Quality Control Monitoring by using Artificial Neural Networks for an
Iberian Ham Industry
J. A Garc
, Bel
en Curto
, Vidal Moreno
and Beatriz Mart
Department of Computer Science and Automatics, University of Salamanca, Faculty of Science, Salamanca, Spain
Technology Meat Station of Castilla y Le
on, Guijuelo, Spain
Monitoring of Food Quality Control, Food Sensory Predictions, Food Machine Learning Estimation.
The iberian ham is a high valued product, due to this fact, it is very important to offer to the costumer a high
quality food product and to ensure its organoleptic properties. Producers have to evaluate, periodically, its
sensorial attributes by a professional tasting panel. Due to high elevated organizational and economics costs,
in addition to, the sensory fatigue and the subjectivity of the panel members, only a few product lots are sam-
In this paper is proposed a cloud manufacturing based platform to monitor the quality of Iberian ham. The
success of this solution is based on cooperation and data exchange between the main agents involved in the
process: quality manager, professional tasters, production manager, inspection authorities, etc. Intelligent al-
gorithms have been embedded into the cloud monitoring platform to predict the ham sensory properties, using
the Near InfraRed Spectroscopy data from the product samples as input.
The key feature of the solution is that the sensory analysis is performed without gathering routinely a professi-
onal tasting panel, but the solution also allows to the quality manager, with advanced visualization techniques,
to monitor what is the merit figure related with a specific type of ham or shoulder. Another important aspect
of the solution is that, due to the huge amount of data coming from the elaboration process itself are available
is possible to fine-tune continuously the machine-learning algorithms to the particular producer and use them
intelligently to increase the competitiveness.
Today, in the food industry the quality control based
on sensory tasting is restricted to the availability of
expert tasters or the consumption of the product by the
clients (Siegrist and Cousin, 2009). Both the custo-
mer and the taster, value the organoleptic characteris-
tics of the product such as the odour, flavour, colour,
texture, etc (Murray et al., 2001). Until now, tasting
has been limited to a restricted set of samples to eva-
luate complete vintages and also advertising samples.
Therefore, the final consumer has not the guarantee
that the product which he has acquired has the same
sensory characteristics as the product which has been
submitted to the sensory test.
Within the Iberian pork industry, ham and shoul-
der are products with high added value (BOE, 2014)
and with well known organoleptic characteristics.
Therefore, the producers of this sector need to fully
ensure the quality of his product to the final consu-
mer. However, to accomplish sensory measures is
very labor-intensive and involves a high economic
cost and time, and in many cases the measures can
be subjective due to the tasters state (tiredness, sen-
sory fatigue, etc). Unfortunately, this means that no
all lots can be assessed with this evaluation procedure.
Therefore, ham industry needs a reliable estimation of
the all sensory parameters of a larger set of lots, that,
perfectly, could reach the totality.
In this regard, the PAT (Process Analytical
Technology) allows to analyse and control, during
the manufacturing process, the quality and safety of
industrial products in order to assure that the final
obtained product fulfills with the quality standards.
PAT allows to increase the efficiency in the on-line
quality control process and during the manufactu-
ring process. In addition, it allows continuous sy-
stem learning, which will facilitate continuous impro-
vement (Grassi and Alamprese, 2017; Rathore and
Kapoor, 2017; Sommeregger et al., 2017). Applying
PAT in food industry along with the Near InfraRed
(NIR) technology is being used recently to monitor
the critical process parameters and the quality attri-
butes (Grassi and Alamprese, 2017). In this way, the
food industry can move from a rigid quality to a flexi-
ble quality (Sommeregger et al., 2017), achieving a
García-Esteban J., Curto B., Moreno V. and Martínez B.
Quality Control Monitoring by using Artificial Neural Networks for an Iberian Ham Industry.
DOI: 10.5220/0006911506280635
In Proceedings of the 15th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2018), pages 628-635
ISBN: 978-989-758-321-6
2018 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
transformation towards the fourth industrial revolu-
tion or industry 4.0 (Oio et al., 2017; Efendioglu and
Woitsch, 2017; Ruohomaa et al., 2018).
In this sense, “Cloud Manufacturing” (CM) is an
essential technology to achieve the monitoring tools
for food companies. CM supports in transforming the
industry from a traditional business model to a more
collaborative, distributed and global business model.
More specifically, CM will allow cooperation and ma-
nufacturing activities at different places, departments
or organizations (Mourtzis and Vlachou, 2016). Con-
secuently, the industry will be more efficient and com-
petitive and products quality improves, which implies
an increase of the consumer satisfaction and, there-
fore, of the sales (Oio et al., 2017). By the hand of the
CM, “Big Data” (BD) appears as a technology which
allows to analyse the large amount of data that a com-
pany can have and use it intelligently to increase its
global competitiveness. Proper use of BD and mat-
hematical models can help notably to the efficiency
increment in the productivity and quality. The appli-
cation of advanced sensor systems and PAT paradigm
in combination with mathematical modelling techni-
ques of BD offer enhanced process understanding, al-
low on-line prediction of critical quality attributes and
subsequently real-time product quality control (Som-
meregger et al., 2017).
In this paper is presented ICatador, a software
platform based on Cloud Manufacturing with which is
possible to monitor the quality control process of Ibe-
rian ham and shoulders. The platform allows, on the
one hand, to keep the traditional control methods (tas-
ters panel) increasing the efficiency and allowing the
cooperation and the information exchange in a distri-
buted way. As an instrumental measurement, is used
Near InfraRed Spectroscopy (NIRS) on the samples,
which is based on the electromagnetic radiation ab-
sorption in the band from 780 to 2500 nm. NIRS is
well-known in material sensing, and its major benefit
is that it doesn’t need any sample preparation, furt-
hermore, it can yield a response on-line. Applications
of NIRS can be found in medical and biomedical stu-
dies, food science, forestry, and the pharmaceutical
and petroleum industries (Balabin and Safieva, 2011).
On the other hand, the platform incorporates Artificial
Neural Networks (ANNs), as a computational intelli-
gent technique (Big Data), to reproduce/predict the
tasters valuations about the food sensory attributes.
ANNs are computational techniques perfectly adap-
ted to discover non-linear trends between variables
(Boccorh and Paterson, 2002; Cancilla et al., 2014),
as is in our case the Near InfraRed spectrum and the
sensory attributes.
The design goal in our strategy of the quality control
monitoring is to transfer the estimation of food or-
ganolectic properties from a tasting room to the pro-
duction chain and to tend towards at-line quality mo-
Figure 1: Conceptual scheme of the monitoring of the tas-
ting process.
The core of quality control monitoring is a set
of algorithms based on machine learning techniques
using a supervised learning set of functions that will
allow to predict the valuations of a professional taster,
having as input data a measurement provided by an
instrument, such as can be a NIR spectrometer (Fig.
1). The ultimate aim is not to replace the tasters pa-
nel, but to accelerate the tasting process by means of
alternative and objective valuation sources.
In our monitoring strategy, materialized in a cloud
platform, different actors have been defined, each
with a different role. All of them collaborate in the
quality control process of the food product, from dif-
ferent places of the company or the other companies.
The monitoring strategy is supported in a cloud plat-
form, which has been called ICatador, whereby the
collaborative process is established.
One of the main actors is the professional taster
(Fig. 1). The tasters will value the product following
a traditional methodology. They will introduce their
Quality Control Monitoring by using Artificial Neural Networks for an Iberian Ham Industry
scores directly through mobile devices, laptop or de-
sktop computers. Initially, this information provided
to the system will be used to train the prediction algo-
rithm (Fig. 1). In successive phases this information
will be used to optimize the algorithms of supervised
The organizer of taste panel will define the sen-
sory profile according to the product type. The sen-
sory profile contains the organoleptic attributes to
be evaluated by professional tasters, such as flavour,
odour, texture, etc. Different profiles can be defined
for each product type and, even, several profiles for
the same product type, if the quality standards are
modified. Automatically, tasters will have, through
the platform, the sensory attributes to be scored. The
procedure, the related legislation, the description of
the attributes, and in general any documentation can
be included by the organizer, and used as help by the
The quality inspection technician (Fig. 1) will be
the responsible of making the instrumental measures
of each ham sample. This task can be carried out
manually in the laboratory or at the line production
through a portable instrument and, in the best case,
automatically at line. Configuring the particular type
of product, the instrumental measures will be regis-
tered in the platform. Automatically, the sensory at-
tributes of this product will be predicted through the
obtained functions via supervised learning. These
functions, one for each attribute, are embedded in the
cloud platform.
Another actor involved in the monitoring strategy
is the responsible of supervising the products quality
from the tasting data (Fig. 1). The cloud platform ma-
kes easier the work of this user, because it allows him
to add, delete or modify products, and it is possible
to update the complete information concerning them,
such as the feeding of the animal of provenance, the
quality and the origin of this animals, date of ripening
start, ... The most remarkable feature of plataform is
that quality manager can check and visualize imme-
diately (Fig. 1), the values predicted by the machine
learning algorithm (artificial tasting) and the valuati-
ons that tasters make over the time. He can compare
the human scores with predictions, so that it can feed-
back the ratings and tune prediction functions to cor-
rect deviations.
To perform a predictive evaluation of the Iberian ham
and shoulder organoleptic characteristics, such as vi-
sual appearance of lean, fat streaks, rancidity of fat,
characteristic odour, texture and flavour, is needed
sensory and instrumental measures of food samples.
The Iberian ham and shoulders with known racial
percentages and different feeding varying from acorn,
meadow feed and normal feed, were elaborated by the
partners of Guijuelo Protected Designation of Origin
(Spain). The sensory and instrumental tests have been
performed on 62 ham and shoulder samples, elabora-
ted by different partners. At the end of the ripening
process, tests were made for both products in labora-
tory (NIR spectroscopy) and sensory room.
In our work, the sensory evaluations were perfor-
med by a panel of 4 members trained in the use of
the QDA (Quantitative Descriptive Analysis) metho-
dology. The tasting panel was trained at the ITACYL
Meat Technology Station. As instrumental measu-
res have been used the data of NIR spectroscopy of
the ham and shoulder samples, which is based on the
electromagnetic radiation absorption in the band from
780 to 2500 nm. The NIRS measurements were obtai-
ned in the Analytical Chemistry laboratory of the Uni-
versity of Salamanca.
3.1 Sensory Measures
The sensory profile of the ham and shoulder were
carried out by a panel formed by 4 members trained
in the use of the QDA methodology (Murray et al.,
2001), which provides an objective description of the
products in terms of perceived sensory attributes.
Panel members were trained (ISO, 1994; ISO,
2003) on the sensory profile of ham during 18 ses-
sions. During the training was encouraged that the
group developed a common vocabulary for the evalu-
ation of the sensory characteristics, which contained
simple and specific terms that made easier the des-
cription of products. In the QDA methodology, re-
ference scales are used to evaluate the texture inten-
sity (Piggot and Mowat, 1991), olfactory-taste para-
meters, and also some foods are proposed as standards
to stablish the scale (B
arcenas et al., 2003). Specifi-
cally, the organiser had set the sensory profile with
23 sensory attributes, framed within of: visual ap-
pearance of lean (intensity, homogeneity and bright-
ness), fat (intensity of colour, brightness, untosity, in-
termuscular quantity), fat streaks (quantity, homoge-
neity, thickness and uniformity), characteristic odour,
ICINCO 2018 - 15th International Conference on Informatics in Control, Automation and Robotics
texture (stickiness, hardness, crumbling, fibrousness,
pastiness and juiciness), flavour (intensity of sweet-
ness and saltiness), characteristic flavour (intensity
and persistence) and, finally, rancidity of fat. Du-
ring the panel training, the evaluators along with the
organizer agreed the established reference standards,
the terminology definitions and the evaluation techni-
ques. For quantification the intensity of each attribute,
6 point scales are used, where “0” is the lack of para-
meter, “1” is the minimal intensity and “5” is the max-
imum intensity for each of the parameters (B
et al., 2003).
In the product evaluation phase, the hams and
shoulders were scored by panelists according to the
description list defined during the training. The sen-
sory measures on which our work is based correspond
to different tasting sessions of the ham and shoulders.
This sensory information constitutes the training
data and the neural networks validation.
3.2 Instrumental Measures
Spectroscopic sensors are optimal instruments for real
time analysis during manufacturing, being rapid, non-
invasive, very flexible, and rugged. NIRS, in particu-
lar, with its ability to fingerprint food materials and to
simultaneously analyze different phenomena, is one
of the predominant e-sensing technologies used in
PAT (Grassi and Alamprese, 2017). The NIR techno-
logy is based on the electromagnetic radiation absorp-
tion in the band from 780 to 2500 nm and provides a
spectrum represented as values of log
where R
is the reflectance against the wavelength. The NIR
spectra were obtained with an analyser Foss NIR 5000
in the band of 1100-2000 nm with a spectral reso-
lution of 2 nm. The recording of NIR spectra (Fig.
2) was performed applying the carbon fiber probe di-
rectly on the ham or shoulder sample, at room tem-
perature (20-23
C). Previously to each record, the
probe window was cleaned to minimize the cross-
The recorded information constitutes the instru-
mental measures which are used for neural network
Figure 2: NIR spectra corresponding to a ham sample.
off-line training and validation. When working on-
line, the NIR records of each ham or shoulder sample
constitute the instrumental measures for the sensory
attributes prediction through ANN. The quality ma-
nager can observe the NIR spectra of each sample to
detect possible errors.
The mathematical model used to estimate the sensory
characteristics of the ham and shoulder samples is
based on ANNs, which are framed within no-lineal
(Della-Lucia and Minim, 2010) statistical data mo-
deling tools. Specifically, a Multi-Layer Perceptron
(MLP) ANN was used, where the processing ele-
ments (PE) are structured in three layers: an input
layer where the instrumental measures will be ente-
red, an intermediate or hidden and an output layer
where the sensory attribute to be evaluated is obtai-
The software application JavaNNS (Java Neural
Network Simulator) has been used as design, trai-
ning and validation tool of the ANN. The ANN mo-
del used is implemented within the application as a
“multi layer perceptron network”. After several tri-
als, the network model is constituted as follows:
An input layer with three processing elements
(PE), whose inputs are the principal components
(PCA) of the NIR spectra. The NIR spectra that
has been recorded ranges from 1100 to 2000 nm
with a spectral resolution of 2 nm. With 3 values
of Principal Components from NIR spectrum, the
99.98% of the spectral variability is expressed. In
this way the spectral information is compressed
and the number of input values is reduced from
451 to 3.
A hidden layer with 5 PE
A output layer with 1 PE which corresponds to the
modelled sensory characteristic.
For each sensory characteristic, a network, with
the previously topology commented, is built. These
ANNs were trained using the 80% of the dataset,
randomly selected for each ANN. Once each ANN
was trained, its accuracy were evaluated on the test
dataset formed by the remaining 20%. These data
were not shown during the training and were selected
randomly for each ANN.
Quality Control Monitoring by using Artificial Neural Networks for an Iberian Ham Industry
ICatador system is developed to be used in cloud from
any computing device (PCs, tablets, smartphones...)
so that it can be accessible from any place and any
ICatador has a MVC (Model-View-Controller) ar-
chitecture. Views are developed in HTML, the con-
trollers in Javascript using the open source frame-
work AngularJS. Finally, data model is developed in
a MySQL database. Model has functionalities which
allow selecting automatically the sensory profile to
which a product belongs, making that this fact fully
transparent to users who only have to introduce or
consult data. For the communication between the da-
tabase and the controllers is used a DAO (Data Access
Object) pattern, which is developed in PHP, making
easier maintenance works.
To make easier the ICatador adaptation to new
sensory profiles or to changes in existing ones, a tem-
plate system has been implemented which will allow
to add, modify or delete sensory profiles according to
the needs of the users. All of it is transparent for them
with a dynamic adaptation.
Graphics and visual representations have been de-
veloped using the D3.js library which allows the
graphic representation from the data. Therefore,
thanks to this library, the system can create different
graphical visualizations (histograms, parallel coordi-
nates, radars) to represent the sensory attributes infor-
mation, either from human or artificial scores.
As main result, ICatador is the cloud platform that
can be used by ham producers and quality regulatory
agencies for on-line quality control of their products,
following the PAT guidelines to monitoring the pro-
duct sensory parameters. The quality manager, tas-
ters, tasting panel organizer and the production line
inspection technician collaborate (Fig. 1), from dif-
ferent places, on the quality control tasks. The con-
tributions of each agent are combined to increase the
efficiency and to fulfil demands of the food industry.
In ICatador, the quality manager introduces the
products (Fig. 3) which will be evaluated, with all
the data concerning them (racial percentages, animal
feeding, elaboration date, ...).
The organizer of the panel draws up the sensory
Figure 3: New product entry in ICatador.
Figure 4: “Odour/Flavour” and “Texture” of sensory profile
used by the tasters.
profile (Fig. 4) with the attributes (fully configura-
ble), identifies the products to be tasted, selects the
tasters according to their characteristics stored in the
Database, chooses the help regulation and documen-
tation and, once the tasting is done, scores the tas-
ters. Through forms (Fig. 4), the tasting panel mem-
bers will enter their evaluations directly into the plat-
form, once they are identified, using mobile devices
such as tablets or smartphones. By the product iden-
tifier, ICatador recovers the sensory profile which the
product belongs. Automatically, the sensory attribu-
tes, that tasters have to enter, will be showed together
with the working procedure and the help descriptions.
ICINCO 2018 - 15th International Conference on Informatics in Control, Automation and Robotics
The quality manager can visualize the attribute sco-
res completed by tasters, individually or on average,
through graphs as a histogram (Fig. 5).
Figure 5: Histogram of the panel scores of a determinate
The spectral analyser generates a file with the
data of the NIR spectroscopy of the sample. The in-
spection technician, in production, incorporates the
file in the ICatador Database, selecting the date and
the product identification. Actually, in our monito-
ring platform this task is performed manually from
the laboratory, or from any place of the manufacturing
plant, but, thanks to actual technological innovation,
this information can be incorporated automatically.
The quality manager can observe the NIR spectra of
each sample (Fig. 2) to detect possible errors.
After the training and validation of each neural
network (which is performed off-line) with Deep-
Learning techniques, the quality manager, in his daily
work, can use ICatador to predict the sensory attribu-
tes of a product. An artificial tasting (tasting based
on ANN) is extraordinarily simple and it can be per-
formed anywhere in which a NIR analyser is availa-
ble. The starting point is the NIR record that is per-
formed of a product sample and, then it is enter to
ICatador through a file. After its loading, when the
option Artificial tasting” is clicked, the platform ICa-
tador calculates the organoleptic attribute predictions
which have been configured for this product by the
quality manager.
Using advanced visualization techniques, the re-
sults, grouped by categories, are displayed through
Figure 6: Artificial tasting (discontinuous line) of the attri-
butes of “texture” against the average of professional tasters
(continuous line).
Figure 7: Artificial tasting (discontinuous line) of the attri-
butes of “Odour/Flavour” against the average of professio-
nal tasters (continuous line).
graphs of parallel coordinates (Fig. 7) or radars. The
quality manager, at a glance, can have the “merit fi-
gure” of the considered sample (Fig. 6) to detect the
deviations from the expected quality. The merit figu-
res are grouped according to the attribute typologies.
The result will be stored in the Database for subse-
quent analysis for the system participants (quality ma-
nagers, production managers, etc.).
In food industry, the quality control operations based
on the sensory analysis are restricted to the availabi-
Quality Control Monitoring by using Artificial Neural Networks for an Iberian Ham Industry
lity of the experts, which implies relatively high eco-
nomic costs, a certain degree of subjectivity associa-
ted with the sensory fatigue, and, in addition of the
arduous organisational task. This restrictions impose
that the quality control are reduced to certain lots.
As an alternative approach, the machine learning
techniques along with instrumental measures appear
as an intelligent solution to obtain a reliable estimate
of the sensory parameters of the food products.
In this work, a monitoring tool supported on a
cloud platform, has been presented. The core of
the ICatador tool is a suite of intelligent algorithms
(ANNs) which calculate organoleptic attributes esti-
mates using NIR spectrometry data from samples as
To make possible that the ICatador monitoring
tool had all the ham and shoulders quality data avai-
lable, the main agents (quality manager, tasters, tas-
ting organizer, quality inspector) collaborate and ex-
change information from different points and, in ad-
dition, instrumental data are systematically incorpora-
ted. This collaborative model based on the data, will
allow to have a wide range of data coming from the
production process itself. In this way, the intelligent
algorithm suite can be tuned and adapted to the pro-
cess itself to avoid deviations.
Through the proposed approach, iberian ham ma-
nufacturing companies can carry out an intelligent
production based on the data, they can get that final
product gathers the same sensory characteristics fixed
in the production goals.
The Icatador platform that has been presented in
this paper combines the use of Information and Com-
munication Technologies with Artificial Intelligence
techniques such as ANNS. Therefore, it can be in-
cluded within the concept of Industry 4.0, and contri-
butes to the digitization of the industry. Its goals are
to improve efficiency, quality control times, flexibility
and enables a distributed responsibilities within of the
industrial sector.
This work has been supported by the General Founda-
tion of the University of Salamanca through the Plan
TCUE 2015-2017, co-financed by European Regional
Development Fund (ERDF) and the Castilla - Le
(1994). ISO 11035: 1994:sensory analysis identification
and selection of descriptors for establishing a sensory
profile by a multidimensional approach.
(2003). ISO 4121: 2003: Sensory analysis – guidelines for
the use of quantitative response scales.
(2014). Real decreto 4/2014, de 10 de enero, por el que se
aprueba la norma de calidad para la carne, el jam
la paleta y la ca
na de lomo ib
Balabin, R. M. and Safieva, R. Z. (2011). Biodiesel classi-
fication by base stock type (vegetable oil) using near
infrared spectroscopy data. Analytica Chimica Acta,
Boccorh, R. K. and Paterson, A. (2002). An artificial neu-
ral network model for predicting flavour intensity in
blackcurrant concentrates. Food Quality and Prefe-
arcenas, P., Elortondo, F. P., and Albisu, M. (2003).
Sensory changes during ripening of raw ewes’ milk
cheese manufactured with and without the addition of
a starter culture. Food Science.
Cancilla, J. C., Wang, S. C., D
ıguez, P., Matute, G.,
Cancilla, J. D., Flynn, D., and Torrecilla, J. S. (2014).
Linking chemical parameters to sensory panel results
through neural networks to distinguish olive oil qua-
lity. Journal of Agricultural and Food Chemistry.
Della-Lucia, S. M. and Minim, L. A. (2010). Redes neurais
artificiais: fundamentos e aplicac¸
oes. In UFV, V. E.,
editor, An
alise sensorial: estudos com consumidores,
chapter 9, pages 258–279. Springer-Verlag New York,
Efendioglu, N. and Woitsch, R. (2017). A modelling met-
hod for digital service design and intellectual property
management towards Industry 4.0 caxman case. Inter-
national Conference on Serviceology.
Grassi, S. and Alamprese, C. (2017). Advances in NIR
spectroscopy applied to process analytical technology
in food industries. Current Opinion in Food Science.
Mourtzis, D. and Vlachou, E. (2016). Cloud-based cyber-
physical systems and quality of services. The TQM
Murray, J. M., Delahunty, C., and Baxter, I. A. (2001).
Descriptive sensory analysis: past, present and future.
Food Research International.
Oio, K.-B., Lee, V.-H., Tan, G. W.-H., Hew, T.-S., and Hew,
J.-J. (2017). Cloud computing in manufacturing: The
next industrial revolution in malaysia? Expert Systems
with Applications.
Piggot, J. R. and Mowat, R. G. (1991). Sensory aspects of
maturation of cheddar cheese by descriptive analysis.
Journal of Sensory Studies.
Rathore, A. S. and Kapoor, G. (2017). Implementation of
quality by design towards processing of food products.
Preparative Biochemistry and Biotechnology.
Ruohomaa, H., Kantola, J., and Salminen, V. (2018). Va-
lue network development in Industry 4.0 environment.
Advances in Intelligent Systems and Computing.
Siegrist, M. and Cousin, M.-E. (2009). Expectations influ-
ence sensory experience in a wine tasting. Appetite.
ICINCO 2018 - 15th International Conference on Informatics in Control, Automation and Robotics
Sommeregger, W., Sissolak, B., Kandra, K., von Stosch,
M., Mayer, M., and Striedner, G. (2017). Quality by
control: Towards model predictive control of mamma-
lian cell culture bioprocesses. Biotechnology Journal.
Quality Control Monitoring by using Artificial Neural Networks for an Iberian Ham Industry