AN IMPROVED MODEL FOR THE
ESTIMATION OF PRESSURE DROP
COEFFICIENT OF A CONICAL DIFFUSER
USING SIMULATED ANNEALING
Introduction
• A conical diffuser is a smoothly
expanding tubular section used to make
the transition from a tube or channel of
smaller cross section to a large one
thereby converting the kinetic energy of
flow into potential energy or of velocity
pressure into static pressure with
minimum total pressure losses
• Diffusers are used extensively in
centrifugal compressors, axial flow
compressors, ramjets, combustion
chambers, inlet portions of jet engines,
etc.
• The drawbacks of some previously existing models for estimating pressure
drop coefficients are that they are implicit in the flow regime and diffuser
geometry parameters, and they require extrapolation or interpolation of
values, combining two or more models and reading data off charts and tables.
These procedures are monotonous and time consuming and is not suitable for
automation.
• Recently, studies aimed at developing a single explicit model for estimating
Statement of pressure drop coefficient in a conical diffuser were conducted and the result
of a performance evaluation of the models developed indicate that some of
these models were okay enough for computer-based simulation.
problem • Mfon et al (2019) developed a single explicit model for estimating the
pressure drop coefficient of a conical diffuser. A performance evaluation of
this model revealed a maximum percentage relative absolute error of 21.86%.
This value is high and I believe that this model can be made more accurate
by utilizing global Optimization algorithms (specifically simulated
annealing), to proffer the best global parameters which will further reduce the
maximum relative errors of the model.
This project aims at improving on the semi-
empirical model for the estimation of pressure
drop coefficient developed by Mfon et al
(2019). This will be achieved by:
• Modifying the parameters of the
aforementioned model by using simulated
Aim and annealing algorithm.
objective • Evaluating the performance of the modified
model.
• Comparing the performance of the modified
model with the existing model.
• Akpan (2019) and Bassey (2019) both developed models for
the estimation of pressure drop coefficient using neural
Other networks, although their methods were different. While Akpan
(2019) developed a new model entirely, Bassey (2019) used
Artificial neural network to develop a new model for the shock
Existing coefficient only, and then substituted it into the model
developed by Mfon et al (2019).
Models for • The accuracies of this models were high but they had similar
the limitations due to the use of artificial neural networks.
• The limitations of these model are related neural networks.
Estimation of Since the neural network models use multiple equations to
generate models, a single explicit model cannot be obtained.
Pressure Drop The process required to retrieve these equations is very time
consuming and cumbersome. Therefore, the models developed
Coefficient. by Akpan (2019) and Bassey (2019) are not single explicit
models.
• Simulated annealing (SA) is a probabilistic technique for
approximating the global optimum of a given function (Simulated
annealing, 2021). It is one of the most flexible techniques
available for solving hard combinatorial problems (A
combinatorial problem consists in finding, among a finite set of
objects, one that satisfies a set of constraints).
Simulated • The name of the algorithm comes from annealing in metallurgy, a
technique involving heating and controlled cooling of a material to
increase the size of its crystals and reduce their defects in order to
annealing produce crystals of high quality.
• In order to apply the simulated annealing method to a specific
problem, the following parameters must be specified: the state
space, the energy (goal) function, E(), the candidate generator
procedure, neighbour(), the acceptance probability function P(),
the annealing schedule, temperature(), and initial temperature,
<init temp> (Simulated annealing, 2021)
SA has many advantages over other optimization
algorithms:
• First, because there is a non-zero probability of
accepting higher cost solutions in the search
process, SA avoids becoming stuck at some local
minima, unlike some greedy approaches.
Simulated • Also, the runtime of SA is controllable through the
cooling schedule. One can even abruptly terminate
annealing this algorithm.
• Finally, there is always a best-known solution
(cont.) available no matter how little time has elapsed in the
search process. With SA, the user can always get a
solution. In general, a longer runtime would result in
a better-quality solution. This flexibility explains
SA's wide popularity.
A Pseudocode of the Simulated Annealing
Algorithm
The materials needed will include:
• MATLAB R2015a
• Data containing the experimental values for the resistance coefficient will be obtained
from the literature.
The optimization process will be carried out using MATLAB. From the optimization toolbox,
simulated annealing will be selected. The simulated annealing algorithm performs the
Materials
following steps:
• The parameters of the algorithm (initial random set of variables, initial temperature,
cooling rate) are initialized and the ‘Energy’ e, of the initial set of variables is calculated.
and methods •
•
A random new set of variables is chosen and its ‘Energy’ e new, is evaluated.
If the new set of variables is an improvement, then it is accepted. If it is not an
improvement, then it is accepted with probability P (e, e new, T), where T is the current
temperature.
• The algorithm systematically lowers the temperature (Updating the temperature), storing
the best point found so far.
• Step 3 and 4 is repeated until it reaches a stopping criterion (i.e., Until “freezing” point is
reached).
