bioprocess modeling and optimization-part 1.pptx

7/27/2019 Bioprocess Modeling and Optimization-Part 1.pptx

1/13

Yusuf Hendrawan, STP. M.App. Life Sc.,

Ph.D

Bioprocess Modeling and Optimization

(Part-1)


2/13

Characteristics of food and bioprocesses

(1) They often involve drastic physical, chemical,and biological transformation of the material,during processing.

(2) Many of the transformations have not been

characterized, primarily because of the following:a) such a large variety of possible materials; b)their biological origin, variabilities are significant,even in the same material; c) because thematerial contains large amount of water, unlesstemperatures are low, there is always evaporationin the food matrix. This evaporation is hard tohandle in physics based models and increasescomplexity of the process; d) many food

processes involve coupling of different physicse. . microwave heatin involves heat transfer


3/13

Real-world

problem

Mathematical

model

Solution to

model

Solution to

real-world

problem

assumptions,

abstraction,data,simplifications

optimization

algorithm

interpretation

makes sense? change

the model,

assumptions?

A schematic view of modeling and optimization


4/13

What is a model? A model is an analog of a physical reality, typically

simpler and idealized. Models can be physical ormathematical and are created with the goal to gaininsight into the reality in a more convenient way.

Advantages:

(1) reduction of the number of experiments, thusreducing time and expenses;

(2) providing great insight into the process that may notbe possible with experimentation;

(3) process optimization; (4) predictive capability i.e. ways of performing what if

scenarios;

(5) providing improved process automation and controlcapabilities.


5/13

Need for understanding the

detailed mechanism

Availability of time and

resources, depending on the

state of a-priori knowledge of

the physiscs

Use fundamental lawsto develop physics-

based model

Validate model against

experimental data

Obtained experimental datato develop observation-based

model

Possibly validate against

additional experimental data

Extract knowledge from the

model using sensitivity

analysis

Use model in optimization

and control

Not reall

y

necessary

constrained

Strong need

available

Use


6/13

Modeling of Bioprocesses

Physics-basedObservation-

based

Techniques that can

be useful in either

model

Microscalee.g. Mol. Dynamics

Mesoscale

Macroscale

Fluid flow

Heat

transfer

Mass

transferHeat &

Mass

transfer

Classical

Statistical

Data

driven

Data

mining

Neural

Network

GeneticAlgorithm

Fractal

Analysis

Fuzzy

Logic

Response

surface

methodMultivariate

Analysis

Monte-

Carlo

Dimension

al AnalysisLinear

Programmi

ng

Kinetics


7/13

Physics-basedPhysics-based models follow fundamental physical laws

such as conservation of mass and energy and Newtons

laws of motion; however, empirical rate laws are needed to

apply the conservation laws at the macroscopic scale. For

example: to obtain temperatures using a physical-based

model, combine of energy with Fouriers law.

The biggest advantages of physics-based models are that

they provide insight into the physical process in a manner

that is more precise and more trustable, and the

parameters in such models are measurable, often using

available techniques.

The Advantages:

(1) They can be exact analogs of the physical process;

(2) They allow in-depth understanding of the physical

process as opposed to treating it as a black box;

(3) They allow us to see the effect of changing parametersmore easily.

The Disadvantages:

(1) High level of specialized technical background is

required;

(2) Generally more work is required to apply to real-life

problems;(3) Often longer development time and more resources are


8/13

Macroscale

Fluid flow

Heat

transfer

Mass

transferHeat &

Mass

transfer

Macroscale models primarily deal with transport

phenomenoa, i.e. fluid flow, heat transfer, and mass

transfer. These physicsbased models based on

fundamental physical laws. Typically, these models consist

of a governing equation that describes the physics of the

process along with equations that describe the condition at

the boundary of the system.

KineticsKinetic models mathematically describe rates of chemical

or microbiological reactions. They generally can be

considered to be physics-based. However, in complex

chemical and microbiological processes, as is true for food

and bioprocesses, the mechanisms are generally hard toobtain and are not always available. The kinetic models for

such systems are more data-driven than fundamental.


9/13

Observation-

based

The physics-based modeling process described before

assumes that a model is known, which is frequently difficult

to achieve in complex processes. Although a physics-

based model may also be adjusted based on measured

data, observation-based models are inferred primarily from

measured data. Observational models are black box

models to different degrees in relation to physics of the

process.

Physics-based models often require more specialized

training and/or longer development time. In some

application, detailed understanding provided by the

physics-based model may not be necessary. For example,

in process control, detailed physics-based models often are

not needed, and observation-based models can suffice.

Observation-based models can be extremely powerful in

providing a practical, useful relationship between input andoutput parameters for complex processes.ClassicalStatistical

The classical statistical models can have a model in mind

before obtaining the measured data. This makes them less

of a black box than models such as neural network or

genetic algorithm that are frequently completely data

driven., no prior assumption is made about the model and

no attempt is made to physical interpret the model

parameters once the model is built.


10/13

Response

surface

method

This is statistical technique that use regression analysis to

develop a relationship between the input and output

parameters by treating it as an optimization problem. This

method is quite popular in food applications.Multivariate

Analysis

MVA is a collection of statistical procedures that involve

observation and analysis of multiple measurements madeon one or several samples of items. MVA techniques are

classified in two categories: dependence and

interdependence methods.

In a dependence technique the dependent variable is

predicted or explained by independent variables.

In an Interdependence technique are not used forprediction purposes and are aimed at interpreting the

analysis output for the best representative model.


11/13

NeuralNetwork

Genetic

Algorithm

Fractal

Analysis

Fuzzy

Logic

Data mining refers to automatic searching of large volumes of

data to establish relationships and identify patterns. To do this,

data mining uses statistical techniques and other computing

methods such as machine learning and pattern recognition. It

can be seen as a meta tool that can combine a number of

modeling tools.

Data

mining

An Artificial Neural Network Model (as opposed to a biologicalneural network) is an interconnected group of functions

(equivalent to neurons or nerve cells in a biological system) that

can represent complex input-output relationships. The power of

neural networks lies in their ability to represent both linear and

nonlinear relationships and in their ability to learn these

relationships directly from the modeled data.Genetic Algorithm are search algorithms in a combinational

optimization problem that mimic the mechanism of the biological

evolution process based on genetic operators.

Fractal analysis uses the concepts from fractal geometry. It has

been primarily used to characterize surface microstructure (such

as roughness) in foods and to relate properties such as texture,oil absorption in frying, or the Darcy permeability of a gel to the

microstructure.Fuzzy logic is derived from fuzzy set theory that permits the

gradual assessment of the membership of elements in relation

to a set in contrast to the classical situation where an element

strictly belongs or does not belong to a set. It seems to be

successful for processes in which human reasoning and


12/13


13/13

bioprocess modeling and optimization-part 1.pptx

Documents