribution not directly matching
machine influenced by the upstream production process and cannot be described the operational states, is abypro-the
cess machine influenced by the upstream production process
same models as processing packaging and bottling machines. All other machines analyzed and cannot be described by
the same models
showcased as processing
a direct number ofpackaging
peaks in the andfrequency
bottling machines.
distribution Allofother machines an-
the measured con-
alyzed
sumption showcased
values, aasdirect
shown, number
e.g., inofFigures
peaks in the 9frequency
7 and distribution
and summarized of the
in Table 7. measured
Correlating
consumption
the frequency values, as shown,
distributions e.g.,the
with in Figures
colored 72D and 9 and
line plots summarized
showed an in Table demand
energy 7. Cor-
relating the frequency distributions with the colored 2D line
level matching the changing operational states. Some machines show multimodal peaks, plots showed an energy de-
mand level matching the changing operational states. Some
which can be related to single states (mainly to downtimes). The multimodal frequency machines show multimodal
peaks, which can
distribution peak beindicates
related toa single states (mainly
machine-specific to downtimes).
shutdown behavior.The In multimodal
order to create fre- a
quency
detailed distribution
description peak indicates
of the machine a machine-specific
consumption behavior, shutdown behavior.
state changes Ininorder to cre-
accordance
ate a detailed
with the elapsed description
time canofbethe machine
utilized. consumption
State changes takebehavior, statetransitions,
place as state changes in as accord-
shown
ance with the
in Figures 12elapsed
and 13.time Theycancan bebeutilized.
described State aschanges take place
reproducible basedason state
thetransitions,
analyzed data. as
The time after a state change also influences the level of effective
shown in Figures 12 and 13. They can be described as reproducible based on the analyzed electrical power and can
be used
data. The to describe
time after athe behavior
state changeduring state changes
also influences in models.
the level For extended
of effective electricalmodeling
power
(e.g.,
and can forbesimulation cases), the
used to describe theduration
behaviorofduringtransitions
state can be determined
changes in models.byFor data analysis
extended
or from the
modeling PLC.
(e.g., for simulation cases), the duration of transitions can be determined by data
analysis The ormeasurements
from the PLC. show the operational state being a major influence on the energy
consumption
The measurementsof packaging show andthebottling machines
operational state in contrast
being to the
a major machine
influence onspeed. While
the energy
machine speed is directly correlated to the operational state
consumption of packaging and bottling machines in contrast to the machine speed. While but seems to be not suitable
for the modeling
machine of the consumption
speed is directly correlated tobehavior (shownstate
the operational in Figure 14), the
but seems tooperational
be not suitablestate
related
for consumption
the modeling of thelevels are assumed
consumption as suitable
behavior (shownto indescribe
Figure 14),thetheenergy consumption
operational state
behavior
related of bottlinglevels
consumption plants.areThe numberasofsuitable
assumed peaks was found asthe
to describe reproducible, regardless
energy consumption
of what of
behavior product
bottling was produced;
plants. The numberhowever, the position
of peaks was found of the
as peaks on the x-axis
reproducible, shifts
regardless
(average consumption in kilowatt). Therefore, the product
of what product was produced; however, the position of the peaks on the x-axis shifts and process conditions do
(average consumption in kilowatt). Therefore, the product and process conditions do notIn
not influence the state dependency in general, but only the value of the energy level.
order to the
influence characterize the energy
state dependency in consumption
general, but only behavior, the system
the value boundaries,
of the energy level. including
In order
the nominal speed of the machine, have to be defined,
to characterize the energy consumption behavior, the system boundaries, including as well as the reference product
the
and process.
nominal speed of the machine, have to be defined, as well as the reference product and
process.
4. Summary and Outlook
PET and returnable glass bottling plants both have a main consumer significantly
influencing the total consumption (bottle-cleaning machines and stretch-blow molders).
Sustainability 2021, 13, 9880 18 of 19

Based on the results, it can be said that the energy consumption of food-packaging and
bottling machines is described by a limited discrete number of energy states related to the
operational state. The product type produced by the machine might influence the discrete
value of the described energy states and can be considered as a parameter in modelling
approaches. The observed energy states can be mapped to common models describing
the operational state behavior and are comparable for all the considered machines. State
changes can take place immediately or in time-dependent state transitions, resulting in
a constant value after a certain time. The machine speed is not directly correlated to
the energy consumption and therefore is not suitable for the energy modeling of the
machines. The basic concept of operational-state-related energy consumption can be used
for future model-based forecasting of the electrical energy consumption of food-packaging
and beverage-bottling machines. The consumption values during non-productive times
are still high and should be analyzed in more detail for optimization purposes in future
research. Potential technical and technological changes, as well as changes in the plant
automation, might lead to reduced energy demand during down times, supporting the
ambitious industry targets on sustainability. For a state-based modeling and simulation
approach, it is assumed that, for most machines, a model can be simplified to three main
consumption states (inactive, standby, and production). The described energy demand
level can be regarded as nearly constant. Furthermore, a state-related model can be used
for cause-related energy analyses (e.g., to find energy optimization potentials related to
downtimes) and future specifications of machines (e.g., decision criteria for improved total
cost of ownership considerations). Future research can focus on operational-state-related
modeling and forecasting in order to develop complex optimization strategies taking into
account both the operational state behavior of the interlinked plants and the specific energy
behavior of the single machines in the system.

Author Contributions: Conceptualization, I.A.O. and T.V.; methodology, I.A.O.; formal analysis,
I.A.O.; investigation, I.A.O.; data curation, I.A.O.; validation, I.A.O.; writing—original draft prepara-
tion, I.A.O.; writing—review and editing, T.V.; visualization, I.A.O.; supervision, T.V.; project admin-
istration, I.A.O. and T.V. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Blüml, S.; Fischer, S. (Eds.) Handbuch der Fülltechnik. Grundlagen und Praxis für das Abfüllen flüssiger Produkte, 2nd ed.; Behr:
Hamburg, Germany, 2009.
2. Olajire, A.A. The brewing industry and environmental challenges. J. Clean. Prod. 2020, 256, 102817. [CrossRef]
3. United Nations. Department of Economic and Social Affairs: THE 17 GOALS. Available online: https://sdgs.un.org/ (accessed
on 18 July 2021).
4. Anheuser-Busch, I. Sustainability Goals. Available online: https://www.ab-inbev.com/sustainability/2025-sustainability-goals/
(accessed on 18 July 2021).
5. Heineken, N.V. Sustainability and Responsibility. Available online: https://www.theheinekencompany.com/sustainability-and-
responsibility (accessed on 18 July 2021).
6. The Coca Cola Company: Sustainable Business. Available online: https://www.coca-colacompany.com/sustainable-business
(accessed on 18 July 2021).
7. Manger, H.-J. Füllanlagen für Getränke. Ein Kompendium zur Reinigungs- Füll- und Verpackungstechnik für Einweg- und Mehrwegflaschen,
Dosen, Fässer und Kegs; VLB (VLB-Fachbücher): Berlin, Germany, 2008.
8. DIN 8784: 2013-09. Getränkeabfüllanlagen—Mindestangaben und Auftragsbezogene Angaben. [CrossRef]
9. Canadian Industry Program for Energy Conservation. Guide to Energy Efficiency Opportunities in the Canadian Brewing
Industry. In Collaboration with the Brewers Association of Canada, 2nd ed.; Canadian Industry Program for Energy Conservation:
Toronto, Canada, 2011; ISBN 978-1-100-20439-0.
10. Galitsky, C.; Martin, N.; Worrell, E.; Lehman, B. Energy Efficiency Improvement and Cost Saving Opportunities for Breweries: An Energy
Star(R) Guide for Energy and Plant Managers; Lawrence Berkeley National Lab. (LBNL): Berkeley, CA, USA, 2003. [CrossRef]
Sustainability 2021, 13, 9880 19 of 19

11. Meißner, S. Regionale Ressourcenvernetzung. Eine Studie am Beispiel einer bayerischen Mittelstandsbrauerei, 2nd ed.; Oekom Verlag:
München, Germany, 2003.
12. Heuven, F.; van Beek, T. Benchmarking der Energie- und Wassereffizienz im Brauereisektor. Brauwelt 2013, 29, 851–853.
13. Donoghue, C.; Jackson, G.; Koop, J.H.; Heuven, A.J.M. The Environmental Performance of the European Brewing Sector. 2012.
Available online: https://brewersofeurope.org/uploads/mycms-files/documents/archives/publications/2012/envi_report_
2012_web.pdf (accessed on 18 July 2021).
14. British Beer & Pub Association: The British Brewing Industry. Thirty Years of Environmental Improvement. 2006. Available online:
https://www.campdenbri.co.uk/sustainablebrewing/EnvironmentArticleBBPA.pdf (accessed on 18 July 2021).
15. Fiederer, E.; Guggeis, H.; Mathey, R.; Stoll, M. Praxisorientierte Ansätze für erfolgreiches Energiemanagement. In Betriebliches
Energiemanagement: Conference Cottbus 6./7.03.2001; VDI Verlag: Düsseldorf, Germany, 2001.
16. Petersen, H. Brauereianlagen. In Planung, Energieversorgung, Energiewirtschaft, Betriebstechnik, Kontrolle, Kennzahlen, 2nd ed.; Carl:
Nürnberg, Germany, 1993.
17. Sattler, P. Energiekennzahlen und -sparpotentiale für Brauereien. O.Ö. Energiesparverband; Wirtschaftskammer OÖ, Ökologische
Betriebsberatung. 2000. Available online: http://www.win.steiermark.at/cms/dokumente/11263981_52485923/5311a767
/Energiekennzahlen%20und%20Sparpotenziale%20in% (accessed on 15 October 2016).
18. Hauser, D.G.; Shellhammer, T.H. An Overview of Sustainability Challenges in Beer Production, and the Carbon Footprint of
Hops Production. MBAA TQ 2019, 56, 4.
19. Schreiner, E. Kraft-Wärme-Kopplung zur Energieversorgung von Flaschenfüllanlagen in Brauereien, Mineralbrunnen und
Erfrischungsgetränkebetrieben. Ph.D. Dissertation, Technische Universität Berlin, Berlin, Germany, 1982.
20. VDMA 8751:2019-03. Packaging machinery (incl. filling machinery)-Specification and measurement of energy and utility
consumption.
21. Osterroth, I.; Holm, A.; Voigt, T. State of the art survey of the energy and media demand of German beverage-bottling plants.
BrewingScience 2017, 70, 86–99. [CrossRef]
22. Dietmair, A.; Verl, A. A generic energy consumption model for decision making and energy efficiency optimization in manufac-
turing. Int. J. Sustain. Eng. 2009, 2, 123–133. [CrossRef]
23. Lees, M.; Ellen, R.; Brodie, P.; Steffens, M.; Newell, B.; Wilkey, D. A Utilities Consumption Model for Real-Time Load Identification
in a Brewery. In Proceedings of the 2009 IEEE International Conference on Industrial Technology, Churchill, VIC, Australia,
10–13 February 2009. [CrossRef]
24. Dash, P.; Satpathy, H.; Liew, A. A real time short-term peak and average load forecasting system using a self-organising fuzzy
neural network. Eng. Appl. Artif. Intell. 1998, 11, 307–316. [CrossRef]
25. Lertpalangsunti, N.; Chan, C. An architectural framework for the construction of hybrid intelligent forecasting sytems: Application
for electricity demand prediction. Eng. Appl. Artif. 1998, 11, 549–565. [CrossRef]
26. Xu, B.; Wang, Y.; Ji, Z. CBR based energy consumption model of cutting period in CNC lathe. In Proceedings of the 2016 35th
Chinese Control Conference (CCC), Chengdu, China, 27–29 July 2016; pp. 9691–9697.
27. Cataldo, A.; Taisch, M.; Stahl, B. Modeling, simulation and evaluation of energy consumptions for a manufacturing production
line. In Proceedings of the IECON 2013—39th Annual Conference of the IEEE Industrial Electronics Society, Vienna, Austria,
10–13 November 2013; pp. 7537–7542.
28. Kuhrke, B. Methoden zur Energie- und Medienbedarfsbewertung Spanender Werkzeugmaschinen. Ph.D. Thesis, Technische
Universität Darmstadt, Darmstadt, Germany, 2011. ISBN 978-3-8442-1461-1.
29. VDMA 34179:2015-03: Measurement instruction to determine the energy- and resource demand of machine tools for mass
production.
30. VDMA 34179:2019-04: Messvorschriften zur Bestimmung des Energie- und Medienbedarfs von Werkzeugmaschinen in der
Serienfertigung.
31. Braun, T.; Hurni, P.; Bernardo, P.; Curado, M. Issues with State-based Energy Consumption Modeling. In Proceedings of the
OMNeT++ Community Summit, Zurich, Switzerland, 3–4 September 2015.
32. Osterroth, I.; Klein, S.; Nophut, C.; Voigt, T. Operational-state-related modeling and simulation of the electrical power demand of
beverage-bottling plants. J. Clean. Prod. 2017, 162C, 587–600. [CrossRef]
33. ANSI/ISA-88.00.01-2010 Batch Control Part 1: Models and Terminology.
34. TR88.00.02-2015: Machine and Unit States: An implementation example of ANSI/ISA-88.00.01.
35. Weihenstephaner Standards. Available online: http://www.weihenstephaner-standards.de (accessed on 23 January 2021).
36. The Organization for Machine Automation and Control: PackML. Available online: https://www.omac.org/packml (accessed on
18 July 2021).
37. DIN 8743:2014-01. 2014: Packaging machines and packaging lines—Key figures to characterise operation behaviour and
requirements for data collection in an acceptance test.