You are on page 1of 4

2017 International Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS)

SENSOR SELECTION IN AN AIRCRAFT ENGINE


USING GENETIC ALGORITHM
M. Lakshmi Dharani, S.J. Venkateswara Rao, D Vishnuvardhan
B. Abdul Rahim Dept. of Electronics and Communication Engineering
Dept. of Electronic and Communication Engineering JNTUA College of Engineering, Kalikiri
Annamacharya Institute of Technology and Sciences India
Rajampet, India. vishnushreyan@yahoo.com
lakshmidharani.dharani@gmail.com,
abdulrahimbepar@gmail.com

Abstract In aircraft engine large numbers of sensors are different components which are required to present
installed to provide health condition information for proper and redundancy of sensors and further require for proper selection
congenial operation. The aircraft engine is associated with to monitor aircraft engine health condition.
different types of sensors and actuators and to overcome the Here the problem is in the aircraft turbo fan engine having
safety critical related problems redundant components are
considerable number of components attached to it has to select
added. Redundancy provides reliability to the system and
improves the fault tolerance capability of an aircraft engine. As a suitable sensor. The engine consists of an inlet which supplies
of now, X-by-wire technology is replacing mechanical air to the fan [3]. The air that leaves the fan is separated into
components with large number of electronic components. The two streams, one moving into the engine core and the other
minimum of informative and cost effective sensors are selected to through the bypass duct [4]. The fan is driven by a low
monitor the optimum performance of the system. This selection is pressure turbine which is connected at the far end. The air that
effectively decided by utilizing optimization techniques. There passes through the engine core is made to pass through
are different types of techniques available for implementation. compressor and which is driven by a high pressure turbine [5].
For complex and large component based systems selection is an The fuel is injected and ignited in the combustor to produce hot
undaunting task and to overcome this problem, here we have
gas that drives the turbines connected [4]. The two air streams
used Genetic Algorithm based technique with problem solving
through travelling salesman problem.
are made to recombine in the augmenter duct, where additional
fuel may be added to increase the temperature. The air is
Index Terms Sensor Selection, Travelling Salesman Problem, passed through the nozzle from the augmenter at a high
Genetic Algorithm. velocity, thereby producing the required thrust, for which
nozzle has an adjustable cross section area. The three state
I. INTRODUCTION variables used are low pressure rotor speed, high pressure rotor
speed and average hot section metal temperature to simulate
Optimization is defined as an act or process or methodology and for which MATLAB is used.
for making a design or system or decision as fully perfect. The problem statement is designed with travelling salesman
Particularly the mathematical procedures involved to find the problem for better optimization and evaluation. The following
maximum of a function in the involved procedure. It is the section discusses the optimization and related techniques for
procedure applied to solve quantitative problems in disciplines the problem statement, particularly concentrating on Genetic
of physics, biology, engineering and economics. In a typical Algorithm. Genetic Algorithm is a search heuristic that mimics
optimization problem the goal is to find the values of the process of natural selection. This heuristic (also sometimes
controllable factors determining the behavior of a design or a called a meta-heuristic) is routinely used to generate useful
system. For example a physical production or an investment solutions to optimization and search problems[1]. Genetic
scheme etc,. The simplest of problems involving functions of a algorithms belong to the larger class of evolutionary algorithms
single variable as input factor and is solved with differential (EA), which generate solutions to optimization problems using
calculus [1]. There are different types of optimization techniques inspired by natural evolution, such as inheritance,
techniques like Ant Colony Algorithm, Partical Swam mutation, selection, and crossover. Genetic algorithms find
Optimization, Genetic Algorithm, Biogeography Based application in bioinformatics, phylogenetics, computational
Optimization, etc,. There are several optimization heuristics science, engineering, economics, chemistry, manufacturing,
can be used for design optimization most of them are often mathematics, physics, pharmacometrics and other fields [2].
developed from greedy methods [2]. Here we have used In computing, scheduling is the method by which threads,
Genetic Algorithm for investigation and would like to modify processes or data flows are given access to system resources
accordingly for better optimum results in sensor selection in (e.g. processor time, communications bandwidth). This is
aircraft engine. The preceding section explains or reviews the usually done to load balance and share system resources
2017 International Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS)

effectively or achieve a target quality of service. The need for a


scheduling algorithm arises from the requirement for most
modern systems to perform multitasking (executing more than
one process at a time) and multiplexing (transmit multiple data
streams simultaneously across a single physical channel)[3].
II. GENETIC ALGORITHM
Although the word optimization shares the same route as
optimal, it is rare for process of optimization to produce a truly
optimal solution. The optimized system or solution will be
optimal to produce or applied for only one application or for
one audience [6]. This may reduce the time at a price of
making it consume more memory. If the application where
memory space is the cost then one might deliberately chooses
algorithm which consumes less memory at cost of other
parameter. Often there is not a single algorithm which fits for
all designs and works well in all the cases. So the designers
make tradeoff in selection and optimize the attributes of great
interest. For perfect results the process has to be halted
otherwise the complete optimal solution to reach is far and Fig 1: Genetic Algorithm Flow Chart
cannot be estimated. Fortunately, it is often the cases that the The figure 1 above presents the basic steps required in
greatest improvements come early in the process [6]. Genetic algorithm and the basic operators that are used are
Genetic Algorithm is a search method based on principles Selection Operator, Crossover Operator and Mutation
of natural selection and genetics as said by Fraser, Operator [8]. The Selection Operator is used to select the fitter
Bremermann and Holland [7]. The Genetic Algorithm encodes chromosomes from the population for later breeding. The
decision variables of a search problem into finite length strings selected chromosomes are called parent chromosomes. The
of alphabets of certain cordiality [6]. The strings which are selection of the chromosomes is done on the basis of some
candidate solutions to the search problem are referred to as fitness criteria [9]. This criterion is usually formed depending
chromosomes, the alphabets are referred to as genes and the on the type of problem. In case of Traveling salesman problem
values of genes are called alleles. For travelling salesman the fitter chromosome would be the one whose value of the
problem the chromosome represents a route and the gene genes would be minimized. The Crossover Operator is another
represents the city. In contrast to traditional optimization basic operator used to combine two parent chromosomes in
techniques Genetic Algorithm works with coding of parameters order to get a new pair of child chromosomes [9]. The point at
rather than the parameters themselves. To evolve good which the chromosomes are changed is called the crossover
solutions and to implement natural selection, we need a site and finally the Mutation operator which performs after
measure for distinguishing good solutions from the bad crossover operation. It is used to preserve the genetic diversity
solutions. The measure can be an objective function that is a of chromosomes at each generation of population. It is carried
mathematical model or a computer simulation, or it can be a out by altering the genes of the chromosomes. Here we
subjective function to choose good solution over the worse. In randomly flip some genes in order to maintain diversity or we
essence, the fitness measure must determine a candidate can also exchange some genes present in a chromosome.
solution's relative fitness, which will subsequently be used to Genetic Algorithm has various advantages which made it
select good solutions [7]. Another important concept of Genetic more popular and some of the advantages and limitations are
Algorithm is the notion of population. Unlike the traditional listed below:
search methods, the genetic algorithms rely on a population of Advantages:
candidate solutions. The population size which usually a user 1. Does not require any derivative information, which may
specified parameter is one of the important factors affecting the not be available for many real world problems [10].
scalability and performance of genetic algorithm [6]. For 2. Compared to traditional methods it is faster and more
example, small population sizes might lead to premature efficient [10].
convergence and yield substandard solutions [7]. On the other 3. Has very good parallel processing capabilities hence can
hand, large population sizes lead to unnecessary expenditure of be used in multiprocessor based large complex systems
valuable computational time. where parallel operations are solicited.
4. Optimizes both continuous and discrete functions and also
multi-objective problems.
5. Provides a list of good solutions and not just a single
solution.
6. Useful when the search space is very large and there are a
large number of parameters are involved.
2017 International Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS)

Limitations:
1. It is not suitable for all problems, especially problems
which are simple and for which the derivative information
is available.
2. Fitness value is calculated repeatedly which might be
computationally expensive for some problems.
3. Being stochastic, there are no guarantees on the
optimality or the quality of the solution [10].
4. If not implemented properly, it may not converge to the
optimal solution.
III. RESULTS
From the results obtained it is clearly seen that the task
allocation or sensor selection for the aircraft engine is
optimized and found to be routed to a minimum level. All the
nodes selected are covered once i.e, for our investigation we
Figure 4. sensor locations and distance matrix for 40 sensors with 100
have considered 50 nodes distributed randomly around an area iterations.
and the salesman visits once with a total distance covered
obtained is 55.4781 units as shown in figure 2. And the best
solution obtained is maintained around 1000 iterations,
further, the iterations result in the same distance as can be seen
from the figure 3.

Figure 5. sensor locations and distance matrix for 45 sensors with 100
iterations.

Figure 2: Optimized task allocation to the sensors

Figure 3: Sensor Locations, Total Routing, Distance matrix and best solution
Figure 6. sensor locations and distance matrix for 50 sensors with 100
for sensor selection .
iterations
2017 International Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS)

REFERENCES
[1] Ranno Agrawal, Genetic Algorithms in data mining,
TABLE 1 Best Solutions for Different number of sensors achieved at IJARCSSE, pp.631-634, September 2015.
iterations [2] B Abdul Rahim and K Soundara Rajan, Genetic Algorithm
No.of sensors Best solution No.of based Task Scheduling in Time Triggered Architectures,
iterations IJVESP, vol 2(4), 2016.
25 sensors 39.9895 96 [3] V Hymavathi, B Abdul Rahim, and Fahimuddin Shaik,
30 sensors 44.1494 346 Biogeography based optimization approach for sensor
35 sensors 45.9269 452 selection in an aircraft engine, IJIRCCE, vol 3(4), pp. 176-
40 sensors 53.3653 819 182, April 2015.
45 sensors 58.7422 988 [4] Manish Kumar, Reliable and Efficient task scheduling
50 sensors 63.7490 912 algorithm based on Genetic Algorithm in cloud computing
environment, IJCCR, November 2015.
IV. CONCLUSIONS [5] Nitin Gupta, Randhir Singh, and Praveen Lehana, Texture
and quality enhancement of plants visual images using
As the aircraft engine is the most critical part in any Genetic Algorithm, IJCCR, March 2015.
aircraft, therefore monitoring the health conditions is solicited
[6] K. Sastry, D. Goldberg, and G. Kendall, Genetic
at any instant of time. The sensors attached for the said job has
Algorithm,pp.97-125.
to be healthy and function properly during the operation. Any
[7] Aparna Viswanath, V. Ramesh, and U.P, Sapna, Task
malfunction and defect in the sensors may lead to catastrophe,
Scheduling in homogeneous multiprocessor systems using
causing economic and life loss. The distribution of sensors over evolutionary techniques, IJETAE, February 2014.
the aircraft engine becomes complex and large; the selection
[8] M. Mayil Vaganan and G.S Geethamani, A Preliminary
needs to be addressed. Here to overcome this problem survey on Genetic Algorithm Techniques, IJCTT,pp.174-
optimization techniques are suggested and for investigation we 178, May 2015
have used meta-heuristic technique like Genetic Algorithm. It [9] Vijay Sharma, Mohit Kumar, and Kirthi Rohilla, Optimal
is adopted and found to be very helpful in attaining the said Design of digital integrator using Genetic Algorithm,
goal. It is found that the results obtained with large number of IJEIT,pp.102-105, May 2014.
sensors is congenial and can be adopted in real time [10] G. G. Priyanks and B Sonali, Genetic Algorithm for task
application. Further it is suggested to investigate in near future Scheduling in distributed heterogeneous system,
with other evolutionary methods or algorithms to achieve IJARCSSE, October 2012.
better solution and fast adoption is suggested with real time
hardware. .
ACKNOWLEDGMENT
The authors would like to thank Annamacharya Institute of
Technology and science, Rajampet for the Support in carrying
out this work.

You might also like