Journal of Hygienic Engineering and Design

Original scientific paper

UDC 664/664.013.8

VIRTUAL CLEANING OF MACHINES AND EQUIPMENT IN FOOD INDUSTRY

Christian Gerhards1*, Maria Schramm1, Markus Kiesel1, Nicolai Beisheim1
1
Albstadt-Sigmaringen University of Applied Sciences,
Anton-Guenther-Str 51, 72488 Sigmaringen, Germany

*
e-mail: gerhardsc@hs-albsig.de

Abstract 1. Introduction
To ensure hygienic production of food products, To ensure hygienic production of food products, ma-
­machines and equipment used must be cleaned regular- chines and equipment used must be cleaned regularly.
ly. The cleaning process causes high costs for personnel, The cleaning process causes high costs for personnel,
cleaning agents, water, and energy. To lower these costs cleaning agents, water, and energy. To lower these
and raise the total availability of the plant, the cleaning costs and raise the total availability of the plant, the
process needs to be optimized. Ideally the cleanability cleaning process needs to be optimized (Gerhards et
can be assessed in a very early stage of construction. al., [7]). Different types of soil need to be considered
(Table 1). Ideally, the cleanability can be assessed in a
A test stand was developed for cleaning experiments.
very early stage of construction.
These are the base of the simulation. In the test stand,
easy handling is possible and individual parameters
Table 1. Different types of soil in food industry (Dürr &
(e.g. pressure and flow rate of the spray jet, angle and
Wildbrett [4])
distance to the object) can be varied. In addition, a
well-defined movement of the object through the Reaction with water Examples
spray jet is feasible. The simulation of the cleaning Dissolving
Salts, acids, lower molecular
process was realized as an extension of the VisualDeci- carbohydrates (e.g. sugar)
sionPlatform virtual reality software (ESI Software Ger- Higher molecular carbohydrates
Swelling
many GmbH). The additional source code was entered (e.g. starch), proteins
via the scripting module. By special algorithms the pa- Emulsifying Fats, oils
rameters obtained in the cleaning experiments can be Suspending Fibers, seeds
integrated into the simulation.
For simulation of the cleaning process the user stands Typical cleaning processes in food industry include man-
in front of a screen on which the apparatus is being ual cleaning using a high pressure water jet (Figure 1).
projected. The movements of the spray lance is detect-
ed and transferred to the computer system. By use of
further input devices the user may “move” during the
cleaning process. The result of the simulation will be
calculated after completion of the cleaning process and
shown visually to the operator. By graduation of color
shades the result of the cleaning is displayed. A dark
blue color represents a good cleaning result. Hence, a
light color indicates where cleaning is not sufficient yet.
The combination of cleaning experiments in a test
stand and simulation of cleaning processes in virtual
reality allows to assess the cleanability of equipment
even before construction.

Key words: Food industry, Cleaning, Virtual reality, Figure 1. Typical cleaning process in food industry
­Hygienic design. (Source: http://walter-cleaningsystems.de/)

3

Although this cleaning procedure is used quite
­frequently, there is little theoretical knowledge on the
effect of water jets on typical soils in food industry
(Schmidt and Cremmling, [10]). Furthermore it is hard
to tell in advance, if machines and equipment can be
cleaned efficiently at all. Therefore it is the objective of
our studies to simulate the cleaning process and assess
the cleanability in virtual reality, which allows the inter-
action of the user with a virtual machine or equipment
in food industry (Figure 2).
Figure 2. Simulation of a manual cleaning process
in the virtual reality lab
(Source: Albstadt-Sigmaringen University)
It is well known that the efficiency of cleaning process-
es depends on many parameters. The original Sinner’s
circle describes the effect of the parameters: tempera-
ture, chemistry, mechanical action, and time, which all
contribute to cleaning. If one of the factors is dimin-
ished (e.g. temperature), others (e.g. time) have to be
increased to achieve the same cleaning effect. The ex-
tended Sinner’s circle (Figure 3) takes further parame-
ters into account: the properties of soil (nature, condi-
tion, quantity), and the characteristics of the material
to be cleaned (form, surface roughness, material).
Figure 3. Extended Sinner’s circle for
cleaning processes (Dürr and Wildbrett [4])
4
Journal of Hygienic Engineering and Design
Various mathematical models have been proposed to
describe the effect of water jets on cleaning processes.
In this study a model using the Weibull distribution (Eq.
1) was chosen for simplicity sake. Originally, it was pro-
posed for describing time effects on cleaning a heat
exchanger (Dürr and Grasshof, [2]). The Weibull distri-
bution has also been used for describing the influenc-
ing parameters in spray cleaning of food processing
equipment (Mauermann et al., [9]). The model allows to
calculate the remaining soil (r) as a function of time (t)
with two parameters, namely the time constant T, and
the steepness parameter b (Equation 1). The cleaning
degree (s) is calculated by equation 2. The parameters
T and b need to be derived from cleaning experiments
by statistical analysis.
r(t) = e (Eq.1)
s(t) = 1− r(t)(Eq.2)
Figure 4 shows examples of the cleaning degree as a
function of time, calculated for various combination of
the parameters T and R (equivalent to b in Eq. 1).
As can be seen, the Weibull distribution is equivalent
to the exponential function, if R (or b, resp.) equals 1.
For smaller values of R, the curve has a steeper ascent
for small values of t, but levels out for higher values. As
the authors remark, this time dependency is typical for
cleaning fat containing soil. If R is greater than 1, a sig-
moid curve can be described, which is typical for many
cleaning processes where soaking matters.
The objective of our studies is a simulation of manual
cleaning processes using a high pressure jet stream.
Thus the interaction of the operator needs to be includ-
ed into the simulation process. This can be achieved
using tools of virtual reality.
Virtual Reality (VR) is a technology that refers to a
computer-generated artificial world. With the help of
three-dimensional visualization (acoustic and haptic
perception, optionally) it imparts the user a ­supposedly
Figure 4. Use of Weibull distribution to describe the
effect of time on the cleaning. T: Time constant;
R: steepness parameter (Dürr and Grasshoff [3])
Journal of Hygienic Engineering and Design

real environment. VR can be used to simulate ma- of the object through the spray jet is feasible. By alter-
chines, equipment, buildings as well as production ing the speed of the linear drive, a different resident
and operating processes. Changes in the geometry time of the object in the spray area can be achieved.
and other parameters such as e.g. motion parameters
The course of experiments is shown in Figure 7.
can be immediately implemented digitally. This does
not only save time in product development, but en- Test plates (200 x 100 mm, Henkel Lohnpoliertechnik
ables to detect product defects and to eliminate them GmbH, Neustadt-Glewe, Germany) made from stain-
before the product is built (Beisheim and Kiesel, [1]). less steel (material No.1.4404, AISI 316L, Ra 0.8 µm)
were soiled manually, using a 10% w/w solution of
The applications of virtual reality are many and com-
organic milk protein (Dr. Ritter Bio-Eiweiss-Konzen-
plex. They include:
trat 85, Allos Hof-Manufaktur GmbH, Bremen) which
•• Visualization and variant testing of machines and was applied by a wire-wound rod (type PA-2428,
equipment BYK-Gardner GmbH, Geretsried, Germany) with 200
•• Preparation of installation processes µm nominal height. The protein film was dried under
•• Investigations of ergonomics and maintenance defined conditions (40 0C, 50 % rH) in a drying chamber­
•• Conduction of operator training
•• Product documentation and presentations for sales
and marketing
For example, VR can be used to optimize a mainte-
nance process of machines and equipment, i.e. to
determine service times of the maintenance process
itself, to design and improve maintenance tools, and
recognize faulty design. Figure 5 shows a maintenance
study using a human model (avatar) in VR, working
with flexible components.
Similarly, the interaction between a person and a ma-
chine can be simulated in the cleaning process of ma-
chines and equipment in food industry.

Figure 6. Test stand for cleaning experiments

(Source: http://walter-cleaningsystems.de/)

Figure 5. Simulation of handling flexible

components by a human model
(Source: Albstadt-Sigmaringen University)

2. Materials and Methods

A test stand was developed for cleaning experiments
and supplied by Walter Geraetebau GmbH, Sachsen-
heim-Ochsenbach, Germany (Figure 6).
This test stand has been designed for easy handling.
All parameters (e.g. pressure and flow rate of the spray
jet, angle and distance to the object) can be varied in- Figure 7. Course of experiments and
dividually. In addition, a well-defined linear movement analysis of test plates (Kiesel, [8])

5

(type 125 SB/+10 JU, Weiss Umwelttechnik GmbH, Rei-
skirchen, Germany) for 4 hours. Preconditioned plates
(two plates each time) were fixed in the test stand and
the water jet was activated. Water pressure and tem-
perature were varied from 20 - 60 bar and 30 - 50 0C. Ex-
perimental parameters were checked before each test
run. The angle and the distance to the test plates were
kept constant at 90° and 150 mm. Then the plates were
moved through the water jet with variable speed from
0.1 to 1 m/s. This corresponds to a residence time of
the plates in the water jet from 1 to 0.1 seconds. After
the first run, the water jet was stopped and the plates
were moved to the initial position. A second run at the
same conditions followed after 5 seconds. Finally the
plates were taken out of the test stand, dried briefly,
and analyzed, using a UV-scanner (Kodak Image Sta-
tion 440CF). The integrated area of grey values detect-
ed by the scanner after cleaning was divided by the
integrated area before the experiment as a measure of
the remaining soil. Details of the analysis are given in
Gerhards et al., [5]. Finally, all data was processed using
the statistical program R (https://www.r-project.org/)
to obtain the parameters of the Weibull distribution.
The simulation of the cleaning process was realized as
a software extension of the VR software “VisualDeci-
sionPlatform” (IC.IDO, ESI GmbH Stuttgart, Germany).
The additional source code was entered via the script-
ing module of the software. By special algorithms the
parameters obtained in the cleaning experiments were
integrated into the simulation.
3. Results and Discussion
Some results of cleaning experiments in the test stand
are given in Figures 8 and 9.
As can be seen there is considerable scatter of data at
short contact times (t < 0.4 s). This is not surprising, as
it is known that protein films are not washed down
evenly due to high coherence of the film. It is rather
a stochastic process with protein being removed in
Figure 8. Influence of temperature (30, 40, and 50 0C)
on cleaning degree at various contact times.
Pressure: 40 bar (Schramm, [11])
6
Journal of Hygienic Engineering and Design
plaques (Gerhards and Schmid, [6]). Nevertheless, data
can be fitted well using the Weibull distribution with
the mean standard error (MSE) ranging from 0.01 to
0.03. At higher contact times (t ≥ 0.4 s) scatter of data
decreases and the effect of temperature and pressure
on the cleaning degree can be seen clearly. It is not sur-
prising that the cleaning degree is higher at elevated
temperatures (Figure 8). Due to increasing molecular
motility at higher temperatures, the diffusion rate of
water into the protein film is increased. This results in
a softening of the protein film which can be washed
off by the water jet more easily. In the same way, the
effect of elevated pressure (60 vs. 20 bar) is visible only
at higher contact times. If contact time is sufficient (t
> 0.6 s), the higher mechanical energy of the water jet
at higher pressure results in a higher cleaning degree
(Figure 9).
The effect of temperature on the parameters yielded
by the Weibull distribution is summarized in Table 2.
Table 2. Effect of temperature on parameters of the
Weibull distribution at a constant pressure of 40 bar
(Schramm, [11])
Temperature (0C) T (s) b
30 0.47 0.86
40 0.33 0.94
50 0.24 1.02
Values of the steepness parameter b are in the order
of 1 at various temperatures from 30 to 50 0C. Thus the
shape of the Weibull distribution is very much similar
to the exponential distribution. The influence of tem-
perature can be seen clearly at the values of the time
constant T (which is the intercept of the Weibull distri-
bution curve with the horizontal dotted line, indicat-
ing a cleaning degree of 63.2%), ranging from 0.24 s
(at 50 0C) to 0.47 s (at 30 0C). The effect of pressure on
the parameters yielded by the Weibull distribution is
summarized in Table 3.
Figure 9. Influence of pressure (20, 40, and 60 bar)
on the cleaning degree at various contact times.
Temperature: 40 0C (Schramm, [11])
Journal of Hygienic Engineering and Design

Table 3. Effect of pressure on parameters of the By use of a special input device the user “moves” during
Weibull distribution at a constant temperature of 40 0C the cleaning process, cleaning the machine from vari-
(Schramm, [11]). ous distances and angles. For this purpose a commer-
Pressure [bar] T [s] b cial dance pad was used as input device (Figure 12).
20 0.26 0.65 The result of the simulation is calculated after comple-
40 0.33 0.94 tion of the cleaning process and shown visually to the
60 0.26 1.08 operator. By graduation of color shades the result of
the cleaning is displayed visually on the components
The variation of the time constant T is small at different of the virtual machine. A dark blue color represents
pressures from 20 to 60 bar. In all experiments the time a good cleaning result. Hence, a light color indicates
constant T was between 0.26 and 0.32 s. However, the where cleaning is not sufficient yet.
effect of pressure can be noticed when considering the
steepness parameter b, which is 0.65 at 20 bar, 0.94 at
40 bar and 1.08 at 60 bar. Thus, at the lowest pressure
of 20 bar the fitted curve of the experimental data dif-
fers considerably from the exponential distribution. It
is the advantage of the Weibull distribution, to take
account of this fact, while use of the exponential dis-
tribution would result in a worse fit of the model with
experimental data.
For the simulation of a cleaning process, the user
stands in front of a screen on which the machine is
being projected. The movements of a modified spray
lance are transferred to the computer program by in- Figure 12. Use of a dance pad for moving within Virtual
frared detection (Figure 10). Reality. 1: turn left. 2: move forward. 3: turn right.
4: move left. 5: move right. 6: look up.
Then the collision of the virtual spray jet with parts of 7: move backward. 8: look down (Kiesel, [8])
the machine is detected by the VR software. Figure 11
shows a person cleaning virtually a meat processing
machine in the VR lab.
4. Conclusions
-  The combination of cleaning experiments in a test
stand and simulation of cleaning processes in virtual
reality allows to assess the cleanability of equipment
even before construction.
-  Further experiments are needed to take other effects
into account for the simulation: e.g. distance and angle
Figure 10. Modified spray lance for virtual cleaning of between the spray nozzle and the surface to be cleaned.
machines and equipment in food industry
(Source: Albstadt-Sigmaringen University)
- In the end, all data derived from experiments in the
test stand can be integrated in the simulation software.

5. References
[1] Beisheim N., Kiesel M. (2013). Production With Virtu-
al Machines and Plants. In: Stjepandić J., Rock G., Bil C.
(Eds.), Concurrent Engineering Approaches for Sustain-
able Product Development in a Multi-Disciplinary Envi-
ronment, Springer London, UK, pp. 803-812.
[2] Dürr H., Grasshoff A. (1999). Milk heat exchanger clean-
ing - modelling of deposit removal. Trans IchemE, Vol. 77
(Part C), pp. 114-118.
[3] Dürr, H., Grasshoff, A. (2001). A mathematical model of
cleaning processes. Tenside Surf. Det., 38, (3), pp. 148-157.
[4] Dürr, H., Wildbrett, G. (2006). Basic procedures for con-
Figure 11. Virtual cleaning of a meat processing tamination and cleaning (in German). In: Wildbrett G.
machine in the virtual reality lab (Ed.), Cleaning and disinfection in the food industry
(Source: Albstadt-Sigmaringen University) (2nd Ed.), Behr’s Verlag, Hamburg, Germany.

7
 Journal of Hygienic Engineering and Design

[5] Gerhards Ch., Kudermann T., Schramm M. and Schmid

A. (2014). Assessing the cleanability of stainless steel sur-
faces - effect of surface roughness and various parameters
on cleaning of protein based soils. Journal of Hygienic
Engineering and Design, Vol. 7, pp. 3-7.
[6] Gerhards Ch., Schmid A. (2013). Assessing the cleanabil-
ity of stainless steel surfaces - development of a testing
method for starch and protein based soils. Journal of Hy-
gienic Engineering and Design, Vol. 3, pp. 9-14.
[7] Gerhards Ch., Schramm M., Schmid A. (2015). Cleaning
stainless steel effectively. FleischWirtschaft International,
(1), pp. 57-61.
[8] Kiesel M. (2014). Implementation of an immersive simula-
tion and derivation of a test rig to collect relevant parame-
ters (in German). Master’s Thesis, Faculty of Engineering,
Albstadt-Sigmaringen University, Germany.
[9] Mauermann M., Köhler H., Eschenhagen U., Bellmann
C., Calvimontes A., Majschak J. P. (2011). Influencing pa-
rameters in spray cleaning of food processing equipment.
11th International Congress on engineering and food,
Food Process Engineering in a Changing World Pro-
ceedings, Athens, Greece.
<URL:http://www.icef11.org/content/papers/hdo/
HDO633.pdf. Accessed 12 May 2016.
[10] Schmidt U., Cremmling K. (1979). Process engineering of
cleaning and disinfection. I. Communication: the effect of
high-pressure cleaning on fat and protein contamination
(in German). FleischWirtschaft, 59, (9), pp. 1355-1358.
[11] Schramm M. (2014). Evaluation and statistical analysis
of cleaning tests with a high-pressure jet (in German).
Master’s Thesis, Faculty of Life Sciences. Albstadt-­
­
Sigmaringen University, Germany.