A-i
MINITAB calculates response tables and generates main effects and
interaction plots for: signal-to-noise ratios (S/N ratios, which provide a measure of
robustness) vs. the control factors.
Minitab can also be used for generating the layout of designs for two-level full
and fractional factorial designs using Stat > DOE > Factorial. Taguchi orthogonal
arrays can be generated using Stat > DOE> Taguchi.
1. Before you begin using MINITAB, you need to complete all pre-experimental
planning. For example, you need to choose control factors for the inner array
and noise factors for the outer array. Control factors are factors you can
control to optimize the process. Noise factors are factors that can influence the
performance of a system but are not under control during the intended use of
the product. Note that while you cannot control noise factors during the
process or product use, you need to be able to control noise factors for
experimentation purposes.
2. Use Create Taguchi Design to generate a Taguchi design (orthogonal array).
Or, use Define Custom Taguchi Design to create a design from data that you
already have in the worksheet. Define Custom Taguchi Design allows you to
specify which columns are your factors and signal factors. You can then easily
analyze the design and generate plots.
A-ii
3. After you create the design, you may use Modify Design to rename the
factors, change the factor levels, add a signal factor to a static design, ignore
an existing signal factor (treat the design as static), and add new levels to an
existing signal factor.
4. After you create the design, you may use Display Design to change the units
(coded or uncoded) in which MINITAB expresses the factors in the
worksheet.
5. Perform the experiment and collect the response data. Then, enter the data in
your MINITAB worksheet.
6. Use ANOVA to analyse the experimental data.
7. Use Predict Results to predict S/N ratios and response characteristics for
selected new factor settings.
a. Choose Stat > DOE > Taguchi > Create Taguchi Design.
b. If you want to see a summary of the Taguchi designs available, click Display
Available Designs. Click OK
A-iii
c. Under Type of Design, choose a design.
d. From Number of factors, choose a number. The choices available will vary
depending on what design you have chosen.
e. Click Designs.
The designs that display depend on the number of factors and levels in your design.
The signal factor values are repeated for every run of the Taguchi design
(orthogonal array).
f. In the Designs box, highlight the design you want to create.
g. Click OK even if you do not change any options. This selects the design and
brings you back to the main dialog box.
Naming factors
k. If you want to analyse a design or see whether or not selected interactions can
be estimated from the design, you must store it in the worksheet. By default,
MINITAB stores the design.
A-v
To select interactions
n. Move the interactions that you want to include in the design from Available
Terms to Selected Terms using the arrow buttons to move the interactions one
at a time, highlight an interaction, then click or to move all of the
interactions, click on or . You can also move an interaction by double-
clicking it.
o. Click OK.
A-vi
For performing Analysis of variance, Choose: Stat > ANOVA. This option allows
performing analysis of variance, testing for equality of variances and generating
various plots. The analysis can be carried out, using the suitable sub-option.
A-vii
Factor Plots subdialog box
A-viii
General Linear Model: TS versus N, S, F
A-ix
APPENDIX – B
NN Toolbox
>>nntool
It can also be open as shown below. It will open NN Network/ Data Manager Screen.
Getting Started
B-i
Design
Let P denote the input and T denote the target/output. In Matlab as per the
guidelines of implementation these are to be expressed in the form of matrices:
P = [0 0 1 1; 0 1 0 1]
T = [0 1 1 0]
To use a network first design it, then train it before start simulation.
B-ii
Providing target data
Create Network
Step-4: Now we try to create a XORNet. For this click on New Network. See
a screen like in the following figure. Now change all the parameters on the
screen to the values as indicated on the following screen:
B-iii
Setting network parameters
Make sure the parameters are as follows:
Network Type = Feedforword Backprop
Train Function = TRAINLM
Adaption Learning Function = LEARNGDM
Performance Function = MSE
Numbers of Layers = 2
Step-7: Now, highlight XORNet with DOUBLE click, then click on Train
button. You will get the following screen indicated in figure.
B-iv
Training network
B-v
Step-8: On Training Info, select P as Inputs, T as Targets.
Goal = 0.000000000000001
Max fail = 50
After, confirming all the parameters have been specified as indented, hit Train
Network.
Training process
Various Plots
Now we can get following plots Performance plot it should get a decaying plot
(since you are trying to minimize the error)
Performance plot
Training State Plot
Regression Plot
B-vi
Performance plot
Plots the training, validation, and test performances given the training record
TR returned by the function train.
Training state plot
B-vii
Regression plot
Now, the XORNet has been trained successfully and is ready for simulation.
B-viii
XORNet Structure
Network simulation
With trained network, simulation is a way of testing on the network to see if it
meets our expectation.
Step-9: Now, create a new test data S (with a matrix [1; 0] representing a set
of two inputs) on the NN Network Manager, follow the same procedure
indicated before (like for input P).
Step-10: HighLight XORNet again with one click, then click on the Simulate
button on the Network Manager. Select S as the Inputs, type in
ORNet_outputsSim as Outputs, then hit the Simulate Network button and
check the result of XORNet_outputSim on the NN Network Manager, by
clicking View.
This concludes the whole process of XOR network design, training & simulation.
B-ix
Simulated result
% Input welding parameters (Tool Rotatinal Speed; Welding speed; Axial force)
p=[1200,1200,1200,1200,1200,1200,1200,1200,1200,1600,1600,1600,1600,1600,
1600,1600,1600,1600,2000,2000,2000,2000,2000,2000,2000,2000,2000;48,48,48,60,
60,60,72,72,72,48,48,48,60,60,60,72,72,72,48,48,48,60,60,60,72,72,72;1.5,2.0,2.5,
1.5, 2.0,2.5,1.5,2.0,2.5,1.5,2.0,2.5,1.5,2.0,2.5,1.5,2.0,2.5,1.5,2.0,2.5,1.5,2.0,2.5];
t=[109.13,125.32,133.05,112.24,122.16,131.48,121.68,137.42,145.61,124.61,136.24,
145.48,122.68,132.61,143.37,124.89,139.61,152.00,115.52,127.89,136.13,131.48,
144.32,155.81,139.32,152.81,163.23];
[pn,minp,maxp,tn,mint,maxt] = premnmx(p,t);
% premnmx preprocesses the network training set by normalizing the input and
targets so that they fall in the interval of [-1,1]
% premnmx (p,t) takes these inputs,
% p-RxQ matrix of input (column) vectors
% t-Sxq matrix of target vectors
B-x
% and return,
% pn=RxQ matrix of normalized input vectors
%minp-Rx3 vector containing miminums for each p
%maxp- Rx3 vector containing miminums for each p
%tn-SxQ matrix of normalized input vectors
%minp-Sx3 vector containing miminums for each t
%maxp- Sx3 vector containing miminums for each t
% Size of input layers
numInputs = size(p,1);
B-xi
% Training the network model
[net,tr]=train(net,p,t);
% Simulation of normalized values
an = sim(net,pn);
% Simulation of output values
a = sim(net,p);
anew = postmnmx(a,mint,maxt);
B-xii