Pattern Recognition By: Artificial Neural Networking (Ann)

PATTERN
RECOGNITION BY
ARTIFICIAL NEURAL NETWORKING (ANN)
Project Assigned
 Detection of faulty flow meters through pattern

recognition using ANN.
 Detection of Oil leaks in pipeline through pattern

recognition using ANN
Pattern Recognition
 Pattern recognition is "the act of taking in raw

data and taking an action based on the category of
the pattern“
 Used to classify data (patterns) based either on prior

knowledge or on statistical information extracted
from the patterns.
Applications of Pattern Recognition
 REACTOR MODELING THROUGH IN SITU ADAPTIVE LEARNING

 PREDICTING PLANT STACK EMISSIONS TO MEET
ENVIRONMENTAL LIMITS
 PREDICTING FOULING / COKING IN FIRED HEATERS
 PREDICTING OPERATIONAL CREDITS
 FORECASTING PRICE CHANGES OF A COMPOSITE BASKET
OF COMMODITIES
 CORPORATE DEMOGRAPHIC TREND ANALYSIS
Wavelet transform
 The input data is collected by SCADA ( supervisory
control and data acquisition)
 This data is fed to Wavelet Transform
 A wavelet is a small wave which oscillates and
decays in the time domain
1 t  
 x  t 
 
CWT x ( , s )   x ( , s )  
  dt
s t  s 
Continuous wavelet
transform of the signal Scale = 1/frequency
x(t) using the analysis
wavelet (.)
 Different wavelets are matched with analysis
window
 Analysis windows of different lengths are used for
different frequencies:
 Analysis of high frequencies Use narrower windows
for better time resolution
 Analysis of low frequencies  Use wider windows for
better frequency resolution
 The function used to window the signal is called
the wavelet
Few wavelets
Haar Wavelet
Mexican Hat
Daubechies-4
Daubechies-10 Daubechies-40
Applications of wavelet transforms
 Compression
 De-noising
 Feature Extraction
 Discontinuity Detection
 Distribution Estimation
 Data analysis
 Biological data
 NDE data
 Financial data
Wavelets at work !!
Principal Component Analysis(PCA)
 Principal component analysis (PCA) -mathematical

procedure that transforms a number of possibly
correlated variables into a smaller number of
uncorrelated variables called principal components
 PCA is mathematically defined as an orthogonal linear
transformation that transforms the data to a new
coordinate system such that the greatest variance by any
projection of the data comes to lie on the first coordinate
(called the first principal component), the second greatest
variance on the second coordinate, and so on
Computing PCA
 Organize the data set- m x n Matrix
where m – No. of variables, N – no of data points
 Calculate the empirical mean – m x 1 matrix
 Calculate the deviations from the mean
 Find the covariance matrix
 Find the eigenvectors and eigenvalues of the covariance
matrix – m x m matrix
 Rearrange the eigenvectors and eigenvalues
-Sort the columns of the eigenvector matrix and eigenvalues
matrix in order of decreasing eigenvalues
 Compute the cumulative energy content for each
eigenvector
 Select a subset of the eigenvectors as basis vectors
 Convert the source data to z-scores
 Project the z-scores of the data onto the new basis
Y = W*Z
Where, Y - PCA matrix
W*- conjugate transpose of Eigen vector matrix
Z – z score matrix
Introduction To ANN
 Artificial neural network (ANN) is a mathematical
model or computational model simulating the structure
and/or functional aspects of biological neural networks
 ANN is an adaptive system -changes its structure based
on external or internal information during learning
Learning
 Neural nets mimic human learning processes.
 Nets are trained iteratively on input data along with
the corresponding target outcomes.
 After a sufficient number of training iterations, nets
learn to recognize patterns creating internal models of
the processes governing the data.
 Two different modes of adaptive learning
- Supervised Learning
- Unsupervised Learning
Supervised Learning
 Supervised learning is a learning technique for deducing a
function from training data.
 The training data consist of pairs of input objects (typically
vectors), and desired outputs
 During the training process, the differences between the
actual output from the net and the desired target outcomes)
are propagated backwards through the net and are used to
update the connecting weights.
 Repeated iterations of this operation result in a converged
set of weights and a net that has been trained to identify and
learn patterns
Unsupervised Learning
 Unsupervised learning involves extraction of

characteristic features from a large number of cases
and the subsequent organization of these cases into
groups sharing similar attributes
 Hence in this case , input data points and the
weight functions are provided .This type of
learning leads to cluster formation
 Different techniques used for unsupervised
learning are Radial Basis Function (RBF) and
Artificial resonance technique (ART)
Clustering at a glance !
Clustering

Pattern Recognition By: Artificial Neural Networking (Ann)

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Pattern Recognition By: Artificial Neural Networking (Ann)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Pattern Recognition By: Artificial Neural Networking (Ann)

Uploaded by

Copyright:

Available Formats

PATTERN

 Detection of faulty flow meters through pattern

 Detection of Oil leaks in pipeline through pattern

 Pattern recognition is "the act of taking in raw

 Used to classify data (patterns) based either on prior

 REACTOR MODELING THROUGH IN SITU ADAPTIVE LEARNING

 Principal component analysis (PCA) -mathematical

 Convert the source data to z-scores

 Project the z-scores of the data onto the new basis

 Unsupervised learning involves extraction of

You might also like