2. Concept
Optimization refers to changing one variable at a time so obtain
remedy of problematic solution.
Need
● Reduce cost of formulation
● Safety and reduce the error
● Save the time
● Reproducibility
● Innovation
3. Fundamentals
● Factor: assigned and independent variables. those are
qualitative and quantitative
● Level: designation assigned to the factor
● Response: outcome of an experiment
● Effect and interaction: visibility of effect is same when the
response occurs. Interaction refers to factors involved that are
dependent on each other
● Coding: process of transforming natural variable into a
nondimensional coded variable
4. ● Experimental design: vital properties of the experimental
performance based on assumptions made from the results of
the execution of experiments into organised away.
Types of experiment depends on model and objective of study.
Designs are based on the randomisation principle, replication
principle and error control principle.
Trials and runs are conducted on the basis of the type of designs
selected.
5. Response surface: it represents data or equations showing
interaction among dependent and independent variables
Example 3D graph between two independent variables and one
response variable
Mathematical models: these are simple algebraic equations
representing relationship among different variables.
Types: a) empirical - linear
b) theoretical - non linear
6. Parameters
Problem types
Constrained- with limitation
Unconstrained- no limitations
Variables
Independent- under control of
formulator
Dependent- not directly under
control, results of any changes
in process and formulation
7. Steps
1. Analyse and define the problem.
2. Depending upon previous knowledge preliminary choice can be
made.
3. Selection of model based on the results of the factor screening.
4. The experiments are designed and executed.
5. Anova is used to analyse the sponses for statistics.
6. The responses are screend by using multiple criteria to get the
values of independent variables.
8. Experimental design
● It gives precise information with minimal experimentation.
● It is a statistical design that describe or advises a set of
combination of variables.
● The number and layout of design points within the experiment
region depends on the number of effects that must be
estimated.
● Design of experiment may be defined as the strategy for setting
up experiments in such a manner that the required in formation
is obtained as efficiently and precisely as possible.
9. Experimental design
1. Completely randomised
2. Randmised block design
3. Factorial designs
a. Full factorial
b. Fractional factorial
4. Response surface design
a. Central composite designs
b. Box behnken designs
5. Adding centre points
6. Three level full factorial design
10. 1. Completely randomised designs- It compares the values of
response variables based on different levels of the primary
factor.
2. Randomised block design-blocking is technique used to reduce
experimental error. In this type there is one factor or variable
that is of primary interest.
3. Response surface design-generally support nonlinear and
quadratic response and capable of detecting curvature.
11. Central composite designs -
● Also known as Box Wilson design
● Combination of two level factorial points axial or star points and
a central point
● For nonlinear response requiring second order models
● Popular in response surface optimization
● Total number of factors combination= 2n+2n+1
12. Box-behnken design
● Requires only three levels (-1,0,1)
● Economical as compared to CCD
● Also called as orthogonal balanced incomplete block design.
Because every effect is not estimated in every block.
● Every effect is measured as equal number of times with a
balanced partition over the different blocks.
13. Objective of box-behnken design -
● The design should contain square the terms products of two
factors, linear terms and an intercept which suitably fit a
quadratic model.
● The ratio of the number of experimental points to the number of
coefficient in the quadratic model should be resonable.
● The estimation variance should more or less depend only on
the distance from the centre and should not vary too much
inside the smallest cube
14. Factorial design
● It let's study the effect that several factors can have on a
response.
● Varying the levels of all factors at the same time let us to study
the interaction between factors.
15. Full factorial design
● Responses are measured at all combinations of the factor
levels.
● Number of runs necessary=L❌F
number of levels❌number of factors
16. Fractional factorial design
● Experiments conducts only fraction of runs are selected
subsets in full factorial design.
● Good choice when resources are limited or the number of
factors in the design is large.
● Number of runs necessary=LF-1
L=Number of levels F=number of factors
● Full factorial designs contains twice as many design points as
the fractional design.
17. Types
of
fractional
factorial
designs
Homogeneous
-large number of factors
Mixed level
-factors having different levels
-higher level interaction can be
assumed to be negligible
Plackett Burman
-popular
-only the main effects are of
interest
-assuming all interactions are
negligible
-used to investigate n-1 variables
in an experiments proposing
experimental designs for more
than 7 factors.
Taguchi
-similar to PB designs
-minimizes variance
-categories of problems
● Static
● Dynamic
Latin square
-one treatment factor of
interest and two or more
blocking factors