Sunteți pe pagina 1din 18

NEURAL NETWORKS IN FABRIC ENGINEERING

BY S.VIGNESHWARAN

CONTENTS
Engineering of knitted fabrics Introduction about ANN Types of neural networks and network selection Engineering of single jersey fabrics
Network creation Network parameters Training of network Network simulation Test set

Conclusion

Engineering of knitted fabrics


A scientific method of predicting the finishing targets Areal Density (GSM) Yarn count & Stitch length Length Courses per unit length Width Wales per unit length Performance - Shrinkage

What is ANN?
An artificial neural network (ANN), usually called "neural network" (NN), is computational model that simulates the structure and/or

functional aspects of biological neural networks It consists of an interconnected group of artificial neurons and processes information using a connectionist approach to computation.

Comparision with biological networks

HUMAN NEURON

ARTIFICIAL NEURON

Application of ANN Prediction (input-output) Pattern recognition Classification Curve fitting

Types of networks

Types of network
(1) Historically, the earliest ANNs are The Perceptron, proposed by the psychologist Frank Rosenblatt (2) The Artron (Statistical Switch-based ANN) to R. Lee (3) The Adaline (Adaptive Linear Neuron), due to B. Widrow, 1960). This artificial neuron is also known as the ALC (adaptive linear combiner). (4) The Madaline (Many Adaline), This is an ANN (network) formulation based on the Adaline above.

Cont..
Three later fundamental networks are: (5) The Back-Propagation network A multi-layer Perceptronbased ANN, giving an elegant solution to hidden-layers learning (6) The Hopfield Network. This network is different from the earlier four ANNs in many important aspects, especially in its recurrent feature of feedback between neurons. (7) The Counter-Propagation Network where Kohonen's SelfOrganizing Mapping (SOM) is utilized to facilitate unsupervised learning (absence of a teacher").

Engineering of single jersey fabric


For the following example back propagation type network is used. Input vectors and the corresponding output vectors are used to train a network until it can approximate a function, associate input vectors with specific output vectors, or classify input vectors in an appropriate way as defined by you. Properly-trained back propagation networks tend to give reasonable answers even when presented with inputs that they have never seen

Training set
count (tex) 30.92 30.92 20.22 20.22 20.22 17.68 17.68 15.03 15.03 15.03 Loop length in cm 0.293 0.32 0.265 0.303 0.314 0.253 0.262 0.231 0.269 0.249 Finish GSM 230 207 160 142 135 147 138 140 122 130 Wales/cm 14.1 12.78 14.71 12.87 12.41 16.21 15.65 18.63 15.97 17.3

Architecture of back propagation network used

Network parameters
Training function used: trainlm Transfer functions used Tansig hidden layer Purelin output layer Number of neurons in hidden layer 4 Performance function mse(mean square error) If the last layer of a multilayer network has sigmoid neurons, then the outputs of the network are limited to a small range.

Screen shot of training

Test set
200 190 180 170 160 150 140 130 120 110 100 100 120 140 160 180 200 Series1 Linear (Series1)

R = 0.960

COUNT (Ne)

REAL TIME PREDICTED LOOP LENGTH VALUES OF GSM VALUE OF GSM

24 30 40

0.32 0.28 0.26

168.2 151.6 124.6

163.4 153.8 126.23

17 16 15 14 13 12 11 10 10 11 12 13 14 15 16 17

R = 0.956

Series1 Linear (Series1)

COUNT (Ne)

LOOP LENGTH

REAL TIME VALUES PREDICTED VALUES OF WPC OF WPC

24 30 40

0.32 0.28 0.26

12.8 14.6 15.6

10.6 12.7 13.1

Conclusion
The supply chain of textile sector involves prediction of the end properties for a given input at most of its facilities(yarn & fabric). ANN widen the scope of forecast being flexible to any application. Since quality of product has become the important factor of survival in the globalised textile market, such an effective forecasting tool will play a key role to achieve the targets in a better way at various applications of the industry.

Thank u!!!!!!
vigneshwaransk@gmail.com

S-ar putea să vă placă și