Soft Computing in Food Processing

1. Dr. R. T. Patil
Former Director CIPHET,
Pr. Scientist,
Central Institute of Agricultural
Engineering, Bhopal
Application of Soft Computing in Food
Processing Sector

2. Status of Food Processing Industries
 Size of food market in India - Rs. 8,60,000 Crores
 Primarily processed food market – Rs. 2,80,000 crore
 Value added processed food market – Rs. 1,80,000 crore
 Investment during the 10th plan is estimated at Rs. 62,105
Crores
 Industry growth rate during the last five years is estimated
at 7.14% against GDP of 6.2%
 Investment required during next ten years – Rs. 1,10,000
crores
Food Quality and Safety Issues are Prime
Important

3. Hard Computing Vs Soft Computing in
the Context of Food Processing
• Hard computing requires a precisely stated analytical
model and often a lot of computation time where as the role
model for soft computing is the human mind.
• In soft computing the tolerance for imprecision and
uncertainty is exploited to achieve tractability, lower cost,
high Machine Intelligence Quotient (MIQ) and economy of
communication
• Hard computing uses two-valued logic; soft
computing can use multivalued or fuzzy logic
• Hard computing requires exact input data; soft
computing can deal with ambiguous and noisy data
• Hard computing produces precise answers; soft
computing can yield approximate answers

4. Soft Computing
•Soft computing can be used to model and analyse very
complex problems where conventional methods have not
been able to produce cost-effective, analytical, or complete
solutions.
•In agricultural and biological engineering, researchers and
engineers have developed methods to analyse the operation
of food processing.
•Fuzzy Logic (FL),
•Artificial Neural Networks (ANNs),
•Genetic Algorithms (Gas),
•Bayesian Inference (BI),
•Decision Tree (DT), and
•Support Vector Machines (SVMs)

5. Food Processing Control
There are many parameters in food industry that must be taken into consideration
in parallel.
 A single sensory property like color or texture can be linked individually to
several dimensions recorded by the human brain.
 The food industry works with non-uniform, variable raw materials that, when
processed, should shaped into a product that satisfies a fixed standard.
 The process control of foods are highly non-linear and variables are coupled.
 In addition to the temperature changes during a heating or cooling process,
there are biochemical (nutrient, color, flavor, etc.) or microbial changes that
should be considered.
 The moisture in food is constantly fluctuating either loss or gain throughout the
process which can affect the flavor, texture, nutrients concentration and other
properties.
 Other properties of foods such as density, thermal and electrical conductivity,
specific heat, viscosity, permeability, and effective moisture diffusivity are often
a function of composition, temperature, and moisture content, and therefore
keep changing during the process.
 The system is also quite non-homogeneous and hence detailed input data are
not available.
 Often, irregular shapes are present.

6. Simple Fuzzy Logic Control for Food
Processing
1. Start with a proportional–integral–derivative (PID) controller.
2. Insert an equivalent, linear fuzzy controller.
3. Make it gradually nonlinear.

7. ANN for Crispness of Snack Foods
• The crispness was evaluated by
acoustic testing.
• The acoustic patterns were
generated by crushing the snack
samples with a pair of pincers
• The inputs for training the NNs
comprised 102 amplitudes of
sound signals in 0–7 kHz
frequency range at the intervals
of about 69 Hz with crispness
grades as outputs
• Probabilistic (PNN) models
showed good performance in
classifying the snack foods into
four grades of crispness.
• The prediction accuracy of
models ranged approximately
from 96 to 98%
Plot of average amplitude of acoustical signal spectrum for
different moisture content of Pringles potato chip samples

8. Meat Quality Using Hyperspectral Imaging
and Support Vector Machines.
• In order to predict the total viable
count (TVC) of bacteria of pork
meat, least square support vector
machines (LS-SVM) was adopted
as the modeling method.
• The prediction model based on
the optimal five wavelengths was
able to predict TVC with r = 0.87
and the result was considerably
better than that of ANNs and MLR
method.
• hyperspectral imaging system
coupled with the modeling method
based on LS-SVM is a valid
means for nondestructive
determination of TVC of pork
meat.

9. Parameter Estimation of Twin-Screw Food Extrusion
Process using Genetic Algorithms
The common approach is to determine
the operating conditions and then to
maintain these values as closely as
possible using various control loops, if
not manual control.
• GAs work with a coding of the
parameter set, not the
parameters themselves.
• Secondly, the algorithms search
from the population of points,
climbing many picks in parallel
• GA only require object function
values to guide their search, but
they have no need for derivative
or other auxiliary information.
• Algorithms use probabilistic
rather than deterministic
transition rules to guide their
search.
• Hence genetic algorithm is robust
and offers advantage over other
more commonly used
optimisation techniques

10. Bayesian Inference to Classify White
Grape Varieties
• The fusion method based on the Bayesian
inference was used to to combine the outputs
of various sensors for white grapes varieties.
• The sensors were aroma sensors, FT-IR and
UV spectrometers.
• Two methods were developed based on the
Bayesian inference: the Bayesian minimum
error fusion rule and the minimum risk rule.
• The effective fusion method lead to a
significant improvement in the grape variety
discrimination: the final misclassification error
was 4.7%, whereas the best individual sensor
(FT-IR) gave a misclassification error as 9.6%.
• Bayesian fusion proved to be very well suited
to the combination of all kinds of analytical
measurements with ability to cope with
sensors providing large, noisy and redundant
data as well as sensors having dissimilar
efficiency levels.

11. Decision Tree to Identify Food Additives
and Processing Aids
Question 1:
Does the definition of food additive
exclude the substance from being
a food additive?
Question 2:
Does use of the substance affect
one or more characteristics of the
food?
Question 3:
Does the substance become part of
the food?
Question 4:
Are residues of the substance in the
food "negligible" in accordance with
this policy?

12. Detection of Plant Diseases
• Electronic nose incorporating artificial intelligence
was used to detect plant disease, specifically basal
stem rot (BSR) disease that is caused by
Ganoderma boninense fungus affecting oil palm
plantations in South East Asia.
• The commercially available electronic nose,
Cyranose 320, as the front end sensors and artificial
neural networks for pattern recognition.
• The odour samples were captured on site and the
classification performed on a PC.
• The system was able to differentiate healthy and
infected oil palm with a high rate of accuracy.

13. Automatic Fruit and Vegetable
Classification from Images
• Face recognition, fingerprinting identification,
image categorization, and DNA sequencing is
high tech application
• The fusion approach was validated using a
multi-class fruit-and-vegetable categorization
task in a semi-controlled environment, such as
a distribution center
• The results show that the solution is able to
reduce the classification error in up to 15
percentage points with respect to the baseline.

14. ANN in image recognition and
classification of crop and weeds
• The images were taken . Colour index values were assigned to
the pixels of the indexed image and used as ANN inputs.
• There were 80 images, 100x100 pixels, for training, and 20
images for testing. Many back propagation ANN models were
developed with different numbers of PEs in their hidden and
various output layers.
• Six different evaluation schemes for two ANN output strategies
were used.
• The performance of the ANNs was compared and the success
rate for the identification of corn was observed to be as high as 80
to 100%, while the success rate for weed classification was as
high as 60 to 80%.
• The results indicated the potential of ANNs for fast image
recognition and classification.
• Fast image recognition and classification can be useful in the
control of real-world, site-specific herbicide application.

15. Soft Computing in Food Processing
Applications
2004 Brudzewski et al. Classification of milk by an electronic nose
2005 Pierna et al. Classification of modified starches
2006 Chen et al. Identification of tea varieties
2006 Onaran et al. Detection of underdeveloped hazenuts from fully developed
nuts
2006 Wang and Paliwal Discrimination of wheat classes
2007 Zhang et al. Differentiate individual fungal infected and healthy wheat
kernels.
2008 Fu et al. Quantification of vitamin C content in kiwifruit
2008 Kovacs et al. Prediction of different concentration classes of instant coffee
with electronic tongue
2008 Li et al. Classification of paddy seeds by harvest year
2008 Sun et al. On-line assessing internal quality of pears
2008 Wu et al. Identification of varieties of Chinese cabbage seeds
2009 Deng et al. Classification of intact and cracked eggs
2012 Jha et al. Method of determining maturity of intact mango in tree

16. Researchable Issues
•Online non destructive measurement of quality of
food grains, fruits and vegetables using NIR sensors
•Electronic nose to assess the quality and authenticity
of food products.
• Electronic tongue - for recognition (identification,
classification, discrimination), quantitative multi-
component analysis and artificial assessment of taste
and flavour of various liquids
•Affordable instrumentation for measurement of
spoilage of grain in bags and silos
•Smart labels of food packets to detect their shelf life
with automatically changing bar codes
•Simple gadgets like pH meter to detect pollutants in
drinking water

17. Conclusions
•No matter which soft computing method is used,
adaptive learning is essential to exploit the potential
synergy between methods.
•Another trend in soft computing applications is
likely to be the fusion of soft computing and hard
computing.
•Hard and soft computing fusion in agricultural and
biological engineering has just begun and hence
shows great potential for future research in this
sector.