Print ISSN: 1681-6900

Online ISSN: 2412-0758

Keywords : neural networks

Control on 3-D Fixable Wing Flutter Using an Adaptive Neural Controller

Mauwafak Ali Tawfik; Mohammed Idris Abu-Tabikh; Hayder Sabah Abd Al-Amir

Engineering and Technology Journal, 2012, Volume 30, Issue 16, Pages 2858-2874

An adaptive neural controller to control on flutter in 3-D flexible wing is
proposed. The aeroelastic model was based on the coupling between structure-of the
equivalent plate (wing) and the aerodynamic model that is based on a hybrid unsteady
panel methodTime domain simulations were used to examine the dynamic aeroelastic
instabilities of the system (e.g. the onset of flutter and limit cycle oscillation). The
structure of the controller consists of two models namely modified Elman neural
network (MENN) and feedforward multi-layer Perceptron (MLP). The MENN model
is trained with off-line and on-line stages to guarantee that the outputs of the model
accurately represent the plunge motion of the wing and this neural model acts as the
identifier. The feedforward neural controller is trained off-line and adaptive weights
are implemented on-line to find the generalized control action (function of addition
lift force), which controls the plunge motion of the wing. The general back
propagation algorithm is used to learn the feedforward neural controller and the
neural identifier. The simulation results show the effectiveness of the proposed
control algorithm; this is demonstrated by the minimized tracking error to zero
approximation with very acceptable settling time.

A Hybrid Neural Based Dynamic Branch Prediction Unit

Gheni A. Ali

Engineering and Technology Journal, 2012, Volume 30, Issue 6, Pages 1066-1081

Modern high performance processor architectures have come to depend upon
highly pipelined operation in order to achieve improvements in operating speed. As a result, the cost associated with flushing the pipeline and refilling it when a branch instruction is mis-predicted can significantly impact processor performance. Many schemes, from the extremely simple to the highly complex, have been proposed to
improve branch prediction accuracy. Conventional two-level branch predictors predict the outcome of a branch either based on the( local branch history) information, comprising the previous outcomes of a single branch (intra-branch correlation), or based on the (global branch history) information, comprising the previous outcomes of all branches (inter-branch correlation). The misprediction
rates for these predictors are very high when they predict branch instructions with hybrid correlations. In this paper we suggest a hybrid perceptron based predictor which employs up to 31-bits of both local and global branch history information to minimize the misprediction rates. The software written for simulation and testing
shows that the suggested hybrid predictor achieves a high accuracy. Our results shows that the best response of the predictor is obtained on history length of 16- bits.

Intrusion Detection and Attack Classifier Based on Three Techniques: A Comparative Study

Adel Sabry Issa; Adnan Mohsin Abdulazeez Brifcani

Engineering and Technology Journal, 2011, Volume 29, Issue 2, Pages 386-412

Different soft-computing based methods have been proposed in recent years
for the development of intrusion detection systems. The purpose of this work is to
development, implement and evaluate an anomaly off-line based intrusion
detection system using three techniques; data mining association rules, decision
trees, and artificial neural network, then comparing among them to decide which
technique is better in its performance for intrusion detection system. Several
methods have been proposed to modify these techniques to improve the
classification process. For association rules, the majority vote classifier was
modified to build a new classifier that can recognize anomalies. With decision
trees, ID3 algorithm was modified to deal not only with discreet values, but also
to deal with numerical values. For neural networks, a back-propagation algorithm
has been used as the learning algorithm with different number of input patterns
(118, 51, and 41) to introduce the important knowledge about the intruder to the
neural networks. Different types of normalization methods were applied on the
input patterns to speed up the learning process. The full 10% KDD Cup 99 train
dataset and the full correct test dataset are used in this work. The results of the
proposed techniques show that there is an improvement in the performance
comparing to the standard techniques, furthermore the Percentage of Successful
Prediction (PSP) and Cost Per Test (CPT) of neural networks and decision trees
are better than association rules. On the other hand, the training time for neural
network takes longer time than the decision trees.

Optimizing Opto-Electronic Cellular Neural Networks Using Bees Swarm Intelligent

Hanan A.R.Akkar

Engineering and Technology Journal, 2010, Volume 28, Issue 21, Pages 6237-6252

This paper presents an application of Bees algorithm to the optimization of cellular neural network for opto-electronics design, where cellular neural networks bees are a large – scale nonlinear analog circuit which processes signals in real time. It is made of massive cells, which communicate with each other directly only
through its nearest neighbors. Each bee cell is made of a linear capacitor, a nonlinear voltage controlled current source, and a few resistive linear circuit elements with photo diode and photo-detector for connections. In this paper application of bee cellular neural networks in pattern recognition is presented with its opto-electronic circuit design. It is found the real opto-electronic arrays, with all
their deficiencies are able to learn and perform various processing tasks well.

Design of a Neural Networks Linearization for Temperature Measurement System Based on Different Thermocouples Sensors Types

Ahmed Sabah Abdul Ameer Al-Araji

Engineering and Technology Journal, 2009, Volume 27, Issue 8, Pages 1622-1639

This paper describes an experimental method for the estimation of nonlinearity,
calibration and testing of the different types of thermocouples (J and K) using modified
Elman recurrent neural networks model based Back-Propagation Algorithms (BPA)
learning. Thermocouples sensors are nonlinear in behavior nature but require an output
that is linear. The linear behavior approximation is accepted, for a given accuracy level,
noise and measurement errors are always present. Therefore, neural networks techniques
are frequently required to minimize these effects. The problem of estimating the sensor’s
input–output characteristics is being increasingly tackled using software techniques such
as Turbo C++ language. A neural networks and a data acquisition parallel port interface
board with designed signal conditioning unit are used for data optimization and to collect
experimental data, respectively. After the successful training completion of the neural
networks, it is then used as a neural linearizer to calculate the temperature from the
thermocouple’s output voltage

Study of Principle Component Analysis and Learning Vector Quantization Genetic Neural Networks

Arif A. Al-Qassar; Mazin Z. Othman

Engineering and Technology Journal, 2009, Volume 27, Issue 2, Pages 321-331

In this work, the Genetic Algorithm (GA) is used to improve the performance of
Learning Vector Quantization Neural Network (LVQ-NN), simulation results show that
the GA algorithm works well in pattern recognition field and it converges much faster
than conventional competitive algorithm. Signature recognition system using LVQ-NN
trained with the competitive algorithm or genetic algorithm is proposed. This scheme
utilizes invariant moments adopted for extracting feature vectors as a preprocessing of
patterns and a single layer neural network (LVQ-NN) for pattern classification. A very
good result has been achieved using GA in this system. Moreover, the Principle
Component Analysis Neural Network (PCA-NN) which its learning technique is
classified as unsupervised learning is also enhanced by hybridization with the genetic
algorithm. Three algorithms were used to train the PCA-NN. These are Generalized
Hebbian Algorithm (GHA), proposed Genetic Algorithm and proposed Hybrid
Neural/Genetic Algorithm (HNGA).