International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-8 Issue-6, March 2020
4843
Retrieval Number: F8189038620/2020©BEIESP
DOI:10.35940/ijrte.F8189.038620
Published By:
Blue Eyes Intelligence Engineering
& Sciences Publication
An Extended Artificial Neural Network Assisted
Hybrid Harris Hawks and Whale Optimizer to Find
Optimal Solution for Engineering Design Problems
Someet Singh, Anuj Jain, Sunil Kumar Mahla
Abstract: The algorithms that have been developed recently
have decorous behavior to solve and find optimum solution to
various optimization problems in search space. Withal such
calculations stuck in issues nearby quest space for compelled
engineering problems. In succession to achieve an optimal
solution a hybrid algorithmic approach is proffered. Artificial
Neural Network (ANN) is considered as better solution for the
known outputs. A hybrid variant of applying ANN on Harris
Hawks and Whale Optimization Algorithm (ANNHHOWOA) is
proposed to achieve effective solution for engineering problems.
The effectiveness of proposed algorithm is tested for various non-
linear, non-convex and standard engineering problems and to
approve consequences of proposed algorithm standard
benchmarks and multidisciplinary design problems have been
considered. The validation endorsed that the results shown by
ANNHHOWOA showed much better results than individual
ANN, HHO and WOA and its effectiveness on multidisciplinary
engineering problems.
Keywords: meta-heuristic algorithms, artificial neural
network.
I. INTRODUCTION
Implementation and use of machine learning and artificial
intelligence has become popular and has been accepted as
best techniques over the last decade. Machine learning and
artificial intelligence are being widely used over the years
and provide efficient solutions to solve real world, discrete,
constrained or unconstrained, linear or nor linear
engineering problems. The continuous research and the
results produced from the experiments done in this field has
resulted that the available methods such as sequential
quadratic programming, quasi-Newton method has not
shown significant and effective behavior to find solutions
for non-continuous, multi-model problems. So, meta-
heuristic algorithms have been taken into consideration to
get accurate and efficient result of the defined problems.
Meta-algorithms have been implemented to achieve efficient
results for the natural, multi-modal and engineering
problems. Machine learning is one of the most exciting
recent technologies in Artificial Intelligence. Heuristics are
strategies to discover great ideal arrangements in a
computational expense without ensuring achievability or
optimality.
Revised Manuscript Received on March 30, 2020.
Someet Singh, School of Electronics and Electrical Engineering,
Lovely Professional University, Punjab, India
Anuj Jain, School of Electronics and Electrical Engineering, Lovely
Professional University, Punjab, India
Sunil Kumar Mahla, Department of Mechanical Engineering,
IKJPTU, Punjab, India
Neural systems are demonstrating strategies equipped for
displaying complex functions. Artificial neural networks are
naturally propelled simulations performed on the workstation
to perform particular tasks such as clustering, classification,
pattern recognition, statistical analysis and data modeling. A
neural system is an interrelated assemblage of basic
processing components or nodes, whose purpose is inexactly
in view of human neuron. Those transforming capacity of
the organize may be saved in the inter-unit association
strengths, alternately weights, gotten toward a transform
from claiming adjustment to, or taking in from, an set about
preparation designs. Neural Networks present an efficient
approach to compute and understand the working of human
brain. It takes numerous sources of info having various
weightings and has one yield which relies upon the
characterized inputs. Neural network is broadly utilized as a
result of its capacity to sum up and to react to unpredicted
inputs. A neural network does not need to be reprogrammed
as it learns itself. Throughout training, neurons are taught to
distinguish different particular designs what's more if on
start alternately not at that design may be gained.
A. Feedforward Neural Network
A feed forward neural network is an artificial neural system
wherein associations among the hubs don't prompt
development of cycle. In feed forward network the
progression of data is unidirectional, forward from the
information hubs to the yield nodes through concealed
layers.
B. Basic Architecture
Neural Network layers are autonomous and can have any
number of hubs. Count of concealed hubs must be greater
than all input hubs. The network must contain at least one
hidden layer. Value of bias nodes is always set to 1.
Figure 1: General Architecture