ARTIFICIAL

INTELLIGENCE IN
FOOD INDUSTRY
SUBMITTED TO  : SIR ANWAR 
SUBMITTED BY: GUL E KAINAT
15-ARID-4939
                                                                                                                                                                          
CONTENTS

• INTRODUCTION
• WHAT IS ARTIFICIAL INTELLIGENCE
• TECHNIQUES
• APPLICATION IN FOOD INDUSTRY
• USES IN FOOD INDUSTRY
• CONCLUSION
• REFERENCES
INTRODUCTION

• The term Artificial Intelligence (or basically AI) was first used in 1956 by John
McCarthy. 
• Early AI programs including Samuel's checker program in 1950.
• Study of how computer systems can stimulate intelligent processes such as
learning, reasoning and understanding information in context.
• Branch of computer science that is concerned with automation of intelligent
behaviour. (Luger and stubblefield , 1993).
APPLICATIONS

•     Medicine

• Gaming • Finance
• Robotics • Telecommunication
• Antivirus • Heavy industrie
• Music  • Speech recognition
• Expert system • Automatic programming
• Natural language processing
o Expert system
An expert system is a computer program designed to act as
an expert in a particular domain.

o Natural language processing

Goal is to enable people and computers to communicte in a
natural language rather in a computer language.

o Speech recognition
Goal is to allow computers to understand human speech.
Computer vision
AI helps computer to understand what they see through attached cameras.

Robotics
A robot is an electro-mechanical device that can be programmed to perform manual tasks. An
intelligent robot includes some kind of sensory apparatus that allows it to respond to  change
in its environment.

Antivirus
Artificial intelligence has played an important role in antivirus detection.

Heavy industries
Huge machines involve risk in their maintenance and working,
 so artificial intelligence becomes an important part in their
operation.
CHALLENGES OF AI

• JUDGING PWER
• INITUITIVE THINKING
• COMPUTING POWER
• TOLERANCE POWER
TECHNIQUES:

FUZZY LOGICS ALGORITHMS DATA FUSION

ARTIFICIAL HYBRID
EC
SENSES SYSTEM

NEURAL
NETWORK
  ARTIFICIAL NEURAL NETWORK

• An ANN is a simplified and computer version of human brain. Human brain has hundreds of thousand
neurons with many interconnected links. The number of neurons and the modality of their relationships
determine the strength of intelligence 

•  The network will be able to predict the outputs which are nearly the same as the expected output of
any given input values. There are three learning algorithms, namely supervised, unsupervised, and
reinforced algorithms (Chaturvedi, 2008). By supervised algorithm a set of known input-output data
pairs (i.e., usually the experimental data) is given to the network. 
FUZZY LOGIC

• Fuzzy logic refers to mathematization of human reasoning. Hence, it is viewed as the

accommodation with the pervasive reality of fuzziness and vagueness in human cognition
(Zadeh, 1975).

• Basically, fuzzy set theory concerns with mathematizing and quantifying natural language.

• Development of a fuzzy system requires determining and adjusting its main components
EVOLUTIONARY COMPUTING
• EC is an apt term for describing a range of problem-solving algorithms and it was inspired by biological
evolution. In other words, EC refers to the biological processes that allow populations of organisms to
adapt to their surrounding environment, namely, genetic inheritance and natural selection.
     (Tettamanzi and Tomassini, 2001). 

• EC has been mostly utilized for solving optimization problems. The main ability of EC is due to its
iterative progress mimicking growth or development in a population.
      (Volna, 2013).
SVM

• SVM is a training algorithm that discovers and learns the classification and regression rules of
the input data. SVMs are based on statistical learning theory which aims at solving the
problem at hand and not the overall more sophisticated one (Boswell, 2002). 
• The seminal components of SVMs are the techniques of mathematical programming and
kernel functions.
• Advantage of SVM is that VC dimension can be calculated explicitly while in other learning
methods, there is no measure (Boswell, 2002).
HYBRID INTELLIGENT SYSTEMS

• Combining two or more intelligent methods can provide hybrid systems capable of
cooperating in human decisions
•  The mixture of EC and ANNs not only brings to the ANNs the capabilities of learning and
evolving, and provides stronger intelligence, but also can solve the issues related to the
design and realization of ANNs (Ding et al., 2013). Evolutionary fuzzy systems are the
successful result of hybridization of fuzzy logic and EC. 
• Hybrid systems benefit from both the management of uncertainty and inherent
interpretability of fuzzy systems with the adaptation capabilities of EC algorithms (Abraham,
2005).
FUSION OF ARTIFICIAL SENSES

• The development of alternative methods for the objective, real-time assessment of food

products in a reliable and cost-effective manner is highly desirable .
     (Ghasemi-Varnamkhasti et al. 2010).
•  In this regard, advances in sensor technology, electronics, biochemistry and artificial
intelligence have made it possible to develop  instruments such as E-nose, E-tongue and CVS
capable of measuring and characterizing  quality factors such as flavor, color, and chemical
components of various products . 
     (Wilson and Baietto, 2009).
DATA FUSION TECHNIQUES

• information from associated databases. Fusing multiple information sources together  produces a more
efficient representation of the data and a higher reliability by using data 199 from multiple distributed
sources (Luzuriaga, 1999).
•  Generally, data fusion methods can  be categorized into three levels; Low-Level Fusion (LLF),
Intermediate-Level Fusion  (ILF) and High-Level Fusion (HLF) (Korel et al. 2001) depending how data
issued from different instruments are inferred from one to the other.  

•  (Korel et al. 2001)

SPECTROSCOPY

• Spectroscopy is the study of the interaction between matter and radiated energy. There are
two main methods of spectroscopy, including Ultraviolet-Visible (UV-Vis) and Infrared (IR)
spectroscopy. While absorption of relatively high-energy UV light results in electronic
excitation, absorption of lower energy IR radiation causes vibrational and rotational excitation
of functional groups of the molecules. Both these methods have been used as supplementary
instruments of fusion technique including artificial senses and spectroscopy methods. 
USES IN FOOD INDUSTRY

Production process Sorting food 

Food quality Food hygeine

PRODUCTION PROCESS
•  ANNs have the advantage of processing noisy and distorted data, as well as the data of high
variation. ANNs not only provide fast processing and monitoring, but also reduce the
production costs . Various applications of ANNs can be divided into two main categories of
classification and prediction.
     (Hosu et al., 2014; Mohammadi et al. 2014).
•  Vibration of commercially available intact and damaged polyethylene crates of beverages
was analyzed (Zacharias et al., 2004). In this work, feed-forward multilayer ANNs
were successfully used for predicting damages correctly in more than 99% of all cases .
 Nourbakhsh et al. (2014) used a three-layer perceptron to predict the permeate flux of red
plum juice during membrane clarification.

In a research on green tea, the extraction of polyphenols was modeled and optimized by a
three-layer ANN (Xi et al., 2013). 

Ding et al. (2015) utilized an electronic nose instrument for rapid detection of penicillin G
in goat milk. They used MLP to predict the penicillin G concentration that led to an
accuracy of 94.9%. 

Viejo et al. (2016) assessed beer quality based on sensory properties including foamability,
bubble size, alcohol content, temperature, carbon dioxide release, and beer color. They
used ANNs for predicting the properties resulted in accuracies of about 92.4%.
 Elman network for the prediction of optimum fermentation time of black tea based on
electronic nose data
 (Ghosh et al., 2017). 

Gomes et al. (2017) determined pH and anthocyanin content of wine grape by using a three-
layer feed-forward network

One of the reasons for the wide use of ANNs in a variety of applications in beverage science is
that they are able to process large quantities of data quickly and deeply. This has made neural
networks a practical tool for the prediction of future cases and automatizing of beverage
production processes.
Fuzzy logic has contributed to beverage science by handling both numerical data and
linguistic knowledge at the same time. Fuzzy logic can develop models representing
the relationships between the customer preferences and sensory features. It should
be noted that the higher the number of decision parameters, the more complex the
design of the fuzzy logic rules. Successively, reaching a decision will be difficult
because of a large number of rules
 (Nasiri and Huang, 2008). 

Fuzzy logic plays a main role in the control of biotechnological and food processes
and has been successfully utilized to address uncertainty matters in the beverage
science.

Dutta et al. (2003b) analyzed different tea samples of various qualities using
electronic nose. 
Many researchers have used hybrid intelligent systems in beverage science and technology.
 Das et al. (1999) trained a neural network using a fuzzified training set for the discrimination of
alcohols and alcoholic beverages. The proposed fuzzy version of neural networks showed
successful improvement in discrimination of alcohols and alcoholic beverages (i.e., 62.5%–95%).

 Zhu et al. (2004) studied the optimization of operating conditions in rice heat blast process for
Chinese rice wine.

Roy et al. (2013) used fuzzy neural networks for black tea classification. The models were
developed based on the fusion of electronic nose and tongue responses. They reported that the
fuzzy neural network was the best suited model for tea classification compared with other
conventional models (Roy et al., 2013). 

 Optimization of whole-milk powder processing variables was done by using a hybrid model of
neural networks and GAs (Koc et al., 2007). 
Villarrubia et al. (2017) proposed a hybrid intelligent model as an accurate, safe, and reliable
method for odor classification in beverage science. 

Recent applications of artificial senses to food analysis using the

fusion approach
  
A combined approach based on a multisensor system, E-nose and E-tongue, to obtain 
chemical information from liquid samples through the analysis of the solution and its 
head space was illustrated by Natale et al. (2000). They tested the combined system in two
different experiments in the fields of clinical (Urine samples) and food (Milk 235 
samples) analysis. 

For characterizing wines, a fusion system combining three sensory modalities: an E-nose, 239 
E-tongue, and an optical system (CVS to measure color) called the “electronic panel,” 240 
was designed by Rodríguez-Méndez et al. (2004). 
In the case of olive oil, Natale et al. (2001) applied two E-nose systems based on
different sensor technologies (quartz resonators and metal-oxide chemoresistors)
to measure quality of farmed and commercial olive oil samples. 

A hybrid E-tongue based on data fusion of two different sensor families was
developed and used by Haddi et al. (2011) to recognize three types of beer. 

Spectroscopy provides a very rapid and non-destructive way to detect spoilage in

poultry (Goodacre 2002).

Extrusion cooking control
 Fuzzy control of a food extruder was suggested in 1986 , and the  applications of
fuzzy logic and expert knowledge in the modelling and control of extrusion
cooking has been described in detail before .
NIR spectroscopy and CVS, two green analytical technologies, were combined to
discriminate tea grades (Chen et al. 2008). Four grades of tea were studied by data
fusion .

Thus, simultaneous utilization of several instruments can provide better performance

than the individual devices. This is due to the fact that when one instrument, such as
E-nose, is used separately, it can only identify a few quality parameters. 

ESs can be useful in, for example, crop planning, process development, process
planning and synthesis, process diagnosis and control, plant supervision, data analysis
and market prediction. They can be very helpful in dangerous situations, can relieve
unskilled labourers from unpleasant and tiresome routines, and they can be built for
round-the-clock operation. 
CHALLENGES AND FUTURE TRENDS

• Artificial senses, CVS, E-nose and E-tongue, combined with multivariate data analysis, provide a non-
destructive, rapid, precise, efficient, repeatable, and consistent technique for food quality assessment.
The main advantage of this approach is the ability to mimic mammalian senses to directly assess
different components, simultaneously and to locate the spatial distribution of the components in the
tested product. CVS can be used for in process analysis of the organoleptic and sensorial attributes of
foodstuffs being processed through various facilities. 
• Further advances in E-nose and E-tongue systems and sensor technology along with high capability of
multivariate data analysis methods will show many interesting and promising perspectives in
monitoring, controlling and automating of food processing systems
•  (Ghasemi-Varnamkhasti et al. 2010). 
CONCLUSION

• Machine learning is improving  the food industry in many ways from how farmers determine
which seed to plant to what toppings you serve on pizza. Artificial intelligence has made
many new advancements and innovations in the food industry, many new varieties of the
same products have produced and products are multivariate . Due to its advanced learning
techniques (fuzzy logic, algorithms, data fusion and many others) it is used in food production
and processing food quality evaluation sorting food, food  safety and improving supply chain.
In near future, Artificial intelligence will soon decide what we eat and how we eat. Computer
vision will monitor food quality in your home. Also help in developing new products
    CHALLENGES

• Computing power
• Tolerance power
• Intuitive thinking
• Judging power

    
 Boswell, D. 2002. Introduction to Support Vector Machines. University of Carlifornia, San
Diego.

 Chaturvedi, D.K., 2008. Soft Computing: Techniques and Its Applications in

ElectricalEngineering. Springer, Warsaw.

 Ding, S., Li, H., Su, C., Yu, J., Jin, F.,2013. Evolutionary artificial neural networks: a
review. Artif. Intell. Rev. 1–10.

  Pannu, A., 2015. Artificial intelligence and its application in different areas. Artif. Intell. 4.

 Tettamanzi, A.G., Tomassini, M., 2001. Soft Computing: Integrating Evolutionary,     

Neural,and Fuzzy Systems. Springer-Verlag, Berlin. 

  Zadeh, L.A., 1975. Fuzzy logic and approximate reasoning. Synthese 30, 407–428. 
Linko, S. (1998). Expert systems—what can they do for the food industry? Trends in
Food Science and Technology, 9, 3-12.

Zadeh, L. (1965) `Fuzzy Sets' in Inform. Control 8, 338-353.

Ghasemi-Varnamkhasti, M., Mohtasebi, S.S., Rodriguez-Mendez, M.L., Lozano, J., Razavi, S.H., &
Ahmadi, H. (2011a). Potential application of electronic nose technology  in brewery. Trends in
Food Science & Technology, 22, 165-174.

Wilson, A.D., and Baietto, M. (2009). Applications and advances in electronic-nose technologies,
Sensors. 5099-5148. 

Laddi, A., Prakash, N. R., Sharma, S., Mondal, H. S., Kumar, A., & Kapur, P. (2012).  Significant
physical attributes affecting quality of Indian black (CTC) tea. Journal of Food Engineering, 113,
69-78.
Hosu, A., Cristea, V.-M., Cimpoiul, C. 2014. Analysis of total phenolic, flavonoids,
anthocyanins and tannins content in Romanian red wines: prediction of antioxidant activities
and classification of wines using artificial neural networks. Food Chem.150,113-118.

Mohammadi, V., Ghasemi-Varnamkhasti, M., Ebrahimi, R., Abbasvali, M. 2014. Ultrasonic

techniques for the milk production industry. Measurement 58, 93–102.

Fernandes, A.M., Franco, C., Mendes-Ferreira, A., Mendes-Faia, A., Da Costa, P.L., Melo-
Pinto, P. 2015. Brix, pH and anthocyanin content determination in whole Port wine grape
berries by hyperspectral imaging and neural networks. Comput. Electron. Agric. 115, 88–96.

Jiang, H., Chen, Q. 2015. Chemometric models for the quantitative descriptive sensory
properties of green tea (Camellia sinensis L.) using Fourier transform near infrared (FT-NIR)
spectroscopy. Food Anal. Methods 8, 954–962.
HHFGFHFJ

THANK YOU!